<%BANNER%>

Information Theoretic Measures and Their Applications to Image Registration and Segmentation

xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20110209_AAAACO INGEST_TIME 2011-02-09T16:05:42Z PACKAGE UFE0015631_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 1087887 DFID F20110209_AABKQM ORIGIN DEPOSITOR PATH wang_f_Page_93.jp2 GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
e6ab889d17d37bd4fdd3dc9eac9d7078
SHA-1
9aae7ce61651629de2bc42b8f514735c51315a55
718319 F20110209_AABKPX wang_f_Page_61.jp2
cb682074a0b97aab385887d7b28d18f7
1ae68805cd5db32cb817a25ce29c115ece0c8820
3686 F20110209_AABKRA wang_f_Page_05thm.jpg
78d3f2c69289a99f425a69cb517c065e
a01709025edb1baa71c68462e8c343ca9e9ed6b3
294389 F20110209_AABKQN wang_f_Page_95.jp2
c92b0192cf75656afe98c984039cc0c9
ffa691c1305965ddc6efee170a0d82666074a6da
1087877 F20110209_AABKPY wang_f_Page_67.jp2
ff7bd1506f66c579cab0e567ec3841e1
b6830c849126c60fc3539884a6cb87b93a1c56c1
8967 F20110209_AABKRB wang_f_Page_43thm.jpg
dfda369fd59cef513edac01fbca01f2e
0362208f0e84448dd3e77c9877659a92299d3aa2
9184 F20110209_AABKQO wang_f_Page_68thm.jpg
f737cf2f80361e002b182e4fa1c934ff
0906ea0c1ca346536c07d59b695eb6f7ac4c2e05
1087898 F20110209_AABKPZ wang_f_Page_68.jp2
4c2bc52b92a4bf22eab6a6b5586040c4
3b9b49191d4636ecd7d7688c1db469ebbbb056ca
8176 F20110209_AABKRC wang_f_Page_75thm.jpg
d7c07c4645ac71fdbcab226c9a657313
da28a144dc854b649b9d18da83d35e3890a38842
32858 F20110209_AABKQP wang_f_Page_30.QC.jpg
7a4b88ea22c908c69875447eb984837d
5497edb7a02a45a0724f024af02b0747011e7ef8
24590 F20110209_AABKRD wang_f_Page_61.QC.jpg
8ac49844600c2a14afc2ffe59d529641
857d3391e2c5378a9c0be3150aab704ef1931a1c
8635 F20110209_AABKQQ wang_f_Page_39thm.jpg
693f844f7f2370fc2d1caf744bcbec6e
9741dd3e326a2de2c400a8e5de4c2b52a2423ea8
503 F20110209_AABKRE wang_f_Page_02thm.jpg
a6012d587071fa96af708a8ca5f8c0dd
acca8c6fc2dbe24c11d5d9e5a835500bf1f1df52
33890 F20110209_AABKQR wang_f_Page_15.QC.jpg
0ebe70808e7ded47b8c72580ffb7b17b
7d2bcaad902df567a1964572b0ce744910b0cf08
31720 F20110209_AABKRF wang_f_Page_45.QC.jpg
36adf03fc6bd4dfb131ef7d05d15eadf
0330718bef85dff89df4a94e511cc0ec984f2a58
2452 F20110209_AABKQS wang_f_Page_01thm.jpg
0b1af3dfe8731d3db6cab6e368a0bc4e
91ea18cb163ac694ae970580767517792aa86acf
37651 F20110209_AABKRG wang_f_Page_50.QC.jpg
090f0ca62e50a3f59ed83c55e38e6322
54dd6ff8892e7bb2c494262a0999ba4f787ff9b9
40061 F20110209_AABKRH wang_f_Page_93.QC.jpg
8eb0ab04d28e4b11362dfd16a2edadf8
12fd058005cb8fd7896c2e7df5dc59e1913d3671
40100 F20110209_AABKQT wang_f_Page_85.QC.jpg
e929f010cc7ff85ac06a98e5b4dded12
6095e61c40e8891e9b4f84daea8eae2bcdec5806
7606 F20110209_AABKRI wang_f_Page_35thm.jpg
d0c749298e9e4cca1280d3c03e84c829
05ca99778e3833e6c00b345a4bfac15ea04068a1
4603 F20110209_AABKQU wang_f_Page_94thm.jpg
62ca1a215db4c56a3ecf0627c436214f
fb98f99827ac232345f32a6b7560cb8195f41f33
30406 F20110209_AABKRJ wang_f_Page_76.QC.jpg
d0d264595b0934186e34dbd5c00f3e17
8b2d8709f52493ac4da0365f379c4a1fe9402ae8
7265 F20110209_AABKQV wang_f_Page_17thm.jpg
894703bdef0837a9b24fc68e3abfb9ab
439dc5a1ef634194f2f2a42137ed079d7bf51768
5100 F20110209_AABKRK wang_f_Page_87.QC.jpg
7857ab88a64eebe8c3b770f153c2e2c8
58abb38fb04f71896e2b8faf492a0165e6b0ba45
41728 F20110209_AABKQW wang_f_Page_27.QC.jpg
48ee55cd998102ff1b5da2dc09b5eede
436547b9074dd356857a549452001e2cce7daf0a
31644 F20110209_AABKRL wang_f_Page_78.QC.jpg
8bc10bff6c884df291eb954369568510
09348936f1731bd40c5b92214d688f98db33e5c7
8867 F20110209_AABKSA wang_f_Page_59thm.jpg
b146447189659cb0ad3884f8197cc004
245875ecaea836c1ab815aa0eb645b7ce01ebb64
8513 F20110209_AABKRM wang_f_Page_86thm.jpg
0681b351e3345fb0e618726d2cbc6fd6
acf42bec17bae0efc5fc52149bf60657779e0b3c
7586 F20110209_AABKQX wang_f_Page_88thm.jpg
6dbb3926becf01cc6a6d74bfc108442b
601620dcbf7975621fe6c73a22fe7447cbe75552
26309 F20110209_AABKSB wang_f_Page_32.QC.jpg
9bce80a4de5deeba0c0be1434610629c
10e081f0e3c41e1f1655b7225b250bb380fa0ffa
8621 F20110209_AABKRN wang_f_Page_49thm.jpg
4a1c1a79707839f4ae9368076e217c74
da4aaea472d014a5d28e165927ba5662c26d9ba2
7452 F20110209_AABKQY wang_f_Page_47thm.jpg
c07c1e28dae5b3fedd0efb903e61e07f
1e0926813f032d4a8d9b8ff67709b65bc453a400
6663 F20110209_AABKSC wang_f_Page_61thm.jpg
1d61a9b2dd7e8e3d09485d0b3c2013c8
146d07a54f5c1c328d2f8bc64944763101224c89
22437 F20110209_AABKRO wang_f_Page_77.QC.jpg
a50daf7ba592aed3dfdef3c83e4e0dc4
1d24f1bf1f1a998ea528fbe3b598a4e31a608bcf
8235 F20110209_AABKQZ wang_f_Page_04thm.jpg
8b5342aef3eafb962d8d351d268ae52b
292fbfc41fad7d191680aa632ea3965658b8b96a
40642 F20110209_AABKSD wang_f_Page_90.QC.jpg
89bd0408627a86cac3aeda6776172e57
6dd575b9594708caccb7429639c346d7aa456f28
6676 F20110209_AABKRP wang_f_Page_23thm.jpg
83890c3912ba04efb3f163c2c1f53641
6346c9c8a1a47c472ab62288832adb390a4336b2
6467 F20110209_AABKSE wang_f_Page_54thm.jpg
f0a4fd835891f2dc66b7630a21b6d4f5
ffd16d60a10e937700ed94dc4a8e2103adccbc46
5589 F20110209_AABKRQ wang_f_Page_34thm.jpg
3a5645437ce27033228bc6cb74c81554
8c40e40fffec7b09671211bb85fa5f1e9f22177d
9159 F20110209_AABKSF wang_f_Page_57thm.jpg
456e9a890bcd3627b8312f934498b916
8d6419384647fdc9550149741a2f613bbac1b0c5
7847 F20110209_AABKRR wang_f_Page_42thm.jpg
f9042b000c58dc1afbcf95c53cb30493
102964855c0f47cc843084cfd631397182054771
35794 F20110209_AABKSG wang_f_Page_29.QC.jpg
42316e6ad1e5883baa3dc6c54caa2582
cdce0568b3d35161e0832d8cd91993d2bf8b73dc
24123 F20110209_AABKRS wang_f_Page_23.QC.jpg
867c6349afbe5a03247b9322ec037d69
8e965b60fc883bb679e2b837c8b96d3ffe8a53b3
7374 F20110209_AABKSH wang_f_Page_13thm.jpg
670d177d3c16589a8e06073e697634de
9416a1b52140908894ee7920804dd28d730a41fa
24879 F20110209_AABKRT wang_f_Page_63.QC.jpg
43dddb31e9309654b7253e40d1f4bbdf
4270a331c9a9d3a8113fd6a5df2e61288d5f5e1b
22128 F20110209_AABKSI wang_f_Page_66.QC.jpg
5f3a243b0080a0da2e40b00cc30a3f84
c34ac48b172cfbfab9392f0d14c241ceb037d8b7
8107 F20110209_AABKRU wang_f_Page_52thm.jpg
e7fd781d46c0d67f6ccca20b3057b6d4
7bd0fc71bd49ed81475f1994673992f59e6b4d44
1620 F20110209_AABKSJ wang_f_Page_03.QC.jpg
1e45450860acf4a6a4e52d0209b5437b
d19973a21737d2f1a98c13668f20e8d2fe2ececa
38980 F20110209_AABKRV wang_f_Page_57.QC.jpg
73c83d18a89e92832c61e782262b699b
ad56e64f4423cb9b438413bf059b1e181606ccd1
9157 F20110209_AABKSK wang_f_Page_41thm.jpg
d3f97033da1733b065348cc8a8d95919
d2f69230cf6a69718276df6332303ec98594e727
8248 F20110209_AABKRW wang_f_Page_20thm.jpg
67a6261df6e0e209b18f927607ac08fe
459a0da01cd3cc737979a6509af828c6c2742bce
7184 F20110209_AABKSL wang_f_Page_14thm.jpg
03b939158c1bfebaeee4934276f9ff3d
c3a1b5a99760d7c2775085cc05d0c4041fbcae3b
5362 F20110209_AABKRX wang_f_Page_22thm.jpg
3e12a4e53c28231d4b4f9c207209ce16
fc8ad3cb93e6bbc109d97b5bd1b6cd85b1c57c9b
8225 F20110209_AABKSM wang_f_Page_80thm.jpg
dd9d27df2ec8162d03e12acdbba842ff
92b99b7cf6e136f65171ec6b3855428abc1af14d
20415 F20110209_AABKTA wang_f_Page_36.QC.jpg
0ea44e868e1fc6468ea86fd8df081365
ee08e4ed0623097b9a54abcefafe8e9d0cae5341
5986 F20110209_AABKSN wang_f_Page_77thm.jpg
d4ab69f2de4bf0b565a38994ee79b26e
5ee5c71f6a9a07774ccb5ea19dbb94154a41747a
2498 F20110209_AABKRY wang_f_Page_95thm.jpg
4f1eb984ba23524abc849bbde6f233f2
0eba443f2485e1852749d9684c0c0500b01d9f3c
37862 F20110209_AABKTB wang_f_Page_43.QC.jpg
912257776142e27d8173e95e7bfba58d
6409b8c7542284837543c47026d2da80d1aa1221
37023 F20110209_AABKSO wang_f_Page_56.QC.jpg
f68782bef497f3eaffb029c85396098e
18b8dc83943063015daf3f7eda328423470d1133
9023 F20110209_AABKRZ wang_f_Page_85thm.jpg
7615951d5b0b7229a285c0c63dea540f
c21e1325c9f76dbf091ae6688ce5bf1a78e0e5b0
33252 F20110209_AABKTC wang_f_Page_67.QC.jpg
a076e980507c00c752fbcfcd055ee77e
2bc55af21d58df6d20818da314e55505ba3df21e
9607 F20110209_AABKSP wang_f_Page_89thm.jpg
f52ecff88a49b39fae7786b32171d202
98f32aa68f2e9752361dfe2584556532d5988620
29602 F20110209_AABKTD wang_f_Page_73.QC.jpg
e8c54a2ed91ad3d2938ffb8ad6793384
f8c2d02cfcdd57e52d893299cce9eb03fc62a125
5908 F20110209_AABKSQ wang_f_Page_36thm.jpg
e230d34b3b3615386d56e942c6bce4c9
fa02f359e15bff0999def987007f684f97e30265
29358 F20110209_AABKTE wang_f_Page_84.QC.jpg
34cfccaeb58427028143f433ee7a1d7f
357673d2ecc4bf60b935c5eadc2e332f550a4ae8
34615 F20110209_AABKSR wang_f_Page_58.QC.jpg
327f5bfcfbcfce96041573a842b3e1a6
37f9b3bd39954645f82eeaee7364d57c45b48989
10105 F20110209_AABKTF wang_f_Page_95.QC.jpg
57de7ddf6aae2af943dd631c9c485f7d
3f5dcff46333e2ad24ad052244065cec8f26d032
7819 F20110209_AABKSS wang_f_Page_76thm.jpg
414858a19397d0524e08523d484f086f
c0d8dc74328be7f94b57edcb86efc3b3e4cacaa6
2250 F20110209_AABKTG wang_f_Page_08thm.jpg
0b33d8a94944278717533acbaa484c73
f82fcd8b08f1fb119898defa776dda77e3e6689f
37623 F20110209_AABKST wang_f_Page_86.QC.jpg
2eafa5af4cc17e193876f7f5a65dcab7
c1ad92fe453c2d3607842f6d5edc147696b9a764
7905 F20110209_AABKTH wang_f_Page_26thm.jpg
99b197f9a132cdf8779ad249a511cafa
8e749d958984dca0ef5378d6d04a5eb87adbf7a6
31919 F20110209_AABKSU wang_f_Page_44.QC.jpg
38b77cdeeaa92fd1087ae3643eb3d499
6695b4b09e035edb81c0cd5ff4567e3f49ae3705
8556 F20110209_AABKTI wang_f_Page_29thm.jpg
448f12732bd5bf3062550ad749e8aeaf
de184dcc7c1c802e1f4d3b4cce1baba2b60e744a
3267 F20110209_AABKSV wang_f_Page_25thm.jpg
11f5f5e7caef17ffccb503ced77127dd
d506084d66cfdedd0a12ce5fcfb400df975749bf
8040 F20110209_AABKTJ wang_f_Page_46thm.jpg
c2edef5b84a29dbbdc764418eb079695
9878be389ce1b06dbcc6fbd78727e03c932f2b68
152312 F20110209_AABKSW UFE0015631_00001.xml FULL
a353e5a5b44c838877bc092abcf6e32b
408f8b7faa31225c30c0b456daae2ad70823dc30
8612 F20110209_AABKTK wang_f_Page_56thm.jpg
b085f8a183a4d3acdc5641c9ac67aa10
d28555c05e5e834577b03c19640bae9e67e64916
40304 F20110209_AABKSX wang_f_Page_12.QC.jpg
0f81eb09ca2b76508b73050b3a70d53a
3af014167c742c6ea3c3120546f4de316a9c56fd
5596 F20110209_AABKTL wang_f_Page_66thm.jpg
f6c7207b741e75b984fa19eff99f6478
e16f56b0aa351904ad79e94c418392379a870b1d
12962 F20110209_AABKSY wang_f_Page_25.QC.jpg
a9a9a182857c966dc35f6d7460022b58
899d6859aaabdc79fed5e4b339107aba197b9069
7629 F20110209_AABKTM wang_f_Page_82thm.jpg
f46c83012e80d0623580f2a59d4fd5c7
8f6dd265b51e4d91ffaa571679c7feee77039746
30066 F20110209_AABKSZ wang_f_Page_31.QC.jpg
dae1b4dad160957407e9280a55e439e5
b5e782893d1f9aec3b14ef349b81a9a57940e789
13213 F20110209_AABJUA wang_f_Page_08.pro
e0e37f568a4a17fd3dfccea386fac76f
17953826763db44808d2f57961d26ebd6bdeaaa8
1087897 F20110209_AABJUB wang_f_Page_71.jp2
2b4496ae46c10c7d26bf3876fa2e4ed4
75fb495d35ac8f35e4be3cfa446d5edae363a392
377 F20110209_AABJUC wang_f_Page_83.txt
5d91202763071ea943b53ded185918d7
814341eecb4245f0d5bf327baf6a19a66cfe6ef4
8697 F20110209_AABJUD wang_f_Page_50thm.jpg
4dd1d65ac4b30770bae5676f33db0e94
da963ff60890bd92d8694808cc5cedf4b9bae078
8711878 F20110209_AABJUE wang_f_Page_85.tif
fb9d2c9bea919df9325776f54ff04cc6
c13cdb34b87284fc0954087bbf12f75689c9dcb3
26027 F20110209_AABJUF wang_f_Page_40.pro
6df484bbb232d46926ba17fed64dc75e
08531897b87eb9dac8dffc4ff45eb7ab62e4d08b
30645 F20110209_AABKAA wang_f_Page_13.QC.jpg
66efaf75fb500a1af04605dce10c89f9
fac4c08ad01ded542de2d67dd0db65388e593309
36118 F20110209_AABJUG wang_f_Page_79.QC.jpg
a7278500415b9643ee994a9a3c565c43
f3e1d9a8fb3d6eb7681c269bc2d2300798b9c09b
1087838 F20110209_AABKAB wang_f_Page_72.jp2
13a80e1270372ec8f231d7e5ce873db8
d4d9a93fed227a1698401690f958388a7dfa6b81
19297 F20110209_AABJUH wang_f_Page_34.QC.jpg
42ee5fc527a41ee8d76ed266b5cbe1e5
85b365cb71764c55132ff4b31b14b9aa37dc107b
2169 F20110209_AABJTT wang_f_Page_46.txt
95f43bfe5cc8de04c07b8bbcd82d4453
405c56c1736f03dbcdcf544b6ba6ef7e5f34d49c
F20110209_AABKAC wang_f_Page_41.tif
e0bb7c00ef49adebc207346cdcc7c4c0
44cb75b76db18a8888227208114e60cd564815d4
F20110209_AABJUI wang_f_Page_59.tif
f300f4e24f1a6aba9114225e95b8c685
8a0d2100287c5a43e22338f1b57fe23df8a82d4c
85047 F20110209_AABJTU wang_f_Page_18.jpg
9108856253e0de2715178ab003ea172d
de388692e16a4504e0a7aaf38cf4f0d8e520c0cf
69674 F20110209_AABKAD wang_f_Page_10.jp2
7af3e7f53754ef2df885f8a233a7aacf
1dc5ef65a333710681e6c6796700d549212f7d52
10135 F20110209_AABJUJ wang_f_Page_10.jpg
889917356fa61f635722094103e316ce
ed48960c8db6c442921f224a4279711e98d73d06
40323 F20110209_AABJTV wang_f_Page_91.QC.jpg
9e7b5c5a56567c28e23b017e8947d844
06616f837293a25164df72122cddef4f0e480bfc
33210 F20110209_AABKAE wang_f_Page_26.QC.jpg
595181cfb26736527287623266c9759f
0634dda019260874303bb46e4df65470f3abdc3a
1087870 F20110209_AABJUK wang_f_Page_90.jp2
7c40665d8e7a87922c498752673ffb81
70ef5379a09a9d37b16cb3b2e3f37663e2fc2a2a
48953 F20110209_AABJTW wang_f_Page_52.pro
0426b311ae1c51bfcbe42f31f1ce0ccc
ba8d1dcbfb3fa244b2e06e889400807ab663f02c
F20110209_AABKAF wang_f_Page_21.tif
36c1c3cd5ca6387cb0090d4a962f75ec
e80013b7bc9d376aa562b61db32c3841ae761a95
53992 F20110209_AABJUL wang_f_Page_46.pro
5b5a9265bddf4bb4f6ef03768f1b8283
c367270b4724a6e2b193b6e1dafbe25a664c143c
8075 F20110209_AABJTX wang_f_Page_72thm.jpg
a4d1144f95078a0159764181c602b9ee
ee600910edc5c2bf7dba68ca2355b940be1ce836
33527 F20110209_AABKAG wang_f_Page_23.pro
dcd8741cd044e5b900c2eaa5654e3f9f
22b8ee6a048d832e8b954bb99357831d9ef638b3
2210 F20110209_AABJUM wang_f_Page_69.txt
1f16e68170ef13bf184ecb8018f6fa4c
46c85c7c1a5dd2a95fafde6c91aaf089cba3022e
F20110209_AABJTY wang_f_Page_90.tif
2d436c9608db6d4ed2013a7f83bf4260
f6606534d3b26a86a88a41c35d41ee77c08debc9
8460 F20110209_AABJVA wang_f_Page_01.pro
9f29b6c678004a80625886a4f88651da
86f95190dbf73e59698ae8e0e662c2fdd7e410d7
1087900 F20110209_AABJUN wang_f_Page_20.jp2
9030da0989620bf54fe03913e4f47bdc
e4b30684237e3b01cb5cfea0b4460f790a044fea
2239 F20110209_AABJTZ wang_f_Page_57.txt
c43315db1efb92ffe6418428ad85831c
2e49292028359d657440fd76386a59bfc8154efc
30655 F20110209_AABKAH wang_f_Page_14.QC.jpg
5c2ce3cb78a36aad9731370ca9df9619
0a6708a433cff85451dde26ccdc4e044947ee435
4184 F20110209_AABJVB wang_f_Page_83thm.jpg
571c24c0456eb1a8714c31c9b5fa2709
a115b3a25f15beb7ef506730fe4833ee745dbb48
8331 F20110209_AABJUO wang_f_Page_58thm.jpg
340cfe663260fa205353e83314c4fd4b
103aea157b912b814696cacd5fd332998d14009e
84328 F20110209_AABKAI wang_f_Page_64.jpg
a54ecf716a9ada7becdf99dbf2dbf3e2
b068cd60ab3cb160b920b399d0fc9e63b57a5fe8
53479 F20110209_AABJVC wang_f_Page_38.pro
06e949b883d613838e98a0e973a118ca
0dedba4c7f591fd45803a8142f80bbd55d55c34e
19307 F20110209_AABJUP wang_f_Page_21.QC.jpg
450aa2e7bb8555dabcb6aa77ec6dd68e
b6285a43f3485c08592f7d666992cb9af826977d
1087867 F20110209_AABKAJ wang_f_Page_92.jp2
784b91c0fd7a8b10dfac52e09a827bd7
fd6f94b4b81ac4b02f00e60077af398f1b053f78
9426 F20110209_AABJVD wang_f_Page_91thm.jpg
81256a126fc38c0696b5d6970c97ecb3
ace9c5ed40d0f9f943d550247e5c064e4ed856f8
28977 F20110209_AABJUQ wang_f_Page_47.pro
821c6809b777ccd230221c551db5240b
a5e3f563637fd76a1f2f68f219426f67c09a308e
60797 F20110209_AABKAK wang_f_Page_12.pro
b279bc1f9c0a10288898ec1a6b43b653
2d86312d3fd7c55b23c92b382bea690dce44bbcb
55984 F20110209_AABJVE wang_f_Page_55.pro
d34efe9c16dffc6fa40886941ff9e999
5b0f9763a65de6cd466d4f584cf41e6a9f1cb9bd
2096 F20110209_AABJUR wang_f_Page_37.txt
07d0feb624be22a44dd4435eae230db8
c4f31394e04f17dd00d755eeb2bad3c1af9fdf28
8455 F20110209_AABKBA wang_f_Page_65thm.jpg
acf504eaa041031f44f534d9b58f9203
d9295776ba69d318c48707f55f6af9bb20e418d8
36059 F20110209_AABKAL wang_f_Page_61.pro
0de6b88f1d68fb36c67228cecf15506c
db8e4974bea9b26f555f57054d032cbb5d4df5ca
86891 F20110209_AABJVF wang_f_Page_24.jpg
1cf155427b4315d499885ec6423b99ee
a0c87953f59b8c91280b463c321263e52f4555a1
35437 F20110209_AABJUS wang_f_Page_04.QC.jpg
ef4b3d2949ddb82ebd3fe5d06f78dce4
4b6159819e18145465f4393fa4db510f12133ff4
102143 F20110209_AABKBB wang_f_Page_39.jpg
8bc77aa93bfef281d0705cf903415e5e
49d4394d7e725c5eb3b27b9aeee41dfa3b7796dd
F20110209_AABKAM wang_f_Page_78.tif
9a9af8d19afb55b1819dfbaabf4ce3b5
9aca9fd271c58c7ec82ecc7338bcc4d61c137620
1390 F20110209_AABJVG wang_f_Page_66.txt
dea9abd169dcffc44074f8e8dc9997c6
0878dbeeafdb2f86bde43d7b77a6a73420e7fef5
1045951 F20110209_AABJUT wang_f_Page_74.jp2
1d7b6df39bdea124d1cb8253280bcd32
30d5aaf84317247de54154ccf730de6bea4f6bc0
121128 F20110209_AABKBC wang_f_Page_41.jpg
89b3f38e3baf1014781b146f1fb127fe
aa1c546c9bd0f7fd61885455fd2c3e2d1e268148
599 F20110209_AABKAN wang_f_Page_08.txt
35dc60ea1495bcbc688b55b691ec98e9
1cd2d6cda9a69ba28b7820914d80321977dcb881
8662 F20110209_AABJVH wang_f_Page_55thm.jpg
4915759d63cc6de5a1c2fd827b6e2874
8ae44964946872ed4040ddf9498901d569ea5b22
1941 F20110209_AABJUU wang_f_Page_88.txt
48ffb1bc0898d7e22f5a1b363694432b
658af0d6d796b4fcd9cb9332910cdf449f9c98a0
5859 F20110209_AABKBD wang_f_Page_87.pro
cb33cbd8dbb682c672416daefe92e739
76c340655f9924fa3fa8571c009ff24c71589167
F20110209_AABKAO wang_f_Page_60.tif
71b8b353efc13ca175854bf8efddf70f
98bd8e8845a07c312b23cbb8e19c68f456fdf7fb
F20110209_AABJVI wang_f_Page_69.tif
2d63a6356ff3275c027622afd6cb08d1
ed7908637e516bde267913ae2c3f6de0380ee3cc
40381 F20110209_AABJUV wang_f_Page_28.QC.jpg
2eb37983b68946e37a2946ed37112398
d84f487f66bad3af0d667fdf9a49bd7eb7149b09
30765 F20110209_AABKBE wang_f_Page_17.QC.jpg
6ac9b13f4235f2d3148fcc56cd84ba4b
a5a7c2df8279d49166d6263c24bbfff287cbe56e
1087876 F20110209_AABKAP wang_f_Page_51.jp2
4c1373a2ded60ffa0cef4db3818b44f4
edc7f47822bcfebe93c11d8d0eb4f44db0aa605c
1970 F20110209_AABJVJ wang_f_Page_78.txt
e687ba1a29b018153fc8fbe6496d6d5d
c7528d20a99e9304acb39e9d6f7622ce7d67968b
59625 F20110209_AABJUW wang_f_Page_09.pro
7e0514f7cce0a0c60f4e63cb7ef6cb2d
43dd469aba5b0d123ba275b7317870d871916bd3
34747 F20110209_AABKBF wang_f_Page_71.QC.jpg
36bfa82547c4f86a3e71821199badf46
f1a1781da68fe84ac3f51e435b9cdd0886f4bfca
F20110209_AABKAQ wang_f_Page_44.tif
d1040e9b701a41fb7de3ac3672418141
547ad8c1caa316c8d8012b34b8d2cf6693a0b3d8
109770 F20110209_AABJVK wang_f_Page_80.jpg
20654241475e1b3433107dd2a5ec7ea6
cabac8fbaf33fb43f7dc130dc714568f46d979b0
34402 F20110209_AABJUX wang_f_Page_75.QC.jpg
bae16689ebce995bc4f04d36b3081c86
fb55c174f4864daaaf7fb30bd72ff2c4e413d3b2
108723 F20110209_AABKBG wang_f_Page_72.jpg
0558fd0e6a685e497f5a679b31ce5723
44f100e4a27f73731ebe112fff177a8b527d6043
39465 F20110209_AABKAR wang_f_Page_39.pro
abb22ca6b8e2dfa53ee805f3fff8ec6b
7f309b35691cf9b854a41ce41761c72927321b7e
842205 F20110209_AABJVL wang_f_Page_64.jp2
5667fa407f063a89e3fcfe18224c6693
2f503d999b09a72eccb070ecc1ff52a9b43e96c6
8258 F20110209_AABJUY wang_f_Page_83.pro
223ca716081cbf4529164c65cf8c9010
c51cf00b263960d22d9178b71c1eb56c910bc148
51038 F20110209_AABKBH wang_f_Page_58.pro
bfd8687292e72038e3a93461d60958d4
cae4d213f1e2c77b6f1d0d4ef04b23a1174a2186
F20110209_AABJWA wang_f_Page_04.tif
e0428726ad40b78405091b69865dba8b
a78b336a44404423e29d655359aa1a8518ace4e7
1464 F20110209_AABKAS wang_f_Page_87thm.jpg
751593af8a7cb9b1f34e85a60b7d3e5c
1cada0943769447a1572a2d44ded23be281b503f
1399 F20110209_AABJVM wang_f_Page_34.txt
104f5a754bdb64c7183732b1dd6e8f81
02cdea68ffd51a764294dd3f6ac77abcf5a2fa79
F20110209_AABJUZ wang_f_Page_24.tif
ddc356eb995414f99190d3fd9bdd7f0e
96311722fffe0c91dcd0eceb419d33b07e7c3c76
950800 F20110209_AABJWB wang_f_Page_84.jp2
a09b4c7b4e52dd91d60b421fc6f02353
6b11b215049a51873cf1e801c522efbf57535670
131071 F20110209_AABKAT wang_f_Page_27.jpg
663a0f88d3138477f6ff69540c940635
99f95901933700cd2069228e29f958efc75f74b5
F20110209_AABJVN wang_f_Page_51.tif
ba819f9463426ecd4f4cd4c3ba2bcaa2
d192cd4f633a87bf33d69ea3c07f4523273b5b2f
F20110209_AABKBI wang_f_Page_20.tif
e12b8b1d9976cc53b924eded4d2019f8
3987a4732878ee00b635febf9bcb8e23f0b703f0
2643531 F20110209_AABJWC wang_f.pdf
2ba79718a6ec78d5560ebb73a0f8d4bc
e95096d7adc49f55d67922e4f191f79a5f5d8b2b
36645 F20110209_AABKAU wang_f_Page_42.pro
324a5cfd64a5b1ad58da36f400a6eea1
ef918d37566c2ce7293ab3324215eed872681b0e
1087888 F20110209_AABJVO wang_f_Page_38.jp2
dbb0e17fc1e71f46f4de2942880b6891
619aac473e7d098ef8cf16be25de9ab6443f23c4
1087895 F20110209_AABKBJ wang_f_Page_12.jp2
70f361ffed0ad936eb0a632713cfa2c4
9644741fe42557678729ff4ab4a7865f1a22b671
559518 F20110209_AABJWD wang_f_Page_22.jp2
fbc585c5b15d3e5da9de44adc5b41af7
f6fbf8880ba826b527a9379898773aff8abb5742
8452 F20110209_AABKAV wang_f_Page_51thm.jpg
c66e701483062203edced3e5854ad4e8
07bf62962e21d80ddad6ad9f2e93335c36f0848e
2733 F20110209_AABJVP wang_f_Page_90.txt
667df6d8511b7282b5c57e896e2d5b43
80050dfdf491b50bb6646a8e5b82e7329fd0d2f2
F20110209_AABKBK wang_f_Page_23.tif
0fd7b3cd207287f559aff6f4eb45533b
10aa82be43139713bd5395812dde25bd987e96d7
899153 F20110209_AABJWE wang_f_Page_18.jp2
1bd14781917135f21b8c061315dffb03
fe2839272df176f1d8b112d3ff84c520c428e7fa
106146 F20110209_AABKAW wang_f_Page_75.jpg
d543ce1ad14252a9e5c66714b4bf6efb
aec4a18196d7911891eb1e61aab20583805a9bca
820361 F20110209_AABJVQ wang_f_Page_66.jp2
479a32148b6c798b94af8bce749133b7
dac1c970a1b6c9b2bf533ecb005607a0a27de84a
1856 F20110209_AABKBL wang_f_Page_24.txt
9751947cdf291db2e3fef6b87dcdfed0
7e269b3e5a69f243644c4938c83e712fc65322a7
607427 F20110209_AABJWF wang_f_Page_36.jp2
2824667865ac32f724b761933a2adf2e
0349d9f70aef2d7ba2eede2e06c0a3fbb19899f8
68604 F20110209_AABKAX wang_f_Page_54.jpg
c66615c82eb9cc7ba4cac401e90e15e9
21b8c41cadb2e59f6b4019a929e8c76ca1865a4a
1979 F20110209_AABJVR wang_f_Page_31.txt
f05c78ba0bfc3df8fb4aae03f1565bc0
4930454cbdfedfe8b85628fe75436e62e9499561
F20110209_AABKCA wang_f_Page_06.tif
4358a38bee140196b443fd416b21cce6
e133fa6c70fb4253809f680d4ee02098fc5f4449
1054023 F20110209_AABKBM wang_f_Page_52.jp2
9f7ea23887905a17cce180fb1562e486
7bfa9b030e8db631cc0a7ff0077cefe8362ce603
1055939 F20110209_AABJWG wang_f_Page_78.jp2
9b30ebcc9799e3a7999f3a644ed6105e
1f7b113de82124d1dfaaa67347f858d3a141d358
38566 F20110209_AABKAY wang_f_Page_92.QC.jpg
20a82c1ad87e8a5dce9404c3eee1f739
3eff2745fc68c20b0af526fbf051e71d5804f4df
9293 F20110209_AABJVS wang_f_Page_92thm.jpg
f7c629dac60e8a449f1b3eb1779fdde4
923c0ee8dcf18439c1ee3169dc5c4a637e7aafc8
8820 F20110209_AABKCB wang_f_Page_48thm.jpg
186a75841147a085985c98e6090cd2ac
a9b1d9f6c9b44d318a0d41147a34e7bb6e1410f0
5867 F20110209_AABKBN wang_f_Page_07thm.jpg
19fad0b751ebace30fbb90393bee9ef7
bd7d6b3d63bd9560d2e8d940f85cc779aed8254d
6371 F20110209_AABJWH wang_f_Page_32thm.jpg
ed1da13f1a578d36953e802895484523
010f87f229654160f3a1a996eda7d46428808ccd
F20110209_AABKAZ wang_f_Page_88.tif
ef030c648da1667cb3c987937b2564b9
130ed41d306a7f2419cfc9d70abf80ddafc3b398
F20110209_AABJVT wang_f_Page_86.tif
e926a254ff6e71939e2a36c05336b54c
29433a2f66fc145645429496355aa3ae39b13d2c
56487 F20110209_AABKCC wang_f_Page_50.pro
88d39ba93e091b58eb91fd25a9f0cfe6
28f7986cc5529e9bff47dc84f43ba0191a7d5dab
9404 F20110209_AABKBO wang_f_Page_93thm.jpg
a76cec01928d12c530ecc4a707063244
38f39ceca9e3f2b6f8de38b1fdbda840d474f33f
41630 F20110209_AABJWI wang_f_Page_35.pro
55522848ff0eef4d09aa8e8a38d354f8
320a1d8b6349de77a9a1853986875aeb906cd96a
8275 F20110209_AABJVU wang_f_Page_78thm.jpg
ba5a4a20b84369d62f5880a50dc1db48
5e47cb54b118cf273ecbe8200da93aa0df94d176
100866 F20110209_AABKCD wang_f_Page_45.jpg
7f9363913813d4e0c360a3a0f1ecbd51
7c1c21379348aa64f90bd1b70d49e3cfa131b63c
98382 F20110209_AABKBP wang_f_Page_82.jpg
fa9a76506366d886b0034120836dda5b
356f23b86d7951feb47def4327c27b5d0d248c06
66714 F20110209_AABJWJ wang_f_Page_93.pro
3cfebfc88ef1e23ee407c0ea990175c9
d45f3a3ff7aaab03196e18fe6c1c67cc7329a6f8
F20110209_AABJVV wang_f_Page_36.tif
ec3dbbca1969113ac1acbffb1fcea66e
19090d76246a14a884f234a8c2a395a126f07e55
2263 F20110209_AABKCE wang_f_Page_50.txt
85e8de32121ca52aeaa71d85b789ea89
c4ff5868400e60e2ea84d80311181f33bdeee73a
126177 F20110209_AABKBQ wang_f_Page_68.jpg
02b11b2558634a27065c94d44c9bd278
60b6ad4f97a65fd5d4361e29f616467580e1e553
37379 F20110209_AABJWK wang_f_Page_64.pro
f2acec04e998321a2587706d0f0d068b
ea03c505b9bbdfbc06bb79f5b22fc78baedf3e7c
5358 F20110209_AABJVW wang_f_Page_21thm.jpg
5e0f31a722f2976e0d822613f468f59b
bce5c8fa2cd83733e4e23e65734b33bf27715a57
49009 F20110209_AABKCF wang_f_Page_30.pro
3fee37ca3693c3b03ac5343e1c6d67d4
5c3be937eae67918cfd324ab92874cbab633c752
7049 F20110209_AABKBR wang_f_Page_11thm.jpg
1f115679bf2a23d650f93a5891e08c77
49ea1cbd64615c9c3427b0b873defab5c0aa9bc6
103682 F20110209_AABJWL wang_f_Page_52.jpg
19b2c4b4d735ae7c790fb70c4e722454
50b73385293489404099f44992d91a7bc39b8e12
246056 F20110209_AABJVX wang_f_Page_01.jp2
92ee576fc4c9e6ee838a8eedeb420791
0ecddee29dd02dc580a1359d7a24dc5633db07a6
28525 F20110209_AABKCG wang_f_Page_18.QC.jpg
d1b8fdf02409bbdb6421d125c4091deb
75534b5ca0e02caaecff7438d869d77aa1ad68d6
F20110209_AABJXA wang_f_Page_30.tif
30aa64c18f5eb5d43068bd2c4c01ebc1
b60d0e152e3084c7b7e5ce0ffffe87a10dbbbf25
F20110209_AABKBS wang_f_Page_10.tif
fa368389b2b044e86f987640ba159c8b
3f01668aa85941b711499b2fbfa71856228e7146
146838 F20110209_AABJWM wang_f_Page_93.jpg
a6b03076f0ff1848187ab24c0d389322
65a1ac216b86ddc66eaf10852cbb5ca70ba5f4e7
22321 F20110209_AABJVY wang_f_Page_54.QC.jpg
ed19c7de3d8713e277f0f1a8b93b5ba3
7ebf663ab4644a7e7e08860b8605736a9e2a1821
1561 F20110209_AABKCH wang_f_Page_79.txt
4b09e2a4f02527546143c3de861fce28
b8d3489eef3ea9392dbbdc4eb3f92c999e9ac5e9
F20110209_AABJXB wang_f_Page_47.tif
76af85ab9f5f7d2ebae5f97ccb3de41d
c9650467e8562f9712790f206afb40c52e3416a4
F20110209_AABKBT wang_f_Page_74.txt
7ddbf24966e942e8d0818b02008f8fc4
fc934951472e08a92f99978f2ab0b39a569694c1
9102 F20110209_AABJWN wang_f_Page_12thm.jpg
da642c1c8f0adf9c488c5afb3b55716d
942045bc30fb78b30879d59b14ec72a353bfe0cf
103634 F20110209_AABJVZ wang_f_Page_30.jpg
56cfff43e83415a8f431589394d70123
b3b39e1763615d7feb33c04bf9a9e06b62c153f7
121305 F20110209_AABKCI wang_f_Page_19.jpg
9b93db0d19aa9660bc36f38928e3800f
b9a9a6b1bdaa88639e8adb5c05e0308e0af43c36
120475 F20110209_AABKBU wang_f_Page_57.jpg
9d17c512c6c5db2b3bfe2b942cad5cd1
cdc2206c8b865ede38a77b6a65d382c60c6af13d
5190 F20110209_AABJWO wang_f_Page_06thm.jpg
19d66f3045dc8ec3b495e02a8d068a18
11a106fe06cca028a8bb3c4293dddf7cdeb0c85a
110714 F20110209_AABJXC wang_f_Page_65.jpg
784b1ae7793dc3ce416b8ba6205bcc90
d77f23bdeb8bc6ad1d51400745c564d07474539f
12995 F20110209_AABKBV wang_f_Page_83.QC.jpg
8a7e971e62976926c7f485f7e89193ef
2c450b80fab6db432e49b54efcc9a40e726668fe
45964 F20110209_AABJWP wang_f_Page_33.pro
f984c63c83510697e621e94a2ceded43
c44fcc70d75140ed1b97eda233ce7458b931d1d2
6843 F20110209_AABKCJ wang_f_Page_09thm.jpg
d1d8d514897ee608d7ccbd389389357c
83b7d16ef40d59975567b22df683aa7c5f6384d8
7684 F20110209_AABJXD wang_f_Page_33thm.jpg
8e03fa1a5dd461c1befa732621ddd829
256a426c3945889454fb692b30f6cbb40ed84a04
1087878 F20110209_AABKBW wang_f_Page_26.jp2
31ab4f192fe110a9618ea2f4300827b5
43743004fd38f71583643a8bd528d2c64596552b
F20110209_AABJWQ wang_f_Page_54.tif
79cb2bef872f0ae46c0bdf08157ed5cf
99a8cc90a706ec7cab59caf48366eaa11d318a6d
33054 F20110209_AABKCK wang_f_Page_52.QC.jpg
a5e3e8d0898c4d58a3f0343ed63a0e54
cf6f84e753bacd8e5362c303ab7854957d1701d9
97271 F20110209_AABJXE wang_f_Page_33.jpg
48199f5ed385d55c742f54fcf77e2a45
96a00d7c4892ca830ded448ca01f6b23a5dc58dc
8447 F20110209_AABKBX wang_f_Page_44thm.jpg
d01192519e5dac1938c3aabc30da57b2
6b6331c84bdc271612c3f19ad4dfc7b5b320627a
1806 F20110209_AABJWR wang_f_Page_17.txt
7200ad88e4b3e4eb0042e6c7d9f6884e
b0cfaa2ad26d0e9773b5b302a9fb8e0d4ff54ee2
F20110209_AABKDA wang_f_Page_87.tif
3293223bb3c2a1d7efda064d78462246
44055f02fc3be83759fd2dfb58aea5faac13dcd0
2386 F20110209_AABKCL wang_f_Page_12.txt
6fa6ad8c39b1e28449887305d8045291
2f6b5e969f598396fdcb00197c1f27259f718dad
F20110209_AABJXF wang_f_Page_29.tif
01c2a14fd525d196e7341f2d71e3c8a7
3c46605599740d4f846ecb5b71df2c34c4bf9e82
150350 F20110209_AABKBY wang_f_Page_90.jpg
7ed91eecae696ae7ee9c38cdbfc34706
59c13de56ad5fbfe5f3837a9f4d188bd400bd490
2307 F20110209_AABJWS wang_f_Page_85.txt
9c104f3df16b5db69ea9809a8f93ea5a
67fbcfd179e33bd91226282adb8008cf5478a603
F20110209_AABKDB wang_f_Page_91.tif
3a8463d21b7610f3f7eaba55026d811e
8a094b2fb3f28f5e355657d71ba1b62cd2289d2f
F20110209_AABKCM wang_f_Page_09.txt
991c07681b725f73ac71300f2e387b88
b6853d570bfce62ca3aedc5268cce94d1502df91
52606 F20110209_AABJXG wang_f_Page_51.pro
d9059a700de82ca6222dea26c8d6c4d8
ed9741f06f676558f165ebf39c05e41a05bea5d3
1119 F20110209_AABKBZ wang_f_Page_53.txt
6f75875e0405ff1d827438069c41c746
f9173fd8b134121592054158a787fb8172236bb8
F20110209_AABJWT wang_f_Page_34.tif
77c5224974d876c143fff8b8d3116a0f
2480da6861c9ef4fffa7b4a337e2eefd82b8444e
F20110209_AABKDC wang_f_Page_25.tif
fa3b2f244a932cd9a1176e3ae0f77e8f
d40ac55fabfecd0304913e3349e0ceeaf22af655
67563 F20110209_AABKCN wang_f_Page_91.pro
8e5633cc81e72ab4078682b60f07cd6d
2d1914d03f958f3a65743ed409c714fa538ad035
F20110209_AABJXH wang_f_Page_53.tif
218f0cb8f0ba6922a9f4d799d0cf9eb3
45c85dee9af3eb0a5604152d6607d220df6f11a1
33645 F20110209_AABJWU wang_f_Page_63.pro
eeddf3d6c14c6457eb3b2139376646ae
21c1a5a7ea343ff0e4e87e49e4b9e0046ae0b35e
F20110209_AABKDD wang_f_Page_27.jp2
59a520a2cb173c854e336587077c4391
12607ab39e0b115f118de0f25092311f6d8932d4
25166 F20110209_AABKCO wang_f_Page_22.pro
0197ad84449d83390149c135ac553aeb
e311bd5dca9adf4f8ed51fda89606ecca19c6de2
949 F20110209_AABJXI wang_f_Page_10thm.jpg
b8b3f4dcd6f198e0399513ab6cf65e3c
583601aa7931f819c073173d4b1c23df1bc3c299
74287 F20110209_AABJWV wang_f_Page_61.jpg
3d5b40a0fba5997f600ef367780d3ee8
00b5e334e06843e59f0d9e61d95a7394081dddcb
561651 F20110209_AABKDE wang_f_Page_21.jp2
b7761c80f818fb55d93a4d30cab52507
6331661945428d5bbaa406c681de64de4c54b133
2732 F20110209_AABKCP wang_f_Page_91.txt
83c3efe7d3e815b54893e7999a460d40
4e5deefd0675228d5aac557937f2539ecdbf3c32
468429 F20110209_AABJXJ wang_f_Page_05.jp2
c2ee2aecba3a39df841d4a4d2b408431
f8b0b84acd642c9d048ea72feda52796c9da15f8
32461 F20110209_AABJWW wang_f_Page_74.QC.jpg
fe33a54bb1d103c01c5afc778355c712
2b2f78537e500062c1400c6c3aecbf00138a5f04
90169 F20110209_AABKDF wang_f_Page_14.jpg
c56be5c5bb2aac651d0891b946728778
efa61bb26955633e8145b9596d5ee2958c2fd8f8
926259 F20110209_AABKCQ wang_f_Page_60.jp2
7e2e5d4ce1833eccf906a897e11f26a0
fa00746f156c20a6e2e2fea314c5592526a02d12
6686 F20110209_AABJXK wang_f_Page_16thm.jpg
63f7c3cc5c3b913b3f173b7266ef7db0
3dabfb8bf34ae4e3da74074081d611f60680da33
1874 F20110209_AABJWX wang_f_Page_45.txt
8b54c73a11f7ca7d3a135ba5f0a2b11c
052d9a299cc101edbec2031d39937c838cfc762d
43940 F20110209_AABKDG wang_f_Page_60.pro
92ab767c1519fa1b55a52a6e2fbabd7e
d91fd4591f811d56c974db8aef56e2fd7c196606
2691 F20110209_AABKCR wang_f_Page_32.txt
86176b6385c32519dfbab427ed3c8095
1e68edea7f99f4c6de228685063287622b36fd27
55347 F20110209_AABJXL wang_f_Page_43.pro
6f8c94df58d04d5ddf634c8bcbc704a7
62ccc24abe287d9a2c908aa4740163c565380f62
38369 F20110209_AABJWY wang_f_Page_19.QC.jpg
93fc6d5bd6bf9152c2a94a3c86ced4dc
ade2467903d03b41ac0a1cf083695def4de31755
1379 F20110209_AABKDH wang_f_Page_54.txt
3a615d440e7d63456e81aa3d0a51f36a
eac2df6409ea5d70aeeedbbb80bc81026d0f5b0b
36832 F20110209_AABJYA wang_f_Page_38.QC.jpg
cfaae018b0f1a9b3e554896ed836ca25
1c07883aeca993591b55e4d2ba3ed94978afde7a
2118 F20110209_AABKCS wang_f_Page_58.txt
02862ccb6f241c88d32268ad79b7f867
28f7d4ccdf7610cdf7c154c93b7e66d266c52da0
60260 F20110209_AABJXM wang_f_Page_34.jpg
891bc7b1029b1d0ba01c0d2d27dd1779
ab94c12db8ea6d2dfd02d7804dc67705d617292c
23384 F20110209_AABJWZ wang_f_Page_69.QC.jpg
85679ec1b26ba2151ccbb2bc418c83f1
e6cd59d2cec0ab5ec3bc2368afa9a499af3d79d8
26626 F20110209_AABKDI wang_f_Page_64.QC.jpg
9877320e86218f27d1090f959cff40b0
3d2ddaada88c157aab7d7f803437c149b9500623
F20110209_AABJYB wang_f_Page_42.tif
f68a8ec07079c14c8879862b1dde1d18
0cd89f5b3364755af60ca7503f164bfbf5a555be
35848 F20110209_AABKCT wang_f_Page_46.QC.jpg
a5ffd6db972e4c77c59324d61926652f
68bc68ea56537cca5bc6977cb386a57b12a7ada8
649391 F20110209_AABJXN wang_f_Page_94.jp2
88bd6d7315ffc99973f7678543b8a3fe
e0e8af3f1fb0dcc3a026d54eeb137243104643f3
9419 F20110209_AABKDJ wang_f_Page_90thm.jpg
83eb2b981fb8a992a118537bb4a52cc1
7c5477272730040db679e48f799158c3326aa55a
7533 F20110209_AABJYC wang_f_Page_81thm.jpg
45c24dca430ed74440f907c27205762f
4e5ec00ef6c1d520a3efb657ed055dddba600b4b
F20110209_AABKCU wang_f_Page_83.tif
2c4457a2d1f149458fc5d995b08245ef
a560c14ed75a2f6d120512b173f7cc3ae8174c78
30037 F20110209_AABJXO wang_f_Page_35.QC.jpg
1353b482c2caaa28ba7008b27b85030f
7d365f17862ac6813e8037995ff1749e39fd4df7
7746 F20110209_AABJYD wang_f_Page_60thm.jpg
c35515ee007e5580bc187235ac0888f1
3cd37c02c206155995a83c6b5d0a0aac0897cc1c
92031 F20110209_AABKCV wang_f_Page_42.jpg
beb32d104c545a07c29aaad88ddc0376
1bfb0c63aefe37191ca695cbd5ac3e366022ff98
2298 F20110209_AABJXP wang_f_Page_49.txt
bea15412a56a402d52705c0b0074d6c1
1886ca6d25dc7b90fcd414c627b3e44f7fdfd8d0
F20110209_AABKDK wang_f_Page_58.tif
abe5dcfde75b71bbd0f61c2ff630b3a1
5ace8f1a0bf3595b3e8c5a0a0b8a0d8dff6c036b
F20110209_AABJYE wang_f_Page_17.tif
444b18f340f037727b2f39b85e2b86ff
30bba6c9bc4fe9d8b90d53d1b3d3c82316de92fd
F20110209_AABKCW wang_f_Page_07.tif
6022b6c7a027e7803fa3047800d605ab
e0255a2c25c02e28416bfefe3c7ec5fb381ec3ad
915627 F20110209_AABJXQ wang_f_Page_09.jp2
1c20667de3e97129a475038909a929b6
0abed614345e7e924fa9e532ece660e1da51de0f
8012 F20110209_AABKDL wang_f_Page_15thm.jpg
f82d788a3b9dadd499d3fcfcb8e48b98
d8a30ed6b7205f798a7508abaec03db8370acee4
F20110209_AABJYF wang_f_Page_63.tif
a8109aad932ff1da96d4709704d2c9d1
b6b1bac23fb788a77edce73cb8eebd6b0f380eba
35550 F20110209_AABKCX wang_f_Page_65.QC.jpg
80e396cf76fdbcd4de2f3059c3ccdb7f
97fd865cac009983faeae194b3c29ce4df57f968
43432 F20110209_AABJXR wang_f_Page_70.jpg
804783c41560ad19b9c3f3cb93a11767
a46e11014cd4f1369b87d1e52f72bac46579f6bb
8839 F20110209_AABKEA wang_f_Page_19thm.jpg
c20672dc8d80367eef77ff2b88c51c72
497447ce1abec4fd4b41422ed5e11460fa7d83c9
F20110209_AABKDM wang_f_Page_02.tif
90fdd636ce22c03b53c08311aa171ee8
0eca3f02e764d80b040604ad635f67f327df29cc
F20110209_AABJYG wang_f_Page_40.tif
680a562a094eceef96194d6b026cc4c1
72fc87661ae348e422fb6e035bb361eb0b030301
125790 F20110209_AABKCY wang_f_Page_85.jpg
05d7b7a78250bbbe3489547f18c9c583
66f5f2eca233fbacaf368cb9b4e431ab92152c78
F20110209_AABJXS wang_f_Page_15.tif
186f397f8d6d099a5453dab6d6747a0d
f90d334d54c65aec95f663bda67467bbade18e82
F20110209_AABKEB wang_f_Page_70.tif
d0717d8aadfc7b8a56c7f1c68a9491e3
77dfa5a0a52caa0d9afed165d60d1aabd6cc794a
837 F20110209_AABKDN wang_f_Page_25.txt
3e4a69ed65748fa7a77b189b157dc934
0047c31c3dbe1f4654e2eebd9ee99e33fa685760
115850 F20110209_AABJYH wang_f_Page_56.jpg
5a3711ba5edce19c6a441731a994e9f2
cb5df6ad2c3f3900f17ff497c532c5c6e1a74b16
967460 F20110209_AABKCZ wang_f_Page_42.jp2
dd1473b86651d4c13578f0213a90a731
924b4489e9260e9ce510dfcf229e3b15f3646011
F20110209_AABJXT wang_f_Page_43.jp2
49e60afdff7766c029d7af251f343b22
944d5adf8ad51bdc07578d2452efe6402f72d52c
F20110209_AABKEC wang_f_Page_84.tif
ed27a79365925a105ab6ac4ba7d2903d
7939ba34ec2a975ebe2b35ad6e17f6d2fad94147
1551 F20110209_AABKDO wang_f_Page_02.QC.jpg
f8752f02150b06700825f305987114a3
be317c4cd2bc68be65be7a914a57e98514a08057
91486 F20110209_AABJYI wang_f_Page_84.jpg
5188a5c62f21763f1e741e641afa55fd
7ab0805fae9ae09adb79a2418e4fd48eae1d05b1
7968 F20110209_AABJXU wang_f_Page_71thm.jpg
6edb83c814cc8471b4ef002c2d762682
21d7b81693e9afe94da4f69a8062ed6eb9a60997
4967 F20110209_AABKED wang_f_Page_53thm.jpg
7bd94d7a64494367fc9e70daab8d948b
8b5251ea890e4cda4eb78b4429df41d70b354217
57494 F20110209_AABKDP wang_f_Page_07.pro
cf37d29ff45dad50edd1eb02902b3dff
93bf77d073aa36d64eeedb8c8480e25b07e74904
8628 F20110209_AABJYJ wang_f_Page_38thm.jpg
96dfd957c27c34398e2c042a0f34e357
3b430faf3ada7ab1c8768b249ea4ae8e65de780b
2021 F20110209_AABJXV wang_f_Page_61.txt
4120a281bf93c928a596e4248fef3420
6cc82f918d6bb4526d37aef224898e531e4b89d1
114822 F20110209_AABKEE wang_f_Page_86.jpg
c6822ffd1e221cb324ec5787f6c46a1b
e0af5299d7b161426e3f31a67fa9f89aa060fa17
29639 F20110209_AABKDQ wang_f_Page_09.QC.jpg
cb05b5027ce552876fd1055e6566d403
9133dfa68bc750d2cba309e3724ebbaf208c7108
7432 F20110209_AABJYK wang_f_Page_73thm.jpg
9a19045cc7a73acf1398278b31540829
113ec59fac0e374f618978838e7c4dd081649f0f
29011 F20110209_AABJXW wang_f_Page_42.QC.jpg
e99a05463472cce7e0e174aa2259af47
ff44f1872d1452caed2271e4c5d0c3ebf80a1f90
30718 F20110209_AABKEF wang_f_Page_82.QC.jpg
f6760d6c991ec97a5c472d9be0499906
c0b452ffa93b3438a88e4fa8c92a67132a72e41d
117145 F20110209_AABJZA wang_f_Page_55.jpg
fb471731124bacdad2109f7bead77f35
0945aa3f947953682868b69678334af192c489b1
2441 F20110209_AABKDR wang_f_Page_27.txt
f21f8c6dc28650378df59b63d0123615
67eafb81623084cffccdadabd1089a1c252ff587
24823 F20110209_AABJYL wang_f_Page_06.QC.jpg
a7568b9631ebb471674701c88a0616d3
fe58cf432d695b97413cdc554407cea75155798d
29996 F20110209_AABJXX wang_f_Page_60.QC.jpg
385ea4ca9deb7c15bc1a5afee3df8bbe
8e622e9f9147be1706ed9580dd8060acde772b00
29399 F20110209_AABKEG wang_f_Page_37.QC.jpg
5953f9b53e89e0bca312b6ec2c9e2ab5
4e60c07708db43eb887934d844b18897a55074de
280 F20110209_AABKDS wang_f_Page_87.txt
c87fd6fdd667968a68c07120419eb2e0
7b8a6b6b1418fbaf7c403478cc7e7c649fcdf867
55096 F20110209_AABJYM wang_f_Page_56.pro
1e88e07bd9b21d3ae593c3705e3e70ee
4d83b62ac0ae93f92baab0f67452646eb92724fc
49772 F20110209_AABJXY wang_f_Page_04.pro
9ce6d1628d7385b0c5605eb23bef5e90
5cded97539fc9511df07313f75a1f9423a1a629a
40293 F20110209_AABKEH wang_f_Page_68.QC.jpg
777aca1f2896bc97c196f18b6b2e18ad
331f8deb72d4340f0c45c891e84fce347c1367b5
2159 F20110209_AABJZB wang_f_Page_51.txt
d630456816d46af079644df1141afc70
a148612ae19669ec7e9df4bb31adca95f5133b81
1777 F20110209_AABKDT wang_f_Page_84.txt
344bda1b9934f4b4e67b2a83d70207fe
67bb8e08fb0bf84d3759afde800ccdcaa7057b20
39870 F20110209_AABJYN wang_f_Page_73.pro
25e036f67cf4f95ed146ff8988163852
ef8d4b356bcb33ca3785dd0e4bf0a4a1f2a30451
38998 F20110209_AABJXZ wang_f_Page_49.QC.jpg
fd192085a40e8060cfe086a58dbd4fa1
84f22feb349951a8d55326ffd7d3fcc41eac397b
1168 F20110209_AABKEI wang_f_Page_16.txt
14d32a9dc21202a21c8bb46f841b61ca
567ee18eaf60e074415e34719561dc66ad53fcaf
1060485 F20110209_AABJZC wang_f_Page_88.jp2
f05577b7c1739119b978a55f9ad4f1d7
44abac6ae2bfefb04015faa65c6908f1ac61a70d
44525 F20110209_AABKDU wang_f_Page_45.pro
43d9bd89343cb5338ff16c35581e3929
88be715da496464233b4b16b8632bca47afb3bb3
57145 F20110209_AABJYO wang_f_Page_21.jpg
1eb7f78250b490374ebf2c71e9c390b2
40030d6f3972bd031adb9434a4e1263863413319
90528 F20110209_AABKEJ wang_f_Page_37.jpg
81d3c779ebe30931d5d5b7ae19d5e66c
b1f39a394ad68b4c01fa63f2d128ed0e64ea838c
751590 F20110209_AABJZD wang_f_Page_63.jp2
06f134f63a999e1566b26ae3eed35cf7
900e2ec31f64575f2f53e1c417e1a89c88b1a830
984057 F20110209_AABKDV wang_f_Page_11.jp2
edc860841a88f76660a2a9ca0ee25082
4facb0193a55f8f93284c50962768cb5ededceb1
F20110209_AABJYP wang_f_Page_61.tif
b795a8e13a879a86902c4ba092df8ed1
2469b8a929db9861f3f76c385b5470a9c8a501ff
34953 F20110209_AABKEK wang_f_Page_72.QC.jpg
33a7391e671eb2975b46dfc7613e43b6
0b811da58ffe8970fb53b9b56528ef6544ddbe18
F20110209_AABJZE wang_f_Page_26.tif
072f5be272b3f1403a5c02e09c7dc5b9
30ab978443c2671f3b7ed4c3452dd602c7565c1f
988261 F20110209_AABKDW wang_f_Page_33.jp2
f8cba4465f8f6aa19c21751e789f44a7
602528f4a122a8ab3619e5b8e248c4a4b1790b4b
12333 F20110209_AABJYQ wang_f_Page_95.pro
eaa3e3e69fbb8237848b3d12e4332a5d
b39518de3f57e2c1c23e7f8e1bdc0ccbd78bf89d
57154 F20110209_AABJZF wang_f_Page_57.pro
f75c31cba70c20f135e2771c01030d59
e6d5dd76e3463108df910f64a8b19023aedca15d
F20110209_AABKDX wang_f_Page_49.tif
c707ef41754210f69cb714e60ddc197d
74cbd5208ac3fb852292f6cafe4f5e791c5b5512
50349 F20110209_AABJYR wang_f_Page_20.pro
36e6fdc25d2e1de03651e13dfe33c9bb
4b59ccc9a315ca55cce35419cd3bf840e71a670f
7702 F20110209_AABKFA wang_f_Page_31thm.jpg
eccc37f3ce5509b3a7846e2611d21507
fe4e6cd4f883e8e5e2b944d20e31953d23f2c600
1171 F20110209_AABKEL wang_f_Page_94.txt
435b794cddaf2eb16fa64b9ef3c3f233
02d382f5e55ef144742735616d830d5b0db9cafa
F20110209_AABJZG wang_f_Page_37.tif
3d38240b651d88f622220e85680ade4a
da3b3f8d3c1d9eaad3dd2bb2bc53a92c2f643d16
20220 F20110209_AABKDY wang_f_Page_05.pro
eb839fbb6cfa4b44e8dc73c59a57edab
5ac10896d2f982f049e66b6446edd55e5fbf2869
57399 F20110209_AABJYS wang_f_Page_48.pro
7c2464c105cb80e4d4fc0abee5cc0d02
1d84b2dab9fe194398260950af595cae75b783f6
F20110209_AABKFB wang_f_Page_65.tif
6a7c9f4373663d49a0d920033a8d8076
2e98d0c8d00330214bf9940c0170037ff76c8c82
34701 F20110209_AABKEM wang_f_Page_20.QC.jpg
b50b5d65011b82c4c1602c864430cf4b
63dcc22e01cd12b54e075f9e3b53d81cba1c35a8
7986 F20110209_AABJZH wang_f_Page_01.QC.jpg
a9cf8fdfb226f62354025b1958e19c36
4ff67ec93aaf66d11a57e8dd440980cc468a6e0d
2106 F20110209_AABKDZ wang_f_Page_65.txt
cf3b846794c2efce5d982e6a2dfe5963
74ba654f0fdd2753291e01ceafeb8040191bea0b
2685 F20110209_AABJYT wang_f_Page_93.txt
f5af2c78744dfb4cd07f6aaeba8ee823
03173c8849d043fbe60fa29ad324b8806413055f
55970 F20110209_AABKFC wang_f_Page_59.pro
658c2be116e223592708816293d740f8
8d6c20f259b6ebb0fb640c81de7ad53c4da7e3ff
128089 F20110209_AABKEN wang_f_Page_28.jpg
93800e38f5b2d17c61ec04ef67ca3026
adbb896259107fe4aeadf6e727bd9d4a19a78795
18296 F20110209_AABJZI wang_f_Page_94.QC.jpg
3c0d4e7da936d1a42119db468c6bcc3e
a5886c8967a36e5f9909d992548d40921993930a
1012080 F20110209_AABJYU wang_f_Page_13.jp2
71aaea6cd915d085ab8e740834049168
7ea607b83061903184155ef213e5774635085bfc
1087303 F20110209_AABKFD wang_f_Page_15.jp2
5b95751a73bb7952d336f2b1efd52d2f
49a87560b9cf29bc06f9801b7cbc27e7104dc075
26664 F20110209_AABKEO wang_f_Page_21.pro
898037224df85e991fa6c0ee570c09ee
a46bb4fab2b0fd0dcc92837eabcd2040f312c1d6
2719 F20110209_AABJZJ wang_f_Page_89.txt
bd2f258af45a858b1760fa329c18966e
a75e6853d5f7b149cd4a923cfd078961746cd8c4
100518 F20110209_AABJYV wang_f_Page_07.jpg
b5605786b387d5736f009e39c97de747
1b4c11a1aeddf756a368d984a29dd65722f3a38e
25385 F20110209_AABKFE wang_f_Page_08.jpg
e3a751a1b38a172b2235d1dba5328f67
ec3eafb946ec61bb843c3e6761e8b4c9fe2e4d20
31554 F20110209_AABKEP wang_f_Page_81.QC.jpg
70b4b86a9eb97495f909867fced0ffea
57dda7165c881ee62e2b935423bbdfbaf4b676f0
3215 F20110209_AABJZK wang_f_Page_70thm.jpg
426ea2e8a7b7007826b57b398b9d4230
44df32f002b078c9ccfedc25d0355c8d18737679
2290 F20110209_AABJYW wang_f_Page_56.txt
91cabc1cb254a51b7ed5e6fce594d709
66ce20b1e2ffa13a6b35d94dcc90bea114bb76c2
1087899 F20110209_AABKFF wang_f_Page_81.jp2
c2b6d0fe2b21f9bc141c496c1dfa15d4
0ea11fbdcf83503e443f226f4eba18be5e641f7e
668034 F20110209_AABKEQ wang_f_Page_54.jp2
290a26795ad7fdf590b7e4f7ff990bc7
b68c1f1269d1d6b4fd1dea9ee862854adb6cf3cd
2313 F20110209_AABJZL wang_f_Page_55.txt
a098b29450b47534e87eaf25f0ad0374
d0851811b7b68fe5dfa03ababaac229a4885d99d
81079 F20110209_AABJYX wang_f_Page_47.jpg
c1b52fabe7abc3aecc8974efaf7f5bd7
2264c816612802a0022c4286408c480c351fa433
30293 F20110209_AABKFG wang_f_Page_62.QC.jpg
1252da8fbcad1ed2c5fea074036c2018
e1699ed56a8f92f64e03c68723d4bbaf7ebd836c
11195 F20110209_AABKER wang_f_Page_70.QC.jpg
59567de2c3acfeb9f2241585e3182aad
12e10d8e438357a039eebb51662d593c0cd7ce23
24798 F20110209_AABJZM wang_f_Page_53.pro
7f38a0e0a3779c351fe8cd519d7fac1d
f153f914e244f9610933b2e228c80d6eecc43ee4
467 F20110209_AABJYY wang_f_Page_01.txt
643c105b963367fa0ee45a2bf3c02e40
7a48d88303b15fb3abbcd20913b6b22fda7ed65e
23486 F20110209_AABKFH wang_f_Page_40.QC.jpg
5ac6e7822566d427e24024901fe78ab1
8148f800d9cae5a1ac758bf2b9d82da88993c0ed
F20110209_AABKES wang_f_Page_27.tif
d4e3fec51325558930f445d09d030eea
30b387778b8c67b8278b753721817e5bd74a2d4c
96886 F20110209_AABJZN wang_f_Page_44.jpg
b9937d6d7c351e91643a0afdf8b8b01d
a0e088513763bc09c65db1e34f1b1d5ce56251bb
919835 F20110209_AABJYZ wang_f_Page_62.jp2
167ccc20d9c45d7574d3dd5cc2d195fe
2995c709593bd31c65500b10ce48338eee331708
7701 F20110209_AABKFI wang_f_Page_37thm.jpg
a322108530a1c1b6c5bde4004a13edd4
478f9e85a5785773776fc487b35d16182c292920
F20110209_AABKET wang_f_Page_73.tif
c0202041162e606649de418e505d1e36
9b5cde30e66d42a48468c168d984b1847eae76bf
190 F20110209_AABJZO wang_f_Page_10.txt
4b1de48bed0f411f9a7fe5ce7a6e5844
2f9b47bb57e3b00eef90273dedcde4acf3db029a
220692 F20110209_AABKFJ wang_f_Page_08.jp2
7ccdf330d12ae1270cc1ece8e5f06f73
3394ccff4da191fc962e5abd0bc22ca6cdfb1895
29702 F20110209_AABKEU wang_f_Page_88.QC.jpg
9c9e6b2b8fa72f5d50699645ac98e51b
da9915a6a94dcde8035a7cd7d5095a6069d7e99f
45168 F20110209_AABJZP wang_f_Page_76.pro
cc80fb6c92c78428d9477066a646beb9
acafca66b5336b4ed98b9fac0147273c8fe85a64
57384 F20110209_AABKFK wang_f_Page_53.jpg
2d85320f8e59f222241b0451a9388992
95e22ca752821630604e625e85b19f1479815941
F20110209_AABKEV wang_f_Page_82.tif
2d89074dd8c091a76ef6be34c6add842
590a25d69e3f847d359bfc1b37b3fa0afb7e69ab
15795 F20110209_AABJZQ wang_f_Page_87.jpg
9703973494ff35cea8cc467237b0ea9c
7dbba259d7fc6f2fd09567e656752e8f3be7015a
1646 F20110209_AABKFL wang_f_Page_42.txt
3174d62bccb6884cd300356059f611aa
97065891f99ac63607554dda371b1eb8667d393f
F20110209_AABKEW wang_f_Page_89.jp2
1e6f7089a5f8572471f5aa297dc7b4f1
d21ad9967ed1e1e6f6c89a41332335c9be19e038
1918 F20110209_AABJZR wang_f_Page_13.txt
deec3564aa2a228cac6dafc5bfaacaa8
50553032125b37d604e1e3652f3d0ed54a69e13f
19220 F20110209_AABKGA wang_f_Page_53.QC.jpg
00587a52143e38883047172b23e24de9
a0a157346df9d3c398fdf4d1b0aa0a4d9f4a12b1
9343 F20110209_AABKEX wang_f_Page_27thm.jpg
8fe6bce81d68abfbaab6c74a799f0133
18f0969d9b6f97761bd8dc06e62ed62a96567cab
F20110209_AABJZS wang_f_Page_67thm.jpg
f175cf46d44da22637f061163692019d
6904677b6916d8d9434810b8ce46b3fd78b9df37
2067 F20110209_AABKGB wang_f_Page_30.txt
b8a44b406b1597de832bd3110728fb04
0e81191a8ddb22779c217f87a6bfe82586358474
111104 F20110209_AABKFM wang_f_Page_51.jpg
b18c7bbcbcd4e43527f472a3223ee9ec
677c8602877d1fe7d9f68ec7da603f17cf7dd1fe
89369 F20110209_AABKEY wang_f_Page_73.jpg
b6d0c5f995f7b971eb73e34799c19ca9
6fa5da52b1792fbff47b25e71e466170d292968f
38982 F20110209_AABJZT wang_f_Page_41.QC.jpg
77859b02f788bbd471898e0c941868b7
bc3adb13f35752792e1b46cb33e3358c1d3723ea
71939 F20110209_AABKGC wang_f_Page_77.jpg
7adfe88eec0a5f61763b7eed687fd665
98543110f3d20510b654d39dfcfc2448b01b7b73
F20110209_AABKFN wang_f_Page_86.jp2
930b7e851dbcca3f184cbef814fffe90
b85d31db7806b1e3ac3d75e34723101370b0ce46
754697 F20110209_AABKEZ wang_f_Page_23.jp2
443d59bc7a1721ed67d669e080635cec
925599f9db322b3396596520c18d1b38b2e88ee2
120574 F20110209_AABJZU wang_f_Page_48.jpg
0165f0ac3eeb85a74b6a206ea2f421c9
c63622ff5f8927f5f51e2a003c0fa64ad62d2388
90084 F20110209_AABKGD wang_f_Page_32.jpg
83a161d8dce120d2a09169f635ebca1b
7f49262f55072a47179ab62489d95ed2d541f6f7
2133 F20110209_AABKFO wang_f_Page_71.txt
d08bbc47f73907364e7537013712ebb6
b58a70bbbf1684729e9a33d10204942639b1c8ef
7639 F20110209_AABJZV wang_f_Page_24thm.jpg
6c932065368f3e2cd7f68b21363ad91c
33b1bc876583f3ecaf3589edb31e9ff1502422c7
42215 F20110209_AABKGE wang_f_Page_83.jpg
21f708a1f92397b01411af7e4ffa2963
abaaad0f11f111ff419bf47b83f4bbd473ca2268
7107 F20110209_AABKFP wang_f_Page_64thm.jpg
a8f1ef8ee78557709affe2dbe2bfc11e
6bf1e73994f87365dbede46b6bc840058091f9c8
52167 F20110209_AABJZW wang_f_Page_71.pro
3dc5d4b2b23157cc63c67dfecc7e159f
7c91e11a82b0dd51ba1afa8e40940e001e7fbbf3
F20110209_AABKGF wang_f_Page_74.tif
37b6d245dacdbd2ce8831739b4f7757f
32482c5f51b7417d2a40e21b7ceaac957c5bff4d
549 F20110209_AABKFQ wang_f_Page_03thm.jpg
65dce6e33b5a8b715b8789d1bc548abc
49c416b58d9c9a1ec06f7ee4f761cabb19a6c11e
1087815 F20110209_AABJZX wang_f_Page_29.jp2
9b67fbc98e1e42d4e40904741e915fbc
25a7d10688a1aeb74f8d8ef1e36cb9a5f0dfd047
23946 F20110209_AABKGG wang_f_Page_16.QC.jpg
ee92909331c8ec01bc259af546017cba
fad106184db52362b7426997e5128ca03b02d7bb
1087890 F20110209_AABKFR wang_f_Page_65.jp2
6d61cc2fdb9e85376f542b476d9e1558
feeede376ee257b74952ba1b419720eed267cbba
8900 F20110209_AABJZY wang_f_Page_79thm.jpg
e1699949af72e84e3ba20a9c958b0135
81a23c46f810f3d69a489e1ba45dcd50f2455aa1
121650 F20110209_AABKGH wang_f_Page_49.jpg
eb6737ab0a958236f7044da18ac4cd71
25818eb007ed170c4eb1af8d09814643724d9b51
32227 F20110209_AABKFS wang_f_Page_77.pro
ac518b45769bb1c5694a36047c6b3f6d
b1c08152cac9f9e22f37f0f4a256218688fd214b
32459 F20110209_AABJZZ wang_f_Page_39.QC.jpg
0ebe68a2b4cd445120bd1fbeefe25901
c60ee23496488571c6650feb90e31ca1672cfaff
F20110209_AABKGI wang_f_Page_35.tif
21ceffe89bbe95a80de44bdc60ca99d9
d9e1da8b7bd10c1ac6e22d3978b8c04d520a97a0
F20110209_AABKFT wang_f_Page_50.tif
8c7283294cb709bb299647c5c5d63919
807e098c4a5571340f1bd30fe329099880bc9440
58887 F20110209_AABKGJ wang_f_Page_85.pro
959509e4dbfbf025f3a167ef820a4169
67f3cad40c20818876ae7e787a2f8d9fd5291e94
1478 F20110209_AABKFU wang_f_Page_23.txt
85745c572479a7a9f295019ae65f1d3b
47a3ae1c1e47ada3e80a73bfcbc80f000c1ad0e5
27445 F20110209_AABKGK wang_f_Page_47.QC.jpg
6652676819130274368552b64a0c5658
1074f10e3c57d18f9a1c6f35709f4debc6898e53
2533 F20110209_AABKFV wang_f_Page_07.txt
51592c1d894cc16e655444fdd8d71aea
32dd2c6bee5e0a5aa4141db6c0db456c5ed71411
2209 F20110209_AABKGL wang_f_Page_06.txt
2602b8c0b1ba98b427074afe6e997669
cb9ac495ac09ec641b459628816eb9ba5923df59
29811 F20110209_AABKFW wang_f_Page_11.QC.jpg
350942fd3ac5bf1621dec804d2fa9fc1
e37098340aec81924e89d9e5a5b8b16692472bbf
52999 F20110209_AABKGM wang_f_Page_65.pro
c171d5bf26882aa70762ea320bc7bc40
6c4fb564767fefb55c6a0a0b9817b373c73f467b
28100 F20110209_AABKFX wang_f_Page_24.QC.jpg
b1ab68184cc00044e2010954156fd987
ee0fae1f6f447ae08b82b3e23aa47fec7e7786ac
F20110209_AABKHA wang_f_Page_80.jp2
7bc5243653f13bf76cf66b5284319b37
68fb4aaa96c6cb3023ca203153844e771961b341
F20110209_AABKFY wang_f_Page_77.tif
e6a55904ea3f82755bda3835fce9512a
c137005feb2bec07423e94ccaae52985a3afcf11
F20110209_AABKHB wang_f_Page_81.tif
457c30c859c59f15aaab3d0389a04485
2f0e3857d2b8f9dabea6d2363c46dadd1b62ea97
F20110209_AABKGN wang_f_Page_67.tif
b0c2789d766ef1cb63636856d37fcb78
abb5ec00aac5c32bf600dd4881e40d4ed5206880
106 F20110209_AABKFZ wang_f_Page_02.txt
18e494ae0f958748defeb0494b816913
a71a498d29fa649d45f0c56e0daf148f8540185f
35880 F20110209_AABKHC wang_f_Page_51.QC.jpg
f5f43359055deccb37ce5f7b34b926be
43b19f10b9e7a374568d82a5a8930be3d73697c0
105802 F20110209_AABKGO wang_f_Page_88.jpg
1800d9722dd8e4ef1e58995e7497996a
5d43af85223c6e15e07ab979ddb6ffade60667fd
28282 F20110209_AABKHD wang_f_Page_07.QC.jpg
df93e42e23a66cf6777ec8624e7b54b7
f8ce675f5735d438ca96075234023596e150c3ed
59954 F20110209_AABKGP wang_f_Page_68.pro
ea360e3fcddab9525f3a9626315faace
4b336f7e5948ec8cb8960feda61f8601ce427c7e
42835 F20110209_AABKHE wang_f_Page_11.pro
1ef277d7b12098e89fda416fcbad5401
93319caf45e1dbad4614ccd1a52d7c9e021cdb77
7168 F20110209_AABKGQ wang_f_Page_45thm.jpg
daec5bb3477fe23c9b4b089b8ab4afcd
51b9b8475f60496783b9d8caa6417aac5a88b8e7
4880 F20110209_AABKHF wang_f_Page_10.pro
7f18e5e395694ce6bda0549768be8d68
0455795cc0af470855318a5ee702ebfc20ed74ae
30950 F20110209_AABKGR wang_f_Page_33.QC.jpg
5a69d8753a21fdeba5c1094929caa945
1d39e4e33fd24596f8bd13a90a867ae7b11df0c1
6365 F20110209_AABKHG wang_f_Page_69thm.jpg
a754e6218a9bdbd26ea5335ff98bee31
bc3eeebad0bf79f68feafb348fd82628ec0156c5
7352 F20110209_AABKGS wang_f_Page_84thm.jpg
1204b026065d87f7ef36649949b691b7
c9ff2b4894793fc2ba7a78fafdf93782312b5e13
41419 F20110209_AABKHH wang_f_Page_24.pro
0f839cc5bc47849abc9f3ed1c156fb18
e86bb65676ef25af76c0f2d0f121b2327695ac4e
F20110209_AABKGT wang_f_Page_11.tif
172cd425e258a3ac7928bcc7e1cd5399
afa5684e35ff8b2d84f31693ae9fd3e9f7f9e25f
112470 F20110209_AABKHI wang_f_Page_67.jpg
7052ffcf932ad715c6f7394ce2c2159b
464d52d4ca113f9caa16638d403ca913ab354736
37431 F20110209_AABKGU wang_f_Page_59.QC.jpg
e00ad4fd400d7e9e648bd9ee907f3adc
07ea3b99bd2fdf7458eb726b9c661ea69230c18e
F20110209_AABKHJ wang_f_Page_31.tif
009f1240d0e3cf7bb7ea8e84ef49680b
2823d24220caf8cc5604d61180c4fa774d295d04
67404 F20110209_AABKGV wang_f_Page_89.pro
299f0841cb70eae6488809bb5d246957
9ba56453b2b6b19772aa61437e871960e2603956
1767 F20110209_AABKHK wang_f_Page_63.txt
9a32c5e30e34712b899976d7fe061649
c6e93528125e25738da00eef9caca2d77afad761
538 F20110209_AABKGW wang_f_Page_95.txt
bfcea273d8b3d71418269c5fec021e53
5622c99f15e7f6c2771ea1b08495615d9f10546c
96169 F20110209_AABKHL wang_f_Page_76.jpg
8bed511ceabfbecd79a10e7bd64fc34c
916c872c3f246457bbaecdb12fa0167762c9b1bb
106894 F20110209_AABKGX wang_f_Page_26.jpg
15a2c278ebe1fb06eba17d9092a9c4f3
b88e04a2fdd7e24c9fa79828420fc7d17e020066
1087836 F20110209_AABKIA wang_f_Page_46.jp2
2f01886a7a1c318fb62471d7f504667a
a8012750860ff734fb6a6e97a74fc82ca57f736b
57049 F20110209_AABKHM wang_f_Page_41.pro
76b39b0cce858d62ca10bebbd7ad1fdf
2fa5810c3f5d825b7366ecf26cb5eb0d17544597
90581 F20110209_AABKIB wang_f_Page_60.jpg
1e56cabdbcba7e72670856b7509f5e77
7241411f10fc69bd3b59689ad863a856e0b390a2
146056 F20110209_AABKHN wang_f_Page_91.jpg
5e797e29f9d28a08f099cce1e296d8e8
cc3c9014f093ea164d6d89ca1a5b16ca12418b9d
6482 F20110209_AABKGY wang_f_Page_40thm.jpg
9a3603653e56674732af9ee657ff29f4
d486ed18b7b6ca7f919464476a3c1e20f92e318e
6977 F20110209_AABKIC wang_f_Page_18thm.jpg
f15cd4e9ddf1880a92c21dc1ab27283b
de3cbfa9a4927507fd5b3344c59e9b7eae614bb4
1087896 F20110209_AABKGZ wang_f_Page_59.jp2
cf8f02f392f6eb398cb9a81dd7272f39
9c85160a6ae137120e2103a7dd15e0e34e9dc18e
35280 F20110209_AABKID wang_f_Page_80.QC.jpg
567ec20de2546607806896a6af976035
9449e48d995ca1e63166e4dd9741f86d77b7a70f
148893 F20110209_AABKHO wang_f_Page_89.jpg
eadc50865cf1de2108dfd658607ae20b
34d9813e7d26d8f4f2bd78fbdb002d515d9d7824
7186 F20110209_AABKIE wang_f_Page_62thm.jpg
7f7b65163cbfc54243d1efb6bd7eaad1
a88ed6e11915996443f6d725c25bcb6ac3a207f8
30640 F20110209_AABKHP wang_f_Page_54.pro
9d09672b0fee8d8dd7aeff1557b7e65f
aef526ca4275c399472e200b1389cc5ad805fed7
104636 F20110209_AABKIF wang_f_Page_15.jpg
28ad255050d2354eefd5865b27122c4a
e413f7f979dd33c17d0bb7d6e23a1a7dc1dc7b89
68098 F20110209_AABKHQ wang_f_Page_94.jpg
8ed27f6e231fb0ca0a6a205c5799cc81
315591e4ab465684cc831230d635fb7cd78372a0
1141 F20110209_AABKIG wang_f_Page_03.pro
68ef18dc57055fab85aa640c77ef465b
610a55599a60ecde7b3680786610a1ea4e907d46
F20110209_AABKHR wang_f_Page_28.tif
b09bb7537e38fe0cdfaacbbe4a1ec602
758e2832bf0d81d35d6994dc7af95d62bf3e77d6
F20110209_AABKIH wang_f_Page_19.tif
ba8b2c47828ad7fb6eb6bec89b7935ce
64d79e39f4c9ade7a5f8a1faddbf7a0f8f05a6ba
40359 F20110209_AABKHS wang_f_Page_89.QC.jpg
16052c1cec793a1dd38278daafcddbe5
9e9576153cac8670a8489a13ebed4af678e8a87e
107998 F20110209_AABKII wang_f_Page_04.jpg
eac094d4e188e4a706c14ddab2da4333
c355ac0f78628bef8c012bb3e2ab94c7738895d3
47526 F20110209_AABKHT wang_f_Page_88.pro
4743275c8c01c2900ed8a5c3ddc0f2e3
856c4f8039787dbbf7c759c2b4225241ceefb8f8
3424 F20110209_AABKIJ wang_f_Page_10.QC.jpg
6571f8a2ad57915508dfe923aa1e4e5e
36f89a9b931c53ad85b0ad377a6228c0830d7a38
951718 F20110209_AABKHU wang_f_Page_14.jp2
522eecb87529314d88ae118067ed95a5
d9672962436d9e7802aaa2b896032aac50c9c979
8218 F20110209_AABKIK wang_f_Page_08.QC.jpg
b41265c55cf9c2352a3631ec7a631a89
3b3912679eb6f55bb08bc7dd2821561f5f7828e5
6661 F20110209_AABKHV wang_f_Page_63thm.jpg
58b78269fd1b66c99102e938b1cbb9dd
3f6c7ed98b54dba39453799194f1595c489df48e
F20110209_AABKIL wang_f_Page_66.tif
55008310c5b00c2346de29aab247987f
39faf7a7512f929a2883cc4d1667285b26919e0b
106603 F20110209_AABKHW wang_f_Page_20.jpg
a5c7546282be5e4449b67d134c69937e
e9bec1f962c4b862e977ec94dd21cfe026578bf8
573967 F20110209_AABKIM wang_f_Page_34.jp2
1b828f6724477bcf3553b00d111950fd
cd98d01d4e9bb8e7adfb5fe2f92fdfc8383a767b
8099 F20110209_AABKHX wang_f_Page_30thm.jpg
65a9665f2196afbd2d8aa890e45f3d8b
475646d94e5d67ad939f64f2b083b8e5992f0126
37577 F20110209_AABKJA wang_f_Page_55.QC.jpg
316f6feda28dc398d03573c0dc1c9513
631aea50288734c20b7ef183bb2e4edbae8a9f0c
28563 F20110209_AABKIN wang_f_Page_94.pro
fa24cba9b42da70cd47d8877c30e8efb
a2070470986a003f74d23edc5d0c0d6669605c2c
44313 F20110209_AABKHY wang_f_Page_17.pro
f38d8e56623891f1c17cf46cfcced914
10fd45dde41a67b1b349f3aeb47325cf7ce89194
F20110209_AABKJB wang_f_Page_22.tif
79d70d529618c597bd68fed17084fca7
661b583a123feda2cdf1d35845ec554cab15ed64
62345 F20110209_AABKIO wang_f_Page_27.pro
6f10d5f1db7e91c5ab7f0828c05d2aec
0621254bd006215805f797061e998e2ae596ceb7
9132 F20110209_AABKHZ wang_f_Page_28thm.jpg
7c624b8fa14b5a506859ce4cc82f650b
8c010c5dfe5fd01b0961ad04e8f73a6c0820092f
25664 F20110209_AABKJC wang_f_Page_34.pro
e5d7652387eda15c1383bc285c75bc82
873b9d666a82e103d3260a9f82f41310f652f8ec
18779 F20110209_AABKJD wang_f_Page_22.QC.jpg
9d02645f3c7e8905e95e19d49d9f7ba0
eaa5f04e682e2cc3b0618b613aeeb685d4ce2121
8129 F20110209_AABKIP wang_f_Page_74thm.jpg
8cad859bcfa1d8a804cde1c3afeddd84
91ce3b62c99b8f8ea76d73d7c38804785e8533fc
15372 F20110209_AABKJE wang_f_Page_05.QC.jpg
2246b280606ab2d25ce876418d16b145
be046d322ff5aef5617fbfd7f5f4dfde1c257bb3
1397 F20110209_AABKIQ wang_f_Page_44.txt
76e30df44ceac08c7080d0d5b4e73791
141fd42f64abd22cd8537852c220c8962d1659dc
F20110209_AABKJF wang_f_Page_03.tif
1694d2d8c9eb42113c3adcd4ad78e520
a3a9adb6c12fe8ff1b3514b44c2451700755f46b
2148 F20110209_AABKIR wang_f_Page_75.txt
2903c73318a9db70454593d960a83cac
2f50cf3658e50d5c5b4371af8f588fc63063ebe5
111739 F20110209_AABKJG UFE0015631_00001.mets
fe9f3f4a86946331b8796ed1e72d1bb0
ce35dae79eb8a26d110eda28b6aa615402537e4b
52052 F20110209_AABKIS wang_f_Page_29.pro
2030937bc1dc13d9d4f7570e68226b89
1090135c49a8a5007c07f0686680212d7459cf1d
1691 F20110209_AABKIT wang_f_Page_40.txt
43f085449cdd1e021731be558cc1b2f0
10324badecdcafb15af9c6a5cead6d146ca729d3
1751 F20110209_AABKIU wang_f_Page_36.txt
f84321f1d1ab0b38f241f4423b0d597a
7eca1a294377c51353480339434122ffd9871ab1
F20110209_AABKJJ wang_f_Page_01.tif
a19a72babf82111bcebf4e79150968bd
9f8cfa89383544454dda6e7cd73d63ac0a7bc7d6
F20110209_AABKIV wang_f_Page_92.tif
f7fb5072b10ce29b249ba3b0e82fe3c8
392e0895637948afe5b68181e68c03b755dc8b1b
F20110209_AABKJK wang_f_Page_05.tif
d390a17516d666a239120a2e599f9587
2c767fa58e9eca669e423815e34904d28dbc31bd
37881 F20110209_AABKIW wang_f_Page_48.QC.jpg
72f963f85394edda189beb5a7014999c
7d0bc7c5506af3f2c1a508fb56713165b26103ed
F20110209_AABKJL wang_f_Page_08.tif
e43998f10a2a3bd0b9a4bd251e37adaf
db67ff730e797d1009597f5d6ef49d7933107e71
31055 F20110209_AABKIX wang_f_Page_95.jpg
d010794ed8c9a5fce00addfb4e6dc217
b97a6a4628179bd558e95ce942855bef78f79776
F20110209_AABKKA wang_f_Page_52.tif
75b79174de4cb71c0e26896f6ca2361e
f73c51a7139084bfc03cf178da84a76b58912472
F20110209_AABKJM wang_f_Page_09.tif
135ade4b70d4910548aee4d43659c98e
bcc349a4433375c16e62565767fbbc99b3a266cd
1087809 F20110209_AABKIY wang_f_Page_55.jp2
5335871c06ad7a7c22fad52c54da701e
bae9a3051d35e2aea1e2eb49c17849c40d67a7db
F20110209_AABKKB wang_f_Page_55.tif
c187efa77af167c1346fe640072d55b1
080db2b32fe1175183757ccf87a44d06eb33552e
F20110209_AABKJN wang_f_Page_12.tif
9623c93d9c6fd4164e1e29fe43ac31cb
ada60f2d47910f65db3885ba6bf32eff6ec71de1
94673 F20110209_AABKIZ wang_f_Page_11.jpg
fec7792d1b14b7cc4d4c88557a083fca
95533a7cc31704f89c3c267165d288946c22d62f
F20110209_AABKKC wang_f_Page_56.tif
e21e00d59d7706c2926851847ecb86f9
7715bba763caac64600153af4315d7dfdc1ddd36
F20110209_AABKJO wang_f_Page_13.tif
e3b2a29ca50ffbdee613b24052c08b7d
3b1bdcab2fa6c27cbda9ac73cd79da5a17208a01
F20110209_AABKKD wang_f_Page_57.tif
8be711c9068f4ba0e402ec771e5582a6
f7f973d2ba1c087e298b1b8699de98f6a1ac74ff
F20110209_AABKJP wang_f_Page_14.tif
03a8440d5a657ffc482f9214b1db1019
9f790feaaeefbb0442e44a24539cac701b6c2825
F20110209_AABKKE wang_f_Page_62.tif
a9a30c81c25abf3a406446b931785c25
483734eb10e721b29a2e0428a3f025ea02e531af
F20110209_AABKKF wang_f_Page_64.tif
6901e6394f0dcf16122aa028158c86d5
7c54130fa3a4247381c3ada7f540ee5c4b28d49e
F20110209_AABKJQ wang_f_Page_16.tif
48123d885b87af2409a698415ec7b157
4dbf4b360d52e7fc6854dfa8797046cbbaae36b1
F20110209_AABKKG wang_f_Page_68.tif
34ad69a4000a51568720feb366aafdfb
8f612f65bbe094c928fef523a206cb01333c66b0
F20110209_AABKJR wang_f_Page_18.tif
89569b87d2a0d1e115a663efa4698c2c
02c2814e7e902e913612a5be0e80d12bac7d35ae
F20110209_AABKKH wang_f_Page_71.tif
e551443171a8ba966c380f613a365779
82ec66b91e9e2a7384cbc897233c05dc7c405ca9
F20110209_AABKJS wang_f_Page_32.tif
5981b7fbb91f693a708374eb9aaef50f
07224392e113d854da17e8cdc362db0f2036b0b4
F20110209_AABKKI wang_f_Page_72.tif
1d9e68fa68ba91ccc8a5899bc3be31d4
d9a2ad70d0d97af9bf6d5d04894cb044df1fd0ba
F20110209_AABKJT wang_f_Page_33.tif
ea6c983dcd65e8bb3a1dc6bbf89c9ada
7fd2c15cca4c81c987dc98e3c6c1776e375b74c4
F20110209_AABKKJ wang_f_Page_75.tif
08de0f7281bf3880ada40eae58dc10b0
22e1f414e6ae5fa4f3df74a295413cf75cabd557
F20110209_AABKJU wang_f_Page_38.tif
53ca7784cc5929c015407ce45d52b51d
572ba63f826b6363f01d7bcec15626e35c4c5166
F20110209_AABKKK wang_f_Page_76.tif
45456572335dbf2abf003fa581a0e5b6
a1ca6b98cfdd0f0de21128ad2356bbb32dd4f8d8
F20110209_AABKJV wang_f_Page_39.tif
9fe6038e549d200868eeb5384ac142ca
b6cdfb3b24b204eb38e387e8ddf0065e9f72a54c
F20110209_AABKKL wang_f_Page_79.tif
4a474164fa5618e9eeedda6a413f7a3b
d322648b734d1262d85c679b81a8fd53b6ca03bb
F20110209_AABKJW wang_f_Page_43.tif
6343647744c67c57117ec70899ad0143
a057a941275373b3bd29f72dd841d7d0f42a0891
F20110209_AABKLA wang_f_Page_21.txt
4bb6484c3ff4768a05b7ae88fb7ef943
bd60fa229c57863378094e29759b420e87b53e33
F20110209_AABKKM wang_f_Page_80.tif
06d919e902cd0153a6bb652be0817419
e594f429342238199345f5b7b408e6882b68a2dc
F20110209_AABKJX wang_f_Page_45.tif
884fd9e78781d7022376ba2b524407c6
f2e9c77e7b4629766b8ea8b8d4dcced1b81f73d6
1386 F20110209_AABKLB wang_f_Page_22.txt
15c842934b0b388377c7559c9f504828
9ab2f36ffcfc0a5633853d24cfc31e9cf77e81b0
F20110209_AABKKN wang_f_Page_89.tif
2757d3490609d730a63a3d7d99e3454f
b1062037f40b60e718e6c0bd3777199ed1f779ef
F20110209_AABKJY wang_f_Page_46.tif
c4a955fbf91685e48f990d9464c93b10
4557397f599028fc40f7fed63a22c2e9d3470d4f
2059 F20110209_AABKLC wang_f_Page_26.txt
e81bab6e386426298bb05f70f9438824
9ebfc6f0fdaed36a400c56671177aec6a6ba9c83
F20110209_AABKKO wang_f_Page_93.tif
e3b62590431e7c24b6fe843d36766592
43a6e4c9bb3ba4346488ec20e45014b47fc693a0
F20110209_AABKJZ wang_f_Page_48.tif
0563d82e5f038aeedc26315fdcdbcedd
cc33e87d4b7ed70af907b90e4bae31b58bb4fad8
2372 F20110209_AABKLD wang_f_Page_28.txt
59277d91147ca46fcb6eb5943fd42437
fc1d76f4f842cecb387271e116cf50929c3ef04b
F20110209_AABKKP wang_f_Page_94.tif
ece4a5d8735db4b616de7cbc28e3ed98
abdeebd028bd06fe73e566e88fcc535e19e514c2
2115 F20110209_AABKLE wang_f_Page_29.txt
ef6946b53935bfb27246261a784fe197
ecdd8c8f08656678ec4bfb7214a08e7808de3106
F20110209_AABKKQ wang_f_Page_95.tif
a764d5555be9745e40941ece62bfb222
66b8cad2b5e1fb39fda8df1537cda758db572287
2054 F20110209_AABKLF wang_f_Page_33.txt
442d454e3dd011a11046bae3633133a2
099b99293a5b9bc0e526def1a37a69a9fc1319ef
F20110209_AABKLG wang_f_Page_35.txt
e69df5c31ca62975ec1dc57f41d42988
7431994f5263c7021764b42dcd5969e83e313465
64 F20110209_AABKKR wang_f_Page_03.txt
3f37fd3b7d933eb713134ec0d6f4deaf
cde6ec012a043fe7d3ceb17e89ddbb4339112823
F20110209_AABKLH wang_f_Page_38.txt
a8c55cd690347dd4197e526b2b79070b
9a2464455ee1ef7b05c3f5fe80fd5762828517eb
2002 F20110209_AABKKS wang_f_Page_04.txt
0368805414284d6a7aad3cc3bb209d97
c442706e2bf82578a67ffcd388435db12480c012
1617 F20110209_AABKLI wang_f_Page_39.txt
731eddba3106d5ff40465fd97eeb3d74
b1bcc18cfe863fcdbb9976ed64e12d4d13090560
818 F20110209_AABKKT wang_f_Page_05.txt
44faa87da01250c6a9dfbafbde6cea80
b823455d7e1c8e273311e2ab28c0476d27384051
2244 F20110209_AABKLJ wang_f_Page_41.txt
bf868b9c50a681f5a57691023e2c4ebf
f8e070d974bbc59e245f8bbb418c1760ebe464fb
1871 F20110209_AABKKU wang_f_Page_11.txt
9f229be9eceb8054e8e5d0e45bc2f7c6
f9c8440edb446e90a11ca61a9b7c87d652029eab
2246 F20110209_AABKLK wang_f_Page_43.txt
c1a2e895672cf419e8495e46d37cb8dc
13fb25a89c09cc514569559194fa77fc1acadda8
1977 F20110209_AABKKV wang_f_Page_14.txt
51273f90136bc212381ef69047fbb3e9
21f6e62340c62591be389ae39ac690f02005ad55
1205 F20110209_AABKLL wang_f_Page_47.txt
e842108679b414bfc648c43a3348ef0f
1b42548b13d410c5662187f8dbce201d95ade5ad
2129 F20110209_AABKKW wang_f_Page_15.txt
67e9954e95c9d190770b5a3c7a84ce32
923fc66c129ece947d30057cb6d16119483831f9
2294 F20110209_AABKLM wang_f_Page_48.txt
be1f03fb79b51bba4f82376e665b61bc
0be2636762cf8fe83e0c8ec9501316cc5f89aaf1
1816 F20110209_AABKKX wang_f_Page_18.txt
041be6f0bac7335c70c0f216cd51ca1a
07431281e207ab640ddd499b607c703bd2c55bad
1578 F20110209_AABKMA wang_f_Page_81.txt
81ec3fe5ce2da97ecca47f11e66f44a8
88f12f5621b5024d68e62497719bcbf11fc05da4
1964 F20110209_AABKLN wang_f_Page_52.txt
8cbabcd8ee342bb9d1d892e18ca71ba6
8b0a9ab140a9b80c2b94d34db421c6f5a8f53373
2279 F20110209_AABKKY wang_f_Page_19.txt
ba0c4c1ee245be599d4ee0501fbf9069
53b1d3647854bd3fb5d6ce17fcc599b152d5045b
1525 F20110209_AABKMB wang_f_Page_82.txt
9c3332fa7bbf9bc639a2051e13aa5660
487b4ac0c1cee4a7a074a7fde314f6942f203bc0
2189 F20110209_AABKLO wang_f_Page_59.txt
14b01a913e7a3f0758251dcd076a1cdc
c6fb78bc99015b85e8cbcd5b337e7ec569caa290
2080 F20110209_AABKKZ wang_f_Page_20.txt
0a1953ea95b16695cc2203d25f6d71b8
29009b0b089844797a81cb10e096c50220922876
2196 F20110209_AABKMC wang_f_Page_86.txt
9ead79b087103ffadb201be648dcc509
21d6fe0c253b7e5dbe57d18f888e56ee6060fa03
2208 F20110209_AABKLP wang_f_Page_60.txt
d8e782f73b7322ab7e213cb55e049574
a019a23366af83337403ad922f22f79dc3cbb2d1
2601 F20110209_AABKMD wang_f_Page_92.txt
d47e8e565d515a72f4629db3a9b23c6b
d914ba1741ab9a0b83801e42f9e6fb370984f348
1836 F20110209_AABKLQ wang_f_Page_62.txt
0c343a5365ad876c75a2e3c39da33f98
ee4c9dbddacf29c9e82583513d6ee536242da92f
989 F20110209_AABKME wang_f_Page_02.pro
1786bc730f158a3e94f8c079806ce38f
f37d027b2d7175a4abd54028ab54ef21e1a338a7
F20110209_AABKLR wang_f_Page_64.txt
94789d29e16d9a798cd4b9f067940d0e
97a554cea21e7ba58c8394b3dd0d1b88d898b7c6
54201 F20110209_AABKMF wang_f_Page_06.pro
146f01f20d18d72c9b95ce5202e63d12
82fa3376bd5108f25dba06223782c4e60faeac20
45621 F20110209_AABKMG wang_f_Page_13.pro
42dcb5c1fe83847a0155ff4c8586bbcf
5e6647b5f61c7adf6b6523a3ae5f19a8b99609ca
2671 F20110209_AABKLS wang_f_Page_67.txt
60e01d27ca14cd9705b9217ca7e14af1
569657ef6b6d941fe79e5f1f471b4fe7aa500bed
43133 F20110209_AABKMH wang_f_Page_14.pro
ba452bcf5f39416c7440763ff38c77ec
a180b4be56aefbbc34b9758b02cd35f8b09ceaef
2344 F20110209_AABKLT wang_f_Page_68.txt
1d210d826db578231af49ef2b6a921c6
8e2f19f1c1bca7c23c0d6df0ff941742a02d27dd
50673 F20110209_AABKMI wang_f_Page_15.pro
045a659d29a28250c01c332a7adf64a0
e756ce9341308708a5e2b1401539706c7d574a9f
694 F20110209_AABKLU wang_f_Page_70.txt
f0098a76365d5bfef8da3952f1bbe944
8450e2785596caa994a855605a66331490903bc7
30933 F20110209_AABKMJ wang_f_Page_16.pro
0ce929ef005af3c2f3f336ba59841c5d
9db4bf028f770c8cb4b648108a8e2e602f876617
41353 F20110209_AABKMK wang_f_Page_18.pro
3490da3d28f32bd3d627ec1cf519a315
45735dafc2d49a3f65071e71400706f356c099c7
2098 F20110209_AABKLV wang_f_Page_72.txt
0e1dcfe77271c6c86722f9046d882cf4
533a61b045519211ff32d407e24cbc2e6d9b5cb7
58008 F20110209_AABKML wang_f_Page_19.pro
64094ca529e727e72d3f966c38c2f127
adf78cd4fc7164ad49fb4f51931fef2ddfce03a6
F20110209_AABKLW wang_f_Page_73.txt
59faaaa3f9e9c63b28777c65a722f7dc
5c0263bc16ff8628ef9de1d33f30b826a1caae1d
52651 F20110209_AABKNA wang_f_Page_72.pro
9c9565bbacb7ae7cc6959a4bcaf301c8
fc8cee7e71a977494c47760968470d5965d8f4e1
17296 F20110209_AABKMM wang_f_Page_25.pro
5b26e918978fc846cbbd551113ccef30
a9bd3d4292d497272896d3c3bf72ddfbe4de2a8c
1969 F20110209_AABKLX wang_f_Page_76.txt
9c319bc6235df341f1d6ffe4786171fc
048426a6da657c093a76f5004e1003637a340df2
47762 F20110209_AABKNB wang_f_Page_74.pro
d66981a35ad8c77e06fb831c4e797b75
120c333fd25a65bd4b8e74ea76d8e060b86bac6e
50134 F20110209_AABKMN wang_f_Page_26.pro
4c9bd633f336aa0b90fcd5229e99fae1
bf137571e76ff071f96baa5fe359f994899a85dd
1848 F20110209_AABKLY wang_f_Page_77.txt
f3fda1d1f3e00963e1d2bc4818ccf989
a319357fb16b4756aee96ce4989a462e1bfb26da
50136 F20110209_AABKNC wang_f_Page_75.pro
6e73281e346093378d85891e612528d2
692fef2d5a9c23a30c889ed4a4544b2150970b38
60573 F20110209_AABKMO wang_f_Page_28.pro
9d2c120bf51c26b392a429195a4298dd
0ac7e5a32aa5a3317b4b4020960edd2d61a33a67
2229 F20110209_AABKLZ wang_f_Page_80.txt
b7fa1787fb9e6a6268a76b602cb6b22e
70e1873e4cc32cdd1b62f7d7b6beda7e7a02c617
48520 F20110209_AABKND wang_f_Page_78.pro
d7b48973fdd502063910c385eb3d1b12
09fd338f0c22093ee6e1ed860210366bf9f54e8e
43255 F20110209_AABKMP wang_f_Page_31.pro
aadb553aa4088912ae96d4ecaca7bcdf
534de0f31325773947c2e94acacc73a3090f3344
39279 F20110209_AABKNE wang_f_Page_79.pro
4857dc5496d6dd6aad3a4de4463d2391
ac41e82af47147ab8928d23069679e53ad55f8aa
45555 F20110209_AABKMQ wang_f_Page_32.pro
976a365f798881610fcbc4b4c883fb18
4073a0fcc6e5d922a78cd2ddb3e837c0245cd3ca
52556 F20110209_AABKNF wang_f_Page_80.pro
816e47a734237774ae021c20c801bce3
ed76c1c3f6adca3f91f803925ee73c6179e24b38
28030 F20110209_AABKMR wang_f_Page_36.pro
8ab463d0a85bf4300e06118cdfa5a06c
364bd27e5883bccfdedde564ed748528a680f073
39222 F20110209_AABKNG wang_f_Page_81.pro
eb0ab41373b8073bd7993d2b7481b2bb
e4aa7e665c12e67b653ef0463a9c6abb2341935d
41923 F20110209_AABKMS wang_f_Page_37.pro
ec87c660e8add4f1b346c82184ab58be
1a38f4a1c4c61dff68a0002d1b6664663dcd18e7
37291 F20110209_AABKNH wang_f_Page_82.pro
d8d0eb7441de37c4a3607110992acacc
b7015d89fc61f6f503cc47cadd0ea339e24ce3f4
41925 F20110209_AABKNI wang_f_Page_84.pro
15efbbb66eddf9566e8993ef58c1f2b0
999f68205aaf94ce8c9bd235f7e9c40d6eae05bc
34714 F20110209_AABKMT wang_f_Page_44.pro
7b3763c9f17debc33ad291b2a32611c8
a459cdd3a83395b5ed4847950604e225ab7f5971
55361 F20110209_AABKNJ wang_f_Page_86.pro
86598375d20955469883a9f6101bde35
133fa73dcbaccb9ce44868f48066a74b164697e9
58374 F20110209_AABKMU wang_f_Page_49.pro
b270df07158807f43aa165f234be8e36
f7610273998fad89d8a0b954505f501a7fed1b84
67696 F20110209_AABKNK wang_f_Page_90.pro
64c101048493419f50f705837fdc2e92
27f8ceacdd48c3535671e2ddd66e221d4ac9715a
38719 F20110209_AABKMV wang_f_Page_62.pro
e67f28d72afdaa2529a2ca7d7486d530
f1b18c65dd1c3a8bc75bd2893d15904473db85ce
64287 F20110209_AABKNL wang_f_Page_92.pro
86222cb27ec6f4c0d5af8cd0e0e29e8d
59c794de522705662d745581cb1f478054a7f41b
34202 F20110209_AABKMW wang_f_Page_66.pro
f4f282afc6b8beb5eaf30519dd14ad97
7f7887255314b2c449902acf3a11f27ee1048b60
27965 F20110209_AABKNM wang_f_Page_01.jpg
22e2a9bc4ffc16937b4b9c7f141d46a0
f05d51b209910a567070c5c575400cda4dfd1e6e
54391 F20110209_AABKMX wang_f_Page_67.pro
51c75a1eaa767a3b57ca478301e91a9e
1f97a219876315c5e7a41e0bca004c25a98d4bf9
95765 F20110209_AABKOA wang_f_Page_31.jpg
8ab99d9c4afbb66851822b2b06fb55e2
b8d6f3def2d3457a289e8fb0d0edd33a2c55faa3
4458 F20110209_AABKNN wang_f_Page_02.jpg
a16fbf4d6df577edc30fa672f7444ac0
98f9e8aaa1c16bf3c8fdcfb8b1974ca3953a89b7
38003 F20110209_AABKMY wang_f_Page_69.pro
6532f73aa62e7fc8ab165d72b6ca5547
82b99342600a06b5fc138db0799d0c9ee8c0a62e
96089 F20110209_AABKOB wang_f_Page_35.jpg
85f9fcf8e1ceaff5eabf395155e62b21
402fb5975d36fa664be4d76c92b83827e4cafdd5
4550 F20110209_AABKNO wang_f_Page_03.jpg
e4c64e5cb003d05046dbcd1003426681
d646c1e74424a779314ec1682875781b231dd109
14761 F20110209_AABKMZ wang_f_Page_70.pro
c8bf387b78ca2e10a16ab575f647cd59
1f3533f85704cc78d28302c64b1968e4e2a13332
61814 F20110209_AABKOC wang_f_Page_36.jpg
c5c953837ae2b870ef19d64a4ecb2795
193072d4994d922ee8254aa9cdaca9321287f423
47191 F20110209_AABKNP wang_f_Page_05.jpg
8472a6fcd63272440a11b678915129b0
a8ebba14014f31bebd4d7b742cc4261e454a98c7
116188 F20110209_AABKOD wang_f_Page_38.jpg
344609cf40aaff20ba4f98aae2600db5
2a98d64efc67ed016daf30af407b9cc2d7b4c29b
84048 F20110209_AABKNQ wang_f_Page_06.jpg
d84ed21fa026f07f833df48976084304
941039ea8a4d6b3bf0700ec0a858d5d6e6db2314
73341 F20110209_AABKOE wang_f_Page_40.jpg
78d6dd15bcfed75ae0b29d7e80c0ae83
cfc3aa3f9396134ad306415df098dbd7466f052c
94963 F20110209_AABKNR wang_f_Page_09.jpg
b83f16bfef141f293438ecced5988c00
35ccb9e7c53d36b1dbedc8ab2a032eb2d4797584
119167 F20110209_AABKOF wang_f_Page_43.jpg
75eb56d764e6ed65da829e64d87c0b1c
11fa18bc137106343cdc809ada818078a628e3ae
127424 F20110209_AABKNS wang_f_Page_12.jpg
0f4c9c6f9d8b85c578bef90efae5b46d
bc1b8d46e848ce997e9f7da6e29e57453b5956d4
114209 F20110209_AABKOG wang_f_Page_46.jpg
83b009c8a8e9e354f15e40bee6f1546f
367b5dcb989f78c5757fe2f3e2c26f9a4d5daf54
98200 F20110209_AABKNT wang_f_Page_13.jpg
ea0fb6a7542df1eccd4fe1ebe46a8062
e015a75e3d7892c2124bc07b3bc5b4494139216f
117374 F20110209_AABKOH wang_f_Page_50.jpg
3f5123d2c166509c2be2d389fcdb30ff
cbeba099e8bc9984488aa617cebe7caf872341c9
105238 F20110209_AABKOI wang_f_Page_58.jpg
9804f9901e8dcc85cdadf95ec9cc3447
340e96725abb4250d26dbf60ca633b24a4ef9865
74207 F20110209_AABKNU wang_f_Page_16.jpg
8aaa4730c1da34745ebe30399a7a9a87
3f7788ca5b716f479cac0115c0024364b24e1596
116925 F20110209_AABKOJ wang_f_Page_59.jpg
6e9156307c1d19db5e4e194e5f138dab
57f2910a8d809836daafa8f204a45faec5a65828
93298 F20110209_AABKNV wang_f_Page_17.jpg
cd985e1c725ad13eda5793ae2bd406ec
cff621c648802a19c3e4256fa84cfd325fba97df
92241 F20110209_AABKOK wang_f_Page_62.jpg
9e12f856377030be0e83e3ded17b27a7
534f5d9764f64799f2cf59105e8733be14cbdc33
56265 F20110209_AABKNW wang_f_Page_22.jpg
4e44601081143a2dfccb84dcda420fb7
4a6a1e733f2892836765c43a0b77811c693fbf37
78510 F20110209_AABKOL wang_f_Page_63.jpg
9bc02bbaed88397338efa71af488ecbd
d77f25184a4e35a22de934e404ef80a8e5472e7c
74704 F20110209_AABKNX wang_f_Page_23.jpg
c2279e6f0c14d70c9376b15369b1aad7
1f7d8e5c4a6ec8efaac64f0cc7469fb38ecef16a
986845 F20110209_AABKPA wang_f_Page_17.jp2
7d8a4e697c7daf82b24bca648923bdb9
1901f2d307227797936e7ab710f80c62f86b5933
77838 F20110209_AABKOM wang_f_Page_66.jpg
a57dd7b1c893b30b813e0938500adef0
87aab6311e2214f0e6dbbea4d408b427c36718fe
38244 F20110209_AABKNY wang_f_Page_25.jpg
c7c87daada9fb2629606723b4b1338de
64d6d11b88ceb7910e737138f4dca16246fb9017
F20110209_AABKPB wang_f_Page_19.jp2
db15ba530e9d38d2fe4382daccd4a5ce
61e3b1f934d5a8b4ca81d4b95029dc4ea012322d
71042 F20110209_AABKON wang_f_Page_69.jpg
2e27f7e4ce93dfd4f653aa0f0b1b5f16
14e7346cdcda52a28637c1d158e6c1b3ac24ed93
112477 F20110209_AABKNZ wang_f_Page_29.jpg
6983aedaf98e504ea2f205a03cea7631
dd0462b3085179ee96c3c50c4917f280a3f77e52
914248 F20110209_AABKPC wang_f_Page_24.jp2
62a089ce44b8d8dbc140c629fd9cf908
10108f7df0f3e8441a9e36c3531cb85f9f1415be
111242 F20110209_AABKOO wang_f_Page_71.jpg
7c67d76653c279f59d1156cb951174eb
201e812de21625bd47e92dcf84941425c5adf715
371795 F20110209_AABKPD wang_f_Page_25.jp2
b23474315ddc733f6e9a75f9c61a6903
a8164ee003b0cc0dba689acdd92c534f4a4896d3
101884 F20110209_AABKOP wang_f_Page_74.jpg
0b91746a15004da1a75f05083977f21a
73ac49abca985de97317a2e144246fcec5903238
F20110209_AABKPE wang_f_Page_28.jp2
fbe17d54ddb3e0eb080b82644e2ab1c7
256f7dd55555408568f893dd4778c10ea4759064
102858 F20110209_AABKOQ wang_f_Page_78.jpg
f6b7475b582fd1fd3fdb28b58da5f551
2298779abd765746c3d9c836f85973f9bbced542
1084974 F20110209_AABKPF wang_f_Page_30.jp2
549791ddfe37d31f48c5e4920dc86d21
d1cf5ab5a73f83633f53f1733aeeda123f71e1f1
111659 F20110209_AABKOR wang_f_Page_79.jpg
26979d824fdd7e62b927393fba76182a
e6a1d9a3af767421d2430e500b9244886a17fd02
948948 F20110209_AABKPG wang_f_Page_31.jp2
a80abc5d51e5664cc100f716f86e38eb
b8e54cbfa260f002d3049d63447c9ffe4e42a56b
100572 F20110209_AABKOS wang_f_Page_81.jpg
85c6b6d033b9f1de4e2a1ec8f43b511d
a86f4b09ec0757621877e9993722f7e442d4d309
926006 F20110209_AABKPH wang_f_Page_32.jp2
047e15e3d64b80b411678bb343f3df44
3d72df18b70b7b52ca589b72339fa4aa24229e15
141187 F20110209_AABKOT wang_f_Page_92.jpg
66055a44536b0c2b02958f18c6a86eed
86ef0980b2431a81fc90175e0b0144fe823bf16b
956051 F20110209_AABKPI wang_f_Page_35.jp2
ef2abdc1868e88adf29144b3ed9dd348
c64599ea3c200e192351d0abad05a539bf95e0b0
23325 F20110209_AABKOU wang_f_Page_02.jp2
fb15641863b3c4bea3191187e57d1ab3
40866101788a1479ab0e2e63213dbce2e7a17901
920264 F20110209_AABKPJ wang_f_Page_37.jp2
5a4f0b9be7276edad9b5239f8d75b9fa
8abdba4f68aa6d9c1975305deffeae6213728748
1036181 F20110209_AABKPK wang_f_Page_39.jp2
b5c9f3b589fbbd4d272acdb25e0dc218
4fbc6dc42f896b085a13b36e8afd321b28f5eb4d
23722 F20110209_AABKOV wang_f_Page_03.jp2
96971ec96ce1fc6ee6a74ad299de6824
bc2918f16db9b75411940a2c00a7cd5196d0d0c5
786955 F20110209_AABKPL wang_f_Page_40.jp2
08f872bd3bf48c1a402277712209639d
48e265148a468289b67ef347fb89251d1e2eee3d
F20110209_AABKOW wang_f_Page_04.jp2
58a38bc9524d414ca571c0e44ea18d6d
3b0180521c9b0c6a18beb4263f7c427fb421e323
811837 F20110209_AABKQA wang_f_Page_69.jp2
f4943a25079d36dc619ec458c5cb4c04
d9a44fd7a81683ea5f948a1929cf9844304ea909
1087894 F20110209_AABKPM wang_f_Page_41.jp2
8ae2e24550f94a270bf40ff4f6cb20cf
5c8573add2ab79b67ea6f6ab46d2cf6b2c247210
716710 F20110209_AABKOX wang_f_Page_06.jp2
1090943092806bb6055636fc5b8f1aa1
c07e34104329846fa27efa6c4346dcedc57a5295
503029 F20110209_AABKQB wang_f_Page_70.jp2
8a21b1f14750888e51c40fad3557335d
087de79437d7d4224a11cea4647a6a9a83d6cb12
1049144 F20110209_AABKPN wang_f_Page_44.jp2
4e89a225518ca6cdbbd6d566543ebdc8
8a514382816617e21f56732b843bfec997a0a38a
853636 F20110209_AABKOY wang_f_Page_07.jp2
deacf7f0045bcc5fa2fafbfeee4b68b9
b27f444094ae1409e23b750edabaf38e494f0653
934729 F20110209_AABKQC wang_f_Page_73.jp2
acb5ca1b0e98aa67dcb9f025e9278227
94989ed69d093e3185364832afc3e822cc20a321
1027969 F20110209_AABKPO wang_f_Page_45.jp2
9d7b8dc9c7bb54a03df45442ef5b102f
d3a61ea6d2663ce2af5463ba5d8ed6dc030a0d39
852922 F20110209_AABKOZ wang_f_Page_16.jp2
e83aa7a08ea9e46affe0ffa8ecac7950
dae8d46aefe2207e6c0aeebd16f027f82b275e12
1087872 F20110209_AABKQD wang_f_Page_75.jp2
45768056c995c89c335d73e250b88297
2f0784b1e34a27e1537ca64796f3005e2b725aa9
856781 F20110209_AABKPP wang_f_Page_47.jp2
afaf4fc4ba4d1dbd322c599fb323269f
1f3d7f30dd7ae0d6ca15bffad1647b369dd1e926
990827 F20110209_AABKQE wang_f_Page_76.jp2
c53b77e36c4011f219c5df4bbc1a9f60
680f332e046d47fbf24681fe76be7284dcb1cbcd
F20110209_AABKPQ wang_f_Page_48.jp2
902ce0da22ed0c4dbafe3c544d297bb0
f876d3c70965c7f268e64a6847a4964644a22c04
691614 F20110209_AABKQF wang_f_Page_77.jp2
de05fcbc1ab9beb1716b164620620c76
fd21e261431c59b998669248fba6e5f3729919c3
1087886 F20110209_AABKPR wang_f_Page_49.jp2
0492c58a46751f24aae863b68dd979b0
48527f351ba243129d11ac3131e7ebf04db7308f
F20110209_AABKQG wang_f_Page_79.jp2
ab9f94a1bc2ded4adbadb81ef202f205
396e3f2918ae3fe0883d32d7f6f035c0169aaeb9
F20110209_AABKPS wang_f_Page_50.jp2
c8cb0aa8a47ab3472b31a6bc37597844
14ecb5a5e3083935e010a3e74fa03697fa72b925
1087847 F20110209_AABKQH wang_f_Page_82.jp2
dc88d1a726dcff3f10d4a1f7fdaa5b32
b2af076c5145fbbab7831a4f61350207e8c8a504
562195 F20110209_AABKPT wang_f_Page_53.jp2
32731b76a0a07151fce670be3a3ea47b
fcbc3ed9ee911b30ac16a90e4b20ef52fd778cb7
448166 F20110209_AABKQI wang_f_Page_83.jp2
6e8503987e7c0ddc45e41a0c5976eb80
de13084b7497180600edb1291eacc7ac4405307e
1087893 F20110209_AABKPU wang_f_Page_56.jp2
284fcf1048ad9c8dfeab84f099ec112d
2896c1b314c585c34a9fa901404f97b0335525b5
1087856 F20110209_AABKQJ wang_f_Page_85.jp2
6b5c53fc864aee714ef4a113dd5909f6
49ebe2a72326ca50d1df098122e402950a548a80
F20110209_AABKPV wang_f_Page_57.jp2
eed2de913becaf8735c4334f386f0435
c113c7de533625482386243cd779b92ca28c5ca5
128191 F20110209_AABKQK wang_f_Page_87.jp2
94acb192aaa1bc3485385581cb9ab39b
2b484d8f4e3d599262fe35efac3d02187d2eaaee
F20110209_AABKQL wang_f_Page_91.jp2
e3bc80326ef61f2db5cd6e9aa0652653
6b725d181537c9440bc704ebe050a2df75d31872
1087803 F20110209_AABKPW wang_f_Page_58.jp2
9991a010e41426bd273057cef8a3ad03
acc5813e7d0ee3f0f3116dcad5bdcba0656cb004



PAGE 4

Iwouldliketorstthankmyadvisor,Dr.BabaC.Vemuri,foreverythinghehasdoneformeduringmydoctoralstudy.Thisdissertationwouldnothavetakenshapewithouthisinvaluableinput.Dr.Vemuriintroducedmetotheeldofmedicalimageanalysis.Hisinsightandexperiencehaveguidedmethroughoutmyresearchduringwhichtimeheprovidednumerousinvaluablesuggestions.Itwasagreatpleasureformetoconductthisdissertationunderhissupervision.IwouldalsoliketothankDr.AnandRangarajan,Dr.SartajSahni,Dr.ArunavaBanerjeeandDr.TanWongfortheirwillingnesstoserveonmycommittee.Inaddition,specialthanksgotoDr.JorgPetersforattendingmyPhDoralexamination.Mydoctoralresearchisahappycooperationwithmanypeople.Dr.Vemurihasbeeninvolvedwiththewholeprocess,Dr.Rangarajanhasguidedmealotinthegroupwisepointregistrationpart,Dr.IlonaSchmalfussandDr.StephanEisenschenkhavekindlyprovidedthedataforhippocampalsegmentationandtaughtmewhatlittleIknowofNeuroscience.IhavealsobenettedfromDr.ThomasE.Davis'sguidancewhenIrstjoinedthelab.IwouldalsoliketothankDr.Banerjeeforstimulatingdebates,andDr.JeffreyHoforhisprofessionaladviceandphilosophicaldiscussions.Thanksalsogoestomyco-authors,Drs.MuraliRaoandYunmeiChen,whoweremyco-authorsonasetofpapersthatintroduced,theconceptofentropybasedonprobabilitydistributionsandseveralpropertiesofthesame.Needlesstosay,IamgratefulforthesupportofmycolleaguesandfriendsattheComputerandInformationScienceandEngineeringDepartmentattheUniversityofFlorida.Dr.ZhizhouWang,Dr.JundongLiu,Dr.TimMcgraw,Dr.EricSpellman,BingJian,SanthoshKodipaka,NicholasLord,NeetiVohra,AngelosBarmpoutis,SenihaEsen iv

PAGE 5

v

PAGE 6

page ACKNOWLEDGMENTS ................................ iv LISTOFTABLES ................................... viii LISTOFFIGURES ................................... ix ABSTRACT ....................................... xi CHAPTER 1INTRODUCTION ................................ 1 1.1ImageandPoint-setRegistration ...................... 1 1.1.1ImageRegistration ......................... 1 1.1.2GroupwisePoint-setsRegistration ................. 3 1.2ImageSegmentation ............................. 4 1.3OutlineofRemainder ............................ 5 2ENTROPYANDRELATEDMEASURES ................... 6 2.1ShannonEntropyandRelatedMeasures .................. 6 2.2CumulativeResidualEntropy:ANewMeasureofInformation ...... 8 2.3PropertiesofCRE .............................. 10 2.3.1CREandEmpiricalCRE ...................... 11 2.3.2RobustnessofCRE ......................... 12 3APPLICATIONSTOMULTIMODALITYIMAGEREGISTRATION ..... 14 3.1RelatedWork ................................ 14 3.2MultimodalImageRegistrationusingCCRE ................ 17 3.2.1TransformationModelforNon-rigidMotion ............ 21 3.2.2MeasureOptimization ........................ 21 3.2.3ComputationofP(i>;k;)and@P(i>;k;) 23 3.2.4AlgorithmSummary ......................... 25 3.3ImplementationResults ........................... 25 3.3.1SyntheticMotionExperiments ................... 26 3.3.1.1Convergencespeed ................... 26 3.3.1.2Registrationaccuracy .................. 28 3.3.1.3Noiseimmunity ..................... 29 3.3.1.4Partialoverlap ...................... 30 vi

PAGE 7

....................... 31 4DIVERGENCEMEASURESFORGROUPWISEPOINT-SETSREGIS-TRATION ..................................... 34 4.1PreviousWork ................................ 36 4.2DivergenceMeasures ............................ 38 4.2.1Jensen-ShannonDivergence ..................... 38 4.2.2CDF-JSDivergence ......................... 40 4.3Methodology ................................ 42 4.3.1EnergyFunctionforGroupwisePoint-setsRegistration ...... 43 4.3.2JSDivergenceinaHypothesisTestingFramework ......... 44 4.3.3UnbiasnessPropertyoftheDivergenceMeasures ......... 45 4.3.4EstimatingJSanditsDerivative ................... 47 4.3.4.1Finitemixturemodels .................. 47 4.3.4.2OptimizingtheJSdivergence .............. 49 4.3.5EstimatingCDF-JSanditsDerivative ............... 50 4.3.5.1OptimizingtheCDF-JSdivergence ........... 52 4.4ExperimentResults ............................. 53 4.4.1JSDivergenceResults ........................ 53 4.4.1.1Alignmentresults .................... 53 4.4.1.2Atlasconstructionresults ................ 55 4.4.2CDF-JSDivergenceResults ..................... 56 5APPLICATIONSTOIMAGESEGMENTATION ................ 59 5.1RelatedWork ................................ 59 5.2Registration+SegmentationModel ..................... 60 5.2.1Gradientows ............................ 63 5.2.2AlgorithmSummary ......................... 66 5.3Results .................................... 66 6CONCLUSIONSANDFUTUREWORK .................... 72 6.1ContributionsoftheDissertation ...................... 72 6.2ImageandPoint-setsRegistration ..................... 72 6.2.1Non-rigidImageRegistration .................... 72 6.2.2GroupwisePoint-setsRegistration ................. 73 6.3ImageSegmentation ............................. 74 REFERENCES ..................................... 76 BIOGRAPHICALSKETCH .............................. 83 vii

PAGE 8

Table page 3ComparisonoftheregistrationresultsbetweenCCREandMIforaxedsyn-theticdeformationeld. ............................. 30 3ComparisonoftotaltimetakentoachieveregistrationbytheCCREwithMI. 31 3ComparisonofthevalueSofseveralbrainstructuresforCCREandMI. .... 33 5Statisticsoftheerrorinestimatednon-rigiddeformation. ............ 68 viii

PAGE 9

Figure page 1Illustrationofgroupwiseregistrationofcorpuscallosumpoint-setsmanuallyextractedfromtheoutercontoursofthebrainimages. .............. 4 3CCRE,MIandNMItracesplottedforthemisalignedMR&CTimagepair .. 20 3ComparisonofconvergencespeedbetweenCCREandMI ........... 27 3PlotdemonstratingthechangeofMeanDeformationErrorforCCREandMIregistrationresultswithtime. ........................... 28 3Resultsofapplicationofouralgorithmtosyntheticdata(seetextfordetails). 28 3RegistrationresultsofMRT1andT2imageslicewithlargenon-overlap. ... 30 3RegistrationresultsofdifferentsubjectsofMR&CTbraindatawithrealnon-rigidmotion.(seetextfordetails. ........................ 32 4Illustrationofcorpuscallosumpoint-setsrepresentedasdensityfunctions. .. 35 4Resultsofrigidregistrationinnoiselesscase.'o'and'+'indicatethemodelandscenepointsrespectively. ........................... 54 4Non-rigidregistrationofthecorpuscallosumpointsets. ............. 54 4Experimentresultsonseven2Dcorpuscollasumpointsets. ........... 55 4Robustnesstooutliersinthepresenceoflargenoise. ............... 57 4Robustnessteston3Dswandata ......................... 57 4Atlasconstructionfromfour3Dhippocampalpointsets. ............ 58 5ModelIllustration ................................. 61 5Illustrationofthevarioustermsintheevolutionofthelevelsetfunction. ... 65 5Resultsofapplicationofouralgorithmtosyntheticdata ............. 67 5ResultsofapplicationofouralgorithmtoapairofslicesfromhumanbrainMRIs ....................................... 69 5CorpusCallosumsegmentationonapairofcorrespondingslicesfromdistinctsubjects. ...................................... 70 ix

PAGE 10

.................................. 71 x

PAGE 11

Informationtheoryhasplayedafundamentalroleinmanyeldsofscienceandengineeringincludingcomputervisionandmedicalimaging.Inthisdissertation,variousinformationtheoreticmeasuresthatareusedinachievingthegoalofsolvingseveralimportantproblemsinmedicalimagingnamely,imageregistration,point-setregistrationandimagesegmentationarepresented. Tomeasuretheinformationcontentinarandomvariable,werstpresentanovelmeasurebasedonitscumulativedistributionthatisdubbedCumulativeResidualEntropy(CRE).Thismeasureparallelsthewell-knownShannonentropybuthasthefollowingadvantages:(1)itismoregeneralthantheShannonentropyasitsdenitionisvalidinthediscreteandcontinuousdomains,(2)itpossessesmoregeneralmathematicalpropertiesand(3)itcanbeeasilycomputedfromsampledataandthesecomputationsasymptoticallyconvergetothetruevalues.BasedonCRE,wedenethecross-CRE(CCRE)betweentworandomvariables,andapplyittosolvetheimagealignmentproblemforparameterizedtransformations.ThekeystrengthsoftheCCREoverusingthenowpopularMutualInformation(basedonShannon'sentropy)betweenimagesbeing xi

PAGE 12

Jensen-Shannon(JS)divergencehaslongbeenknownasameasureofcohesionbetweenmultipleprobabilitydensities.Similartotheideaofdeninganentropymeasurebasedondistributions,wederivedaJSdivergencebasedonprobabilitydistributionsanddubitastheCDF-JSdivergence.WethenapplytheJSandtheCDF-JSdivergencetothegroupwisepoint-setregistrationproblem,whichinvolvessimultaneouslyregisteringmultipleshapes(representedaspoint-sets)forconstructinganatlas.Estimatingameaningfulaverageormeanshapefromasetofshapesrepresentedbyunlabeledpoint-setsisachallengingproblem,sincethisusuallyinvolvessolvingforpointcorrespondenceunderanon-rigidmotionsetting.ThenovelandrobustalgorithmweproposeavoidsthecorrespondenceproblembyminimizingtheCDF-JS/JSdivergencebetweenthepoint-setsrepresentedasprobabilitydistribution/densityfunctions.Thecostfunctionsarefullysymmetricwithnobiastowardanyofthegivenshapestoberegisteredandwhosemeanisbeingsought.WeempiricallyshowthatCDF-JSismorerobusttonoiseandoutliersthanJSdivergence.Ouralgorithmcanbeespeciallyusefulforcreatingatlasesofvariousshapespresentinimagesaswellasforsimultaneouslyregistering3Drangedatasetswithouthavingtoestablishanycorrespondence. Inthecontextofimagesegmentation,wedevelopedanovelmodel-basedseg-mentationtechniquethatinvolvessegmentingthenovel3Dimagedatabynon-rigidlyregisteringanatlastoit.Thekeycontributionheretothesolutionofthisproblemisthatwepresentanovelvariationalformulationoftheregistrationassistedimagesegmenta-tiontask,whichleadstosolvingacoupledsetofnonlinearPDEsthataresolvedusingefcientnumericalschemes.Oursegmentationalgorithmisadeparturefromearliermethodsinthatwehaveauniedvariationalprinciplewhereinnon-rigidregistrationandsegmentationaresimultaneouslyachieved;unlikeprevioussolutionstothisproblem,ouralgorithmcanaccommodateimagepairswithverydistinctintensitydistributions. xii

PAGE 13

In1948,motivatedbytheproblemofefcientlytransmittinginformationoveranoisycommunicationchannel,ClaudeShannonintroducedarevolutionarynewprobabilisticwayofthinkingaboutcommunicationandsimultaneouslycreatedthersttrulymathematicaltheoryofentropy.Hisideascreatedasensationandwererapidlydevelopedtocreatetheeldofinformationtheory,whichemploysprobabilityandergodictheorytostudythestatisticalcharacteristicsofdataandcommunicationsystems.Sincethen,informationtheoryhasplayedafundamentalroleinmanyeldsofscienceandengineeringincludingcomputervisionandmedicalimaging.Inthisdissertation,weendeavortodevelopnovelinformationtheoreticmethodswiththeapplicationtomedicalimageanalysis. Weexaminetwoapplicationsinparticular,image(point-set)registrationandimagesegmentation.Intherstoftheseapplications,wefollowapromisingavenueofworkinusingaprobabilitydensityordistributionfunctionasthesignatureofagivenobject(imageorpoint-set).Thenbyoptimizingcertaininformationtheoreticmeasuresbetweenthesefunctions,weachievethedesiredregistration.Inthesegmentationapplication,weconsideranatlasbasedapproach,inwhichsegmentationandregistrationaresimultaneouslyachievedbysolvinganovelvariationalprinciple. 1

PAGE 14

totheunknownparameterizedtransformationtobedetermined,deneamatchmetricM(I1(x;y);I2(x0;y0))andoptimizeMoverallT. Thefundamentalcharacteristicofanyimageregistrationtechniqueisthetypeofspatialtransformationormappingusedtoproperlyoverlaytwoimages.Thetransforma-tioncanbeclassiedintoglobalandlocaltransformations.Aglobaltransformationisgivenbyasingleequationwhichmapstheentireimage.Localtransformationsmaptheimagedifferentlydependingonthespatiallocationandarethusmoredifculttoexpresssuccinctly.Themostcommonglobaltransformationsarerigid,afneandprojectivetransformations. Atransformationiscalledrigidifthedistancebetweenpointsintheimagebeingtransformedispreserved.Arigidtransformationcanbeexpressedas whereu(x;y)andv(x;y)denotethedisplacementatpoint(x;y)alongtheXandYdirections;istherotationangle,and(dx;dy)thetranslationvector. Atransformationiscalledafnewhenanystraightlineintherstimageismappedontoastraightlineinthesecondimagewithparallelismbeingpreserved.In2D,theafnetransformationcanbeexpressedas where(a11a12a21a22)denotesanarbitraryreal-valuedmatrix.Scalingtransformation,whichhasatransformationmatrixofs100s2andshearingtransformation,whichhasamatrix(1s301)aretwoexamplesofafnetransformation,wheres1,s2ands3arepositiverealnumbers.

PAGE 15

Amoreinterestingcase,ingeneral,isthatofaplanarsurfaceinmotionviewedthroughapinholecamera.Thismotioncanbedescribedasa2Dprojectivetransforma-tionoftheplane. wherea0;:::;a7aretheglobalparameters. Whenaglobaltransformationdoesnotadequatelyexplaintherelationshipofapairofinputimages,alocaltransformationmaybenecessary.Registeringanimagepairobtainedatdifferenttimeswithsomeportionofthebodyexperiencinggrowth,orregisteringtwoimagesfromdifferentpatients,fallintothislocaltransformationregistrationcategory.Amotioneldisusuallyusedtodescribethechangeinlocaltransformationproblem. GivenNpoint-sets,whicharedenotedbyfXp;p2f1;:::;Ngg,eachpoint-setXpconsistsofpointsfxpi2RD;i2f1;:::;npggandnpisthenumberofpointscontainedinpoint-setXp.Thetaskofmultiplepointpatternmatchingorpoint-setregistrationiseithertoestablishaconsistentpoint-to-pointcorrespondencebetweenthesepoint-setsortorecoverthespatialtransformationwhichyieldsthebestalignment.Forexample,wearegivenagroupofcorpuscallosumpoint-setsfromthebrainimagescan,whichisshownintheleftcolumnofFigure 1 .Allthepoint-setsareregisteredsimultaneouslytothepoint-setsshownintherightcolumninasymmetricmanner,meaningthattheregistration

PAGE 16

resultisnotbiasedtowardsanyoftheoriginalpoint-set.WewilldiscusstheseissuesingreaterdetailinChapter 4 Figure1: Illustrationofgroupwiseregistrationofcorpuscallosumpoint-setsmanuallyextractedfromtheoutercontoursofthebrainimages.

PAGE 17

tosegmentobjectsinmultipleimagespriortoregistration.ThisisespeciallytrueinimagesfromdifferentmodalitiessuchasCTandMRI. Image-guidedsurgeryisanotherimportantapplicationthatneedsimagesegmen-tation.Recentadvancesintechnologyhavemadeitpossibletoacquireimagesofthepatientwhilethesurgeryisin-progress.Thegoalisthentosegmentrelevantregionsofinterestandoverlaythemonanimageofthepatienttohelpguidethesurgeoninhis/herwork. Segmentationisthereforeaveryimportanttaskinmedicalimaging.However,manualsegmentationisnotonlyatediousandtimeconsumingprocess,butisalsoinaccurate.Segmentationbyexpertshasshowntobevariableupto20%.Itisthereforedesirabletousealgorithmsthatareaccurateandrequireaslittleuserinteractionaspossible.

PAGE 18

1 ]inthecontextofcommunicationtheory.TheentropyShannonproposedisameasureofuncertaintyinadiscretedistri-butionbasedontheBoltzmanentropyofclassicalstatisticalmechanics.TheShannonEntropyofadiscretedistributionFisdenedby Sincethen,thisconceptandvariantsthereofhavebeenextensivelyutilizedinnumerousapplicationsofscienceandengineering.Todate,oneofthemostwidelybenetingapplicationhasbeeninnancialanalysis[ 2 ],datacompression[ 3 ],statistics[ 4 ],andinformationtheory[ 5 ]. Thismeasureofuncertaintyhasmanyimportantpropertieswhichagreewithourintuitivenotionofrandomness.Wementionthree:(1)Itisalwayspositive.(2)Itvanishesifandonlyifitisacertainevent.(3)Entropyisincreasedbytheadditionofanindependentcomponent,anddecreasedbyconditioning.However,extensionofthisnotiontocontinuousdistributionposessomechallenges.AstraightforwardextensionofthediscretecasetocontinuousdistributionsFwithdensityfcalleddifferentialentropyreads However,thisdenitionraisesthefollowingconcerns,1)Firstofall,itisdenedbasedonthedensityoftherandomvariable,whichingeneralmayormaynotexist,e.g.,for 6

PAGE 19

caseswhenthecumulativedistributionfunction(cdf)isnotdifferentiable.Itwouldnotbepossibletodenetheentropyofarandomvariableforwhichthedensityfunctionisundened;2)Secondly,theShannonentropyofadiscretedistributionisalwayspositive,whilethedifferentialentropyofacontinuousvariablemaytakeanyvalueontheextendedrealline;3)Shannonentropycomputedfromsamplesofarandomvariablelacksthepropertyofconvergencetothedifferentialentropy,i.e.evenwhenthesamplesizegoestoinnity,theShannonentropyestimatedfromthesesampleswillnotconvergetodifferentialentropy[ 5 ].Theconsequenceofwhichisthatitisimpossible,ingeneral,toapproximatethedifferentialentropyofacontinuousvariableusingtheentropyofempiricaldistributions;4)Considerthefollowingsituation:SupposeXandYaretwodiscreterandomvariablesrepresentingtheheightofagroupofpeople,withXtakingonvaluesf5:1;5:2;5:3;5:4;5:5g,eachwithaprobability1=5andYtakingonvaluesf5:1;5:2;5:3;5:4;7:5g(withYaoMinginthisgroup)againeachwithprobability1=5.TheinformationcontentmeasuredinthesetworandomvariablesusingShannonentropyisthesame,i.e.,Shannonentropydoesnotbringoutanydifferencesbetweenthesetwocases.However,ifthetworandomvariablesrepresentedthewinningchancesinabasketballgame,theinformationcontentinthetworandomvariableswouldbeconsideredasbeingdramaticallydifferent.NeverthelessShannonentropyfailstomakeanydistinctionwhatsoeverbetweenthem.Foradditionaldiscussiononsomeoftheseissuesthereaderisreferredto[ 6 ]. InthisworkweproposeanalternativemeasureofuncertaintyinarandomvariableXandcallittheCumulativeResidualEntropy(CRE)ofX.ThemainobjectiveofourstudyistoextendShannonentropytorandomvariableswithcontinuousdistributions.Theconceptweproposedovercomestheproblemsmentionedabove,whileretainingmanyoftheimportantpropertiesofShannonentropy.Forinstance,botharedecreasedbyconditioning,whileincreasedbyindependentaddition.Theybothobeythedata

PAGE 20

processinginequality,etc.However,thedifferentialentropydoesnothavethefollowingimportantpropertiesofCRE. 1. CREhasconsistentdenitionsinboththecontinuousanddiscretedomains; 2. CREisalwaysnon-negative; 3. CREcanbeeasilycomputedfromsampledataandthesecomputationsasymptoti-callyconvergetothetruevalues. ThebasicideainourdenitionistoreplacethedensityfunctionwiththecumulativedistributioninShannon'sdenition 2 .Thedistributionfunctionismoreregularthanthedensityfunction,becausethedensityiscomputedasthederivativeofthedistribution.Moreover,inpracticewhatisofinterestand/ormeasurableisthedistributionfunction.Forexample,iftherandomvariableisthelifespanofamachine,thentheeventofinterestisnotwhetherthelifespanequalst,butratherwhetherthelifespanexceedst.Ourdenitionalsopreservesthewellestablishedprinciplethatthelogarithmoftheprobabilityofaneventshouldrepresenttheinformationcontentintheevent.ThediscussionsaboutthepropertiesofCREinthenextfewsections,wetrust,areconvincingenoughforfurtherdevelopmentoftheconceptofCRE. 7 ]. whereRN+=xi2RN;xi0.

PAGE 21

CREcanberelatedtothewell-knownconceptofmeanresiduallifefunctioninReliabilityEngineeringwhichisdenedas: F(t)(2) ThemF(t)isoffundamentalimportanceinReliabilityEngineering&isoftenusedtomeasuredeparturefromexponentiality.CREcanbeshowntobetheexpectationofmF(t)[ 8 ],i.e. Nowwegiveafewexamples. Considerageneraluniformdistributionwiththedensityfunction: ThenitsCREiscomputedasfollows a)log(1x a)dx=1 4a Theexponentialdistributionwithmean1=hasthedensityfunction: Correspondingly,theCREoftheexponentialdistributionis

PAGE 22

TheGaussianprobabilitydensityfunctionis wheremisthemeanand2isthevariance. Thecumulativedistributionfunctionis: ); whereerfcistheerrorfunction: )logerfc(xm )dx We'llnowstatesimportantpropertiesthatarerelatedtotheapplicationofCREtoimageregistration.Foracompletelistofproperties,wereferthereaderstoamorecomprehensivetreatmentin[ 7 ]. 7 ].

PAGE 23

SimilartothecaseofShannon'sentropy,ifXandYareindependentrandomvariables,E(X;Y)=E(jXj)E(X)+E(jYj)E(Y).Moregenerally, 7 ]. Conditionalentropyisafundamentalconceptininformationtheory.WenowdenetheconceptofconditioninginthecontextofCRE. AsintheShannonentropycase,conditioningreducesCRE. E[E(XjY)]E(X)(2) 2 .Incontrast,inthisregardCREdoesnotdifferentiatebetweendiscreteandcontinuouscases,asshownbythefollowingtheorem:

PAGE 24

Shannonentropycomputedfromsamplesofarandomvariablelacksthepropertyofconvergencetothedifferentialentropy(seeEqn. 2 foradenition).Incontrast,theCRE,E(x)computedfromthesamplesconvergestothecontinuouscounterpart.Thisissummarizedinthefollowingtheorem. 7 ]fortheproof. Thisisapowerfulpropertyandasaconsequenceofit,wecancomputeCREofanrandomvariablefromthesampleswhichwouldconvergetothetrueCREoftherandomvariable.NotethatXkcanbesamplesofacontinuousrandomvariable.

PAGE 25

2 )andYnasabove.SupposeYn!0inprobability.Then 3 3 Theorems( 3 )and( 4 )areveryimportantpropertiesastheyprovethattheCREisrobusttonoisewhichisnotthecasefordifferentialentropy.Intuitively,therobustnessofCREmaybeattributedtotheuseofCDFasopposedtoaPDFinitsdenition,i.e.,anintegralformulationasopposedtoadifferentialformulationanditiswellknownthattheformerismorerobustcomparedtothelater.

PAGE 26

Matchingtwoormoreimagesundervaryingconditionsillumination,pose,acquisitionparametersetc.isubiquitousinComputerVision,medicalimaging,geographicalinformationsystemsetc.Inthepastseveralyears,informationtheoreticmeasureshavebeenverywidelyusedindeningcostfunctionstobeoptimizedinachievingamatch.Anexampleproblemcommontoalltheaforementionedareasistheimageregistrationproblem.Inthefollowing,wewillreviewtheliteratureonexistingcomputationalalgorithmsthathavebeenreportedforachievingmultimodalityimageregistration,withthefocusonthenon-rigidregistrationmethods.Wewillpointouttheirlimitationsandhencemotivatetheneedforanewandefcientcomputationalalgorithmforachievingourgoal. Severalfeature-basedmethodsinvolvedetectingsurfaceslandmarks[ 9 10 11 12 ],edges,ridges,etc.MostoftheseassumeaknowncorrespondencewiththeexceptionoftheworkinChuietal.[ 9 ],JianandVemuri[ 13 ],Wangetal.[ 14 ]andGuoetal.[ 15 ].WorkreportedinIraniandAnandan[ 16 ]usestheenergy(squaredmagnitude)inthedirectionalderivativeimageasarepresentationschemeformatchingachievedusingthe 14

PAGE 27

SSDcostfunction.Recently,Liuetal.[ 17 ]reportedtheuseoflocalfrequencyinarobuststatisticalframeworkusingtheintegralsquarederrora.k.a.,L2E.TheprimaryadvantageofL2Eoverotherrobustestimatorsinliteratureisthattherearenotuningparametersinit.TheideaofusinglocalphasewasalsoexploitedbyMellorandBrady[ 18 ],whousedmutualinformation(MI)tomatchlocal-phaserepresentationofimagesandestimatedthenon-rigidregistrationbetweenthem.However,robustnesstosignicantnon-overlapintheeldofview(FOV)ofthescannerswasnotaddressed.Formoreonfeature-basedmethods,wereferthereadertotherecentsurveybyZitovaandFlusser[ 19 ]. Inthecontextofdirectmethods,theprimarymatchingtechniquesforintra-modalityregistrationinvolvetheuseofnormalizedcross-correlation,modiedSSD,and(normalized)mutualinformation(MI).Ruiz-Alzolaetal.[ 20 ]presentedauniedframeworkfornon-rigidregistrationofscalar,vectorandtensordatabasedontemplatematching.Forscalarimages,thecostfunctionistheextensionofmodiedSSDusingadifferentdenitionofinnerproducts.Howeverthismodelcanonlybeusedonimagesfromthesamemodalityasitassumessimilarintensityvaluesbetweenimages.In[ 21 22 ],alevel-setbasedimageregistrationalgorithmwasintroducedthatwasdesignedtonon-rigidlyregistertwo3Dvolumesfromthesamemodalityofimaging.Thisalgorithmwascomputationallyefcientandwasusedtoachieveatlas-basedsegmentation.Directmethodsbasedontheoptical-owestimationformalargeclassforsolvingthenon-rigidregistrationproblem.Hellieretal.[ 23 ]proposedaregistrationmethodbasedonadenserobust3-Destimationoftheopticalowwithapiecewiseparametricdescriptionofthedeformationeld.Theiralgorithmisunsuitableformulti-modalimageregistrationduetothebrightnessconstancyassumption.Variantsofopticalow-basedregistrationthataccommodateforvaryingilluminationmaybeusedforinter-modalityregistrationandwereferthereaderto[ 24 25 ]forsuchmethods.Guimondetal.,[ 26 ]reportedamulti-modalbrainwarpingtechniquethatusesThirion'sDemonsalgorithm[ 27 ]withanadaptiveintensitycorrection.Thetechniquehoweverwasnottestedforrobustness

PAGE 28

withrespecttosignicantnon-overlapintheFOVs.Morerecently,Cuzoletal.[ 28 ]introducedanewnon-rigidimageregistrationtechniquewhichbasicallyinvolvesaHelmoholtzdecompositionoftheoweldwhichisthenembeddedintothebrightnessconstancymodelofopticalow.TheHelmholtzdecompositionallowsonetocomputelargedisplacementswhenthedatacontainssuchdisplacements.Thistechniqueisaninnovationonaccommodatingforlargedisplacementsandnotonethatallowsforintermodalitynon-rigidregistration.Formoreonintra-modalitymethods,wereferthereadertothecomprehensivesurveys[ 29 19 ]. Apopularframeworkfordirectmethodsisbasedontheinformationtheoreticmeasures[ 30 ],amongthem,mutualinformation(MI)pioneeredbyViolaandWells[ 31 ]andCollignonetal.,[ 32 ]andmodiedinStudholmeetal.,[ 33 ]hasbeeneffectiveintheapplicationofimageregistration.Reportedregistrationexperimentsintheseworksarequiteimpressiveforthecaseofrigidmotion.Theproblemofbeingabletohandlenon-rigiddeformationsintheMIframeworkisaveryactiveareaofresearchandsomerecentpapersreportingresultsonthisproblemare[ 18 34 35 36 37 38 39 40 41 42 ].In[ 34 ],Mattesetal.,andin[ 35 ],Rueckertetal.,presentedmutualinformationbasedschemesformatchingmulti-modalimagepairsusingB-Splinestorepresentthedeformationeldonaregulargrid.Guetter[ 43 ]recentlyincorporatedalearnedjointintensitydistributionintothemutualinformationformulation,inwhichtheregistrationisachievedbysimultaneouslyminimizingtheKLdivergencebetweentheobservedandlearnedintensitydistributionsandmaximizingthemutualinformationbetweenthereferenceandalignmentimages.Recently,D'Agostinoetal.,[ 44 ]presentedaninformationtheoreticapproachwhereintissueclassprobabilitiesofeachimagebeingregisteredareusedtomatchoverthespaceoftransformationsusingadivergencemeasurebetweentheidealcase(wheretissueclasslabelsbetweenimagesatcorrespondingvoxelsaresimilar)andactualjointclassdistributionsofbothimages.Thisworkexpectsasegmentationofeitheroneoftheimagesbeingregistered.Computationalefciencyand

PAGE 29

accuracy(intheeventofsignicantnon-overlaps)areissuesofconcerninmostifnotalltheMI-basednon-rigidregistrationmethods. Finally,someregistrationmethodsunderthedirectapproachareinspiredbymodelsfrommechanics,eitherfromelasticity[ 45 46 ],oruidmechanics[ 47 48 ].Fluidmechanics-basedmodelsaccommodateforlargedeformations,butarelargelycomputationallyexpensive.Christensen[ 49 ]recentlydevelopedaninterestingversionofthesemethods,wherethedirectdeformationeldandtheinversedeformationeldarejointlyestimatedtoguaranteethesymmetryofthedeformationwithrespecttopermutationofinputimages.Amoregeneralandmathematicallyrigoroustreatmentofthenon-rigidregistrationwhichsubsumestheuid-owmethodswaspresentedinTrouve[ 50 ].Allthesemethodshoweverareprimarilyapplicabletointra-modalityandnotinter-modalityregistration. whereh(X)isthedifferentialentropyoftherandomvariableXandisgivenbyh(x)=R1p(x)lnp(x)dx,wherep(x)istheprobabilitydensityfunctionandcanbeestimatedfromtheimagedatausinganyoftheparametricandnonparametricmethods.Thereason

PAGE 30

fordeningMIintermsofdifferentialentropyasopposedtoShannonentropyistofacilitatetheoptimizationofMIwithrespecttotheregistrationparametersusinganyofthegradientbasedoptimizationmethods.NotethatMIdenedusingtheShannon'sentropyindiscreteformwillnotconvergetocontinuouscasedenedhereduetothefactthatShannon'sentropydoesnotconvergetothedifferentialentropy(see[ 5 ]). Wenowdenethecross-CRE(CCRE)usingCREdenedinEqn. 2 Wewillusethisquantityasamatchingcriterionintheimagealignmentproblem.Morespecically,letIT(x)beatestimagewewanttoregistertoareferenceimageIR(x).Thetransformationg(x;)describesthedeformationfromVTtoVR,whereVTandVRarecontinuousdomainsonwhichITandIRaredened,isthesetofthetransformationparameterstobedetermined.Weposethetaskofimageregistrationasanoptimizationproblem.ToalignthereferenceimageIR(x)withthetransformedtestimageIT(g(x;)),weseekthesetofthetransformationparametersthatmaximizesC(IT;IR)overthespaceofsmoothtransformationsi.e., ThecomputationofCCRErequiresestimatesofthemarginalandjointprobabilitydistributionsoftheintensityvaluesofthereferenceandtestimages.Wedenotep(l;k;)asthejointprobabilityof(ITg(x;);IR.LetpT(l;)andpR(k)representthemarginalprobabilityforthetestimageandreferenceimagesrespectively,LTandLRarethediscretesetsofintensitiesassociatedwiththetestimageandreferenceimagerespectively.Then,wecanrewritetheCCREITg(x;);IRasfollows:CITg(x;);IR=E(IT)E[E(ITg(x;)=IR)]=X2LTZ1pT(l;)dlloghZ1pT(l;)dli

PAGE 31

LetP(i>;)=R1pT(l;)dlandP(i>;k;)=R1p(l;k;)dl.UsingthefactthatpT(l;)=Pk2LRp(l;k;),wehaveP(i>;)=Pk2LRP(i>;k;).Eqn.( 3 )canbefurthersimplied,whichleadsto, ToillustratethedifferencebetweenCCREandthenowpopularinformationtheoreticcostfunctionssuchasMI&NMI,wechoosetoplotthesefunctionsagainstaparameterofthetransformation,forillustrativepurposes,saytherotations.TheimagepairweusedhereisMR&CTimagesthatwereoriginallyaligned,andtheMRandCTdataintensitiesrangefrom0-255withthemean55.6and60.6respectively.ThecostfunctionsarecomputedovertherotationanglethatwasappliedtotheCTimagetomisalignitwithrespecttotheMRimage.IneachplotoftheFigure 3 theX-axisshowsthe3DrotationangleaboutZaxis,whiletheY-axisshowsthevaluesofCCRE,MIandNMIcomputedfromthemisaligned(byarotation)imagepairs.Thesecondrowshowsazoom-inviewoftheplotsoverasmallerregion,soastogetadetailedviewofthecostfunction.Thefollowingobservationsaremadefromthisplot:

PAGE 32

Figure3: CCRE,MIandNMItracesplottedforthemisalignedMR&CTimagepairwheremisalignmentisgeneratedbyarotationoftheCTimage.Firstrow:overtherange40to40.Secondrow:zoominviewbetween0:5to0:5,wherethearrowsintherstrowsignifytheposition.Notethatallthreecostfunctionareimplementedwithtri-linearinterpolation.Thirdrow:Threecostfunctionsimplementedwithpartialvolumeinterpolation[ 32 ]. 1. SimilartoMIandNMI,themaximumofCCREoccursat0ofrotation,whichconrmsthatournewinformationmeasureneedstobemaximizedinordertondoptimumtransformationbetweentwomisalignedimages. 2. TheCCREshowsmuchlargerrangeofvaluesthanMI&NMI.Thisfeatureplaysanimportantroleinthenumericaloptimizationsinceitleadstoamorestablenumericalimplementationbyavoidingcancelation,roundoffetc.thatoftenplaguearithmeticoperationswithsmallernumericalvalues.

PAGE 33

3. Uponcloserinspection,weobservethatCCREismuchsmootherthantheMIandNMIfortheMR&CTdatapair,thereforeveriesthatCCREismoreregularthanotherinformationtheoreticmeasures. ThebasicideaofthecubicB-splinedeformationistodeformanobjectbymanipu-latinganunderlyingmeshofcontrolpointsi.Thedeformationgisdenedbyasparseregularcontrolpointgrid.In3Dcase,thedeformationatanypointx=[x;y;z]TinthetestimagecanbeinterpolatedwithalinearcombinationofcubicB-splineconvolutionkernel. where(3)(x)=(3)(x)(3)(y)(3)(z)and4isspacingofthecontrolgrid.jistheexpansionB-splinecoefcientscomputedfromthesamplevaluesoftheimage.Fortheimplementationdetails,wereferthereadertoForsey[ 51 ]andMattes[ 34 ].

PAGE 34

EachcomponentofthegradientcanbefoundbydifferentiatingEqn.( 3 )withrespecttoatransformationparameters.WeconsiderthetwotermsinEqn.( 3 )separatelywhencomputingthederivative.FortherstterminEqn.( 3 ),wehave, @hX2LTZ1pT(l;)dllogZ1pT(l;)dli=X2LTlogP(i>;)+1@P(i>;) whereP(i>;)=R1pT(l;)dl,and Thederivativeofthesecondtermisgivenby, @hXk2LRpR(k)X2LTZ1p(l;k;) whereP(i>;k;)=R1p(l;k;)dl,and Combiningthederivativesofthetwotermstogether,andusingthefactthat

PAGE 35

wehavetheanalyticgradientofCCRE, ComparingtheexpressionsforCCREandderivativeofCCRE wenotethatthetwoformulasin( 3 )aresimilartoeachotherandtheysharethecommontermlogP(i>;k;)

PAGE 36

images.Let(0)beazero-ordersplineParzenwindow(centeredunitpulse)and(3)beacubicsplineParzenwindow,thesmoothedjointprobabilityof(IR;ITg)isgivenby whereisanormalizationfactorthatensuresPp(l;k)=1,andIR(x)andIT(g(x;)aresamplesofthereferenceandinterpolatedtestimagesrespectively,whichisnormal-izedbytheminimumintensityvalue,f0R,f0T,andtheintensityrangeofeachbin,4bR,4bT. SinceP(i>;k;)=R1p(l;k;)dl,wehavethefollowing, where()isthecumulativeresidualfunctionofcubicsplinekerneldenedasfollows, 22 3v+v3 22 3v+v3

PAGE 37

Notethatd(u) du=(3)(u),wecanthentakethederivativeofEqn. 3 withrespectto,andweget where@IT(t) 3 )andEqn.( 3 )respectively. 3 ),wecanthenusetheQuasi-Newtonmethodtonumericallysolvetheoptimizationproblem.

PAGE 38

showthatCCREisnotonlymorerobust,butalsoconvergesfasterthanothers.WebeginbyapplyingCCREtoregisterimagepairsforwhichthegroundtruthwasavailable. ThedataweuseforthisexperimentarecorrespondingslicesfromanMRT1andT2imagepair,whichisfromthebrainwebsiteattheMontrealNeurologicalInstitute[ 52 ].Theyareoriginallyalignedwitheachother.Thetwoimagesaredenedona1mmisotropicvoxelgridintheTalairachspace,withdimension(256256).Wethenapplyaknownnon-rigidtransformationtotheT2imageandthegoalistorecoverthisdeformationbyapplyingourregistrationmethod.Themutualinformationschemewhichwearegoingtocomparewithisoriginallyreportedinliterature[ 34 ][ 53 ],inwhichtheexplicitgradientformsarepresentedandthusallowingfortheapplicationofgradientbasedoptimizationmethods.

PAGE 39

Figure3: Upperleft,MRT1imageassourceimage;Upperright,deformedMRT2imageastargetimage;Lowerleftandright,resultsofestimatedtransformationsusingCCREandMIappliedtothesourcerespectively.Bothalgorithmsrunfor30secondsusingthesamegradientdescenttechnique. ThesourceandtargetimagepairalongwiththeresultsofestimatedtransformationusingCCREandMIappliedtothesourceareshowninFigure 3 .Asevidentvisually,weobservethattheresultgeneratedbyCCREismoresimilarinshapewiththetargetimagethantheoneproducedbyMI. QuantitativeassessmentofaccuracyoftheregistrationispresentedsubsequentlyinFigure 3 ,whereweplottedthechangeofmeandeformationerror(MDE)obtainedfortheCCRE-basedalgorithmandtheMI-basedalgorithmrespectively.MDEisdenedasdm=1

PAGE 40

Figure3: PlotdemonstratingthechangeofMeanDeformationErrorforCCREandMIregistrationresultswithtime.SolidlineshowstheMDEforCCREregistrationresult,whiledottedlineillustratestheMDEforMIresult. empiricallyvalidatesthefasterconvergencespeedofCCREbasedalgorithmovertheMI-basedalgorithm. 3.3.1.2 depictstheresults Figure3: Resultsofapplicationofouralgorithmtosyntheticdata(seetextfordetails). obtainedforthisimagepair.whichisorganizedasfollows,fromlefttoright:therst

PAGE 41

rowdepictsthesourceimagewiththetargetimagesegmentationsuperposedtodepicttheamountofmis-alignment,theregisteredsourceimagewhichisobtainedusingouralgorithmsuperposedwiththetargetsegmentation,followedbythetargetimage;secondrowdepictsgroundtruthdeformationeldwhichweusedtogeneratethetargetimagefromtheMRT2image,theestimatednon-rigiddeformationeldfollowedbyhistogramoftheestimatedmagnitudeerror.Notethattheerrordistributionismostlyconcentratedinthesmallerrorrangeindicatingtheaccuracyofourmethod.Asameasureofaccuracyofourmethod,wealsoestimatedtheaverage,,andthestandarddeviation,,oftheerrorintheestimatednon-rigiddeformationeld.Theerrorwasestimatedastheanglebetweenthegroundtruthandestimateddisplacementvectors.Theaverageandstandarddeviationare1.5139and4.3211(indegrees)respectively,whichisquiteaccurate. 3.3.1.3Noiseimmunity Themeanmagnitudeofthesyntheticmotionis4.37pixel,withthestandarddeviationat1.8852.Table 3 showtheregistrationresultsforthetwoschemes.Fromthetable,weobservethattheMIfailswhenthestandarddeviationofthenoiseisincreasedto40,whileCCREistolerantuntil66,asignicantdifferencewhencomparedtotheMI.

PAGE 42

Table3: ComparisonoftheregistrationresultsbetweenCCREandMIforaxedsyntheticdeformationeld. CCRE MI MDEStandardDeviation 1:08160:9345 1:38841:4538 19 1:13811:1702 1:48711:5052 30 1:19751:3484 1:52041:5615 40 1:37911:9072 66 ThisexperimentconclusivelydepictsthatCCREhasmorenoiseimmunitythanMIwhendealingwiththenon-rigidmotion. 3 depictsanexampleofregistrationoftheMRT1andT2datasetswithlargenonoverlap.TheleftimageoftheguredepictstheMRT1brainscanasthesourceimage,andtherightimageshowstheMRT2dataasthetarget.NotethattheFOVforthedatasetsaresignicantlynonoverlapping.Thenonoverlapwassimulatedbycutting66%oftheMRT1image(Sourceimage).Themiddlecolumndepictsthetransformedsourceimagealongwithanedgemapofthetarget(DeformedMRT2image)superimposedonthetransformedsource.Asisevident,theregistrationisvisuallyquiteaccurate. Figure3: RegistrationresultsofMRT1andT2imageslicewithlargenon-overlap.(left)MRT1sourceimagebeforeregistration;(right)DeformedT2targetimage;(mid-dle)thetransformedMRimagesuperimposedwithedgemapfromtargetimage.

PAGE 43

34 ]tothesesamedatasets.TheCTimageisofsize(512;512;120)whiletheMRimagesizeis(512;512;142),andthevoxeldimensionsare(0:46;0:46;1:5)mmand(0:68;0:68;1:05)forCTandMRrespectively.Theregistrationwasperformedonreducedvolumes(210210120)withthecontrolknotsplacedevery161616voxels.TheprogramwaswrittenintheC++programminglanguage,andallexperimentswererunona2.6GHZPentiumPC. Table3: ComparisonoftotaltimetakentoachieveregistrationbytheCCREwithMI. 1 2 3 4 5 6 7 8 CCRETime(s) 4827 3452 4345 4038 3910 4510 5470 3721 MITime(s) 9235 6344 10122 17812 12157 11782 13157 10057 WehaveusedasetofeightvolumesofCTdatasetsandthetaskwastoregistertheseeightvolumestotheMRdatachosenasthetargetimageforallregistrations,byusingbothCCREandMIalgorithms.NotethatallCT&MRvolumesarefromdifferentsubjectsandthuscontainsrealnon-rigidmotion.Theparametersusedwithbothalgorithmswereidentical.Forbothalgorithms,theoptimizationofthecostfunctionswashaltedwhenimprovementsofatleast0:0001inthecostfunctioncouldnotbedetected.ThetimerequiredforregisteringalldatasetsforouralgorithmaswellasMImethodaregiveninTables 3 .Thistableshowsthat,onaverage,ourCCREalgorithmisabout2.5timesfasterthanthetraditionalMIapproachforthissetofexperiments.Forbrevity,weonlyshowoneregistrationresultinFigure 3 .Here,onesliceofthevolumeisshownonrstrowwiththesourceCTimageatleftandreferenceimageatright.ThemiddleimageshowthetransformedCTimageslicesuperimposedwithedgemapfromtargetimage.Onthesecondrow,thesourceimagesuperimposedwithedgemapfrom

PAGE 44

targetimageisshownontheleft,whileshowninthemiddleandrightarethesurfacesreconstructedfromthetransformedsourceusingCCREmethodandthetargetMRimagerespectively.Fromthisgure,wecanseethatthesourceandtargetimagedepictconsiderablenon-rigidchangesinshape,neverthelessourmethodwasabletoregisterthesetwoimagesquiteaccurately.Tovalidatetheconformityofthetworeconstructedsurfaces,werandomlysample30pointsfromthesurfaceofthetransformedsourceusingCCRE,andthenestimatethedistancesofthesepointstothesurfaceofthetargetMRvolume.Theaverageofthesedistancesisabout0:47mm,whichindicatesaverygoodagreementbetweentwosurfaces.TheresemblanceofthereconstructedshapesfromtransformedsourcewiththetargetindicatesthatourCCREalgorithmsucceededinmatchingthesourceCTvolumetothetargetMRimage. Figure3: RegistrationresultsofdifferentsubjectsofMR&CTbraindatawithrealnon-rigidmotion.(seetextfordetails. Theaccuracyoftheinformationtheoreticbasedalgorithmfornon-rigidregistrationproblemswasassessedquantitativelybymeansofanregion-basedsegmentationtask[ 54 ].ROIs(wholebrain,eyes)weresegmentedautomaticallyintheseeightCTdatasets

PAGE 45

usedasthesourceimageandbinarymaskswerecreated.ThedeformationeldsbetweentheCTandMRvolumewerecomputedandusedtoprojectthemasksfromeachoftheCTtotheMRvolume.ContoursweremanuallydrawnonafewsliceschosenatrandominMRvolume(fourslices/volume).ManualcontoursonMRandcontoursobtainedautomaticallywerethencomparedusinganacceptedsimilarityindexdenedastwotimesthenumberofpixelsintheintersectionofthecontoursdividedbythesumofthenumberofpixelswithineachcontour[ 41 ].Thisindexvariesbetweenzero(completedisagreement)andone(completeagreement)andissensitivetobothdisplacementanddifferencesinsizeandshape.Table 3 listsmeanvaluesforthesimilarityindexforeachstructure.Itiscustomarilyacceptedthatavalueofthesimilarityindexabove0.80indicatesaverygoodagreementbetweencontours.Ourresultsarewellabovethisvalue.Forcomparisonpurpose,wealsocomputedthesameindexfortheMImethod.WecanconcludefromthetablethatourCCREcanachievebetterregistrationaccuracythantheMIforthetaskofnon-rigidregistrationofrealmulti-modelimages. Table3: ComparisonofthevalueSofseveralbrainstructuresforCCREandMI. Volume 1 2 3 4 5 6 7 8 WholeBrain 0.987 0.996 0.974 0.962 0.975 0.967 0.988 0.981 CCRE LeftEye 0.925 0.935 0.925 0.907 0.875 0.890 0.834 0.871 RightEye 0.840 0.940 0.891 0.872 0.851 0.829 0.910 0.921 WholeBrain 0.986 0.981 0.976 0.96 0.950 0.961 0.942 0.952 MI LeftEye 0.911 0.893 0.904 0.791 0.853 0.810 0.851 0.853 RightEye 0.854 0.917 0.889 0.814 0.849 0.844 0.897 0.854

PAGE 46

MatchingpointpatternsisubiquitousinmanyeldsofEngineeringandSciencee.g.,medicalimaging,sportsscience,archaeology,andothers.Pointsetsarewidelyusedincomputervisiontorepresentboundarypointsofshapescontainedinimagesoranyothersalientfeaturesofobjectscontainedinimages.Giventwoormoreimagesrepresentedusingthesalientfeaturescontainedtherein,mostoftenthannot,oneisinterestedinmatchingthese(feature)pointpatternstodeterminealinearoranonlineartransformationbetweenthecoordinatesofthefeaturepointsets.Suchtransformationscapturethechangesinthepatterngeometrycharacterizedbythegivenfeaturepointset. Theprimarytechnicalchallengeinusingpoint-setrepresentationsofshapesisthecorrespondenceproblem.Typicallycorrespondencescanbeestimatedoncethepoint-setsareproperlyalignedwithappropriatespatialtransformations.Iftheobjectsathandaredeformable,theadequatetransformationwouldobviouslybeanon-rigidspatialmapping.Solvingfornon-rigiddeformationsbetweenpoint-setswithunknowncorrespondenceisahardproblem.Infact,manycurrentmethodsonlyattempttosolveforafnetransformationforthealignment[ 55 ].Furthermore,wealsoencountertheissueofthebiasproblemingroupwisepoint-setsregistration.Ifonearbitrarilychoosesanyoneofthegivendatasetsasareference,theestimatedregistrationtransformationwouldbebiasedtowardthischosenreferenceanditwouldbedesirabletoavoidsuchabias.Thequestionthatarisesis:Howdowealignallthepoint-setsinasymmetricmannersothatthereisnobiastowardanyparticularpoint-set? Toovercometheseaforementionedproblems,wepresentanovelapproachtosimultaneouslyregistermultiplepoint-setsandconstructtheatlas.Theideaistomodeleachpointsetbyakernelprobabilitydensityordistribution,thenquantifythedistance 34

PAGE 47

betweentheseprobabilitydensitiesordistributionsusinginformation-theoreticmeasures.Figure 4 illustratethisidea,wheretherightcolumnofthegureisthedensityfunctioncorrespondingtothecorpuscallosumpoint-setsshownintheleft.Thedistanceis Figure4: Illustrationofcorpuscallosumpoint-setsrepresentedasdensityfunctions. optimizedoveraspaceofcoordinatetransformationsyieldingthedesiredregistrations.Itisobviousthatonceallthepointsetsaredeformedintothesameshape,thedistancemeasurebetweenthesedistributionsshouldbeminimizedsinceallthedistributionareidenticaltoeachother.Weimposeregularizationoneachdeformationeldtopreventover-deformingofeachpoint-sets(e.g.allthepoint-setsmaydeformintoasingledatapoint). Therestofthechapterisorganizedasfollows:webeginbyreviewingalltherelatedliteratures,whichisfollowedbyadescriptionofthedivergencemeasuresweusedforquantifythedistancebetweendensitiesordistributions.Wethenpresentthedetailsof

PAGE 48

ourenergyfunction,andtheempiricalwayofestimatingthecostfunctionsandtheirderivatives.Finallywewillshowtheexperimentalresultsattheendofthischapter. Theworkpresentedin[ 56 ]isarepresentativemethodusinganintrinsiccurveparameterizationtoanalyzedeformableshapes.Shapesarerepresentedaselementsofinnite-dimensionalspacesandtheirpairwisedifferencearequantiedusingthelengthsofgeodesicsconnectingthemonthesespaces,theintrinsicmean(Karchermean)canbecomputedasapointonthemanifold(ofshapes)whichminimizethesumofsquaregeodesicdistancebetweenthisunknownpointtoeachindividualshape,whichliesonthemanifold.Howeverthecurvesarelimitedbyclosedcurves,andithasnotbeenextendedtothe3Dsurfaceshapes.Formethodsusingintrinsiccurveorsurfacerepresentations[ 56 57 58 ],furtherstatisticalanalysisontheserepresentationsismuchmoredifcultthananalysisonthepointrepresentation,buttherewardmaybehigherduetotheuseofintrinsichigherorderrepresentation. Amongthesemethodsusingpoint-setsparameterization,theideaofusingnon-rigidspatialmappingfunctions,specicallythin-platesplines[ 59 60 61 ],toanalyzedeformableshapehasbeenwidelyadopted.Bookstein'sworkin[ 59 ],successfullyinitiatedtheresearcheffortsontheusageofthin-platesplinestomodelthedeformationofshapes.Thismethodislandmark-based,itavoidsthecorrespondenceproblemsince

PAGE 49

theplacementofcorrespondingpointsisdrivenbythevisualperceptionofexperts,howeveritsuffersfromthethetypicalproblembesettinglandmarkmethods,e.g.inconsistency.Severalsignicantarticlesonrobustandnon-rigidpointsetmatchinghavebeenpublishedbyRangaranjanandcollaborators[ 62 60 63 ]usingthin-platesplines.Intheirrecentwork[ 60 ],theyattempttoextendtheirworktotheconstructionofanmeanshapefromasetofunlabeledshapeswhicharerepresentedbyunlabeledpoint-sets.Themainstrengthoftheirworkistheabilitytojointlydeterminethecorrespondencesandnon-rigidtransformationbetweeneachpointsetstotheemergingmeanshapeusingdeterministicannealingandsoft-assign.However,intheirwork,thestabilityoftheregistrationresultisnotguaranteedinthecaseofdatawithoutliers,andhenceagoodstoppingcriterionisrequired.Unliketheirapproach,wedonotneedtorstsolveacorrespondenceprobleminordertosubsequentlysolveanon-rigidregistrationproblem. Theactiveshapemodelproposedin[ 64 ]utilizedpointstorepresentdeformableshapes.Theirworkpioneeredtheeffortsinbuildingpointdistributionmodelstounder-standdeformableshapes[ 64 65 ].Objectsarerepresentedascarefully-denedlandmarkpointsandvariationofshapesaremodeledusingaprincipalcomponentanalysis.Theselandmarkpointsareacquiredthroughamoreorlessmanuallandmarkingprocesswhereanexpertgoesthroughallthesamplestomarkcorrespondingpointsoneachsample.Itisarathertediousprocessandaccuracyislimited.Inrecentwork[ 66 ],theauthorsattempttoovercomethislimitationbyattemptingtoautomaticallysolveforthecorrespondencesinanon-rigidsetting.Theresultingalgorithmisverysimilartotheearlierworkin[ 58 ]andisrestrictedtocurves.Theworkin[ 55 ]alsouses2Dpointstolearnshapestatistics,whichisquitesimilartotheactiveshapemodelmethodexceptthatmoreattentionhasbeenpaidtothesamplepoint-setsgenerationprocessfromtheshape.Unlikeourmethod,thetransformationbetweencurvesarelimitedbyrigidmapping,andprocessisnotsymmetric.

PAGE 50

Thereareseveralpapersinthepoint-setsalignmentliteraturewhichbearcloserelationtoourresearchreportedhere.Forinstance,TsinandKanade[ 67 ]proposedakernelcorrelationbasedpointsetregistrationapproachwherethecostfunctionisproportionaltothecorrelationoftwokerneldensityestimates.Itissimilartoourworksincewetoomodeleachofthepointsetsbyakerneldensityfunctionandthenquantifythe(dis)similaritybetweenthemusinganinformation-theoreticmeasure,followedbyanoptimizationofa(dis)similarityfunctionoveraspaceofcoordinatetransformationsyieldingthedesiredtransformation.Thedifferenceliesinthefactthatdivergencemeasuresusedinourworkisalotmoregeneralthantheinformation-theoreticmeasureusedin[ 67 ],andcanbeeasilyextendedtomultiplepoint-sets.Morerecently,in[ 68 ],Glaunesetal.convertthepointmatchingproblemintoanimagematchingproblembytreatingpointsasdeltafunctions.Thentheyliftthesedeltafunctionsanddiffeomorphicallymatchthem.Themainproblemforthistechniqueisthattheyneeda3Dspatialintegralwhichmustbenumericallycomputed,whilewedonotneedthisduetotheempiricalcomputationofthedivergencemeasures.Wewillshowitintheexperimentalresultsthatourmethod,whenappliedtomatchpoint-sets,achievesverygoodperformanceintermsofbothrobustnessandaccuracy.

PAGE 51

(KL)divergence.TheKLdivergence(alsoknownastherelativeentropy)betweentwodensitiespandqisdenedas Toillustratethis,considertheMorsecode,designedtosendmessagesinEnglish.TheMorsecodeencodestheletterEwithasingledotandtheletterQwithasequenceoffourdotsanddashes.BecauseEisusedfrequentlyinEnglishandQseldom,thismakesforefcienttransmission.HoweverifonewantedtousetheMorsecodetosendmessagesinChinesepinyin,whichmightuseQmorefrequently,hewouldndthecodelessefcient.IfweassumecontrafactuallythattheMorsecodeisoptimalforEnglish,thisdifferenceinefciencyistheredundancy. NoticethatKLdivergenceisnotsymmetricandapopularwaytosymmetrizeitis 2(DKL(pkq)+DKL(qkp)) 69 ],servesasameasureofcohesionbetweenmultipleprobabilitydistributions.Ithasbeenusedbysomeresearchersasadissimilaritymeasureforimageregistrationandretrievalapplications[ 70 71 ]withverygoodresults.Ithasmanydesirableproperties,tonameafew,1)ThesquarerootofJS-divergence(inthecasewhenitsparameterisxedto0:5)isametric[ 72 ];2)JS-divergencerelatestootherinformation-theoreticfunctionals,such

PAGE 52

astherelativeentropyortheKullbackdivergence,andhenceitsharestheirmathematicalpropertiesaswellastheirintuitiveappeal;3)ThecompareddistributionsusingtheJS-divergencecanbeweighted,whichallowsonetotakeintoaccountthedifferentsizesofthepointsetsamplesfromwhichtheprobabilitydistributionsarecomputed;4)TheJS-divergencemeasurealsoallowsustohavedifferentnumberofclustercentersineachpoint-set.ThereisnorequirementthattheclustercentersbeincorrespondenceasisrequiredbyChuietal[ 73 ].Givennprobabilitydensityfunctionspi,i2f1;:::;ng,theJS-divergenceofpiisdenedby where=f1;2;:::;nji>0;Pi=1garetheweightsoftheprobabilitydensityfunctionspiandH(pi)istheShannonentropy.ThetwotermsontherighthandsideofEquation( 4 )aretheentropyofp:=Pipi(the-convexcombinationofthepis)andthesameconvexcombinationoftherespectiveentropies.WecanshowthatJS-divergencecanbederivedfromtheKLdivergence where2(0;1)isaxedparameter;wewillalsoconsideritsstraightforwardgeneral-izationtondistributions. 2 ,wedenedanentropymeasurewhichisbasedonprobabilitydis-tributioninsteadofdensityfunction.Thedistributionfunctionismoreregularbecauseitisdenedinanintegralformunlikethedensityfunction,whichisthederivativeofthedistribution.ThedenitionofCumulativeResidualEntropyalsopreservesthewellestablishedprinciplethatthelogarithmoftheprobabilityofaneventshouldrepresenttheinformationcontentintheevent.CREisshowntobemoreimmunetonoiseand

PAGE 53

outliers.Basedonthisidea,wecandeneaKL-divergencemeasurebetweenCumulativeDistributionFunctions(CDFs), Pr(X2>x)dx(4) FollowthesamerelationshipbetweenJensen-ShannondivergenceandKLdivergence,wecanderivetheso-calledCDF-JSdivergencefromthedenitionofCDF-KLdivergence,(denotedasJ),theresultofwhichisshowninthefollowingtheorem, 2 )ofChapter 2

PAGE 54

Pr(X>x)dx+(1)ZPr(X2>x)lnPr(X2>x) Pr(X>x)dx=ZPr(X1>x)lnPr(X1>x)dx+(1)ZPr(X2>x)lnPr(X1>x)dxZhPr(X1>x)+(1)Pr(X2>x)ilnPr(X>x)dx=E(P1)(1)E(P2)ZhPr(X1>x)+(1)Pr(X2>x)ilnPr(X>x)dx(4) where,Pisthedistributionfunctioncorrespondingtothedensityfunctionp=p1+(1)p2,whichistheconvexcombinationofthetwoprobabilitydensities,therefore Consequently,CDF-JSdivergencefortworandomvariablecanberewrittenas

PAGE 55

non-rigidregistrationfollowingwhichanatlasisconstructedfromtheregistereddata.However,inourwork,theatlasemergesasabyproductofthenon-rigidregistration.Thebasicideaistomodeleachpointsetbyaprobabilitydensityordistributionfunction,thenquantifythedistancebetweenthesefunctionsusinganinformation-theoreticmeasure.Thedistancemeasureisoptimizedoveraspaceofcoordinatetransformationsyieldingthedesiredtransformations.Wewillbeginbypresentingtheenergyfunctionforsolvingthegroupwisepoint-setsregistrationproblem. In( 4 ),Disthedivergencemeasureformultipledistributions,whichweproposetouseeitherJSdivergenceorCDF-JSdivergence.TheweightparameterisapositiveconstanttheoperatorLdeterminesthekindofregularizationimposed.Forexample,Lcouldcorrespondtoathin-platespline,aGaussianradialbasisfunction,etc.EachchoiceofLisinturnrelatedtoakernelandametricofthedeformationfromandtoZ. Followingtheapproachin[ 73 ],wechoosethethin-platespline(TPS)torepresentthenon-rigiddeformation.Givenncontrolpointsx1;:::;xninRd,ageneralnon-rigid

PAGE 56

mappingf:Rd!Rdrepresentedbythin-platesplinecanbewrittenanalyticallyas:f(x)=WU(x)+Ax+tHereAx+tisthelinearpartoff.Thenonlinearpartisdeterminedbyadnmatrix,W.AndU(x)isann1vectorconsistingofnbasisfunctionsUi(x)=U(x;xi)=U(kxxik)whereU(r)isthekernelfunctionofthin-platespline.Forexample,ifthedimensionis2(d=2)andtheregularizationfunctionalisdenedonthesecondderivativesoff,wehaveU(r)=1=(8)r2ln(r). Therefore,thecostfunctionfornon-rigidregistrationcanbeformulatedasanenergyfunctionalinaregularizationframework,wheretheregularizationterminequation 4 isgovernedbythebendingenergyofthethin-platesplinewarpingandcanbeexplicitlygivenbytrace(WKWT)whereK=(Kij),Kij=U(pi;pj)describestheinternalstructureofthecontrolpointsets.Notethelinearpartcanbeobtainedbyaninitialafneregistration,thenanoptimizationcanbeperformedtondtheparameterW. wherexkconsistsofpointsfxai;i2f1;:::;nag;a2f1;:::;Ngg,whichisthepooleddataofallthesamples.Incontrasttothetypicalstatisticaltestrelativetoathreshold,weseekthemaximumofthelikelihoodrationinEqn.( 4 ).ThefollowingtheoremshowstherelationshipbetweenJensen-Shannondivergenceandthelikelihoodration.

PAGE 57

4 )isequivalenttominimizingtheJensen-ShannondivergencebetweentheNprobabilitydensitiespa,a2f1;:::;Ng. Weseektomaximizetheprobabilitythatthesamplesaredrawnfromthemixtureratherthanfromseparatemembersofthefamily(p1;p2;:::;pN).Inthecontextofgroupwisematchingofpoint-sets,thismakeseminentsensesincemaximizingtheaboveratioistantamounttoincreasingthechancethatalloftheobservedpoint-setsarewarpedversionsofthesameunderlyingwarpedandpooleddatamodel.Thenotionofthepooleddatamodelisdenedasfollows.Inourprocessofgroupwiseregistration,thewarpingdoesnothaveaxedtargetdataset.Instead,thewarpingisbetweentheinputdatasetsandanevolvingtargetwhichwecallthepooledmodel.Thetargetevolvestoafullyregisteredpooleddatasetattheendoftheoptimizationprocess.Thepooledmodelthenconsistsofinputdatasetswhichhaveundergonegroupwisematchingandarenowfullyregisteredwitheachother.TheconnectiontotheJS-divergencearisesfromthefactthatthenegativelogarithmoftheaboveratio(Eqn. 4 )asymptoticallyconvergestotheJS-divergencewhenthesamplesareassumedtobedrawnfromthemixturePaapa. AssumethatwearegivenNpoint-sets,fromwhichwearegoingtoconstructtheatlas,wecanthendividetheNpoint-setsintomsubsets(generallymN),thereforewecan

PAGE 58

constructmatlasesfromeachsubsetsusingouralgorithms,andallthepoint-setsinsideeachsubsetsareregistered.Thenwecaneitherconstructasingleatlasfromthesematlaspoint-sets,orwecanfurtherdividematlaspoint-setsintoevensmallersubsets,andfollowthesameprocessuntilasingleatlasisconstructed.Theremainingquestioniswhethertheatlasthusobtainedisbiasedornot?Thefollowingtheoremwillleadustotheanswer. 4 )bysimplealgebraicoperations. Inourregistrationalgorithm,allthepoint-setsarerepresentedasprobabilitydistributions,andtheatlasthusconstructedcanbeconsideredasconvexcombinationofthesedistributions.Therefore,wecantreatPasandSisasthedistributionscorrespondingtothepoint-setsandtheconstructedatlasesfromthesubsetsrespectively.ThereforefromTheorem 7 ,weknowthattherelationshipinEqn.( 4 )holdsbetweentheJSdivergenceofthePasandSis.NoticethattherighthandsideofEqn.( 4 )istheJS/CDF-JSdivergencesofthedistributionsinallthesubsets,whichareminimizedineachstepsofthehierarchicalmethodweproposed.Intuitively,ifthesepointsetsarealignedproperly,

PAGE 59

thecorrespondingdistributionfunctionsshouldbestatisticallysimilar.Thereforethedivergencesofallthesubsetsshouldbezeroallveryclosetozero,whichmeanstherighthandsideofEqn.( 4 )iszero.Consequently,theJS/CDF-JSdivergenceofthePasanddivergenceoftheSisareequaltoeachother,thereforeminimizingJS/CDF-JSdivergenceofalltheresultantatlaspoint-setsisequivalenttominimizingdivergenceoftheoriginalpoint-sets,implyingthatthereisnobiastowardanyparticularpartitioningofthepoint-sets. Havingintroducedthecostfunctionandthetransformationmodel,nowthetaskistodesignanefcientwaytoestimateempiricaldivergencemeasuresbetweenmultipledensitiesordistributionsandderivetheanalyticgradientoftheestimateddivergenceinordertoachievetheoptimalsolutionefciently.WedesigntwocompletedifferentapproachesforestimatingJSdivergenceandCDF-JSdivergence.WeusenitemixturemodelforestimatingJSdivergenceandtheparzenwindowtechniqueforCDF-JSdivergence,thedetailsofwhichwillbeintroducednext. 4.3.4.1Finitemixturemodels 74 ]isdenedasaconvexcombinationofGaussiancomponentdensities. Tomodeleachpoint-setasaGaussianmixture,wedeneasetofclustercenters,oneforeachpoint-set,toserveastheGaussianmixturecenters.Sincethefeaturepoint-setsareusuallyhighlystructured,wecanexpectthemtoclusterwell.Furthermorewecangreatlyimprovethealgorithmefciencybyusinglimitednumberofclusters.Notethatwecanchoosetheclustercenterstobethepoint-setitselfifthesizeofpoint-setisquitesmall.Theclustercenterpoint-setsaredenotedbyfVp;p2f1;:::;Ngg.Eachpoint-setVpconsistsofpointsfvpi2RD;i2f1;:::;Kpgg.NotethatthereareKppointsineachVp,

PAGE 60

andthenumberofclustersforeachpoint-setmaybedifferent(inourimplementation,thenumberofclusterswereusuallychosentobeproportionaltothesizeofthepoint-sets).Theclustercentersareestimatedbyusingaclusteringprocessovertheoriginalsamplepointsxpi,andweonlyneedtodothisoncebeforetheprocessofjointatlasestimationandpoint-setsregistration.Inourimplementation,weutilizedeterministicannealing(DA)procedurewithitsprovenbenetofrobustnessinclustering[ 75 ].Webeginbyspecifyingthedensityfunctionofeachpointset. InEquation( 4 ),theoccupancyprobabilitywhichisdifferentforeachdatapoint-setisdenotedbyp.Thecomponentdensitiesp(xjvpa)is (2)D 2aexp1 2xvpaT1axvpa(4) ProbabilityofthepointsetXpcomingfromthismixtureisthen Later,wesettheoccupancyprobabilitytobeuniformandmakethecovariancematricesatobeproportionaltotheidentitymatrixinordertosimplifyatlasestimationprocedure. Forsimplicity,wechoosei=1 (2)D 2aexp1 2fj(xji)fp(vpa)T1afj(xji)fp(vpa)(4) Wherefa;a2f1;:::;Kggisthesetofclustercovariancematrices.Forthesakeofsimplicityandeaseofimplementation,weassumethattheoccupancyprobabilitiesare

PAGE 61

uniform(pa=1 4 )asfollows, Foreachtermintheequation,wecanestimatetheentropyusingtheweaklawoflargenumbers,whichisgivenby,H(X1 @1;@JS @2;:::;@JS @N]

PAGE 62

EachcomponentofthegradientmaybefoundbydifferentiatingEqn.( 4 )withrespecttothetransformationparameters.Inordertocomputethisgradient,let'srstcalculatethederivativeofQxjipwithrespecttol, (2)D 22jFjpj2(Fjp@fj(xji) (2)D 22jFjpj2(Fjp@fp(vpa) (2)D 22jFjpj2(Fjp[@fp(vpa) @l=n1

PAGE 63

whereaisanormalizationfactorthatensuresRRp(l;k)dldk=1,[l;k]arethecoordinatevaluesintheXandYaxisrespectively,thetransformedpointcoordinates[fa(xai;a);fa(yai;a)]isnormalizedbytheminimumcoordinatevalue,x0,y0,andtherangeofeachbin,4bX,4bY.Fromthedensityfunction,wecancalculatethecumulativeresidualdistributionfunctionbytheformulaPa(l>;k>;a)=R1R1pa(l;k;a)dldk,where2Lx;2Ly,Lx;LyarediscretesetsofcoordinatevaluesintheXandYaxisrespectively.Toexpressitinfurtherdetail,wehavethefollowing, du=(3)(u). Havingspeciedthedistributionfunctionofthedata,wecanthenrewriteEqn.( 4 )asfollows,(Forsimplicity,wechoosea=1

PAGE 64

wherePisthecumulativeresidualdistributionfunctionforthedensityfunction1 EachcomponentofthegradientmaybefoundbydifferentiatingEqn( 4 )withrespecttothetransformationparameters.Itcanbeeasilyshownthat@P(;;a)

PAGE 65

experimentalresultsonpoint-setalignmentbetweentwogivenpoint-setsaswellasatlasconstructionfrommultiplepoint-setsinthenextsection. 4 .Thetoprowshowstheregistrationresultfora2Drealrangedatasetofaroad(whichwasalsousedinTsinandKanade'sexperiments[ 67 ]).Theguredepictstherealdataandtheregistered(usingrigidmotion).Topleftframecontainstwounregisteredpointsetssuperposedoneachother.Toprightframecontainsthesamepointsetsafterregistrationusingouralgorithm.A3Dhelixexampleispresentedinthesecondrow(withthesamearrangementasthetoprow).WealsotestedourmethodagainsttheKCmethod[ 67 ]andtheICPmethods,asexpected,ourmethodandKCmethodexhibitamuchwiderconvergencebasin/rangethantheICPandbothachieveveryhighaccuracyinthenoiselesscase. Wealsoappliedouralgorithmtonon-rigidlyregistermedicaldatasets(2Dpoint-sets).Figure 4 depictssomeresultsofourregistrationmethodappliedtoasetof2Dcorpuscallosumsliceswithfeaturepointsmanuallyextractedbyhumanexperts.Registrationresultisshownintheleftcolumnwiththewarpingof2Dgridundertherecoveredmotionwhichisshowninthemiddlecolumn.Ournon-rigidalignmentperformswellinthe

PAGE 66

Figure4: Resultsofrigidregistrationinnoiselesscase.'o'and'+'indicatethemodelandscenepointsrespectively. presenceofnoiseandoutliers(Figure 4 rightcolumn).Forthepurposeofcomparison,wealsotestedtheTPS-RPMprogramprovidedin[ 62 ]onthisdataset,andfoundthatTPS-RPMcancorrectlyregisterthepairwithoutoutliers(Figure 4 topleft)butfailedtomatchthecorruptedpair(Figure 4 topright). Figure4: Non-rigidregistrationofthecorpuscallosumdata.Leftcolumn:twoman-uallysegmentedcorpuscallosumslicesbeforeandafterregistration;Middlecolumn:warpingofthe2Dgridusingtherecoveredmotion;Topright:samesliceswithonecor-ruptedbynoiseandoutliers,beforeandafterregistration.

PAGE 67

Figure4: Experimentresultsonseven2Dcorpuscollasumpointsets.Thersttworowsandtheleftimageinthirdrowshowthedeformationofeachpoint-settotheat-las,superimposedwithinitialpointset(showin'o')anddeformedpoint-set(shownin'*').Middleimageinthethirdrow:Theestimatedatlasisshownsuperimposedoverallthepoint-sets.Right:Theestimatedatlasisshownsuperimposedoverallthedeformedpoint-sets. Wemanuallyextractedpointsontheoutercontourofthecorpuscallosumfromsevennormalsubjects,(asshownFigure 4 ,indicatedbyo).Therecovereddeformationbetweeneachpoint-setandthemeanshapearesuperimposedonthersttworowsinFigure 4 .Theresultingatlas(meanpoint-set)isshowninthirdrowofFigure 4 ,andissuperimposedoverallthepoint-sets.Aswedescribedearlier,alltheseresultsarecomputedsimultaneouslyandautomatically.Thisexampleclearlydemonstratethatour

PAGE 68

jointmatchingandatlasconstructionalgorithmcansimultaneouslyalignmultipleshapes(modeledbysamplepoint-sets)andcomputeameaningfulatlas/meanshape. 4 ,whichconrmthatCDF-JSisindeedmorerobustinthepresenceofhighnoiseandoutlierlevel.A3DexampleisalsopresentedinFigure 4 Next,wepresentgroupwiseregistrationresultson3Dhippocampalpoint-sets.Four3Dpoint-setswereextractedfromepilepsypatientswithleftanteriortemporallobefociidentiedwithEEG.Aninteractivesegmentationtoolwasusedtosegmentthehippocampusfromthe3DbrainMRIscansof4subjects.Thepoint-setsdifferinshape,withthenumberofpoints450;421;376;307ineachpoint-setrespectively.Intherst

PAGE 69

Figure4: Robustnesstooutliersinthepresenceoflargenoise.Errorsinestimatedrigidtransformvs.proportionofoutliers(()=())forbothourmethodandKCmethod. Figure4: Robustnessteston3Dswandata.'o'and'+'indicatethemodelandscenepointsrespectively.Notethatthescenepoint-setiscorruptedbynoiseandoutliers. fourimagesofFigure 4 ,therecoverednonrigiddeformationbetweeneachhippocampalpoint-settotheatlasisshownalongwithasuperimpositiononalloftheoriginaldatasets.InsecondrowoftheFigure 4 ,wealsoshowthescatterplotoforiginalpoint-setsalongwithallthepoint-setsafterthenon-rigidwarping.Anexaminationofthetwoscatterplotsclearlyshowstheefcacyofourrecoverednon-rigidwarping.Notethatvalidationofwhatanatlasshapeoughttobeintherealdatacaseisadifcultproblemandwerelegateitsresolutiontoafuturepaper.

PAGE 70

Figure4: Atlasconstructionfromfour3Dhippocampalpointsets.Therstrowandtheleftimageinsecondrowshowsthedeformationofeachpoint-settotheatlas(rep-resentedasclustercenters),superimposedwithinitialpointset(showingreen'o')anddeformedpoint-set(showninred'+').Leftimageinthesecondrow:Scatterplotoftheoriginalfourhippocampalpoint-sets.Right:Scatterplotofallthewarpedpoint-sets.

PAGE 71

InMedicalImagingapplications,segmentationcanbeadauntingtaskduetopossiblylargeinhomogeneitiesinimageintensitiesacrossanimagee.g.,inMRimages.Theseinhomogeneitiescombinedwithvolumeaveragingduringtheimagingandpossiblelackofpreciselydenedshapeboundariesforcertainanatomicalstructurescomplicatesthesegmentationproblemimmensely.Onepossiblesolutionforsuchsituationsisatlas-basedsegmentation.Theatlasonceconstructedcanbeusedasatemplateandcanberegisterednon-rigidlytotheimagebeingsegmented(henceforthcalledatargetimage)therebyachievingthedesiredsegmentation.Manyofthemethodsthatachieveatlas-basedsegmentationarebasedonatwostageprocessinvolving,(i)estimatingthenon-rigiddeformationeldbetweentheatlasimageandthetargetimageandthen,(ii)applyingtheestimateddeformationeldtothedesiredshape/atlastoachievethesegmentationofthecorrespondingstructure/sinthetargetimage.Inthischapter,wedevelopanoveltechniquethatwillsimultaneouslyachievethenon-rigidregistrationandsegmentation.Thereisavastbodyofliteratureforthetasksofregistrationandsegmentationindependentlyhowever,methodsthatcombinethemintoonealgorithmarefarandfewinbetween.Inthefollowing,wewillbrieyreviewthefewexistingmethodsthatattempttoachievesimultaneousregistrationandsegmentation. 76 ]developedaminmaxentropyframeworktorigidlyregister&segmentportalandCTdatasets.In[ 77 ],Yezzietal.,presentavariationalprincipleforachievingsimultaneousregistrationandsegmentation,however,theregistrationpartislimitedtorigidmotions.AsimilarlimitationappliestothetechniquepresentedbyNobleetal.,in[ 78 ].Avariational 59

PAGE 72

principleinalevel-setbasedformulationwaspresentedinPargioset.al.,[ 79 ],forsegmentationandregistrationofcardiacMRIdata.Theirformulationwasagainlimitedtorigidmotionandtheexperimentswerelimitedto2Dimages.InFischletal.,[ 80 ],aBayesianmethodispresentedthatsimultaneouslyestimatesalinearregistrationandthesegmentationofanovelimage.Notethatlinearregistrationdoesnotinvolvenon-rigiddeformations.Thecaseofjointregistrationandsegmentationwithnon-rigidregistrationhasnotbeenaddressedadequatelyinliteraturewiththeexceptionoftherecentworkreportedinSoattoandYezzi[ 81 ]andVemurietal.,[ 82 ].However,thesemethodscanonlyworkwithimagepairsthatarenecessarilyfromthesamemodalityortheintensityprolesarenottoodisparate. Inthispaper,wepresentauniedvariationalprinciplethatwillsimultaneouslyregistertheatlasshape(contour/surface)tothenovelbrainimageandsegmentthedesiredshape(contour/surface)inthenovelimage.Inthiswork,theatlasservesinthesegmentationprocessasapriorandtheregistrationofthispriortothenovelbrainscanwillassistinsegmentingit.Anotherkeyfeature/strengthofourproposedregistration+segmentationschemeisthatitaccommodatesforimagepairshavingverydistinctintensitydistributionsasinmultimodalitydatasets.Moredetailsonthisarepresentedinsection 5.2

PAGE 73

Figure5: ModelIllustration Where,thersttermdenotesthesegmentationfunctional.~Cistheboundarycontour(surfacein3D)ofthedesiredanatomicalshapeinI2.Thesecondtermmeasuresthedistancebetweenthetransformedatlasv(C)andthecurrentsegmentation~Cinthenovelbrainimagei.e.,thetargetimageandthethirdtermdenotesthenon-rigidregistrationfunctionalbetweenthetwoimages.Ourjointregistration&segmentationmodelisillustratedinFigure 5.2 Forthesegmentationfunctional,weuseapiecewiseconstantMumfordShahmodel,whichisoneofthewell-knownvariationalmodelsforimagesegmentation,whereinitisassumedthattheimagetobesegmentedcanbemodeledbypiece-wiseconstantregions,aswasdonein[ 54 ].Thisassumptionsimpliesourpresentationbutourmodelitselfcanbeeasilyextendedtothepiecewisesmoothregionscase.Additionally,sinceweareonlyinterestedinsegmentingadesiredanatomicalshape(e.g.,thehippocampus,thecorpuscallosum,etc.),wewillonlybeconcernedwithabinarysegmentationi.e.,twoclassesnamely,voxelsinsidethedesiredshapeandthosethatareoutsideit.Theseassumptionscanbeeasilyrelaxedifnecessarybutatthecostofmakingtheenergyfunctionalmorecomplicatedandhencecomputationallymorechallenging.Thesegmentationfunctionaltakesthefollowingform:

PAGE 74

Where,istheimagedomainandisaregularizationparameter.u=uiifx2~Cinandu=uoifx2~Cout.~Cinand~Coutdenotetheregionsinsideandoutsideofthecurve,~CrepresentingthedesiredshapeboundariesinI2. Forthenon-rigidregistrationtermintheenergyfunction,weusetheinformationtheoretic-basedcriteria,crosscumulativeresidualentropy(CCRE)whichweintroducedinChapters 2 .CCREwasshowntooutperformMutualInformationbasedregistrationinthecontextofnoiseimmunityandconvergencerange,motivatingustopickthiscriteriaovertheMI-basedcostfunction.Thenewregistrationfunctionalisdenedby where,cross-CREC(I1;I2)isgivenby, withE(I1)=RR+P(jI1j>)logP(jI1j>)dandR+=(x2R;x0).v(x)isasbeforeandistheregularizationparameterandjjjjdenotesFrobeniusnorm.UsingaB-splinerepresentationofthenon-rigiddeformation,oneneedonlycomputethiseldatthecontrolpointsoftheB-splinesandinterpolateelsewhere,thusaccruingcomputationaladvantages.Usingthisrepresentation,wehavederivedanalyticexpressionsforthegradientoftheenergywithrespecttotheregistrationparameters.Thisinturnmakesouroptimizationmorerobustandefcient. Inorderfortheregistrationandthesegmentationtermstotalktoeachother,weneedaconnectiontermandthatisgivenby where,Ristheregionenclosedby~C,v(C)(x)istheembeddingsigneddistancefunctionofthecontourv(C),whichcanbeusedtomeasurethedistancebetweenv(C)and~C.Thelevel-setfunction:R2!Rischosensothatitszerolevel-setcorrespondstothe

PAGE 75

transformedtemplatecurvev(C).LetEdist:=dist(v(C);~C),onecanshowthat@Edist @t=v(C)(~C)N(5) Notonlydoesthesigneddistancefunctionrepresentationmakeiteasierforustoconvertthecurveevolutionproblemtothelevel-setframework,italsofacilitatesthematchingoftheevolvingcurve~Candthetransformedtemplatecurvev(C),andyetdoesnotrelyonaparametricspecicationofeither~Corthetransformedtemplatecurve.Notethatsincedist(v(C);~C)isafunctionoftheunknownregistrationvandtheunknownsegmentation~C,itplaysthecrucialroleofconnectingtheregistrationandthesegmentationterms. Combiningthesethreefunctionalstogether,wegetthefollowingvariationalprincipleforthesimultaneousregistration+segmentationproblem:

PAGE 76

andmerges),theabilitytodealwiththeformationofcuspsandcorners,whichareextremelycommonincurveevolution,andthenumericalstabilityandefciencyaffordedinitsimplementation.Forourmodelwheretheequationfortheunknowncurve~Ciscoupledwiththeequationsforv(x),uo;ui,itisconvenientforustousethelevelsetapproachasproposedin[ 54 ]. TakingthevariationofE(:)withrespectto~Candwritingdownthegradientdescentleadstothefollowingcurveevolutionequation: @t=h(I2ui)2+(I2uo)2+1+2v(C)(~C)iN(5) Notethatequation( 5 )isusedinthederivation.Equation( 5 )inthelevel-setframeworkisgivenby: @t=(I2ui)2+(I2uo)2+1rr whereuianduoarethemeanvaluesinsideandoutsideofthecurve~CintheimageI2.Todrivethecurvetowardsthetemplate'slevel-setfunctionv( @t=h(I2ui)2+(I2uo)2+1rr AsillustratedinFigure 5 ,thetwoparameters1and2areusedtobalancetheinuenceoftheshapedistancemodelandtheregion-basedmodel.Notethat(~C)=0atanylocationofthecurvebythedenitionoflevel-setfunction,thisaddedtermdoesnotaffectthecurveevolutionequation[ 83 ]. Asmentionedbefore,weuseaB-splinebasistorepresentthedisplacementvectoreldv(x;),whereisthetransformationparametersoftheB-splinebasis. @=2@RRv(C)(x)dx

PAGE 77

Figure5: Illustrationofthevarioustermsintheevolutionofthelevelsetfunction.Toupdate,wecombinethestandardregionbasedupdatetermS,andlevelsetfunctioncorrespondingtotheshapedistanceterm. Thersttermofequation( 5 )canberewrittenasfollows: where@v(C) 5 )hasbeenderivedinEqn.( 3 )ofthechapter 3 ..Wesimplystatetheresultherewithoutthederivationsforthesakeofbrevity, whereP(i>;k;)andP(i>;)arethejointandmarginalcumulativeresidualdistributionsrespectively.pI2(k)isthedensityfunctionofimageI2.ThelasttermofEqn.( 5 )leadsto, whereboththematricesrvand@v

PAGE 78

Substitutingequations( 5 ),( 5 )and( 5 )respectivelybackintotheequation( 5 ),wegettheanalyticalgradientofourenergyfunctionwithrespecttotheB-splinetransformationparameters.Wethensolveforthestationarypointofthisnonlinearequationnumericallyusingaquasi-Newtonmethod. 52 ].Theywereoriginallyalignedwitheachother.WeusetheMRT1imageasthesourceimageandthetargetimagewasgeneratedfromtheMRT2imagebyapplyingaknownnon-rigidtransformationthatwasprocedurallygeneratedusingkernel-basedsplinerepresentations(cubicB-Spline).Thepossiblevaluesofeachdirectionindeformationvaryfrom15to15inpixels.Inthis

PAGE 79

Figure5: Resultsofapplicationofouralgorithmtosyntheticdata(seetextfordetails). case,wepresenttheerrorintheestimatednon-rigiddeformationeld,usingouralgorithm,asanindicatoroftheaccuracyofestimateddeformations. Figure 5 depictstheresultsobtainedforthisimagepair.WiththeMRT1imageasthesourceimage,thetargetwasobtainedbyapplyingasyntheticallygeneratednon-rigiddeformationeldtotheMRT2image.Noticethesignicantdifferencebetweentheintensityprolesofthesourceandtargetimages.Figure 5 isorganizedasfollows,fromlefttoright:therstrowdepictsthesourceimagewiththeatlas-segmentationsuperposedinred,theregisteredsourceimagewhichisobtainedusingouralgorithmfollowedbythetargetimagewiththeunregisteredatlas-segmentationsuperposedtodepicttheamountofmis-alignment;secondrowdepictsgroundtruthdeformationeldwhichweusedtogeneratethetargetimagefromtheMRT2image,followedbytheestimatednon-rigiddeformationeldandnallythesegmentedtarget.Asvisuallyevident,theregistration+segmentationarequiteaccuratefromavisualinspectionpointofview.Asameasureofaccuracyofourmethod,weestimatedtheaverage,,andthestandarddeviation,,oftheerrorintheestimatednon-rigiddeformationeld.Theerrorwasestimatedastheanglebetweenthegroundtruthandestimateddisplacementvectors.The

PAGE 80

Table 5 depictsstatisticsoftheerrorinestimatednon-rigiddeformationwhencomparedtothegroundtruth.Forthemeangroundtruthdeformation(magnitudeofthedisplacementvector)inColumn-1ofeachrow,5distinctdeformationeldswiththismeanaregeneratedandappliedtothetargetimageofthegivensource-targetpairtosynthesize5pairsofdistinctdatasets.Thesepairs(oneatatime)areinputtoouralgorithmandthemean()ofthemeandeformationerror(MDE)iscomputedoverthevepairsandreportedinColumn-2ofthetable.MDEisdenedas Table5: Statisticsoftheerrorinestimatednon-rigiddeformation. 2.4 0.5822 0.0464 3.3 0.6344 0.0923 4.5 0.7629 0.0253 5.5 0.7812 0.0714

PAGE 81

Figure5: ResultsofapplicationofouralgorithmtoapairofslicesfromhumanbrainMRIs(seetextfordetails). inthesourceimage.Figure 5 isorganizedasfollows,fromlefttoright:therstrowdepictsthesourceimagewiththeatlas-segmentationsuperposedinblackfollowedbythetargetimagewiththeunregisteredatlas-segmentationsuperposedtodepicttheamountofmis-alignment;secondrowdepictstheestimatednon-rigidvectoreldandnallythesegmentedtarget.Asevidentfromgures 5 ,theaccuracyoftheachievedregistration+segmentationvisuallyverygood.Notethatthenon-rigiddeformationbetweenthetwoimagesinthesetwoexamplesisquitelargeandourmethodwasabletosimultaneouslyregisterandsegmentthetargetdatasetsquiteaccurately. ThesecondrealdataexampleisobtainedfromtwobrainMRIsofdifferentsubjectsandmodalities,thesegmentationofthecerebelluminthesourceimageisgiven.Weselectedtwocorrespondingslicesfromthesevolumedatasetstoconducttheexperiment.Notethateventhoughthenumberofslicesforthetwodatasetsarethesame,theslicesmaynotcorrespondtoeachotherfromananatomicalpointofview.However,forthepurposesofillustrationofouralgorithm,thisisnotverycritical.Weusethecorrespondingsliceofthe3Dsegmentationofthesourceasouratlas-segmentation.Theresultsofanapplication

PAGE 82

Figure5: CorpusCallosumsegmentationonapairofcorrespondingslicesfromdis-tinctsubjects. ofouralgorithmareorganizedasbeforeingure 5 .Onceagain,asevident,thevisualqualityofthesegmentationandregistrationareveryhigh. Finallywepresenta3Drealdataexperiment.Inthisexperiment,theinputisapairof3Dbrainscanswiththesegmentationofthehippocampusinoneofthetwoimages(labeledtheatlasimage)beingobtainedusingthewellknownPCAontheseveraltrainingdatasets.Eachdatasetcontains19slicesofsize256x256.Thegoalwasthentoautomaticallyndthehippocampusinthetargetimagegiventheinput.Figure 5 depictstheresultsobtainedforthisimagepair.Fromlefttoright,therstimageshowsthegiven(atlas)hippocampussurfacefollowedbyonecross-sectionofthissurfaceoverlaidonthesourceimageslice;thethirdimageshowsthesegmentedhippocampussurfacefromthetargetimageusingouralgorithmandnallythecross-sectionofthesegmentedsurfaceoverlaidonthetargetimageslice.Tovalidatetheaccuracyofthesegmentationresult,werandomlysampled120pointsfromthesegmentedsurfaceandcomputedtheaveragedistancefromthesepointstothegroundtruthhandsegmentedhippocampalsurfaceinthetargetimage.Thehandsegmentationwasperformedbyanexpertneuroanatomist.The

PAGE 83

Figure5: Hippocampalsegmentationusingouralgorithmonapairofbrainscansfromdistinctsubjects.(seetextfordetails) averageandstandarddeviationoftheerrorintheaforementioneddistanceinestimatedhippocampalshapeare0.8190and0.5121(invoxels)respectively,whichisveryaccurate.

PAGE 84

Wedemonstratedtheirapplicationstothefollowingmedicalimageanalysisproblems, Ourcontributionstoeachofthesetopicsaresummarizedinthefollowingsections. 6.2.1Non-rigidImageRegistration 84 ]tomeasurethesimilaritybetweentwoimages.Thematchingmeasureisdenedbasedonanewinformationmeasure,namelycumulativeresidualentropy(CRE),whichisdenedbasedontheprobabilitydistributionsinsteadofprobabilitydensities,thereforeCCREisvalidforbothdiscreteandcontinuousdomain.Furthermore,CCREalsoinheritstherobustnesspropertyoftheCREmeasure.In[ 84 ],wepresentedresultsofrigidandafneregistrationunderavarietyofnoiselevelsandshowedsignicantlysuperiorperformanceoverMI-basedmethods. 72

PAGE 85

TheCross-CREbetweentwoimagestoberegisteredismaximizedoverthespaceofsmoothandunknownnon-rigidtransformations,whichisrepresentedbyatri-cubicBSplinesplacedonaregulargird.Theanalyticgradientofthismatchingmeasureisthenderivedinthispapertoachieveefcientandaccuratenon-rigidregistration.ItturnsoutthatthegradientoftheCCREhasasimilarformulationwiththecostfunction,whichgreatlysavesmemoryspaceintheoptimizationprocess.ThematchingcriterionisoptimizedusingQuasi-Newtonmethodtorecoverthetransformationparameters. Thekeystrengthsofourproposednon-rigidregistrationschemearedemonstratedthroughtheregistrationofthesyntheticaswellasrealdatasetsfrommulti-modality(MRT1andT2weighted,MR&CT)imagingsources.ItisshowedthatourCCREnotonlycanaccommodateimagestoberegisteredofvaryingcontrast+brightness,butitisalsorobustinthepresenceofnoise.CCREconvergesfasterwhencomparedwithotherinformationtheory-basedregistrationmethods.FinallywealsoshowedthatCCREiswellsuitedforsituationswherethesourceandthetargetimageshaveFOVswithlargenon-overlappingregions(whichisquitecommoninpractice).ComparisonsweremadebetweenCCREandtraditionalMI[ 34 51 ],whichwasdenedusingtheShannonentropy.AlltheexperimentsdepictedsignicantlybetterperformanceofCCREovertheMI-basedmethodscurrentlyusedinliterature. Ourfutureworkwillfocusonextendingthetransformationsmodeltotheonethatpermitsthespatialadaptationofthetransformation'scompliance,whichwillallowustoreducethenumberofdegreesoffreedomintheoveralltransformation.Validationofnon-rigidregistrationonrealdatawiththeaidofsegmentationsandlandmarksobtainedmanuallyfromagroupoftrainedanatomistsarethegoalsofourongoingwork.

PAGE 86

sets,weproposedseveraldivergencemeasures,therstofwhichistheJensen-Shannondivergence.Sinceitlacksrobustness,wedevelopanovelmeasurebasedontheircumulativedistributionfunctionsthatwedubastheCDF-JSdivergence.ThemeasureparallelsthewellknownJensen-Shannondivergence(denedforprobabilitydensityfunctions)butismoreregularthantheJSdivergencesinceitsdenitionisbasedonCDFsasopposedtodensityfunctions.Asaconsequence,CDF-JSismoreimmunetonoiseandstatisticallymorerobustthantheJS. Ourproposedmethodsdonotrequireanyknowledgeofcorrespondencebetweentheinputpoint-sets,andthereforethesepoint-setsneednothavethesamecardinality.Oneothersalientfeatureofourproposedalgorithmsisthatwegetaprobabilisticatlasasthebyproductoftheregistrationprocess.Ouralgorithmcanbeespeciallyusefulforcreatingatlasesofvariousshapespresentinimagesaswellasforsimultaneously(rigidlyornon-rigidly)registering3Drangedatasetswithouthavingtoestablishanycorrespondence. Ourfutureworkwillfocusonusingmaximumlikelihoodestimation(MLE)toautomaticallydetermineweightingcoefcientsinthedivergencemeasuresandsmoothingterm;Wearealsoattemptingtoextendourtechniquestodiffeomorphicpoint-setsmatching.

PAGE 87

estimatesoftheregistrationinthesyntheticdatacase.Theaccuracyasevidentintheseexperimentsisquitesatisfactory.Ourfutureeffortswillfocusonadaptingouralgorithm+softwarefortheclinicuse.

PAGE 88

[1] C.E.Shannon,Amathematicaltheoryofcommunication,BellSystemTechnicalJournal,pp.379and623,1948. [2] W.F.Sharpe,Investments.London:PrenticeHall,1985. [3] D.Salomon,DataCompression.NewYork:Springer,1998. [4] S.Kullback,InformationTheoryandStatistics.NewYork:Wiley,1959. [5] T.M.CoverandJ.A.Thomas,ElementsofInformationTheory.NewYork:Wiley,1991. [6] G.Jumarie,RelativeInformation.NewYork:Springer,1990. [7] M.Rao,Y.Chen,B.C.Vemuri,andF.Wang,Cumulativeresidualentropy,anewmeasureofinformation,IEEETransactionsonInformationTheory,vol.50,no.6,pp.1220,June2004. [8] M.AsadiandY.Zohrevand,Onthedynamiccumulativeresidualentropy,UnpublishedManuscript,2006. [9] H.Chui,L.Win,R.Schultz,J.Duncan,andA.Rangarajan,Auniednon-rigidfeatureregistrationmethodforbrainmapping,MedicalImageAnalysis,vol.7,no.2,pp.112,2003. [10] N.Paragios,M.Rousson,andV.Ramesh,Non-rigidregistrationusingdistancefunctions,Comput.Vis.ImageUnderst.,vol.89,no.2-3,pp.142,2003. [11] M.A.Audette,K.Siddiqi,F.P.Ferrie,andT.M.Peters,Anintegratedrange-sensing,segmentationandregistrationframeworkforthecharacterizationofintra-surgicalbraindeformationsinimage-guidedsurgery,Comput.Vis.ImageUnderst.,vol.89,no.2-3,pp.226,2003. [12] A.Leow,P.M.Thompson,H.Protas,andS.-C.Huang,Brainwarpingwithimplicitrepresentations.inInternationalSymposiumonBiomedicalImaging,2004,pp.603. [13] B.JianandB.C.Vemuri,Arobustalgorithmforpointsetregistrationusingmixtureofgaussians.inIEEEInternationalConferenceonComputerVision,2005,pp.1246. 76

PAGE 89

[14] F.Wang,B.C.Vemuri,A.Rangarajan,I.M.Schmalfuss,andS.J.Eisenschenk,Simultaneousnonrigidregistrationofmultiplepointsetsandatlasconstruction,inEuropeanConferenceonComputerVision,2006,pp.551. [15] S.J.H.Guo,A.Rangarajan,Anewjointclusteringanddiffeomorphismestimationalgorithmfornon-rigidshapematching,inIEEEComputerVisionandPatternRecognition,2004,pp.16. [16] M.IraniandP.Anandan,RobustMulti-sensorImageAlignment,inInternationalConferenceonComputerVision,Bombay,India,1998,pp.959. [17] J.Liu,B.C.Vemuri,andJ.L.Marroquin,Localfrequencyrepresentationsforrobustmultimodalimageregistration,IEEETransactionsonMedicalImaging,vol.21,no.5,pp.462,2002. [18] M.MellorandM.Brady,Non-rigidmultimodalimageregistrationusinglocalphase,inMedicalImageComputingandComputer-AssistedIntervention,Saint-Malo,France,Sep2004,pp.789. [19] B.ZitovaandJ.Flusser,Imageregistrationmethods:asurvey.ImageVisionComput.,vol.21,no.11,pp.977,2003. [20] J.Ruiz-Alzola,C.-F.Westin,S.K.Wareld,A.Nabavi,andR.Kikinis,Nonrigidregistrationof3dscalarvectorandtensormedicaldata,inThirdInternationalConferenceonMedicalImageComputingandComputer-AssistedIntervention,A.M.DiGioiaandS.Delp,Eds.,Pittsburgh,October112000,pp.541. [21] L.Marroquin,B.Vemuri,S.Botello,F.Calderon,andA.Fernandez-Bouzas,Anaccurateandefcientbayesianmethodforautomaticsegmentationofbrainmri,inIEEETransactionsonMedicalImaging,2002,pp.934. [22] B.C.Vemuri,J.Ye,Y.Chen,andC.M.Leonard,Alevel-setbasedapproachtoimageregistration,inIEEEWorkshoponMathematicalMethodsinBiomedicalImageAnalysis,2000,pp.86. [23] P.Hellier,C.Barillot,E.Mmin,andP.Prez,Hierarchicalestimationofadensedeformationeldfor3drobustregistration,IEEETransactionsonMedicalImaging,vol.20,no.5,pp.388,May2001. [24] R.SzeliskiandJ.Coughlan,Spline-basedimageregistration,Int.J.Comput.Vision,vol.22,no.3,pp.199,March1997. [25] S.H.LaiandM.Fang,Robustandefcientimagealignmentwithspatially-varyingilluminationmodels,inIEEEConferenceonComputerVisionandPatternRecognition,1999,pp.II:167. [26] A.Guimond,A.Roche,N.Ayache,andJ.Menuier,Three-DimensionalMultimodalBrainWarpingUsingtheDemonsAlgorithmandAdaptiveIntensity

PAGE 90

Corrections,IEEETransactionsonMedicalImaging,vol.20,no.1,pp.58,2001. [27] J.-P.Thirion,Imagematchingasadiffusionprocess:ananalogywithmaxwell'sdemons,MedicalImageAnalysis,vol.2,no.3,pp.243,1998. [28] A.Cuzol,P.Hellier,andE.Memin,Anovelparametricmethodfornon-rigidimageregistration,inProc.InformationProcessinginMedicalImaging(IPMI'05),ser.LNCS,G.ChristensenandM.Sonka,Eds.,no.3565,GlenwoodSpringes,Colorado,USA,July2005,pp.456. [29] A.W.TogaandP.M.Thompson,Theroleofimageregistrationinbrainmapping,ImageVisionComput.,vol.19,no.1-2,pp.3,2001. [30] E.D'Agostino,F.Maes,D.Vandermeulen,andP.Suetens,Non-rigidatlas-to-imageregistrationbyminimizationofclass-conditionalimageentropy.inMedicalImageComputingandComputer-AssistedIntervention,2004,pp.745. [31] P.A.ViolaandW.M.Wells,Alignmentbymaximizationofmutualinformation,inIEEEInternationalConferenceonComputerVision,MIT,Cambridge,1995. [32] A.Collignon,F.Maes,D.Delaere,D.Vandermeulen,P.Suetens,andG.Marchal,Automatedmultimodalityimageregistrationbasedoninformationtheory,Proc.InformationProcessinginMedicalImaging,pp.263,1995. [33] C.Studholme,D.Hill,andD.J.Hawkes,Automated3DregistrationofMRandCTimagesinthehead,MedicalImageAnalysis,vol.1,no.2,pp.163,1996. [34] D.Mattes,D.R.Haynor,H.Vesselle,T.K.Lewellen,andW.Eubank,Pet-ctimageregistrationinthechestusingfree-formdeformations.IEEETransactionsonMedicalImaging,vol.22,no.1,pp.120,2003. [35] D.Rueckert,A.F.Frangi,andJ.A.Schnabel,Automaticconstructionof3dstatisticaldeformationmodelsofthebrainusingnon-rigidregistration.IEEETransactionsonMedicalImaging,vol.22,no.8,pp.1014,2003. [36] G.Hermosillo,C.Chefd'hotel,andO.Faugeras,Variationalmethodsformultimodalimagematching,Int.J.Comput.Vision,vol.50,no.3,pp.329,2002. [37] D.Rueckert,L.I.Sonoda,C.Hayes,D.L.G.Hill,M.O.Leach,andD.J.Hawkes,Nonrigidregistrationusingfree-formdeformations:Applicationtobreastmrimages,IEEETransactionsonMedicalImaging,vol.18,no.8,pp.712,August1999. [38] M.E.LeventonandW.E.L.Grimson,Multimodalvolumeregistrationusingjointintensitydistributions,inMedicalImageComputingandComputer-AssistedIntervention(MICCAI),Cambridge,MA,1998,pp.1057.

PAGE 91

[39] T.Gaens,F.Maes,D.Vandermeulen,andP.Suetens,Non-rigidmultimodalimageregistrationusingmutualinformation,inProc.ConferenceonMedicalImageComputingandCompter-AssistedIntervention(MICCAI),1998,pp.1099. [40] D.Loeckx,F.Maes,D.Vandermeulen,andP.Suetens,Nonrigidimageregistrationusingfree-formdeformationswithalocalrigidityconstraint.inMedicalImageComputingandComputer-AssistedIntervention,2004,pp.639. [41] G.K.Rohde,A.Aldroubi,andB.M.Dawant,Theadaptivebasesalgorithmforintensitybasednonrigidimageregistration.IEEETransactionsonMedicalImaging,vol.22,no.11,pp.1470,2003. [42] V.Duay,P.-F.D'Haese,R.Li,andB.M.Dawant,Non-rigidregistrationalgorithmwithspatiallyvaryingstiffnessproperties.inInternationalSymposiumonBiomedicalImaging,2004,pp.408. [43] C.Guetter,C.Xu,F.Sauer,andJ.Hornegger,Learningbasednon-rigidmulti-modalimageregistrationusingkullback-leiblerdivergence.inMedicalImageComputingandComputer-AssistedIntervention,2005,pp.255. [44] E.D'Agostino,F.Maes,D.Vandermeulen,andP.Suetens,Aninformationtheoreticapproachfornon-rigidimageregistrationusingvoxelclassprobabilities,MedicalImageAnalysis,vol.10,no.3,pp.413,2006. [45] C.Davatzikos,Spatialtransformationandregistrationofbrainimagesusingelasticallydeformablemodels,Comput.Vis.ImageUnderst.,vol.66,no.2,pp.207,1997. [46] J.C.Gee,M.Reivich,andR.Bajcsy,Elasticallydeforming3datlastomatchanatomicalbrainimages,J.Comput.Assist.Tomogr.,vol.17,no.2,pp.225,1993. [47] M.Bro-NielsenandC.Gramkow,Fastuidregistrationofmedicalimages,inProc.ofthe4thInternationalConferenceonVisualizationinBiomedicalComputing.London,UK:Springer-Verlag,1996,pp.267. [48] G.E.Christensen,R.D.Rabbitt,andM.I.Miller,Deformabletemplatesusinglargedeformationkinematics,IEEETransactionsOnImageProcessing,vol.5,no.10,pp.1435,October1996. [49] X.Geng,D.Kumar,andG.E.Christensen,Transitiveinverse-consistentmanifoldregistration.inProc.InformationProcessinginMedicalImaging,2005,pp.468. [50] A.Trouve,Diffeomorphismsgroupsandpatternmatchinginimageanalysis,Int.J.Comput.Vision,vol.28,no.3,pp.213,1998. [51] D.R.ForseyandR.H.Bartels,Hierarchicalb-splinerenement,ComputerGraphics,vol.22,no.4,pp.205,1988.

PAGE 92

[52] C.Cocosco,V.Kollokian,R.-S.Kwan,andA.Evans,Brainweb:onlineinterfacetoa3-dmrisimulatedbraindatabase,1997,lastaccessed:July2005.[Online].Available: http://www.bic.mni.mcgill.ca/brainweb/ [53] P.ThevenazandM.Unser,Optimizationofmutualinformationformultiresolutionimageregistration,IEEETransactionsonImageProcessing,vol.9,no.12,pp.2083,December2000. [54] T.ChanandL.Vesse,Anactivecontourmodelwithoutedges,inIntl.Conf.onScale-spaceTheoriesinComputerVision,1999,pp.266. [55] N.Duta,A.K.Jain,andM.-P.Dubuisson-Jolly,Automaticconstructionof2dshapemodels,IEEETransactionsPatternAnal.Mach.Intell.,vol.23,no.5,pp.433,2001. [56] E.Klassen,A.Srivastava,W.Mio,andS.H.Joshi,Analysisofplanarshapesusinggeodesicpathsonshapespaces.IEEETransactionsPatternAnal.Mach.Intell.,vol.26,no.3,pp.372,2003. [57] T.B.Sebastian,P.N.Klein,B.B.Kimia,andJ.J.Crisco,Constructing2dcurveatlases,inIEEEWorkshoponMathematicalMethodsinBiomedicalImageAnalysis,Washington,DC,USA,2000,pp.70. [58] H.Tagare,Shape-basednonrigidcorrespondencewithapplicationtoheartmotionanalysis.IEEETransactionsonMedicalImaging,vol.18,no.7,pp.570,1999. [59] F.L.Bookstein,Principalwarps:Thin-platesplinesandthedecompositionofdeformations.IEEETransactionsPatternAnal.Mach.Intell.,vol.11,no.6,pp.567,1989. [60] H.Chui,A.Rangarajan,J.Zhang,andC.M.Leonard,Unsupervisedlearningofanatlasfromunlabeledpoint-sets.IEEETransactionsPatternAnal.Mach.Intell.,vol.26,no.2,pp.160,2004. [61] S.Belongie,J.Malik,andJ.Puzicha,Shapematchingandobjectrecognitionusingshapecontexts,IEEETransactionsPatternAnal.Mach.Intell.,vol.24,no.4,pp.509,2002. [62] H.ChuiandA.Rangarajan,Anewalgorithmfornon-rigidpointmatching.inIEEEComputerVisionandPatternRecognition,2000,pp.2044. [63] H.Guo,A.Rangarajan,S.Joshi,andL.Younes,Non-rigidregistrationofshapesviadiffeomorphicpointmatching.inInternationalSymposiumonBiomedicalImaging,2004,pp.924. [64] T.F.Cootes,C.J.Taylor,D.H.Cooper,andJ.Graham,Activeshapemodels:theirtrainingandapplication,Comput.Vis.ImageUnderst.,vol.61,no.1,pp.38,1995.

PAGE 93

[65] Y.WangandL.H.Staib,Boundaryndingwithpriorshapeandsmoothnessmodels,IEEETransactionsonPatternAnalysisandMachineIntelligence,vol.22,no.7,pp.738,2000. [66] A.Hill,C.J.Taylor,andA.D.Brett,Aframeworkforautomaticlandmarkidenticationusinganewmethodofnonrigidcorrespondence.IEEETransactionsPatternAnal.Mach.Intell.,vol.22,no.3,pp.241,2000. [67] Y.TsinandT.Kanade,Acorrelation-basedapproachtorobustpointsetregistration.inEuropeanConferenceonComputerVision,2004,pp.558. [68] J.Glaunes,A.Trouve,andL.Younes,Diffeomorphicmatchingofdistributions:Anewapproachforunlabelledpoint-setsandsub-manifoldsmatching.inIEEEComputerVisionandPatternRecognition,2004,pp.712. [69] J.Lin,Divergencemeasuresbasedontheshannonentropy,IEEETransactionsInformationTheory,vol.37,pp.145,1991. [70] A.Hero,O.M.B.Ma,andJ.Gorman,Applicationsofentropicspanninggraphs,IEEETransactionsSignalProcessing,vol.19,pp.85,2002. [71] Y.He,A.Ben-Hamza,andH.Krim,Ageneralizeddivergencemeasureforrobustimageregistration,IEEETransactionsSignalProcessing,vol.51,pp.1211,2003. [72] D.M.EndresandJ.E.Schindelin,Anewmetricforprobabilitydistributions,IEEETransactionsInformationTheory,vol.49,pp.1858,2003. [73] H.ChuiandA.Rangarajan,Anewpointmatchingalgorithmfornon-rigidregistration,ComputerVisionandImageUnderstanding(CVIU),vol.89,pp.114,2003. [74] G.McLachlanandK.Basford,MixtureModel:InferenceandApplicationstoClustering.NewYork:MarcelDekker,1988. [75] A.L.Yuille,P.Stolorz,andJ.Utans,Statisticalphysics,mixturesofdistributions,andtheemalgorithm,NeuralComput.,vol.6,no.2,pp.334,1994. [76] R.Bansal,L.Staib,Z.Chen,A.Rangarajan,J.Knisely,R.Nath,andJ.Duncan.,Entropy-based,multiple-portal-to-3dctregistrationforprostateradiotherapyusingiterativelyestimatedsegmentation,inMedicalImageComputingandComputer-AssistedIntervention,1999,pp.567. [77] A.Yezzi,L.Zollei,andT.Kapur,Avariationalframeworkforjointsegmentationandregistration,inIEEEWorkshoponMathematicalMethodsinBiomedicalImageAnalysis,2001,pp.388. [78] P.WyattandJ.Noble,Mrf-mapjointsegmentationandregistration,inMedicalImageComputingandComputer-AssistedIntervention,2002,pp.580.

PAGE 94

[79] N.Paragios,M.Rousson,andV.Ramesh,Knowledge-basedregistration&segmentationoftheleftventricle:Alevelsetapproach.inWACV,2002,pp.37. [80] B.Fischl,D.Salat,E.Buena,andM.A.et.al.,Wholebrainsementation:Automatedlabelingoftheneuroanatomicalstructuresinthehumanbrain,inNeuron,vol.33,2002,pp.341. [81] S.SoattoandA.J.Yezzi,Deformotion:Deformingmotion,shapeaverageandthejointregistrationandsegmentationofimages,inEuropeanConferenceonComputerVision,2002,pp.32. [82] B.C.Vemuri,Y.Chen,andZ.Wang,Registrationassistedimagesmoothingandsegmentation,inEuropeanConferenceonComputerVision,2002,pp.546. [83] T.ZhangandD.Freedman,Trackingobjectsusingdensitymatchingandshapepriors.inIEEEInternationalConferenceonComputerVision,2003,pp.1056. [84] F.Wang,B.C.Vemuri,M.Rao,andY.Chen,Anew&robustinformationtheoreticmeasureanditsapplicationtoimagealignment.inProc.InformationProcessinginMedicalImaging,2003,pp.388.

PAGE 95

FeiWangwasborninYanCheng,JiangSu,P.R.China.HereceivedhisBachelorofSciencedegreefromtheUniversityofScienceandTechnologyofChina,P.R.China,in2001.HeearnedhisMasterofScienceandDoctorofPhilosophydegreefromtheUniversityofFloridainDecember2002andAugust2006respectively.Hisresearchinterestsincludemedicalimaging,computervision,patternrecognition,computergraphicsandshapemodeling. 83


Permanent Link: http://ufdc.ufl.edu/UFE0015631/00001

Material Information

Title: Information Theoretic Measures and Their Applications to Image Registration and Segmentation
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015631:00001

Permanent Link: http://ufdc.ufl.edu/UFE0015631/00001

Material Information

Title: Information Theoretic Measures and Their Applications to Image Registration and Segmentation
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015631:00001


This item has the following downloads:


Full Text












INFORMATION THEORETIC MEASURES AND THEIR APPLICATIONS TO
IMAGE REGISTRATION AND SEGMENTATION

















By

FEI WANG


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2006


































Copyright 2006

by

Fei Wang
















For my wife, Lin, and my parents.















ACKNOWLEDGMENTS

I would like to first thank my advisor, Dr. Baba C. Vemuri, for everything he has

done for me during my doctoral study. This dissertation would not have taken shape

without his invaluable input. Dr. Vemuri introduced me to the field of medical image

analysis. His insight and experience have guided me throughout my research during

which time he provided numerous invaluable suggestions. It was a great pleasure

for me to conduct this dissertation under his supervision. I would also like to thank

Dr.Anand Rangarajan, Dr. Sartaj Sahni, Dr. Arunava Banerjee and Dr. Tan Wong for their

willingness to serve on my committee. In addition, special thanks go to Dr. Jorg Peters

for attending my PhD oral examination.

My doctoral research is a happy cooperation with many people. Dr. Vemuri has been

involved with the whole process, Dr. Rangarajan has guided me a lot in the groupwise

point registration part, Dr. Ilona Schmalfuss and Dr. Stephan Eisenschenk have kindly

provided the data for hippocampal segmentation and taught me what little I know of

Neuroscience. I have also benefitted from Dr. Thomas E. Davis's guidance when I first

joined the lab. I would also like to thank Dr. Banerjee for stimulating debates, and Dr.

Jeffrey Ho for his professional advice and philosophical discussions. Thanks also goes

to my co-authors, Drs. Murali Rao and Yunmei Chen, who were my co-authors on a set

of papers that introduced, the concept of entropy based on probability distributions and

several properties of the same.

Needless to say, I am grateful for the support of my colleagues and friends at the

Computer and Information Science and Engineering Department at the University of

Florida. Dr. Zhizhou Wang, Dr. Jundong Liu, Dr. Tim Mcgraw, Dr. Eric Spellman, Bing

Jian, Santhosh Kodipaka, Nicholas Lord, Neeti Vohra, Angelos Barmpoutis, Seniha Esen










Yuksel, Ozlem Subakan, Ritwik Kumar, Evren Ozarslan, Ajit Rajwade, Adrian Peter, Dr.

Jie Zhang and Dr. Hongyu Guo all deserve thanks.

And finally, most importantly, I thank my family. I thank my mother and father

for everything and my brother, too. And of course I thank my dearest Lin for her

understanding and love during the past few years. Their support and encouragement are

my source of strength.

This research was supported in part by the grants NIH RO1-NS42075 and NIH

R01-NS046812. I would also like to acknowledge travel support (for attending various

conferences to present research papers) from the IEEE Computer Society, the Department

of Computer and Information Science and Engineering and the College of Engineering of

the University of Florida.


















TABLE OF CONTENTS


ACKNOW LEDGMENTS ................................

LIST OF TABLES .................

LIST OF FIGURE S . . . . . . . . .

A B STR A C T . . . . . . . . . .

CHAPTER

1 INTRODUCTION ................................

1.1 Image and Point-set Registration ................... .
1.1.1 Im age R registration . . . . . . .
1.1.2 Groupwise Point-sets Registration ................
1.2 Im age Segm entation . . . . . . . .
1.3 Outline of Rem ainder . . . . . . .

2 ENTROPY AND RELATED MEASURES .........

2.1 Shannon Entropy and Related Measures .................
2.2 Cumulative Residual Entropy: A New Measure of Information ......
2.3 Properties of CRE . . . . . . . .
2.3.1 CRE and Empirical CRE ......................
2.3.2 Robustness of CRE .........................

3 APPLICATIONS TO MULTIMODALITY IMAGE REGISTRATION ..


3.1 Related W ork ......... .........
3.2 Multimodal Image Registration using CCRE . .
3.2.1 Transformation Model for Non-rigid Motion .
3.2.2 Measure Optimization . . . .
3.2.3 Computation of P(i > A, k; p) and P(i>,k;p)
3.2.4 Algorithm Summary . . . .
3.3 Implementation Results . . . . .
3.3.1 Synthetic Motion Experiments . . .
3.3.1.1 Convergence speed . ......
3.3.1.2 Registration accuracy . .....
3.3.1.3 Noise immunity . ........
3.3.1.4 Partial overlap . . . .


. . 14
. . 17
. . 2 1
. . 2 1
. . 2 3
. . 2 5
. . 2 5
. . 26
. . 26
. . 2 8
. . 29
. . 3 0


page












3.3.2 Real Data Experiments .........

4 DIVERGENCE MEASURES FOR GROUPWISE POINT-SETS REGIS-
TRATION .................


4.1 Previous W ork . . . . . . .
4.2 Divergence M measures . . . . . .
4.2.1 Jensen-Shannon Divergence . . . .
4.2.2 CDF-JS Divergence . ..............
4.3 Methodology . . .......
4.3.1 Energy Function for Groupwise Point-sets Registration
4.3.2 JS Divergence in a Hypothesis Testing Framework ..
4.3.3 Unbiasness Property of the Divergence Measures .
4.3.4 Estimating JS and its Derivative . . . .
4.3.4.1 Finite mixture models . .......
4.3.4.2 Optimizing the JS divergence . .
4.3.5 Estimating CDF-JS and its Derivative . .....
4.3.5.1 Optimizing the CDF-JS divergence . .
4.4 Experiment Results . .......
4.4.1 JS Divergence Results . . . . .
4.4.1.1 Alignment results . . . .
4.4.1.2 Atlas construction results . ......
4.4.2 CDF-JS Divergence Results . . . .

5 APPLICATIONS TO IMAGE SEGMENTATION . . .

5.1 Related W ork ............
5.2 Registration+Segmentation Model .. ............
5.2.1 Gradient flow s . . . . . .
5.2.2 Algorithm Summary . . . . .
5.3 R results . . . . . . . .

6 CONCLUSIONS AND FUTURE WORK ......

6.1 Contributions of the Dissertation .......
6.2 Image and Point-sets Registration .. ............
6.2.1 Non-rigid Image Registration . . . .
6.2.2 Groupwise Point-sets Registration .. .........
6.3 Image Segmentation . . . . . .

REFERENCES ...............

BIOGRAPHICAL SKETCH .. ....................


. . 36
. . 38
. . 38
. . 40
. . 42
. . 43
. . 44
. . 45
. . 47
. . 47
. . 49
. . 50
. . 52
. . 53
. . 53
. . 53
. . 55
. . 56
















LIST OF TABLES
Table page

3-1 Comparison of the registration results between CCRE and MI for a fixed syn-
thetic deform ation field . . . . . . . 30

3-2 Comparison of total time taken to achieve registration by the CCRE with MI. 31

3-3 Comparison of the value S of several brain structures for CCRE and MI. ... 33

5-1 Statistics of the error in estimated non-rigid deformation. . 68
















LIST OF FIGURES
Figure page

1-1 Illustration of groupwise registration of corpus callosum point-sets manually
extracted from the outer contours of the brain images. . . . 4

3-1 CCRE, MI and NMI traces plotted for the misaligned MR & CT image pair 20

3-2 Comparison of convergence speed between CCRE and MI . ..... 27

3-3 Plot demonstrating the change of Mean Deformation Error for CCRE and MI
registration results with time . . . . . . 28

3-4 Results of application of our algorithm to synthetic data (see text for details). .28

3-5 Registration results of MR T1 and T2 image slice with large non-overlap. 30

3-6 Registration results of different subjects of MR & CT brain data with real non-
rigid motion. (see text for details . . . . . . 32

4-1 Illustration of corpus callosum point-sets represented as density functions. 35

4-2 Results of rigid registration in noiseless case. 'o' and '+' indicate the model
and scene points respectively . . . . . . 54

4-3 Non-rigid registration of the corpus callosum pointsets. ........... 54

4-4 Experiment results on seven 2D corpus collasum point sets. ............ ..55

4-5 Robustness to outliers in the presence of large noise ............. .. .57

4-6 Robustness test on 3D swan data .................. ...... .. 57

4-7 Atlas construction from four 3D hippocampal point sets. . . . 58

5-1 M odel Illustration . . . . . . . . 61

5-2 Illustration of the various terms in the evolution of the level set function . 65

5-3 Results of application of our algorithm to synthetic data . ..... 67

5-4 Results of application of our algorithm to a pair of slices from human brain
M RIs . . . . . . ....... ... . 69

5-5 Corpus Callosum segmentation on a pair of corresponding slices from distinct
subjects . . . . . . . . . 70










5-6 Hippocampal segmentation using our algorithm on a pair of brain scans from
distinct subjects . . . . . . . . 71















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

INFORMATION THEORETIC MEASURES AND THEIR APPLICATIONS TO
IMAGE REGISTRATION AND SEGMENTATION

By

Fei Wang

August 2006

Chair: Baba C. Vemuri
Major Department: Computer and Information Sciences and Engineering

Information theory has played a fundamental role in many fields of science and

engineering including computer vision and medical imaging. In this dissertation, various

information theoretic measures that are used in achieving the goal of solving several

important problems in medical imaging namely, image registration, point-set registration

and image segmentation are presented.

To measure the information content in a random variable, we first present a novel

measure based on its cumulative distribution that is dubbed Cumulative Residual Entropy

(CRE). This measure parallels the well-known Shannon entropy but has the following

advantages: (1) it is more general than the Shannon entropy as its definition is valid

in the discrete and continuous domains, (2) it possesses more general mathematical

properties and (3) it can be easily computed from sample data and these computations

asymptotically converge to the true values. Based on CRE, we define the cross-CRE

(CCRE) between two random variables, and apply it to solve the image alignment

problem for parameterized transformations. The key strengths of the CCRE over using

the now popular Mutual Information (based on Shannon's entropy) between images being










aligned are that the former has significantly larger tolerance to noise and a much larger

convergence range over the field of parameterized transformations.

Jensen-Shannon (JS) divergence has long been known as a measure of cohesion

between multiple probability densities. Similar to the idea of defining an entropy measure

based on distributions, we derived a JS divergence based on probability distributions and

dub it as the CDF-JS divergence. We then apply the JS and the CDF-JS divergence to

the groupwise point-set registration problem, which involves simultaneously registering

multiple shapes (represented as point-sets) for constructing an atlas. Estimating a

meaningful average or mean shape from a set of shapes represented by unlabeled point-

sets is a challenging problem, since this usually involves solving for point correspondence

under a non-rigid motion setting. The novel and robust algorithm we propose avoids the

correspondence problem by minimizing the CDF-JS/JS divergence between the point-sets

represented as probability distribution/density functions. The cost functions are fully

symmetric with no bias toward any of the given shapes to be registered and whose mean

is being sought. We empirically show that CDF-JS is more robust to noise and outliers

than JS divergence. Our algorithm can be especially useful for creating atlases of various

shapes present in images as well as for simultaneously registering 3D range data sets

without having to establish any correspondence.

In the context of image segmentation, we developed a novel model-based seg-

mentation technique that involves segmenting the novel 3D image data by non-rigidly

registering an atlas to it. The key contribution here to the solution of this problem is that

we present a novel variational formulation of the registration assisted image segmenta-

tion task, which leads to solving a coupled set of nonlinear PDEs that are solved using

efficient numerical schemes. Our segmentation algorithm is a departure from earlier

methods in that we have a unified variational principle wherein non-rigid registration and

segmentation are simultaneously achieved; unlike previous solutions to this problem, our

algorithm can accommodate image pairs with very distinct intensity distributions.















CHAPTER 1
INTRODUCTION

In 1948, motivated by the problem of efficiently transmitting information over

a noisy communication channel, Claude Shannon introduced a revolutionary new

probabilistic way of thinking about communication and simultaneously created the first

truly mathematical theory of entropy. His ideas created a sensation and were rapidly

developed to create the field of information theory, which employs probability and

ergodic theory to study the statistical characteristics of data and communication systems.

Since then, information theory has played a fundamental role in many fields of science

and engineering including computer vision and medical imaging. In this dissertation, we

endeavor to develop novel information theoretic methods with the application to medical

image analysis.

We examine two applications in particular, image (point-set) registration and

image segmentation. In the first of these applications, we follow a promising avenue of

work in using a probability density or distribution function as the signature of a given

"object" (image or point-set). Then by optimizing certain information theoretic measures

between these functions, we achieve the desired registration. In the segmentation

application, we consider an atlas based approach, in which segmentation and registration

are simultaneously achieved by solving a novel variational principle.

1.1 Image and Point-set Registration

We start with the image registration problem and then move on to the point-set

registration.

1.1.1 Image Registration

The image registration problem is defined as follows: Given a pair of images

Ii(x, y) and I2 (x', y'), where (x', y') = T(x, y)' and T is the matrix corresponding










to the unknown parameterized transformation to be determined, define a match metric

M(Ii(x, y), I2(x', y')) and optimize M over all T.

The fundamental characteristic of any image registration technique is the type of

spatial transformation or mapping used to properly overlay two images. The transforma-

tion can be classified into global and local transformations. A global transformation is

given by a single equation which maps the entire image. Local transformations map the

image differently depending on the spatial location and are thus more difficult to express

succinctly. The most common global transformations are rigid, affine and projective

transformations.

A transformation is called rigid if the distance between points in the image being

transformed is preserved. A rigid transformation can be expressed as


(1-1)
u(x, y) (cos(O)x sin(O)y + dx) x

v(x,y) = (sin(0)x + cos(0)y + dy) y

where u(x, y) and v(x, y) denote the displacement at point (x, y) along the X and Y

directions; 0 is the rotation angle, and (di, dy) the translation vector.

A transformation is called affine when any straight line in the first image is mapped

onto a straight line in the second image with parallelism being preserved. In 2D, the

affine transformation can be expressed as

u(x, y) = (alix a12y + dx) x
(1-2)
v(x, y) =(a21x a22y dy) y

where (2`1 '2 ) denotes an arbitrary real-valued matrix. Scaling transformation, which
has a transformation matrix of (3 ) and shearing transformation, which has a matrix

( 8 ) are two examples of affine transformation, where si,S2 and s3 are positive real
numbers.










A more interesting case, in general, is that of a planar surface in motion viewed

through a pinhole camera. This motion can be described as a 2D projective transforma-

tion of the plane.


F(r Y) ao+al+a2 x
a l (1-3)

aV' 6x+axy+l Y

where ao, ..., a7 are the global parameters.

When a global transformation does not adequately explain the relationship of a

pair of input images, a local transformation may be necessary. Registering an image

pair obtained at different times with some portion of the body experiencing growth,

or registering two images from different patients, fall into this local transformation

registration category. A motion field is usually used to describe the change in local

transformation problem.

1.1.2 Groupwise Point-sets Registration

Point-set representations of image data, e.g., feature points, are commonly used

in many applications and the problem of registering them frequently arises in a variety

of these application domains. Extensive studies on the point set registration and related

problems can be found in a rich literature covering both theoretical and practical issues

relating to computer vision and pattern recognition.

Given N point-sets, which are denoted by {XP,p E {1,..., N}}, each point-set XP

consists of points {xj E RD, i E {1,..., ip} } and np is the number of points contained

in point-set XP. The task of multiple point pattern matching or point-set registration is

either to establish a consistent point-to-point correspondence between these point-sets or

to recover the spatial transformation which yields the best alignment. For example, we are

given a group of corpus callosum point-sets from the brain image scan, which is shown

in the left column of Figure 1-1. All the point-sets are registered simultaneously to the

point-sets shown in the right column in a symmetric manner, meaning that the registration










result is not biased towards any of the original point-set. We will discuss these issues in

greater detail in Chapter 4.


~*** '*







i*~

***


*45*




I I


Figure 1-1: Illustration of groupwise registration of corpus callosum point-sets manually
extracted from the outer contours of the brain images.


1.2 Image Segmentation

Image segmentation plays a crucial role in many medical imaging applications by

automating or facilitating the delineation of anatomical structures. The segmentation of

structure from 2D and 3D images is an important first step in analyzing medical data. For

example, it is necessary to segment the brain in an MR image, before it can be rendered in

3D for visualization purposes. Segmentation can also be used to automatically detect the

head and abdomen of a fetus from an ultrasound image. The boundaries can then be used

to get quantitative estimates of organ sizes and provide aid in any necessary diagnoses.

Another important application is registration. It may be easier, or at least less error prone










to segment objects in multiple images prior to registration. This is especially true in

images from different modalities such as CT and MRI.

Image-guided surgery is another important application that needs image segmen-

tation. Recent advances in technology have made it possible to acquire images of the

patient while the surgery is in-progress. The goal is then to segment relevant regions of

interest and overlay them on an image of the patient to help guide the surgeon in his/her

work.

Segmentation is therefore a very important task in medical imaging. However,

manual segmentation is not only a tedious and time consuming process, but is also

inaccurate. Segmentation by experts has shown to be variable up to 20%. It is therefore

desirable to use algorithms that are accurate and require as little user interaction as

possible.

1.3 Outline of Remainder

In the next chapter, we rigorously define a novel measure of information in a random

variable based on its cumulative distribution that we dub as cumulative residual entropy

(CRE). We also connect the measure to the mean residual life function in reliability

engineering. Thereafter follows the chapter on using the measure for multimodal image

registration. In Chapter 4, we present a simultaneous groupwise point-sets registration

and atlas construction algorithm, in which we minimize the proposed divergence

measures between point sets represented as probability densities or distributions. Based

on these new measures, we propose a novel variational principle in Chapter 5 for

solving the registration assisted image segmentation problem. Lastly, we end with some

concluding points and thoughts for future work.















CHAPTER 2
ENTROPY AND RELATED MEASURES

2.1 Shannon Entropy and Related Measures

The concept of entropy is central to the field of information theory and was origi-

nally introduced by Shannon in his seminal paper [1] in the context of communication

theory. The entropy Shannon proposed is a measure of uncertainty in a discrete distri-

bution based on the Boltzman entropy of classical statistical mechanics. The Shannon

Entropy of a discrete distribution F is defined by


H(F) pilogpi, (2-1)


Since then, this concept and variants thereof have been extensively utilized in numerous

applications of science and engineering. To date, one of the most widely benefiting

application has been in financial analysis [2], data compression [3], statistics [4], and

information theory [5].

This measure of uncertainty has many important properties which agree with

our intuitive notion of randomness. We mention three: (1) It is always positive. (2) It

vanishes if and only if it is a certain event. (3) Entropy is increased by the addition of

an independent component, and decreased by conditioning. However, extension of this

notion to continuous distribution poses some challenges. A straightforward extension of

the discrete case to continuous distributions F with density f called differential entropy

reads

(F) f(x)log f(x)dx (2-2)

However, this definition raises the following concerns, 1) First of all, it is defined based

on the density of the random variable, which in general may or may not exist, e.g., for










cases when the cumulative distribution function (cdj) is not differentiable. It would not

be possible to define the entropy of a random variable for which the density function

is undefined; 2) Secondly, the Shannon entropy of a discrete distribution is always

positive, while the differential entropy of a continuous variable may take any value

on the extended real line; 3) Shannon entropy computed from samples of a random

variable lacks the property of convergence to the differential entropy, i.e. even when the

sample size goes to infinity, the Shannon entropy estimated from these samples will not

converge to differential entropy [5]. The consequence of which is that it is impossible,

in general, to approximate the differential entropy of a continuous variable using the

entropy of empirical distributions; 4) Consider the following situation: Suppose X and

Y are two discrete random variables representing the height of a group of people, with

X taking on values {5.1,5.2, 5.3, 5.4, 5.5}, each with a probability 1/5 and Y taking on

values {5.1, 5.2, 5.3, 5.4, 7.5} (with Yao Ming in this group) again each with probability

1/5. The information content measured in these two random variables using Shannon

entropy is the same, i.e., Shannon entropy does not bring out any differences between

these two cases. However, if the two random variables represented the winning chances

in a basketball game, the information content in the two random variables would be

considered as being dramatically different. Nevertheless Shannon entropy fails to make

any distinction whatsoever between them. For additional discussion on some of these

issues the reader is referred to [6].

In this work we propose an alternative measure of uncertainty in a random variable

X and call it the Cumulative Residual Entropy (CRE) of X. The main objective of our

study is to extend Shannon entropy to random variables with continuous distributions.

The concept we proposed overcomes the problems mentioned above, while retaining

many of the important properties of Shannon entropy. For instance, both are decreased

by conditioning, while increased by independent addition. They both obey the data










processing inequality, etc. However, the differential entropy does not have the following

important properties of CRE.

1. CRE has consistent definitions in both the continuous and discrete domains;

2. CRE is always non-negative;

3. CRE can be easily computed from sample data and these computations asymptoti-

cally converge to the true values.

The basic idea in our definition is to replace the density function with the cumulative

distribution in Shannon's definition 2-1. The distribution function is more regular than

the density function, because the density is computed as the derivative of the distribution.

Moreover, in practice what is of interest and/or measurable is the distribution function.

For example, if the random variable is the life span of a machine, then the event of

interest is not whether the life span equals t, but rather whether the life span exceeds

t. Our definition also preserves the well established principle that the logarithm of

the probability of an event should represent the information content in the event. The

discussions about the properties of CRE in the next few sections, we trust, are convincing

enough for further development of the concept of CRE.

2.2 Cumulative Residual Entropy: A New Measure of Information

In this section, we define an alternate measure of uncertainty in a random variable

and then derive some properties about this new measurement. We do not delve into the

proofs but refer the reader to a more comprehensive mathematical treatment in [7].

Definition: Let X be a random vector in RN and X = (X, X2, ..., XN), F(A) :

P(IXI > A) is the cumulative residual distribution, where A = (A1, ....AN) and IX| > A
means IXil > Aj. F(A) is also called survival function in the Reliability Engineering

literature. We define the cumulative residual entropy (CRE) ofX, by


S(X) F(A) logF(A)dA (2-3)


where R (x, R" ; > 0).









CRE can be related to the well-known concept of mean residual life function in

Reliability Engineering which is defined as:

mF(t) = E(X tIX > t) t F x (2-4)
F (t)

The mF (t) is of fundamental importance in Reliability Engineering & is often used to

measure departure from exponentiality. CRE can be shown to be the expectation of mF(t)

[8], i.e.
S(x) = E(mF(x)) (2-5)

Now we give a few examples.

Example 1: (CRE of the uniform distribution)

Consider a general uniform distribution with the density function:

1 0 p(x) (2-6)
0 o.w

Then its CRE is computed as follows

S(X) P(IX > x)logP(IX| > x)dx

(1 X) log( )dx
o a a
1
-a (2-7)
4

Example 2: (CRE of the exponential distribution)

The exponential distribution with mean 1/A has the density function:

p(x) = Xe-X (2-8)

Correspondingly, the CRE of the exponential distribution is

S(x) e-X log e-Adt

te dt
JoO









1
S(2-9)

Example 3: (CRE of the Gaussian Distribution)

The Gaussian probability density function is

1 (x m)2
p(x) exp[- 2 (2-10)

where m is the mean and o2 is the variance.

The cumulative distribution function is:

F(x) = erfc( (2-11)

where erfc is the error function:

erfc(x) -- exp(-t2/2)dt.

Then the CRE of the Gaussian distribution is:

S(x) = erfc( [- ) log [erfc(- m)]dx (2-12)

We'll now states important properties that are related to the application of CRE

to image registration. For a complete list of properties, we refer the readers to a more

comprehensive treatment in [7].

2.3 Properties of CRE

The traditional Shannon entropy of a sum of independent variables is larger than that

of either. We have analogously the following theorem:

Theorem 1 For any non-negative and independent variables X and Y,

max (S(X), S(Y)) < S(X + Y)


Proof: For a proof, see [7].









Similar to the case of Shannon's entropy, if X and Y are independent random

variables, S(X, Y) = E(IXI)(X) + E(IYI)S(Y). More generally,

Proposition 1 IfXi are independent, then


S(X) E(X) S(X)
i i~ j

For a proof, see [7].

Conditional entropy is a fundamental concept in information theory. We now define

the concept of conditioning in the context of CRE.

Definition: Given random vectors X and Y E RN, we define the conditional CRE

S(X|Y) by:

S(XIY) J P(IX| > xY) logP(|X| > x lY)dx (2-13)

As in the Shannon entropy case, conditioning reduces CRE.

Proposition 2 For any X and Y

E[S(XIY)] < S(X) (2-14)

Equality holds iff X is independent of Y.

2.3.1 CRE and Empirical CRE

Next theorem shows one of the salient feature of CRE. In the discrete case Shannon

entropy is always non-negative, and equals zero if and only if the random variable is a

certain event. However, this is not valid for the Shannon entropy in the continuous case

as defined in Eqn. 2-2. In contrast, in this regard CRE does not differentiate between

discrete and continuous cases, as shown by the following theorem:

Theorem 2 (X) > 0 and equality holds if and only if P[XI = A] = 1 for some vector

A, ie. IXi = Ai i il i probability 1.










Shannon entropy computed from samples of a random variable lacks the property

of convergence to the differential entropy (see Eqn. 2-2 for a definition). In contrast, the

CRE, S(x) computed from the samples converges to the continuous counterpart. This is

summarized in the following theorem.


Proposition 3 (Weak Convergence). Let the random vectors Xk converge in distribution

to the random vector X; by this we mean


lim E[(Xk)] = E[p(X)] (2-15)
k->oo

for all bounded continuous functions 0 on RN, if all the Xk are bounded in LP for some

p > N, then

lim S(Xk) = S(X) (2-16)
k->oo

Proof: Refer to [7] for the proof.

This is a powerful property and as a consequence of it, we can compute CRE of an

random variable from the samples which would converge to the true CRE of the random

variable. Note that Xk can be samples of a continuous random variable.

2.3.2 Robustness of CRE

We now investigate the robustness (or the lack thereof) of differential entropy and

prove that while differential entropy is not robust with respect to small perturbations,

CRE on the contrary is quite robust. This property plays a key role in demonstrating the

noise immunity of CCRE over MI depicted in the experiments in the next Chapter.


Theorem 3 Let X be a discrete R. V, taking value (x, x2, ..., XN), 1' i/h probabilities

Pl,P2,...,PN

p(X = x) = p 1 < i < N (2-17)

X has .V\Mlnn,, entropy: H(X) = pi log p. Let Y, have density f, and be

independent of X. Z, = X + Y, is no longer discrete, and has a density. Let X be as in










(2-17) and Y, as above. Suppose Y, -- 0 in probability. Then


h(X + Y,) -oo (2-18)




Theorem 4 For X and Y, as defined in Theorem 3,


lim S(X + Y) S(X) (2-19)
Yn-O




Proof: This is a direct consequence of the Proposition 3.

Theorems (3) and (4) are very important properties as they prove that the CRE is

robust to noise which is not the case for differential entropy. Intuitively, the robustness

of CRE maybe attributed to the use of CDF as opposed to a PDF in its definition, i.e., an

integral formulation as opposed to a differential formulation and it is well known that the

former is more robust compared to the later.















CHAPTER 3
APPLICATIONS TO MULTIMODALITY IMAGE REGISTRATION

Matching two or more images under varying conditions illumination, pose,

acquisition parameters etc. is ubiquitous in Computer Vision, medical imaging,

geographical information systems etc. In the past several years, information theoretic

measures have been very widely used in defining cost functions to be optimized in

achieving a match. An example problem common to all the aforementioned areas is the

image registration problem. In the following, we will review the literature on existing

computational algorithms that have been reported for achieving multimodality image

registration, with the focus on the non-rigid registration methods. We will point out their

limitations and hence motivate the need for a new and efficient computational algorithm

for achieving our goal.

3.1 Related Work

Non-rigid image registration methods in literature to date may be classified into

feature-based and "direct" methods. Most feature-based methods are limited to determin-

ing the registration at the feature locations and require an interpolation at other locations.

If however, the transformation/registration between the images is a global transformation

e.g., rigid, affine etc. then, there is no need for an interpolation step. In the non-rigid case

however, interpolation is required. Also, the accuracy of the registration is dependent on

the accuracy of the feature detector.

Several feature-based methods involve detecting surfaces landmarks [9, 10, 11, 12],

edges, ridges, etc. Most of these assume a known correspondence with the exception

of the work in Chui et al.[9], Jian and Vemuri [13], Wang et al.[14] and Guo et al. [15].

Work reported in Irani and Anandan [16] uses the energy (squared magnitude) in the

directional derivative image as a representation scheme for matching achieved using the










SSD cost function. Recently, Liu et al. [17] reported the use of local frequency in a robust

statistical framework using the integral squared error a.k.a., L2E. The primary advantage

of L2E over other robust estimators in literature is that there are no tuning parameters in

it. The idea of using local phase was also exploited by Mellor and Brady [18], who used

mutual information (MI) to match local-phase representation of images and estimated

the non-rigid registration between them. However, robustness to significant non-overlap

in the field of view (FOV) of the scanners was not addressed. For more on feature-based

methods, we refer the reader to the recent survey by Zitova and Flusser [19].

In the context of "direct" methods, the primary matching techniques for intra-

modality registration involve the use of normalized cross-correlation, modified SSD,

and (normalized) mutual information (MI). Ruiz-Alzola et al.[20] presented a unified

framework for non-rigid registration of scalar, vector and tensor data based on template

matching. For scalar images, the cost function is the extension of modified SSD using a

different definition of inner products. However this model can only be used on images

from the same modality as it assumes similar intensity values between images. In [21,

22], a level-set based image registration algorithm was introduced that was designed to

non-rigidly register two 3D volumes from the same modality of imaging. This algorithm

was computationally efficient and was used to achieve atlas-based segmentation. Direct

methods based on the optical-flow estimation form a large class for solving the non-rigid

registration problem. Hellier et al.[23] proposed a registration method based on a dense

robust 3-D estimation of the optical flow with a piecewise parametric description of

the deformation field. Their algorithm is unsuitable for multi-modal image registration

due to the brightness constancy assumption. Variants of optical flow-based registration

that accommodate for varying illumination maybe used for inter-modality registration

and we refer the reader to [24, 25] for such methods. Guimond et al., [26] reported a

multi-modal brain warping technique that uses Thirion's Demons algorithm [27] with

an adaptive intensity correction. The technique however was not tested for robustness










with respect to significant non-overlap in the FOVs. More recently, Cuzol et al. [28]

introduced a new non-rigid image registration technique which basically involves a

Helmoholtz decomposition of the flow field which is then embedded into the brightness

constancy model of optical flow. The Helmholtz decomposition allows one to compute

large displacements when the data contains such displacements. This technique is

an innovation on accommodating for large displacements and not one that allows for

intermodality non-rigid registration. For more on intra-modality methods, we refer the

reader to the comprehensive surveys [29, 19].

A popular framework for "direct" methods is based on the information theoretic

measures [30], among them, mutual information (MI) pioneered by Viola and Wells [31]

and Collignon et al., [32] and modified in Studholme et al., [33] has been effective in

the application of image registration. Reported registration experiments in these works

are quite impressive for the case of rigid motion. The problem of being able to handle

non-rigid deformations in the MI framework is a very active area of research and some

recent papers reporting results on this problem are [18, 34, 35, 36, 37, 38, 39, 40, 41,

42]. In [34], Mattes et al., and in [35], Rueckert et al., presented mutual information

based schemes for matching multi-modal image pairs using B-Splines to represent the

deformation field on a regular grid. Guetter [43] recently incorporated a learned joint

intensity distribution into the mutual information formulation, in which the registration

is achieved by simultaneously minimizing the KL divergence between the observed

and learned intensity distributions and maximizing the mutual information between

the reference and alignment images. Recently, D'Agostino et al., [44] presented an

information theoretic approach wherein tissue class probabilities of each image being

registered are used to match over the space of transformations using a divergence measure

between the ideal case (where tissue class labels between images at corresponding voxels

are similar) and actual joint class distributions of both images. This work expects a

segmentation of either one of the images being registered. Computational efficiency and










accuracy (in the event of significant non-overlaps) are issues of concern in most if not all

the MI-based non-rigid registration methods.

Finally, some registration methods under the direct approach are inspired by

models from mechanics, either from elasticity [45, 46], or fluid mechanics [47, 48].

Fluid mechanics-based models accommodate for large deformations, but are largely

computationally expensive. Christensen [49] recently developed an interesting version

of these methods, where the direct deformation field and the inverse deformation field

are jointly estimated to guarantee the symmetry of the deformation with respect to

permutation of input images. A more general and mathematically rigorous treatment

of the non-rigid registration which subsumes the fluid-flow methods was presented in

Trouve [50]. All these methods however are primarily applicable to intra-modality and

not inter-modality registration.

3.2 Multimodal Image Registration using CCRE

Based on CRE, cross-CRE (CCRE) between two random variables was defined,

and applied to solve the image alignment problem, which is defined as: Given a pair

of images Il(x) and I2(x'), where (x') = T(x)t and T is the matrix corresponding

to the unknown parameterized transformation to be determined, define a match metric

MA(I1 (x), 12(x')) and maximize/minimize M over all T. The class of transformations can

be rigid, affine, projective or non-rigid transformations. Several matching criteria have

been proposed in the past, some of which were reviewed earlier. Amongst them, mutual

information is very popular and is defined as follows for the continuous random variable

case,


MI(X, Y) = h(X) + h(Y) h(X, Y) (3-1)


where h(X) is the differential entropy of the random variable X and is given by h(x) =

f p(x)lnp(x)dx, where p(x) is the probability density function and can be estimated
from the image data using any of the parametric and nonparametric methods. The reason










for defining MI in terms of differential entropy as opposed to Shannon entropy is to

facilitate the optimization of MI with respect to the registration parameters using any

of the gradient based optimization methods. Note that MI defined using the Shannon's

entropy in discrete form will not converge to continuous case defined here due to the fact

that Shannon's entropy does not converge to the differential entropy (see [5]).

We now define the cross-CRE (CCRE) using CRE defined in Eqn. 2-3.

C(X, Y) = S(X) E[S(X/Y)], (3-2)

We will use this quantity as a matching criterion in the image alignment problem.

More specifically, let Ir(x) be a test image we want to register to a reference image

IR(x). The transformation g(x; tt) describes the deformation from VT to VR, where
VT and VR are continuous domains on which IT and Ip are defined, it is the set of the

transformation parameters to be determined. We pose the task of image registration as

an optimization problem. To align the reference image IR(x) with the transformed test

image Ir(g(x; tt)), we seek the set of the transformation parameters tt that maximizes

C(IT, IR) over the space of smooth transformations i.e.,

p= argmax C (IT g(x; lt),In) (3-3)

The computation of CCRE requires estimates of the marginal and joint probability

distributions of the intensity values of the reference and test images. We denote p(l, k; tt)

as the joint probability of (IT o g(x; 1t), IR). Let pr(1; 1t) and pR(k) represent the

marginal probability for the test image and reference images respectively, LT and LR

are the discrete sets of intensities associated with the test image and reference image

respectively. Then, we can rewrite the CCRE (IT o g(x; p), IR) as follows:


C(I, o g(x; p), IR) = E(IT) E[S(IT 0 g(x; p)/IR)]

E 1rP(l; )dl log [ pT(l; )dl
AELT










+ J p(k) p(l, k; ) dl log [ p(l, k; ) dl] (3-4)
kELR AELT A PR() pR( k)

Let P(i > A; p) pTr(l; p)dl and P(i > A, k; p) f p(l, k; t)dl. Using the fact

that pT(l; A) kELR p(l, k; tt), we have P(i > A; p) = kELR P(i > A, k; p). Eqn.
(3-4) can be further simplified, which leads to,


C (T o g (x; A), IR)

P(i > A; ) logP(i > A; t) + P(i > A, k;) log P(i > A, k; )
AELT kELR AELT P(k)
P(i > A,k; ) logP(i > A; ) (3-5)
AELT kELR

+ P(>A,k; p) log P(i > A, k; )
PR(k)
kELR AELT

C E P(>AXk ;P) [1 P(i > A, k; logP( >A; )]
ALT AcLp pR (k)
E P(i > A, k; p) log P(i > k; ) (3-6)
AELT kELR

To illustrate the difference between CCRE and the now popular information theoretic

cost functions such as MI & NMI, we choose to plot these functions against a parameter

of the transformation, for illustrative purposes, say the rotations. The image pair we
used here is MR & CT images that were originally aligned, and the MR and CT data

intensities range from 0-255 with the mean 55.6 and 60.6 respectively. The cost functions

are computed over the rotation angle that was applied to the CT image to misalign it with
respect to the MR image. In each plot of the Figure 3-1 the X-axis shows the 3D rotation

angle about Z axis, while the Y-axis shows the values of CCRE, MI and NMI computed
from the misaligned (by a rotation) image pairs. The second row shows a zoom-in view

of the plots over a smaller region, so as to get a detailed view of the cost function. The

following observations are made from this plot:













CCRE MI Normalized MI
18
05
16 k
16 0 45

14 04 11
0 35
12 03

10 025 1 05
-40 -20 0 20 40 -40 -20 0 20 40 -40 -20 0 20 40

CCRE MI Normalized MI
01 04338 11117

305 0 4336
1 1117
0 4334
16
195 04332 1 1116
95 0 433
1 1115
,99 0 4328
985 04326 11115
-04 -02 0 02 04 -04 -02 0 02 04 -04 -02 0 02 04
CCRE MI Normalized MI
1598 04056 11029

1597 4054

1596 1 1028
0 405

0 4048

15 94 04046 11027
-04 -02 0 02 -04 -02 0 02 04 -04 -02 0 02 04


Figure 3-1: CCRE, MI and NMI traces plotted for the misaligned MR & CT image pair
where misalignment is generated by a rotation of the CT image. First row: over the range
-400 to 400. Second row: zoom in view between -0.50 to 0.50, where the arrows in the
first row signify the position. Note that all three cost function are implemented with tri-
linear interpolation. Third row: Three cost functions implemented with partial volume
interpolation [32].



1. Similar to MI and NMI, the maximum of CCRE occurs at 0 of rotation, which


confirms that our new information measure needs to be maximized in order to find


optimum transformation between two misaligned images.


2. The CCRE shows much larger range of values than MI & NMI. This feature plays


an important role in the numerical optimization since it leads to a more stable


numerical implementation by avoiding cancelation, round off etc. that often plague


arithmetic operations with smaller numerical values.










3. Upon closer inspection, we observe that CCRE is much smoother than the MI and

NMI for the MR& CT data pair, therefore verifies that CCRE is more regular than

other information theoretic measures.

3.2.1 Transformation Model for Non-rigid Motion

We model the non-rigid deformation field between two 3D image pairs using a

cubic B-splines basis in 3D. B-splines have a number of desirable properties for use

in modeling the deformation field. (1) Splines provide inherent control of smoothness

(degree of continuity). (2) B-splines are separable in multiple dimensions which provides

computational efficiency. Another feature of B-splines that is useful in a non-rigid

registration system is the "local control". Changing the location of a single control point

modifies only a local neighborhood of the control point.

The basic idea of the cubic B-spline deformation is to deform an object by manipu-

lating an underlying mesh of control points 7y. The deformation g is defined by a sparse

regular control point grid. In 3D case, the deformation at any point x = [x, y, z]T in the

test image can be interpolated with a linear combination of cubic B-spline convolution

kernel.
g(x) 6jo(3) ) (3-7)
g6 (3-7)


where /(3)(x) = (3) (x)3(3) (y) (3)(z) and Ap is spacing of the control grid. 6j is the

expansion B-spline coefficients computed from the sample values of the image. For the

implementation details, we refer the reader to Forsey[51] and Mattes [34].

3.2.2 Measure Optimization

Calculation of the gradient of the energy function is necessary for its efficient and

robust maximization. The gradient of CCRE is given as,

VC ac (3 8)
VC = 0[.. ] (3-8)









Each component of the gradient can be found by differentiating Eqn. (3-4) with respect
to a transformation parameters. We consider the two terms in Eqn. (3-4) separately when
computing the derivative. For the first term in Eqn. (3-4), we have,


(IT) 0 [ pr(l; t)dl x log pr(1; t)dl
Ott 0A A


-
AELT

where P(i > A; p) = jfpr(;

aP(i


AtELT
log (P(i > A; p)) + x P A )


i)dl, and

> A; p) pT(1, ) dl
A IJ Op


(3-9)




(3-10)


The derivative of the second term is given by,

OE[S(I o g(x; tp)/lp)]


aR(k) p(l x log (j p(l k;) l)
>: PR AELT(k) pR(k)
kELR ACLT
S5 (log P(i > A, k; ) + aP(i > A, k; )
= : 2^(log + 1)
kELR AELT p (k) Ott

where P(i > A, k; pt) = fA p(1, k; pt)dl, and


OP(i > A, k; tt)
OLtt


Combining the derivatives of the two terms together, and using the fact that

Opr(l; A) aEkCLR p(l, k; p)
Ott Ott


(3-11)


(3-12)


(3-13)


"0p(, k; 9p )lk
I dl









we have the analytic gradient of CCRE,

C (IT o g(x; l), IR)

S[log P> A + kELR P(i > A, k; )
[log P(i > A; p) + 1] O P -
AELT
[logP(i > A, k; ) aP(i > A, k;)
kELR AELT


S [logP(i > A;. ) +]OP( > A,k;
AELT kELR
+ [log P(i> A, k; ) 8P(i > A,k; i)
pR(k) Ott
AELT kcLR
^ ~l^ P(i> A,k; l) 9P(i
= [o^g f log P(i > A; p)]x


aELT kL9k
note that in the derivation, we use the fact that P(i > A; p) = CkELR
Comparing the expressions for CCRE and derivative of CCRE


(3-14)


> A, k; )





P(i > A, k; ).


{C(I o g(x; ),IR) ELT kE L log pP(i>;) x P (i> A, k; t) ( )
ac (x;),I P(i>A,k;p) P(i>A,k;p)
SZAELT ZE kfL ELRo pR(k)P(i>A;p) )

we note that the two formulas in (3-15) are similar to each other and they share the

common term log p(ki>Ap) From a computational viewpoint, this is quite beneficial

since the common term can not only save memory space, but also make the calculation
of gradient more efficient. From the formulation, we can also see that calculation of

CCRE and derivative of CCRE require us to find a method to estimate P(i > A, k; p) and
aP(i ,k;) We will address the computation of these terms in the next subsection.

3.2.3 Computation of P(i > A, k; p) and 'P(i>i,k;p)
ap
We will use the parzen window technique to estimate the cumulative distribution

function and its derivative. The calculation of P(i > A, k; p) requires estimate of the

cumulative probability distributions of the intensity values of the reference and test









images. Let 3(0) be a zero-order spline Parzen window (centered unit pulse) and p(3) be a

cubic spline Parzen window, the smoothed joint probability of (IR, IT o g) is given by


p(1, k; p) a k() (k I- (x) (3) IT (g (X f) (3-16)
XEV

where a is a normalization factor that ensures pp(l, k) = 1, and IR(x) and Ir(g(x; t)

are samples of the reference and interpolated test images respectively, which is normal-

ized by the minimum intensity value, f fo, and the intensity range of each bin, AbR,

AbT.

Since P(i > A, k; tt) = fA p(l, k; pt)dl, we have the following,

P(i > A,k; ) p(l,k; t)dl

) O IR ( X ) --
ay~i( IR(x) f Igx 0)) IT (g(Xo
0 ( AbR A AbT '

aY3(o)(k IR(x)- f ( ( iT3(9 (x fIT) (3-17)
AbR AbT
zEV

where 4 () is the cumulative residual function of cubic spline kernel defined as

follows,

4) j) / (3) (u)

1.0 v < -2

1.0 (+2)4 -2 < v < -1
24
1 2 v - 1 2 3 0< <1
2 3 3 8
(v-2)4 < v < 2
1 24
0 v>2










Note that d(u= -(3) (u), we can then take the derivative of Eqn. 3-17 with respect to

pi, and we get

P(i > A, k; p) /a (o) Ix) fR (g (x; t)) f
t bT AbR Ab )
xzEV
SlT(t) A))Og(x; )
a t t O(x;

AbT^ vbbR AbI
xzV
X (OIT(t) 8g(x; A) (3-19)
x ((3-19)
\at t=g(x; Op)/ ) d

where a') is the image gradient.

3.2.4 Algorithm Summary

The registration algorithm can be summarized as follows,

1 For the current deformation field, interpolate the test image by IT o g(x; pA).

Calculate P(i > A, k; p) and 'P(i>,k;,) using Eqn. (3-17) and Eqn. (3-19)

respectively.

2 Compute P(i > A; p) as ZkELR P(i > A, k; p;), which is used to calculate the

common term in both CCRE and gradient of CCRE, i.e., log P(i>Ak )
PR(k) P(i>A;)
3 Compute the energy function and its gradient using the formulas given in Eqn.

(3-15), we can then use the Quasi-Newton method to numerically solve the

optimization problem.

4 Update the deformation field g(x; pt). Stop the registration process if the differ-

ence in consecutive iterates is less than c = 0.01, a pre-chosen tolerance, otherwise

go to Step 1.

3.3 Implementation Results

In this section, we present the results of applying our non-rigid registration algorithm

to several data sets. The results are presented for synthetic as well as real data. The first

set of experiment was done with synthetic motion. We show the advantage of using the

CCRE measure in comparison to other information theoretic registration methods. We










show that CCRE is not only more robust, but also converges faster than others. We begin

by applying CCRE to register image pairs for which the ground truth was available.

3.3.1 Synthetic Motion Experiments

In this section, we demonstrate the robustness property of CCRE and will make

a case for its use over Mutual Information in the alignment problem. The case will be

made via experiments depicting faster convergence speed and superior performance

under noisy inputs in matching the image pairs misaligned by a synthetic non-rigid

motion, Additionally we will depict a larger capture range over MI-based methods in the

estimation of the motion parameters.

The data we use for this experiment are corresponding slices from an MR T1 and

T2 image pair, which is from the brainweb site at the Montreal Neurological Institute

[52]. They are originally aligned with each other. The two images are defined on a

1mm isotropic voxel grid in the Talairach space, with dimension (256 x 256). We then

apply a known non-rigid transformation to the T2 image and the goal is to recover this

deformation by applying our registration method. The mutual information scheme which

we are going to compare with is originally reported in literature [34] [53], in which the

explicit gradient forms are presented and thus allowing for the application of gradient

based optimization methods.

3.3.1.1 Convergence speed

In order to compare the convergence speed of CCRE versus MI, we design the

experiment as follows: with the MR T1 & T2 image pair as our data, we choose the MR

Tl image as the source, the target image was obtained by applying a known smooth non-

rigid transformation that was procedurally generated. Notice the significant difference

between the intensity profiles of the source and target images. For comparison purposes,

we use the same gradient descent optimization scheme, and let the two registration

methods run for the same amount of time, and show the registration result visually and

quantitatively.










Source Image Target Image










Transformed Source using CCRE Transformed Source using MI









Figure 3-2: Upper left, MR T1 image as source image; Upper right, deformed MR T2
image as target image; Lower left and right, results of estimated transformations using
CCRE and MI applied to the source respectively. Both algorithms run for 30 seconds
using the same gradient descent technique.


The source and target image pair along with the results of estimated transformation

using CCRE and MI applied to the source are shown in Figure 3-2. As evident visually,

we observe that the result generated by CCRE is more similar in shape with the target

image than the one produced by MI.

Quantitative assessment of accuracy of the registration is presented subsequently

in Figure 3-3, where we plotted the change of mean deformation error (MDE) obtained

for the CCRE-based algorithm and the MI-based algorithm respectively. MDE is defined

as dm = car R I 9go (xi) g(x) 11, where go(xi) and g(xi) are the ground truth

and estimated displacements respectively at voxel xi. |. | denotes the Euclidean norm,

and R is the volume of the region of interest. In both cases mean deformation error

are decreasing with time, but the solid line is decreasing faster than the dotted line.

For example, it takes about 5 minutes for MI to reach the error level inside 1.2 while

CCRE only requires about half of the time as that required by MI to get to the level. This














4 MI Results
CCRE Results
S35




05
25-

2-
15- -- G




0 1 2 3 4 5 6 7 8
Time (minutes)


Figure 3-3: Plot demonstrating the change of Mean Deformation Error for CCRE and
MI registration results with time. Solid line shows the MDE for CCRE registration result,
while dotted line illustrates the MDE for MI result.


empirically validates the faster convergence speed of CCRE based algorithm over the

MI-based algorithm.

3.3.1.2 Registration accuracy

Using the same experiment setting as in the previous experiment, we present the

registration error for our algorithm in the estimated non-rigid deformation field as an

indicator of the accuracy of estimated deformations. Figure 3.3.1.2 depicts the results












800

600

400

200

0
0 2 4 6


Figure 3-4: Results of application of our algorithm to synthetic data (see text for details).



obtained for this image pair. which is organized as follows, from left to right: the first










row depicts the source image with the target image segmentation superposed to depict

the amount of mis-alignment, the registered source image which is obtained using

our algorithm superposed with the target segmentation, followed by the target image;

second row depicts ground truth deformation field which we used to generate the target

image from the MR T2 image, the estimated non-rigid deformation field followed by

histogram of the estimated magnitude error. Note that the error distribution is mostly

concentrated in the small error range indicating the accuracy of our method. As a measure

of accuracy of our method, we also estimated the average, p, and the standard deviation,

o, of the error in the estimated non-rigid deformation field. The error was estimated as the

angle between the ground truth and estimated displacement vectors. The average and

standard deviation are 1.5139 and 4.3211 (in degrees) respectively, which is quite

accurate.

3.3.1.3 Noise immunity

In the next experiment, we compare the robustness of the two methods (CCRE,

MI) in the presence of noise. Still selecting the MR T image slice from the previous

experiment as our source image, we generate the target image by applying a fixed smooth

synthetic deformation field. We conduct this experiment by varying the amount of

Gaussian noise added and then for each instance of the added noise, we register the

two images using the two techniques. We expect both schemes are going to fail at some

level of noise. ("failed" here means that the optimization algorithm primarily diverged).

By comparing the noise magnitude of the failure point, we can show the degree to

which these methods are tolerant. The numerical schemes we used to implement these

registrations are all based on BFGS quasi-Newton algorithm.

The mean magnitude of the synthetic motion is 4.37 pixel, with the standard

deviation at 1.8852. Table3-1 show the registration results for the two schemes. From the

table, we observe that the MI fails when the standard deviation of the noise is increased

to 40, while CCRE is tolerant until 66, a significant difference when compared to the MI.










Table 3-1: Comparison of the registration results between CCRE and MI for a fixed
synthetic deformation field.

CCRE MI
a MDE Standard Deviation MDE Standard Deviation
10 1.0816 0.9345 1.3884 1.4538
19 1.1381 1.1702 1.4871 1.5052
30 1.1975 1.3484 1.5204 1.5615
40 1.3373 1.6609 FAIL
60 1.3791 1.9072
66 FAIL


This experiment conclusively depicts that CCRE has more noise immunity than MI when

dealing with the non-rigid motion.

3.3.1.4 Partial overlap

Figure 3-5 depicts an example of registration of the MR T and T2 data sets with

large nonoverlap. The left image of the figure depicts the MR T1 brain scan as the source

image, and the right image shows the MR T2 data as the target. Note that the FOV for the

data sets are significantly nonoverlapping. The nonoverlap was simulated by cutting 66%

of the MR T1 image (Source image). The middle column depicts the transformed source

image along with an edge map of the target (Deformed MR T2 image) superimposed on

the transformed source. As is evident, the registration is visually quite accurate.











Figure 3-5: Registration results of MR T1 and T2 image slice with large non-overlap.
(left) MR T1 source image before registration; (right) Deformed T2 target image; (mid-
dle) the transformed MR image superimposed with edge map from target image.










3.3.2 Real Data Experiments

In this section, we present the performance of our method on a series of CT &

MR data containing real non-rigid misalignments. For the purpose of comparison, we

also apply traditional MI implemented as was presented in Mattes et al. [34] to these

same data sets. The CT image is of size (512, 512, 120) while the MR image size is

(512, 512, 142), and the voxel dimensions are (0.46, 0.46, 1.5)mm and (0.68, 0.68, 1.05)

for CT and MR respectively. The registration was performed on reduced volumes

(210 x 210 x 120) with the control knots placed every 16 x 16 x 16 voxels. The

program was written in the C++ programming language, and all experiments were run on

a 2.6GHZ Pentium PC.

Table 3-2: Comparison of total time taken to achieve registration by the CCRE with MI.

1 2 3 4 5 6 7 8
CCRE Time (s) 4827 3452 4345 4038 3910 4510 5470 3721
MITime(s) 9235 6344 10122 17812 12157 11782 13157 10057


We have used a set of eight volumes of CT data sets and the task was to register

these eight volumes to the MR data chosen as the target image for all registrations,

by using both CCRE and MI algorithms. Note that all CT & MR volumes are from

different subjects and thus contains real non-rigid motion. The parameters used with both

algorithms were identical. For both algorithms, the optimization of the cost functions was

halted when improvements of at least 0.0001 in the cost function could not be detected.

The time required for registering all data sets for our algorithm as well as MI method

are given in Tables 3-2. This table shows that, on average, our CCRE algorithm is about

2.5 times faster than the traditional MI approach for this set of experiments. For brevity,

we only show one registration result in Figure 3-6. Here, one slice of the volume is

shown on first row with the source CT image at left and reference image at right. The

middle image show the transformed CT image slice superimposed with edge map from

target image. On the second row, the source image superimposed with edge map from










target image is shown on the left, while shown in the middle and right are the surfaces

reconstructed from the transformed source using CCRE method and the target MR

image respectively. From this figure, we can see that the source and target image depict

considerable non-rigid changes in shape, nevertheless our method was able to register

these two images quite accurately. To validate the conformity of the two reconstructed

surfaces, we randomly sample 30 points from the surface of the transformed source

using CCRE, and then estimate the distances of these points to the surface of the target

MR volume. The average of these distances is about 0.47mm, which indicates a very

good agreement between two surfaces. The resemblance of the reconstructed shapes

from transformed source with the target indicates that our CCRE algorithm succeeded in

matching the source CT volume to the target MR image.






















Figure 3-6: Registration results of different subjects of MR & CT brain data with real
non-rigid motion. (see text for details.


The accuracy of the information theoretic based algorithm for non-rigid registration

problems was assessed quantitatively by means of an region-based segmentation task

[54]. ROIs (whole brain, eyes) were segmented automatically in these eight CT data sets










used as the source image and binary masks were created. The deformation fields between

the CT and MR volume were computed and used to project the masks from each of the

CT to the MR volume. Contours were manually drawn on a few slices chosen at random

in MR volume (four slices/volume). Manual contours on MR and contours obtained

automatically were then compared using an accepted similarity index defined as two

times the number of pixels in the intersection of the contours divided by the sum of the

number of pixels within each contour [41]. This index varies between zero (complete

disagreement) and one (complete agreement) and is sensitive to both displacement and

differences in size and shape. Table 3-3 lists mean values for the similarity index for

each structure. It is customarily accepted that a value of the similarity index above 0.80

indicates a very good agreement between contours. Our results are well above this value.

For comparison purpose, we also computed the same index for the MI method. We can

conclude from the table that our CCRE can achieve better registration accuracy than the

MI for the task of non-rigid registration of real multi-model images.

Table 3-3: Comparison of the value S of several brain structures for CCRE and MI.

Volume 1 2 3 4 5 6 7 8
Whole Brain 0.987 0.996 0.974 0.962 0.975 0.967 0.988 0.981
CCRE Left Eye 0.925 0.935 0.925 0.907 0.875 0.890 0.834 0.871
Right Eye 0.840 0.940 0.891 0.872 0.851 0.829 0.910 0.921
Whole Brain 0.986 0.981 0.976 0.96 0.950 0.961 0.942 0.952
MI Left Eye 0.911 0.893 0.904 0.791 0.853 0.810 0.851 0.853
Right Eye 0.854 0.917 0.889 0.814 0.849 0.844 0.897 0.854















CHAPTER 4
DIVERGENCE MEASURES FOR GROUPWISE POINT-SETS REGISTRATION

Matching point patterns is ubiquitous in many fields of Engineering and Science e.g.,

medical imaging, sports science, archaeology, and others. Point sets are widely used in

computer vision to represent boundary points of shapes contained in images or any other

salient features of objects contained in images. Given two or more images represented

using the salient features contained therein, most often than not, one is interested in

matching these (feature) point patterns to determine a linear or a nonlinear transformation

between the coordinates of the feature point sets. Such transformations capture the

changes in the pattern geometry characterized by the given feature point set.

The primary technical challenge in using point-set representations of shapes is

the correspondence problem. Typically correspondences can be estimated once the

point-sets are properly aligned with appropriate spatial transformations. If the objects

at hand are deformable, the adequate transformation would obviously be a non-rigid

spatial mapping. Solving for non-rigid deformations between point-sets with unknown

correspondence is a hard problem. In fact, many current methods only attempt to solve

for affine transformation for the alignment [55]. Furthermore, we also encounter the issue

of the bias problem in groupwise point-sets registration. If one arbitrarily chooses any one

of the given data sets as a reference, the estimated registration transformation would be

biased toward this chosen reference and it would be desirable to avoid such a bias. The

question that arises is: How do we align all the point-sets in a symmetric manner so that

there is no bias toward any particular point-set?

To overcome these aforementioned problems, we present a novel approach to

simultaneously register multiple point-sets and construct the atlas. The idea is to model

each point set by a kernel probability density or distribution, then quantify the distance










between these probability densities or distributions using information-theoretic measures.

Figure 4-1 illustrate this idea, where the right column of the figure is the density function

corresponding to the corpus callosum point-sets shown in the left. The distance is













*











Figure 4-1: Illustration of corpus callosum point-sets represented as density functions.


optimized over a space of coordinate transformations yielding the desired registrations.

It is obvious that once all the point sets are deformed into the same shape, the distance

measure between these distributions should be minimized since all the distribution are

identical to each other. We impose regularization on each deformation field to prevent

over-deforming of each point-sets (e.g. all the point-sets may deform into a single data

point).

The rest of the chapter is organized as follows: we begin by reviewing all the related

literatures, which is followed by a description of the divergence measures we used for

quantify the distance between densities or distributions. We then present the details of










our energy function, and the empirical way of estimating the cost functions and their

derivatives. Finally we will show the experimental results at the end of this chapter.

4.1 Previous Work

Extensive studies on the atlas construction for deformable shapes can be found in

literature covering both theoretical and practical issues relating to computer vision and

pattern recognition. According to the shape representation, they can be classified into

two distinct categories. One is the methods dealing with shapes represented by feature

point-sets, and everything else is in the other category including those shapes represented

as curves and surfaces of the shape boundary, and these curves and surfaces may be

either intrinsically or extrinsically parameterized (e.g. using point locations and spline

coefficients).

The work presented in [56] is a representative method using an intrinsic curve

parameterization to analyze deformable shapes. Shapes are represented as elements of

infinite-dimensional spaces and their pairwise difference are quantified using the lengths

of geodesics connecting them on these spaces, the intrinsic mean (Karcher mean) can

be computed as a point on the manifold (of shapes) which minimize the sum of square

geodesic distance between this unknown point to each individual shape, which lies on the

manifold. However the curves are limited by closed curves, and it has not been extended

to the 3D surface shapes. For methods using intrinsic curve or surface representations

[56, 57, 58], further statistical analysis on these representations is much more difficult

than analysis on the point representation, but the reward maybe higher due to the use of

intrinsic higher order representation.

Among these methods using point-sets parameterization, the idea of using non-

rigid spatial mapping functions, specifically thin-plate splines [59, 60, 61], to analyze

deformable shape has been widely adopted. Bookstein's work in [59], successfully

initiated the research efforts on the usage of thin-plate splines to model the deformation

of shapes. This method is landmark-based, it avoids the correspondence problem since










the placement of corresponding points is driven by the visual perception of experts,

however it suffers from the the typical problem besetting landmark methods, e.g.

inconsistency. Several significant articles on robust and non-rigid point set matching have

been published by Rangaranjan and collaborators [62, 60, 63] using thin-plate splines.

In their recent work [60], they attempt to extend their work to the construction of an

mean shape from a set of unlabeled shapes which are represented by unlabeled point-sets.

The main strength of their work is the ability to jointly determine the correspondences

and non-rigid transformation between each point sets to the emerging mean shape using

deterministic annealing and soft-assign. However, in their work, the stability of the

registration result is not guaranteed in the case of data with outliers, and hence a good

stopping criterion is required. Unlike their approach, we do not need to first solve a

correspondence problem in order to subsequently solve a non-rigid registration problem.

The active shape model proposed in [64] utilized points to represent deformable

shapes. Their work pioneered the efforts in building point distribution models to under-

stand deformable shapes [64, 65]. Objects are represented as carefully-defined landmark

points and variation of shapes are modeled using a principal component analysis. These

landmark points are acquired through a more or less manual landmarking process where

an expert goes through all the samples to mark corresponding points on each sample. It is

a rather tedious process and accuracy is limited. In recent work [66], the authors attempt

to overcome this limitation by attempting to automatically solve for the correspondences

in a non-rigid setting. The resulting algorithm is very similar to the earlier work in [58]

and is restricted to curves. The work in [55] also uses 2D points to learn shape statistics,

which is quite similar to the active shape model method except that more attention has

been paid to the sample point-sets generation process from the shape. Unlike our method,

the transformation between curves are limited by rigid mapping, and process is not

symmetric.










There are several papers in the point-sets alignment literature which bear close

relation to our research reported here. For instance, Tsin and Kanade [67] proposed

a kernel correlation based point set registration approach where the cost function is

proportional to the correlation of two kernel density estimates. It is similar to our

work since we too model each of the point sets by a kernel density function and then

quantify the (dis)similarity between them using an information-theoretic measure,

followed by an optimization of a (dis)similarity function over a space of coordinate

transformations yielding the desired transformation. The difference lies in the fact that

divergence measures used in our work is a lot more general than the information-theoretic

measure used in [67], and can be easily extended to multiple point-sets. More recently,

in [68], Glaunes et al. convert the point matching problem into an image matching

problem by treating points as delta functions. Then they "lift" these delta functions and

diffeomorphically match them. The main problem for this technique is that they need

a 3D spatial integral which must be numerically computed, while we do not need this

due to the empirical computation of the divergence measures. We will show it in the

experimental results that our method, when applied to match point-sets, achieves very

good performance in terms of both robustness and accuracy.

4.2 Divergence Measures

In probability theory and information theory, divergence measures generally stands

for those measures that quantify "distance" between probability distributions. If there are

multiple distributions, the divergence will serve as a measure of cohesion between these

distributions. Since we are dealing with groupwise point-sets, which will be represented

as multiple probability densities/distributions, we will focus on these divergence measures

between multiple distributions.

4.2.1 Jensen-Shannon Divergence

There are many information and divergence measures exist in the literature on

information theory and statistics. The most famous one among them are Kullback-Leiber










(KL) divergence. The KL divergence (also known as the relative entropy) between two

densities p and q is defined as


DKL(p q) = p() og Xdx

It is convex in p, non-negative (though not necessarily finite), and is zero if and only if

p = q. In information theory, it has an interpretation in terms of the length of encoded

messages from a source which emits symbols according to a probability density function.

While the familiar Shannon entropy gives a lower bound on the average length per

symbol a code can achieve, the KL-divergence between p and q gives the penalty (in

length per symbol) incurred by encoding a source with density p under the assumption

that it really has density q; this penalty is commonly called redundancy.

To illustrate this, consider the Morse code, designed to send messages in English.

The Morse code encodes the letter "E" with a single dot and the letter "Q" with a

sequence of four dots and dashes. Because "E" is used frequently in English and "Q"

seldom, this makes for efficient transmission. However if one wanted to use the Morse

code to send messages in Chinese pinyin, which might use "Q" more frequently, he would

find the code less efficient. If we assume contrafactually that the Morse code is optimal

for English, this difference in efficiency is the redundancy.

Notice that KL divergence is not symmetric and a popular way to symmetrize it is


J(p,q) = (DKL(p lq) + DKL(qllp))
2

which is called the J-divergence. Jensen-Shannon (JS) divergence, first introduced in [69],

serves as a measure of cohesion between multiple probability distributions. It has been

used by some researchers as a dissimilarity measure for image registration and retrieval

applications [70, 71] with very good results. It has many desirable properties, to name a

few, 1) The square root of JS-divergence (in the case when its parameter is fixed to 0.5)

is a metric [72]; 2) JS-divergence relates to other information-theoretic functionals, such










as the relative entropy or the Kullback divergence, and hence it shares their mathematical

properties as well as their intuitive appeal; 3) The compared distributions using the

JS-divergence can be weighted, which allows one to take into account the different sizes

of the point set samples from which the probability distributions are computed; 4) The

JS-divergence measure also allows us to have different number of cluster centers in each

point-set. There is no requirement that the cluster centers be in correspondence as is

required by Chui et al [73]. Given n probability density functions pi, i E {1,..., n}, the

JS-divergence of pi is defined by


JS-(p,, P2, ..,Pn) = H( 7ipi) 7iH(pi) (4-1)

where r = {1rl, 2, ..., ,rn > 0, 7r = 1} are the weights of the probability density

functions pi and H(pi) is the Shannon entropy. The two terms on the right hand side

of Equation (4-1) are the entropy of p := ripi (the 7r- convex combination of the

pis ) and the same convex combination of the respective entropies. We can show that
JS-divergence can be derived from the KL divergence


JS (pI, p2) = aKL(pl, ap, + (1 a)p2) + (1 a)KL(p2, api + (1 a)P2) (4-2)

where a E (0, 1) is a fixed parameter; we will also consider its straightforward general-

ization to n distributions.

4.2.2 CDF-JS Divergence

In Chapter 2, we defined an entropy measure which is based on probability dis-

tribution instead of density function. The distribution function is more regular because

it is defined in an integral form unlike the density function, which is the derivative of

the distribution. The definition of Cumulative Residual Entropy also preserves the well

established principle that the logarithm of the probability of an event should represent

the information content in the event. CRE is shown to be more immune to noise and









outliers. Based on this idea, we can define a KL-divergence measure between Cumulative
Distribution Functions (CDFs),
Definition: Let Pr(Xi > x) and Pr(X2 > x) be the cumulative residual distribution of
two random variables X1 and X2 respectively, we define the CDF-KL divergence by
/ Pr(Xi > x)
/CD(Pl, P2) Pr(XPr(X > x n (d (4-3)
Pr(X2 > x)2

Follow the same relationship between Jensen-Shannon divergence and KL divergence, we
can derive the so-called CDF-JS divergence from the definition of CDF-KL divergence,
(denoted as J), the result of which is shown in the following theorem,

Theorem 5 Given N probability distributions Pi, i E {1,..., n}, the CDF-JS divergence
of Pi is given by

J(P1, P2, ...P), ) ( 7iP,)- i7ri(Pi) (4-4)

where T = {7l, 72, ..., n7ri > 0, Y T = 1} are the weights of the probability
distributions Pi and S is the Cumulative Residual Entropy defined in Eqn. (2-3) of
Chapter 2.









Proof: Without loss of generality, we can prove it for two random variable case, for
which the CDF-JS can be written as follows,


J(PI, P2)

Sa/CD(P,P) + ( a)/CD(P2, P)

/ a Pr(Xi > x) InPr(Xl > x)dx + (1t
[ Pr(X > x)

Sa Pr(Xi > x) In Pr(Xi > x)dx + (1 -

a Pr(Xi > x) + (1- a) Pr(X2

S-aS(Pi) (1- a)S(P2)

a [ Pr(Xi > x) + (1 a) Pr(X2
(7


Pr(X2> x)
-a) Pr(X2 > x)In Pr(X > x)

a) Pr(X2 > x) In Pr(Xi > x)dx

> )] In Pr(X > x)dx



> )] In Pr(X > x)dx


where, P is the distribution function corresponding to the density function
p = oap + (1 a)p2, which is the convex combination of the two probability densities,
therefore


Pr(X > x)


Sp(x)dx
J x
japi(x) + (1 -a)p2(x)dx
JOO


Sa Pr(Xi > x) + (1 a) Pr(X2 > x)

Consequently, CDF-JS divergence for two random variable can be rewritten as


-oa(Pi) (1 a)S(P2)


I P(X > x)InP(X > x)dx


S(P) aS(P) (1 a)S(P2)


4.3 Methodology
In this section, we present the details of the proposed simultaneous atlas construction and
non-rigid registration method. Note that atlas construction normally requires the task of


(4-5)


(4-6)


J(Pl, P2)


(4-7)










non-rigid registration following which an atlas is constructed from the registered data.

However, in our work, the atlas emerges as a byproduct of the non-rigid registration. The

basic idea is to model each point set by a probability density or distribution function, then

quantify the distance between these functions using an information-theoretic measure.

The distance measure is optimized over a space of coordinate transformations yielding the

desired transformations. We will begin by presenting the energy function for solving the

groupwise point-sets registration problem.

4.3.1 Energy Function for Groupwise Point-sets Registration

We use the following notation: The data point-sets are denoted by {XP,p E {1, ..., N}}.

Each point-set XP consists of points {x( E RD, i E {1,..., ip} }. The atlas points-set is

denoted by Z. Assume that each point set XP is related to Z via a function fP, Ap is the

set of the transformation parameters associated with each function fP. To compute the

mean shape from these point-sets and register them to the emerging mean shape, we need

to recover these transformation parameters to construct the mean shape. This problem can

modeled as an optimization problem with the objective function being the JS-divergence

or CDF-JS divergence between the distributions of the deformed point-sets, represented

as Pi = p(f (X')), the atlas construction problem can now be formulated as,
N
min (P, P2..., PN)+A I|Lfi 2
/ i1 (4-8)



In (4-8), D is the divergence measure for multiple distributions, which we propose to use

either JS divergence or CDF-JS divergence. The weight parameter A is a positive constant

the operator L determines the kind of regularization imposed. For example, L could

correspond to a thin-plate spline, a Gaussian radial basis function, etc. Each choice of L

is in turn related to a kernel and a metric of the deformation from and to Z.

Following the approach in [73], we choose the thin-plate spline (TPS) to represent the

non-rigid deformation. Given n control points xl,..., x, in Rd, a general non-rigid










mapping f : RI -- Rd represented by thin-plate spline can be written analytically as:

f(x) = WU(x) + Ax + t Here Ax + t is the linear part of f. The nonlinear part is

determined by a d x n matrix, W. And U(x) is an n x 1 vector consisting of n basis

functions Ui(x) = U(x, xi) = U(||x xil|) where U(r) is the kernel function of

thin-plate spline. For example, if the dimension is 2 (d = 2) and the regularization

functional is defined on the second derivatives of f, we have U(r) = 1/(87)r21n(r).

Therefore, the cost function for non-rigid registration can be formulated as an energy

functional in a regularization framework, where the regularization term in equation 4-8 is

governed by the bending energy of the thin-plate spline warping and can be explicitly

given by trace(WKWT) where K = (Kij), Kij = U(pi, pj) describes the internal

structure of the control point sets. Note the linear part can be obtained by an initial affine

registration, then an optimization can be performed to find the parameter W.

4.3.2 JS Divergence in a Hypothesis Testing Framework

In this section we show that the Jensen-Shannon divergence can be interpreted in the the

frame work of statistical hypothesis testing. To see this, we construct a likelihood ratio

between i.i.d. samples drawn from a mixture (a 7rapa) and i.i.d. samples drawn from a

heterogeneous collection of densities (pi, p2, PN) with the samples being indexed by

the specific member distribution in the family from which they are drawn. Assume that ni

samples are drawn from pi, n2 from p2 etc. Let the total number of pooled samples be
def N
defined as M l ea na. The likelihood ratio then is,

A -1M 1 a-1 7aPa(Xk) (4-9)
N Ir-9a
lla=1 1ka=Pa(Xla)

where xk consists of points {x, i E {1, ..., na}, a E {1,..., N}}, which is the pooled data

of all the samples. In contrast to the typical statistical test relative to a threshold, we seek

the maximum of the likelihood ration in Eqn. (4-9). The following theorem shows the

relationship between Jensen-Shannon divergence and the likelihood ration.










Theorem 6 Given N probability density functions pa, a E {1,..., N}, maximizing the

h)jplthei ratio in Eqn. (4-9) is equivalent to minimizing the Jensen-.\ln11,11ii divergence

between the N probability densities p,, a E {1,..., N}.

Proof: The proof follows by taking logarithm of the likelihood ratio, and then using the

weak law of large numbers, we can show that the log-likelihood ratio is the negative of

Jensen-Shannon divergence. m

We seek to maximize the probability that the samples are drawn from the mixture rather

than from separate members of the family (p, p2, ..., PN). In the context of groupwise

matching of point-sets, this makes eminent sense since maximizing the above ratio is

tantamount to increasing the chance that all of the observed point-sets are warped

versions of the same underlying warped and pooled data model. The notion of the pooled

data model is defined as follows. In our process of groupwise registration, the warping

does not have afixed target data set. Instead, the warping is between the input data sets

and an evolving target which we call the pooled model. The target evolves to a fully

registered pooled data set at the end of the optimization process. The pooled model then

consists of input data sets which have undergone groupwise matching and are now fully

registered, iith each other. The connection to the JS-divergence arises from the fact that

the negative logarithm of the above ratio (Eqn.4-9) asymptotically converges to the

JS-divergence when the samples are assumed to be drawn from the mixture 7rpPa.

4.3.3 Unbiasness Property of the Divergence Measures

Typically we are required to construct an atlas from very large number of point-sets, and

this process will usually take a long time since the computational complexity grows

polynomially with the increase of number of point-sets (N) that we want to register.

However the following hierarchical method will significantly reduce the computational

complexity.

Assume that we are given N point-sets, from which we are going to construct the atlas,

we can then divide the N point-sets into m subsets (generally m < N), therefore we can










construct m atlases from each subsets using our algorithms, and all the point-sets inside

each subsets are registered. Then we can either construct a single atlas from these m atlas

point-sets, or we can further divide m atlas point-sets into even smaller subsets, and

follow the same process until a single atlas is constructed. The remaining question is

whether the atlas thus obtained is biased or not? The following theorem will lead us to the

answer.


Theorem 7 Given N probability distributions Pa, a E {1,..., N}, each having a weight

7Ta in the JS divergence or CDF-divergence. Ifwe further divide the N distributions into

m subsets, such that ith subset contains ni distributions Pa, a {k ), k) ,..., ki) }, and

ni n N. Assume Si is the convex combination of the all the distributions in the ith

subset, ii i/t the weights k(), where 3 = k), i.e. Si, 1 k )Pk/. We then

have the following relationship between the JS divergence of the Pas and divergence of

the Sis

D,(P, P2," ,PN)- D3(Si,S, S-, ,S,)
Tn (4-10)
ADiV ) (P {,,P,..," ,P,. k)
i= 1


Proof: It is trivial to derive the relationship in Eqn. (4-10) by simple algebraic

operations. m

In our registration algorithm, all the point-sets are represented as probability distributions,

and the atlas thus constructed can be considered as convex combination of these

distributions. Therefore, we can treat Pas and Sis as the distributions corresponding to

the point-sets and the constructed atlases from the subsets respectively. Therefore from

Theorem 7, we know that the relationship in Eqn. (4-10) holds between the JS divergence

of the Pas and Sis. Notice that the right hand side of Eqn. (4-10) is the JS/CDF-JS

divergences of the distributions in all the subsets, which are minimized in each steps of

the hierarchical method we proposed. Intuitively, if these point sets are aligned properly,










the corresponding distribution functions should be statistically similar. Therefore the

divergences of all the subsets should be zero all very close to zero, which means the right

hand side of Eqn. (4-10) is zero. Consequently, the JS/CDF-JS divergence of the Pas and

divergence of the Sis are equal to each other, therefore minimizing JS/CDF-JS divergence

of all the resultant atlas point-sets is equivalent to minimizing divergence of the original

point-sets, implying that there is no bias toward any particular partitioning of the

point-sets.

Having introduced the cost function and the transformation model, now the task is to

design an efficient way to estimate empirical divergence measures between multiple

densities or distributions and derive the analytic gradient of the estimated divergence in

order to achieve the optimal solution efficiently. We design two complete different

approaches for estimating JS divergence and CDF-JS divergence. We use finite mixture

model for estimating JS divergence and the parzen window technique for CDF-JS

divergence, the details of which will be introduced next.

4.3.4 Estimating JS and its Derivative

4.3.4.1 Finite mixture models

Considering the point set as a collection of Dirac Delta functions, it is natural to think of a

finite mixture model as representation of a point set. As the most frequently used mixture

model, a Gaussian mixture [74] is defined as a convex combination of Gaussian

component densities.

To model each point-set as a Gaussian mixture, we define a set of cluster centers, one for

each point-set, to serve as the Gaussian mixture centers. Since the feature point-sets are

usually highly structured, we can expect them to cluster well. Furthermore we can greatly

improve the algorithm efficiency by using limited number of clusters. Note that we can

choose the cluster centers to be the point-set itself if the size of point-set is quite small.

The cluster center point-sets are denoted by {V, p {1,..., N} }. Each point-set VP

consists of points {v E 7RD, i E {1,..., KP} }. Note that there are KP points in each VP,










and the number of clusters for each point-set may be different (in our implementation, the

number of clusters were usually chosen to be proportional to the size of the point-sets).

The cluster centers are estimated by using a clustering process over the original sample

points x', and we only need to do this once before the process of joint atlas estimation

and point-sets registration. In our implementation, we utilize deterministic annealing

(DA) procedure with its proven benefit of robustness in clustering [75]. We begin by
specifying the density function of each point set.
KP
pp(x) = oap(x I ) (4-11)
a=l
In Equation (4-11), the occupancy probability which is different for each data point-set is

denoted by ao'. The component densities p(x vI) is

1 pTy1 ,(X_ p
p(x ) = 1 exp 1 (x v (x v)) (4-12)
(27r)'Z a 2
Probability of the point set XP coming from this mixture is then
up up KP
Pr(XP|VP, aP) pp(x4) H ap(x |va) (4-13)
i=l i=1 a=
Later, we set the occupancy probability to be uniform and make the covariance matrices

EY to be proportional to the identity matrix in order to simplify atlas estimation
procedure.

For simplicity, we choose 3i = Vi {= 1, 2,..., N}. Let

Qp: := a=1 aCPr(fJ(x') fP(vP)) be a mixture model containing component densities
Pr(fJ(x) f"(vP)),

eX1 ( 1(fJ(J') fP()P)T al (4f S (p;\
p(f (x ) fP(v)) = -exp ( ) (v) (-1 '(x) f(V))
(27 )T 2
(4-14)

Where {j, a E {1,..., K}} is the set of cluster covariance matrices. For the sake of

simplicity and ease of implementation, we assume that the occupancy probabilities are









uniform (a = -) and the covariance matrices ES are isotropic, diagonal, and identical

[(Y = cr2ID)]. Having specified the density function of the data, we can then rewrite
Equation (4-8) as follows,

JS~(PI,P2, PN)
1 1p
-{[H( j P) Y (P )]
N (4-15)
+[H( i)- H(P2)]

S... [H( Pi) H(PN)

For each term in the equation, we can estimate the entropy using the weak law of large

numbers, which is given by,

S_ l 1i Q Qi + ... + QN 1t '
H(Z Pi) H(P)) log N 1 log Q
ni NQni
in 1 1
-1 t log 1~ Q
ni
= i Q1z + Q +" + Qx

Combining these terms we have,

JS(P,P2, ...,PN)
1 2
i1 n2 X?
1 VNQ1z 1 VQ2
Slog I + log 2 2 (4-16)
S 1 + Q2 +... + N i Q1 + 2z +...+ N

+..+1 flog N
nN" Q, +Q' +...+Q J


4.3.4.2 Optimizing the JS divergence

Computation of the gradient of the energy function is necessary in the minimization

process when employing a gradient-based scheme. If this can be done in analytical form,

it leads to an efficient optimization method. We now present the analytic form of the

gradient of the JS-divergence (our cost function):

OJS OJS OJS
VJS= [ ,- ."' (4-17)
Bp," 8w2 OtN









Each component of the gradient maybe found by differentiating Eqn. (4-16) with respect
to the transformation parameters. In order to compute this gradient, let's first calculate the
derivative of Qp with respect to pl,

1 K ( 1 2F2F afj() if *I- )
S(2) 3K a exp (- 22 ) (F T P L if j
1 ( 1 F. F. fp'
e(2) 3K aexp -2 2 P, )I if p 4j (4-18)
O(2)T,3K a-
I K I F" 2(Fexp( F [afP a f (X)] if I p j

where Fj := f (xj) fP(v). Based on this, it is straight forward to derive the gradient
of the JS-divergence with respect to the transformation parameters tl1, which is given by

a ( nl Xiii
OPS 1 1 1Q
7=1 Qo O + QX + QX
1 2

11 1





y)1 2 N
) +1, 1+...+Q
LO1 iOQZ +I T'" lI J-



Q, + Q +... + Q J


4.3.5 Estimating CDF-JS and its Derivative
We will use the parzen window technique to estimate the cumulative distribution function
and its derivative. The calculation of CDF-JS divergence requires the estimation of the
cumulative probability distributions of each point-set. Without lost of generality, we only
discuss the derivation for the 2D case, which can be extended to 3D case easily. For each
point xa with the coordinates [xa, ya] in the point set Xa, it is transformed by the function

fa to fa(x, pa) [f(x, ~a), fa(', a)]. Let 3(3) be a cubic spline Parzen window.
The smoothed probability density function pa(1, k; ,a) of the point-set










Xa, a {1,..., N}} is given by

p (l, k;a) -(3) faXa ,,a) 0
Ai (4-20)

3(3) (k- fa(li,,,) yo

where a is a normalization factor that ensures ff p(l, k)dldk = 1, [1, k] are the
coordinate values in the X and Y axis respectively, the transformed point coordinates

[fa(xa, na), fa"(, 1)] is normalized by the minimum coordinate value, xo, yo, and the
range of each bin, Abx, Aby. From the density function, we can calculate the cumulative
residual distribution function by the formula
Pa(1 > A, k > 7; ,a) JfAO pa(l, k; ta)dldk, where A e L,7 c Ly, L, Ly are
discrete sets of coordinate values in the X and Y axis respectively. To express it in further
detail, we have the following,
oa o [0 fala ,a\ 0

ia('"!(l: fa(, -^a) ^ On
pa (Axy _,; a) /E3) I- (Xli, p a) \ d
/ () k-f (, ) yo dk


Abx Aby

where I)(.) is the cumulative residual function of cubic spline kernel which is defined as

follows,




Note that d( -P() (3) ().
Having specified the distribution function of the data, we can then rewrite Eqn. (4-8) as
follows, (For simplicity, we choose a = Va = {1, 2,..., N}. )
N N
J(Pl,P2, ...,PN) P = ( 7raPa)- 7rag(Pa)
a=1 a=1 (4-21)
A P log P + a Plog p
A a A









where P is the cumulative residual distribution function for the density function

k Zp"(1, k; gf), which can be expressed as
N nct fa(acfl a 0
P(A,, ;a) ZaEZ A p )
a= i (4-22)

Aby

4.3.5.1 Optimizing the CDF-JS divergence

We now present the analytic form of the gradient of the CDF-JS divergence (our cost
function):

VJ = [ (4-23)
[al' av2 9'aN

Each component of the gradient maybe found by differentiating Eqn (4-21) with respect
to the transformation parameters. It can be easily shown that =P(A'"Y;) 1 IPc(A,;ip)
0aW N 0pa
Based on these facts, it is straight forward to derive the gradient of the CDF-JS
divergence with respect to the transformation parameters pa, which is given by

o > ZHE l + logP] &P(a,;;r)o
A -
+ [1 + lo gP (4-24)


N EEE opa
a A 7


As a byproduct of the groupwise registration using the CDF-JS divergence, we get the

atlas of the given population of data sets, which is simply obtained by substituting the

estimated transformation functions fa in to the formula for the atlas p(A) = N I TTaPa.
Note that our algorithm can be applied to yield a biased registration in situations that

demand such a solution. This is achieved by fixing one of the data sets (say the reference)

and estimating the transformation from this to the novel scene data. We will present










experimental results on point-set alignment between two given point-sets as well as atlas

construction from multiple point-sets in the next section.

4.4 Experiment Results

We now present experimental results using JS divergence and CDF-JS divergence for

point-sets registration. And the results will be demonstrated on both synthetic and real

data sets.

4.4.1 JS Divergence Results

To demonstrate the robustness and accuracy of our algorithm, we show the alignment

results by applying the JS-divergence to the point-set matching problem. Then, we will

present the atlas construction results in the second part.

4.4.1.1 Alignment results

First, to test the validity of our approach, we perform a set of exact rigid registration

experiments on both synthetic and real data sets without noise and outliers. Some

examples are shown in Figure 4-2. The top row shows the registration result for a 2D real

range data set of a road (which was also used in Tsin and Kanade's experiments [67]).

The figure depicts the real data and the registered (using rigid motion). Top left frame

contains two unregistered point sets superposed on each other. Top right frame contains

the same point sets after registration using our algorithm. A 3D helix example is

presented in the second row (with the same arrangement as the top row). We also tested

our method against the KC method [67] and the ICP methods, as expected, our method

and KC method exhibit a much wider convergence basin/range than the ICP and both

achieve very high accuracy in the noiseless case.

We also applied our algorithm to non-rigidly register medical datasets (2D point-sets).

Figure 4-3 depicts some results of our registration method applied to a set of 2D corpus

callosum slices with feature points manually extracted by human experts. Registration

result is shown in the left column with the warping of 2D grid under the recovered motion

which is shown in the middle column. Our non-rigid alignment performs well in the










After registration


A\


40 -20
Initial setup


-30 -20 -10 0 10
After registration


C "isf~


Figure 4-2: Results of rigid registration in noiseless case. 'o' and '+' indicate the model
and scene points respectively.


presence of noise and outliers (Figure 4-3 right column). For the purpose of comparison,

we also tested the TPS-RPM program provided in [62] on this data set, and found that

TPS-RPM can correctly register the pair without outliers (Figure 4-3 top left) but failed

to match the corrupted pair (Figure 4-3 top right).


Initial Setup

++t+t!+



0.4 0.6 0.8 1
After registration




0.4 0.6 0.8 1


Original point set Initial Setup
0.2 0.2 0 ++ +++14
0 1 : t 0.1 ; :


0.4 0.6 0.8 1 0.4 0.6 0.8 1
Deformed point set After registration

0.2 4 ;0. 0.2 0 6

4 0 6 08 1 02 0
0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1


Figure 4-3: Non-rigid registration of the corpus callosum data. Left column: two man-
ually segmented corpus callosum slices before and after registration; Middle column:
warping of the 2D grid using the recovered motion; Top right: same slices with one cor-
rupted by noise and outliers, before and after registration.


Initial setup
0*











4.4.1.2 Atlas construction results

In this section, we begin with a simple but demonstrative example of our algorithm for 2D

atlas estimation. The structure we are interested in this experiment is the corpus callosum

as it appears in MR brain images. Constructing an atlas for the corpus callosum and

subsequently analyzing the individual shape variation from "normal" anatomy has been

regarded as potentially valuable for the study of brain diseases such as agenesis of the

corpus callosum(ACC), and fetal alcohol syndrome(FAS).

it Set1 Point Set2 Point Set3
S0.2 0.2
0.1 0.1
0 L 0
-0.1 -0.1
0.8 1 0.4 0.6 0.8 1 0.4 0.6 0.8 1
it Set4 Point Set5 Point Set6
0.2 | ,- 0.2
0.1 0.1
0 0
-0.1 0.1
0.8 1 0.4 0.6 0.8 1 0.4 0.6 0.8 1
it Set7 Point-sets Before Registration Deformed Point-sets
0.2 ,.4 -- 0.2
0.1 0.1

0.8 1 0.4 0.6 0.8 1 0.4 0.6 0.8 1

Figure 4-4: Experiment results on seven 2D corpus collasum point sets. The first two
rows and the left image in third row show the deformation of each point-set to the at-
las, superimposed with initial point set (show in 'o') and deformed point-set (shown in
'*'). Middle image in the third row: The estimated atlas is shown superimposed over all
the point-sets. Right: The estimated atlas is shown superimposed over all the deformed
point-sets.


We manually extracted points on the outer contour of the corpus callosum from seven

normal subjects, (as shown Figure 4-4, indicated by "o"). The recovered deformation

between each point-set and the mean shape are superimposed on the first two rows in

Figure 4-4. The resulting atlas (mean point-set) is shown in third row of Figure 4-4, and

is superimposed over all the point-sets. As we described earlier, all these results are

computed simultaneously and automatically. This example clearly demonstrate that our










joint matching and atlas construction algorithm can simultaneously align multiple shapes

(modeled by sample point-sets) and compute a meaningful atlas/mean shape.

4.4.2 CDF-JS Divergence Results

First, to see how the CDF-JS method behaves in the presence of noise and outliers, we

designed the following procedure to generate a corrupted template point set from a model

set. For a model set with n points, we control the degree of corruption by (1) discarding a

subset of size (1 p)n from the model point set, (2) applying a rigid transformation (R,t)

to the template, (3) perturbing the points of the template with noise (of strength e), and (4)

adding (7 p)n spurious, uniformly distributed points to the template. Thus, after

corruption, a template point set will have a total of -rn points, of which only pn

correspond to points in the model set. Since ICP is known to be sensitive to outliers, we

only compare our method with the more robust Jensen-Shannon divergence method in

terms of the sensitivity of noise and outliers. The comparison is done via a set of 2D

experiments.At each of several noise levels and outliers ,ienugitl, we generate five

models and six corrupted templates from each modelfor a total of 30 pairs at each noise

and outlier strength setting. For each pair, we use our algorithm and the JS method to

estimate the known rigid transformation which was partially responsible for the

corruption. Results show when the noise level is low, both JS and CDF-JS have strong

resistance to outliers. However, we observe that when the noise level is high, CDF-JS

method exhibits stronger resistance to outliers than the JS method, as shown in Figure

4-5, which confirm that CDF-JS is indeed more robust in the presence of high noise and

outlier level. A 3D example is also presented in Figure 4-6.

Next, we present groupwise registration results on 3D hippocampal point-sets. Four 3D

point-sets were extracted from epilepsy patients with left anterior temporal lobe foci

identified with EEG. An interactive segmentation tool was used to segment the

hippocampus from the 3D brain MRI scans of 4 subjects. The point-sets differ in shape,

with the number of points 450, 421, 376, 307 in each point-set respectively. In the first








57



RMS errors in translation RMS errors in rotation
10* 0.1
-s--JS --JS
8 CDF-JS 0.08 CDF-JS

6 0.06

4 / / 0.04 -
SV e-- '
2 0.02 -

0 O0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Strength of outlier Strenath of outlier

Figure 4-5: Robustness to outliers in the presence of large noise. Errors in estimated rigid
transform vs. proportion of outliers ((7 p)/(p)) for both our method and KC method.

Initial setup After registration



200
150
+ 1004
+ 50
SI 0+ 200
+ 200 ++ + + +
150

S100 50 .
0 5


Figure 4-6: Robustness test on 3D swan data. 'o' and '+' indicate the model and scene
points respectively. Note that the scene point-set is corrupted by noise and outliers.


four images of Figure 4-7, the recovered nonrigid deformation between each

hippocampal point-set to the atlas is shown along with a superimposition on all of the

original data sets. In second row of the Figure 4-7, we also show the scatter plot of

original point-sets along with all the point-sets after the non-rigid warping. An

examination of the two scatter plots clearly shows the efficacy of our recovered non-rigid

warping. Note that validation of what an atlas shape ought to be in the real data case is a

difficult problem and we relegate its resolution to a future paper.




























Pointset 2










Pooled Pointsets


Pointset 3










Deformed Pointsets


.r .


Figure 4-7: Atlas construction from four 3D hippocampal point sets. The first row and
the left image in second row shows the deformation of each point-set to the atlas (rep-
resented as cluster centers), superimposed with initial point set (show in green 'o') and
deformed point-set (shown in red '+'). Left image in the second row: Scatter plot of the
original four hippocampal point-sets. Right: Scatter plot of all the warped point-sets.


Pointset 1










Pointset 4




d .,:.















CHAPTER 5
APPLICATIONS TO IMAGE SEGMENTATION

In Medical Imaging applications, segmentation can be a daunting task due to possibly

large inhomogeneities in image intensities across an image e.g., in MR images. These

inhomogeneities combined with volume averaging during the imaging and possible lack

of precisely defined shape boundaries for certain anatomical structures complicates the

segmentation problem immensely. One possible solution for such situations is atlas-based

segmentation. The atlas once constructed can be used as a template and can be registered

non-rigidly to the image being segmented (henceforth called a target image) thereby

achieving the desired segmentation. Many of the methods that achieve atlas-based

segmentation are based on a two stage process involving, (i) estimating the non-rigid

deformation field between the atlas image and the target image and then, (ii) applying the

estimated deformation field to the desired shape/atlas to achieve the segmentation of the

corresponding structures in the target image. In this chapter, we develop a novel

technique that will simultaneously achieve the non-rigid registration and segmentation.

There is a vast body of literature for the tasks of registration and segmentation

independently however, methods that combine them into one algorithm are far and few in

between. In the following, we will briefly review the few existing methods that attempt to

achieve simultaneous registration and segmentation.

5.1 Related Work

In one of the earliest attempts at joint registration & segmentation, Bansal et al., [76]

developed a minmax entropy framework to rigidly register & segment portal and CT data

sets. In [77], Yezzi et al., present a variational principle for achieving simultaneous

registration and segmentation, however, the registration part is limited to rigid motions. A

similar limitation applies to the technique presented by Noble et al., in [78]. A variational










principle in a level-set based formulation was presented in Pargios et. al., [79], for

segmentation and registration of cardiac MRI data. Their formulation was again limited

to rigid motion and the experiments were limited to 2D images. In Fischl et al., [80], a

Bayesian method is presented that simultaneously estimates a linear registration and the

segmentation of a novel image. Note that linear registration does not involve non-rigid

deformations. The case of joint registration and segmentation with non-rigid registration

has not been addressed adequately in literature with the exception of the recent work

reported in Soatto and Yezzi [81] and Vemuri et al., [82]. However, these methods can

only work with image pairs that are necessarily from the same modality or the intensity

profiles are not too disparate.

In this paper, we present a unified variational principle that will simultaneously register

the atlas shape (contour/surface) to the novel brain image and segment the desired shape

(contour/surface) in the novel image. In this work, the atlas serves in the segmentation

process as a prior and the registration of this prior to the novel brain scan will assist in

segmenting it. Another key feature/strength of our proposed registration+segmentation

scheme is that it accommodatesfor image pairs having very distinct intensity

distributions as in multimodality data sets. More details on this are presented in section

5.2.

5.2 Registration+Segmentation Model

We now present our formulation of joint registration & segmentation model. Let li be the

atlas image containing the atlas shape C, 12 the novel image that needs to be segmented

and v be the vector field, from 12 to I1 i.e., the transformation is centered in 12, defining

the non-rigid deformation between the two images. The variational principle describing

our formulation of the registration assisted segmentation problem is given by:


minE(v, C) = Seg(2, C) + dst(v(C), C) + Reg( 1, 2, v).


(5-1)







61


Atlas Image Target Image



v C)









Figure 5-1: Model Illustration


Where, the first term denotes the segmentation functional. C is the boundary contour

(surface in 3D) of the desired anatomical shape in I2. The second term measures the

distance between the transformed atlas v(C) and the current segmentation C in the novel

brain image i.e., the target image and the third term denotes the non-rigid registration

functional between the two images. Our joint registration & segmentation model is

illustrated in Figure 5.2.

For the segmentation functional, we use a piecewise constant Mumford Shah model,

which is one of the well-known variational models for image segmentation, wherein it is

assumed that the image to be segmented can be modeled by piece-wise constant regions,

as was done in [54]. This assumption simplifies our presentation but our model itself can

be easily extended to the piecewise smooth regions case. Additionally, since we are only

interested in segmenting a desired anatomical shape (e.g., the hippocampus, the corpus

callosum, etc.), we will only be concerned with a binary segmentation i.e., two classes

namely, voxels inside the desired shape and those that are outside it. These assumptions

can be easily relaxed if necessary but at the cost of making the energy functional more

complicated and hence computationally more challenging. The segmentation functional

takes the following form:


Seg(12, C) (2 )2 dx +a ds (5-2)
J/o Jo









Where, Q is the image domain and a is a regularization parameter. u = ui if x C C, and

u = Uo if E Cout. C~ and Cot denote the regions inside and outside of the curve, C
representing the desired shape boundaries in I2.

For the non-rigid registration term in the energy function, we use the information

theoretic-based criteria, cross cumulative residual entropy (CCRE) which we introduced
in Chapters 2. CCRE was shown to outperform Mutual Information based registration in

the context of noise immunity and convergence range, motivating us to pick this criteria

over the MI-based cost function. The new registration functional is defined by

Reg(li, 12, V) (C(II V X)), 12 )) (l Vv(x) 2)) (5-3)

where, cross-CRE C(II, 12) is given by,


C(l, I2) (1) E[S(I1/I2)] (5-4)

with S(11) f+ P(|II| > A)log P(|II > A)dA and R+ (x e R; x > 0). v(x) is as

before and p is the regularization parameter and I I denotes Frobenius norm. Using a

B-spline representation of the non-rigid deformation, one need only compute this field at
the control points of the B-splines and interpolate elsewhere, thus accruing computational

advantages. Using this representation, we have derived analytic expressions for the

gradient of the energy with respect to the registration parameters. This in turn makes our
optimization more robust and efficient.

In order for the registration and the segmentation terms to "talk" to each other, we need a
connection term and that is given by

dist(v(C), C) = PO(c)(x)dx (5-5)

where, R is the region enclosed by C, Ov(c) (x) is the embedding signed distance function

of the contour v(C), which can be used to measure the distance between v(C) and C.

The level-set function 0 : R2 IR is chosen so that its zero level-set corresponds to the










transformed template curve v(C). Let Edist = dist(v(C), C), one can show that

O= v(C) (C)N where N is the normal to C. The corresponding curve evolution
equation given by gradient descent is then

dC
at v(c)(C)N (5-6)

Not only does the signed distance function representation make it easier for us to convert

the curve evolution problem to the level-set framework, it also facilitates the matching of

the evolving curve C and the transformed template curve v(C), and yet does not rely on a

parametric specification of either C or the transformed template curve. Note that since

dist(v(C), C) is a function of the unknown registration v and the unknown segmentation

C, it plays the crucial role of connecting the registration and the segmentation terms.

Combining these three functionals together, we get the following variational principle for

the simultaneous registration+segmentation problem:

minE(C, v, Uo, ui) = (12 )2dx + a1 ds d dist(v(C), C)
p Jc (5-7)
a3C(II(V(X)), 12(X)) + a4 IVv(x) 12dX.

a are weights controlling the contribution of each term to the overall energy function and

can be treated as unknown constants and either set empirically or estimated during the

optimization process. This energy function is quite distinct from those used in methods

existing in literature because it is achieving the Mumford-Shah type of segmentation in an

active contour framework jointly with non-rigid registration and shape distance terms. We

are now ready to discuss the level-set formulation of the energy function in the following

section.

5.2.1 Gradient flows

The level set method have been used extensively for implementing the curve evolution

based segmentation, primarily due to its many advantages over the competing approaches.

These include the ability to elegantly handle changes in the topology of the curve (splits









and merges), the ability to deal with the formation of cusps and corners, which are
extremely common in curve evolution, and the numerical stability and efficiency afforded
in its implementation. For our model where the equation for the unknown curve C is
coupled with the equations for v(x), Uo, ui, it is convenient for us to use the level set
approach as proposed in [54].
Taking the variation of E(.) with respect to C and writing down the gradient descent
leads to the following curve evolution equation:

C= [-(2 i)2 + (12 u0)2 + a1K a2v(C)(C')] N (5-8)

Note that equation (5-6) is used in the derivation. Equation (5-8) in the level-set
framework is given by:


=at h (12 i2 u)2 V + a2 O 2v(C)() |V~ (5-9)
where ui and no are the mean values inside and outside of the curve C in the image I2. To
drive the curve towards the template's level-set function v(c) more efficiently, rather than
just having the zero level-sets match, we can add another term 0(C) into the level-set
evolution equation giving us,

=[ (12 )2 +(12 )2 + V .
(5-10)
+a2 (v)(C) (C'))] V7.

As illustrated in Figure 5-2, the two parameters ac and a2 are used to balance the
influence of the shape distance model and the region-based model. Note that (C) = 0 at
any location of the curve by the definition of level-set function 0, this added term does
not affect the curve evolution equation [83].
As mentioned before, we use a B-spline basis to represent the displacement vector field

v(x, p), where p is the transformation parameters of the B-spline basis.

aE a fR v(c)(x) dx aC(I1(v(x)),2 (x)) a a Vv(x) 2dx
S/ 2 0/3 p+4 (5-11)











v(C)
v(C)

-2( (C) -( C -..
Next Step in the Evolution




Current Shape

S"
Region Based Term

Figure 5-2: Illustration of the various terms in the evolution of the level set function <.
To update 0, we combine the standard region based update term S, and level set function
corresponding to the shape distance term.

The first term of equation(5- 1) can be rewritten as follows:

a SfR v(c) (x) dx f a9v(c) (x) dx
ap 0j (G (5-12)
j v(c) v, v(x, ) dx

where o ) is the directional derivative in the direction of v(x, p). The second term of

Eqn. (5-11) has been derived in Eqn. (3-15) of the chapter 3. We simply state the result

here without the derivations for the sake of brevity,

C (I2, 1 o v(x; p)) P(i > A, k; ) OP(i> A,k;p)13)
= log (5-13)
p AIl keI2 p12(k)P(i > A;p) p

where P(i > A, k; p) and P(i > A; p) are the joint and marginal cumulative residual

distributions respectively. p12 (k) is the density function of image I2. The last term of Eqn.

(5-11)leads to,
Sf IVv(x)||2dx V 9v
2 Vv *- dx (5-14)

where both the matrices Vv and 'v are vectorized before the dot product is computed.
a-/1










Substituting equations (5-12), (5-13) and (5-14) respectively back into the equation

(5-11), we get the analytical gradient of our energy function with respect to the B-spline

transformation parameters p. We then solve for the stationary point of this nonlinear

equation numerically using a quasi-Newton method.

5.2.2 Algorithm Summary

Given the atlas image I1 and the unknown subject's brain scan 12, we want the

segmentation result C in 12. Initialize C in 12 to C and set the initial displacement field to

zero.
1. For fixed C, update deformation field using gradient-based numerical method for

one step.

2. For fixed deformation field v, evolve 0 in 12 and thereby update C as the zero

level-set of 0.

3. Stop the registration process if the difference in consecutive iterates is less than

e = 0.01, a pre-chosen tolerance, else go to Step 1.

5.3 Results

In this section, we present several examples results from an application of our algorithm.

The results are presented for synthetic as well as real data. The first three experiments

were performed in 2D, while the fourth one was performed in 3D. Note that the image

pairs used in all these experiments have significantly different intensity profiles, which is

unlike any of the previous methods, reported in literature, used for joint registration and

segmentation. The synthetic data example contains a pair of MR T1 and T2 weighted

images which are from the MNI brainweb site [52]. They were originally aligned with

each other. We use the MR T1 image as the source image and the target image was

generated from the MR T2 image by applying a known non-rigid transformation that was

procedurally generated using kernel-based spline representations (cubic B-Spline). The

possible values of each direction in deformation vary from -15 to 15 in pixels. In this




























Figure 5-3: Results of application of our algorithm to synthetic data (see text for details).

case, we present the error in the estimated non-rigid deformation field, using our

algorithm, as an indicator of the accuracy of estimated deformations.

Figure 5-3 depicts the results obtained for this image pair. With the MR T1 image as the

source image, the target was obtained by applying a synthetically generated non-rigid

deformation field to the MR T2 image. Notice the significant difference between the

intensity profiles of the source and target images. Figure5-3 is organized as follows, from

left to right: the first row depicts the source image with the atlas-segmentation superposed

in red, the registered source image which is obtained using our algorithm followed by the

target image with the unregistered atlas-segmentation superposed to depict the amount of

mis-alignment; second row depicts ground truth deformation field which we used to

generate the target image from the MR T2 image, followed by the estimated non-rigid

deformation field and finally the segmented target. As visually evident, the

registration+segmentation are quite accurate from a visual inspection point of view. As a

measure of accuracy of our method, we estimated the average, p, and the standard

deviation, a, of the error in the estimated non-rigid deformation field. The error was

estimated as the angle between the ground truth and estimated displacement vectors. The










average and standard deviation are 1.5139 and 4.3211 (in degrees) respectively, which is

quite accurate.

Table 5-1 depicts statistics of the error in estimated non-rigid deformation when

compared to the ground truth. For the mean ground truth deformation (magnitude of the

displacement vector) in Column-1 of each row, 5 distinct deformation fields with this

mean are generated and applied to the target image of the given source-target pair to

synthesize 5 pairs of distinct data sets. These pairs (one at a time) are input to our

algorithm and the mean (p) of the mean deformation error (MDE) is computed over the

five pairs and reported in Column-2 of the table. MDE is defined as

Table 5-1: Statistics of the error in estimated non-rigid deformation.

ig ip of MDE a of MDE
2.4 0.5822 0.0464
3.3 0.6344 0.0923
4.5 0.7629 0.0253
5.5 0.7812 0.0714


dm = car(R) R I vo(x) v(x) 11, where vo(x0 ) v(xa) is the ground truth and

estimated displacements respectively at voxel xi. I1.1 denotes the Euclidean norm, and R

is the volume of the region of interest. Column-3 depicts the standard deviation of the

MDE for the five pairs of data in each row. As evident, the mean and the standard

deviation of the error are reasonably small indicating the accuracy of our joint registration

+ segmentation algorithm. Note that this testing was done on a total of 20 image pairs

(=40) as there are 5 pairs of images per row.

For the first real data experiment, we selected two image slices from two different

modalities of brain scans. The two slices depict considerable changes in shape of the

ventricles, the region of interest in the data sets. One of the two slices was arbitrarily

selected as the source and segmentation of the ventricle in the source was achieved using

an active contour model. The goal was then to automatically find the ventricle in the

target image using our algorithm given the input data along with the segmented ventricles






















A,, ?..^ J




Figure 5-4: Results of application of our algorithm to a pair of slices from human brain
MRIs (see text for details).

in the source image. Figure 5-4 is organized as follows, from left to right: the first row

depicts the source image with the atlas-segmentation superposed in black followed by the

target image with the unregistered atlas-segmentation superposed to depict the amount of

mis-alignment; second row depicts the estimated non-rigid vector field and finally the

segmented target. As evident from figures 5-4, the accuracy of the achieved

registration+segmentation visually very good. Note that the non-rigid deformation

between the two images in these two examples is quite large and our method was able to

simultaneously register and segment the target data sets quite accurately.

The second real data example is obtained from two brain MRIs of different subjects and

modalities, the segmentation of the cerebellum in the source image is given. We selected

two "corresponding" slices from these volume data sets to conduct the experiment. Note

that even though the number of slices for the two data sets are the same, the slices may

not correspond to each other from an anatomical point of view. However, for the purposes

of illustration of our algorithm, this is not very critical. We use the corresponding slice of

the 3D segmentation of the source as our atlas-segmentation. The results of an application






70



















Figure 5-5: Corpus Callosum segmentation on a pair of corresponding slices from dis-
tinct subjects.

of our algorithm are organized as before in figure 5-5. Once again, as evident, the visual

quality of the segmentation and registration are very high.

Finally we present a 3D real data experiment. In this experiment, the input is a pair of 3D

brain scans with the segmentation of the hippocampus in one of the two images (labeled

the atlas image) being obtained using the well known PCA on the several training data

sets. Each data set contains 19 slices of size 256x256. The goal was then to automatically

find the hippocampus in the target image given the input. Figure 5-6 depicts the results

obtained for this image pair. From left to right, the first image shows the given (atlas)

hippocampus surface followed by one cross-section of this surface overlaid on the source

image slice; the third image shows the segmented hippocampus surface from the target

image using our algorithm and finally the cross-section of the segmented surface overlaid

on the target image slice. To validate the accuracy of the segmentation result, we

randomly sampled 120 points from the segmented surface and computed the average

distance from these points to the ground truth hand segmented hippocampal surface in the

target image. The hand segmentation was performed by an expert neuroanatomist. The

































Figure 5-6: Hippocampal segmentation using our algorithm on a pair of brain scans from
distinct subjects. (see text for details)

average and standard deviation of the error in the aforementioned distance in estimated

hippocampal shape are 0.8190 and 0.5121(in voxels) respectively, which is very accurate.















CHAPTER 6
CONCLUSIONS AND FUTURE WORK

6.1 Contributions of the Dissertation

We have introduced a variety of information theoretic measures and showed various

applications. The novel information measures we presented in this dissertation include

Entropy defined on distributions, Cumulative Residual Entropy (CRE)

Cross-Cumulative Residual Entropy (CCRE)

CDF based Kullback-Leiber (KL) divergence

CDF based Jensen-Shannon (JS) divergence

We demonstrated their applications to the following medical image analysis problems,

Non-rigid image registration.

Simultaneous groupwise point-sets registration and atlas construction.

Atlas based image segmentation.

Our contributions to each of these topics are summarized in the following sections.

6.2 Image and Point-sets Registration

6.2.1 Non-rigid Image Registration

For non-rigid image registration, we presented a novel way to register multi-modal

datasets based on the matching criterion called cross cumulative residual entropy(CCRE)

[84] to measure the similarity between two images. The matching measure is defined

based on a new information measure, namely cumulative residual entropy (CRE), which

is defined based on the probability distributions instead of probability densities, therefore

CCRE is valid for both discrete and continuous domain. Furthermore, CCRE also inherits

the robustness property of the CRE measure. In [84], we presented results of rigid and

affine registration under a variety of noise levels and showed significantly superior

performance over MI-based methods.










The Cross-CRE between two images to be registered is maximized over the space of

smooth and unknown non-rigid transformations, which is represented by a tri-cubic

BSplines placed on a regular gird. The analytic gradient of this matching measure is then

derived in this paper to achieve efficient and accurate non-rigid registration. It turns out

that the gradient of the CCRE has a similar formulation with the cost function, which

greatly saves memory space in the optimization process. The matching criterion is

optimized using Quasi-Newton method to recover the transformation parameters.

The key strengths of our proposed non-rigid registration scheme are demonstrated

through the registration of the synthetic as well as real data sets from multi-modality (MR

Tl and T2 weighted, MR & CT) imaging sources. It is showed that our CCRE not only

can accommodate images to be registered of varying contrast+brightness, but it is also

robust in the presence of noise. CCRE converges faster when compared with other

information theory-based registration methods. Finally we also showed that CCRE is well

suited for situations where the source and the target images have FOVs with large

non-overlapping regions (which is quite common in practice). Comparisons were made

between CCRE and traditional MI [34, 51], which was defined using the Shannon

entropy. All the experiments depicted significantly better performance of CCRE over the

MI-based methods currently used in literature.

Our future work will focus on extending the transformations model to the one that permits

the spatial adaptation of the transformation's compliance, which will allow us to reduce

the number of degrees of freedom in the overall transformation. Validation of non-rigid

registration on real data with the aid of segmentations and landmarks obtained manually

from a group of trained anatomists are the goals of our ongoing work.

6.2.2 Groupwise Point-sets Registration

We presented a novel and robust algorithm for the groupwise non-rigid registration of

multiple unlabeled point-sets with no bias toward any of the given point-sets. To quantify

the divergence between multiple probability distributions estimated from the given point










sets, we proposed several divergence measures, the first of which is the Jensen-Shannon

divergence. Since it lacks robustness, we develop a novel measure based on their

cumulative distribution functions that we dub as the CDF-JS divergence. The measure

parallels the well known Jensen-Shannon divergence (defined for probability density

functions) but is more regular than the JS divergence since its definition is based on CDFs

as opposed to density functions. As a consequence, CDF-JS is more immune to noise and

statistically more robust than the JS.

Our proposed methods do not require any knowledge of correspondence between the

input point-sets, and therefore these point-sets need not have the same cardinality. One

other salient feature of our proposed algorithms is that we get a probabilistic atlas as the

byproduct of the registration process. Our algorithm can be especially useful for creating

atlases of various shapes present in images as well as for simultaneously (rigidly or

non-rigidly) registering 3D range data sets without having to establish any

correspondence.

Our future work will focus on using maximum likelihood estimation (MLE) to

automatically determine weighting coefficients in the divergence measures and smoothing

term; We are also attempting to extend our techniques to diffeomorphic point-sets

matching.

6.3 Image Segmentation

In the part of image segmentation, we presented a novel variational formulation of the

joint (non-rigid) registration and segmentation problem which requires the solution to a

coupled set of nonlinear PDEs that are solved using efficient numerical schemes. Our

work is a departure from earlier methods in that we presented a unified variational

principle wherein non-rigid registration and segmentation are simultaneously achieved.

Unlike earlier methods presented in literature, a key feature of our algorithm is that it can

accommodate for image pairs having distinct intensity distributions. We presented several

examples (twenty) on synthetic and (three) real data sets along with quantitative accuracy







75


estimates of the registration in the synthetic data case. The accuracy as evident in these

experiments is quite satisfactory. Our future efforts will focus on adapting our

algorithm+software for the clinic use.
















REFERENCES


[1] C. E. Shannon, "A mathematical theory of communication," Bell System Technical
Journal, pp. 379-423 and 623-656, 1948.

[2] W. F. Sharpe, Investments. London: Prentice Hall, 1985.

[3] D. Salomon, Data Compression. New York: Springer, 1998.

[4] S. Kullback, Information Theory and Statistics. New York: Wiley, 1959.

[5] T. M. Cover and J. A. Thomas, Elements ofInformation Theory. New York: Wiley,
1991.

[6] G. Jumarie, Relative Information. New York: Springer, 1990.

[7] M. Rao, Y Chen, B. C. Vemuri, and F. Wang, "Cumulative residual entropy, a new
measure of information," IEEE Transactions on Information Theory, vol. 50, no. 6,
pp. 1220-1228, June 2004.

[8] M. Asadi and Y Zohrevand, "On the dynamic cumulative residual entropy,"
Unpublished Manuscript, 2006.

[9] H. Chui, L. Win, R. Schultz, J. Duncan, and A. Rangarajan, "A unified non-rigid
feature registration method for brain mapping," Medical Image Analysis, vol. 7,
no. 2, pp. 112-130, 2003.

[10] N. Paragios, M. Rousson, and V. Ramesh, "Non-rigid registration using distance
functions," Comput. Vis. Image Underst., vol. 89, no. 2-3, pp. 142-165, 2003.

[11] M. A. Audette, K. Siddiqi, F. P. Ferrie, and T. M. Peters, "An integrated
range-sensing, segmentation and registration framework for the characterization of
intra-surgical brain deformations in image-guided surgery," Comput. Vis. Image
Underst., vol. 89, no. 2-3, pp. 226-251, 2003.

[12] A. Leow, P. M. Thompson, H. Protas, and S.-C. Huang, "Brain warping with
implicit representations." in International Symposium on Biomedical Imaging, 2004,
pp. 603-606.

[13] B. Jian and B. C. Vemuri, "A robust algorithm for point set registration using
mixture of gaussians." in IEEE International Conference on Computer Vision, 2005,
pp. 1246-1251.










[14] F. Wang, B. C. Vemuri, A. Rangarajan, I. M. Schmalfuss, and S. J. Eisenschenk,
"Simultaneous nonrigid registration of multiple point sets and atlas construction," in
European Conference on Computer Vision, 2006, pp. 551-563.

[15] S. J. H. Guo, A. Rangarajan, "A new joint clustering and diffeomorphism estimation
algorithm for non-rigid shape matching," in IEEE Computer Vision andPattern
Recognition, 2004, pp. 16-22.

[16] M. Irani and P. Anandan, "Robust Multi-sensor Image Alignment," in International
Conference on Computer Vision, Bombay, India, 1998, pp. 959-965.

[17] J. Liu, B. C. Vemuri, and J. L. Marroquin, "Local frequency representations for
robust multimodal image registration," IEEE Transactions on Medical Imaging,
vol. 21, no. 5, pp. 462-469, 2002.

[18] M. Mellor and M. Brady, "Non-rigid multimodal image registration using local
phase," in Medical Image Computing and Computer-Assisted Intervention,
Saint-Malo, France, Sep 2004, pp. 789-796.

[19] B. Zitova and J. Flusser, "Image registration methods: a survey." Image Vision
Comput., vol. 21, no. 11, pp. 977-1000, 2003.

[20] J. Ruiz-Alzola, C.-F. Westin, S. K. Warfield, A. Nabavi, and R. Kikinis, "Nonrigid
registration of 3d scalar vector and tensor medical data," in ThirdInternational
Conference on Medical Image Computing and Computer-Assisted Intervention,
A. M. DiGioia and S. Delp, Eds., Pittsburgh, October 11-14 2000, pp. 541-550.

[21] L. Marroquin, B. Vemuri, S. Botello, F. Calderon, and A. Fernandez-Bouzas, "An
accurate and efficient bayesian method for automatic segmentation of brain mri," in
IEEE Transactions on Medical Imaging, 2002, pp. 934-945.

[22] B. C. Vemuri, J. Ye, Y Chen, and C. M. Leonard, "A level-set based approach to
image registration," in IEEE Workshop on Mathematical Methods in Biomedical
Image Analysis, 2000, pp. 86-93.

[23] P. Hellier, C. Barillot, E. Mmin, and P. Prez, "Hierarchical estimation of a dense
deformation field for 3d robust registration," IEEE Transactions on Medical
Imaging, vol. 20, no. 5, pp. 388-402, May 2001.

[24] R. Szeliski and J. Coughlan, "Spline-based image registration," Int. J Comput.
Vision, vol. 22, no. 3, pp. 199-218, March 1997.

[25] S. H. Lai and M. Fang, "Robust and efficient image alignment with spatially-varying
illumination models," in IEEE Conference on Computer Vision and Pattern
Recognition, 1999, pp. II: 167-172.

[26] A. Guimond, A. Roche, N. Ayache, and J. Menuier, "Three-Dimensional
Multimodal Brain Warping Using the Demons Algorithm and Adaptive Intensity










Corrections," IEEE Transactions on Medical Imaging, vol. 20, no. 1, pp. 58-69,
2001.

[27] J.-P. Thirion, "Image matching as a diffusion process: an analogy with maxwell's
demons," Medical Image Analysis, vol. 2, no. 3, pp. 243-260, 1998.

[28] A. Cuzol, P. Hellier, and E. Memin, "A novel parametric method for non-rigid image
registration," in Proc. Information Processing in Medical Imaging (IPMI'05), ser.
LNCS, G. Christensen and M. Sonka, Eds., no. 3565, Glenwood Springes,
Colorado, USA, July 2005, pp. 456-467.

[29] A. W. Toga and P. M. Thompson, "The role of image registration in brain mapping,"
Image Vision Comput., vol. 19, no. 1-2, pp. 3-24, 2001.

[30] E. D'Agostino, F. Maes, D. Vandermeulen, and P. Suetens, "Non-rigid
atlas-to-image registration by minimization of class-conditional image entropy." in
Medical Image Computing and Computer-Assisted Intervention, 2004, pp. 745-753.

[31] P. A. Viola and W. M. Wells, "Alignment by maximization of mutual information,"
in IEEE International Conference on Computer Vision, MIT, Cambridge, 1995.

[32] A. Collignon, F. Maes, D. Delaere, D. Vandermeulen, P. Suetens, and G. Marchal,
"Automated multimodality image registration based on information theory," Proc.
Information Processing in Medical Imaging, pp. 263-274, 1995.

[33] C. Studholme, D. Hill, and D. J. Hawkes, "Automated 3D registration of MR and
CT images in the head," Medical Image Analysis, vol. 1, no. 2, pp. 163-175, 1996.

[34] D. Mattes, D. R. Haynor, H. Vesselle, T. K. Lewellen, and W. Eubank, "Pet-ct image
registration in the chest using free-form deformations." IEEE Transactions on
Medical Imaging, vol. 22, no. 1, pp. 120-128, 2003.

[35] D. Rueckert, A. F. Frangi, and J. A. Schnabel, "Automatic construction of 3d
statistical deformation models of the brain using non-rigid registration." IEEE
Transactions on Medical Imaging, vol. 22, no. 8, pp. 1014-1025, 2003.

[36] G. Hermosillo, C. Chefd'hotel, and 0. Faugeras, "Variational methods for
multimodal image matching," Int. J Comput. Vision, vol. 50, no. 3, pp. 329-343,
2002.

[37] D. Rueckert, L. I. Sonoda, C. Hayes, D. L. G. Hill, M. O. Leach, and D. J. Hawkes,
"Nonrigid registration using free-form deformations: Application to breast mr
images," IEEE Transactions on Medical Imaging, vol. 18, no. 8, pp. 712-721,
August 1999.

[38] M. E. Leventon and W. E. L. Grimson, "Multimodal volume registration using joint
intensity distributions," in Medical Image Computing and Computer-Assisted
Intervention (MICCAI), Cambridge, MA, 1998, pp. 1057-1066.










[39] T. Gaens, F. Maes, D. Vandermeulen, and P. Suetens, "Non-rigid multimodal image
registration using mutual information," in Proc. Conference on Medical Image
Computing and Compter-Assisted Intervention (MICCAI), 1998, pp. 1099-1106.

[40] D. Loeckx, F. Maes, D. Vandermeulen, and P. Suetens, "Nonrigid image registration
using free-form deformations with a local rigidity constraint." in Medical Image
Computing and Computer-Assisted Intervention, 2004, pp. 639-646.

[41] G. K. Rohde, A. Aldroubi, and B. M. Dawant, "The adaptive bases algorithm for
intensity based nonrigid image registration." IEEE Transactions on Medical
Imaging, vol. 22, no. 11, pp. 1470-1479, 2003.

[42] V. Duay, P.-F. D'Haese, R. Li, and B. M. Dawant, "Non-rigid registration algorithm
with spatially varying stiffness properties." in International Symposium on
Biomedical Imaging, 2004, pp. 408-411.

[43] C. Guetter, C. Xu, F. Sauer, and J. Hornegger, "Learning based non-rigid
multi-modal image registration using kullback-leibler divergence." in Medical Image
Computing and Computer-Assisted Intervention, 2005, pp. 255-262.

[44] E. D'Agostino, F. Maes, D. Vandermeulen, and P. Suetens, "An information
theoretic approach for non-rigid image registration using voxel class probabilities,"
Medical Image Analysis, vol. 10, no. 3, pp. 413-431, 2006.

[45] C. Davatzikos, "Spatial transformation and registration of brain images using
elastically deformable models," Comput. Vis. Image Underst., vol. 66, no. 2, pp.
207-222, 1997.

[46] J. C. Gee, M. Reivich, and R. Bajcsy, "Elastically deforming 3d atlas to match
anatomical brain images," J Comput. Assist. Tomogr., vol. 17, no. 2, pp. 225-236,
1993.

[47] M. Bro-Nielsen and C. Gramkow, "Fast fluid registration of medical images," in
Proc. of the 4th International Conference on Visualization in Biomedical
Computing. London, UK: Springer-Verlag, 1996, pp. 267-276.

[48] G. E. Christensen, R. D. Rabbitt, and M. I. Miller, "Deformable templates using
large deformation kinematics," IEEE Transactions On Image Processing, vol. 5,
no. 10, pp. 1435-1447, October 1996.

[49] X. Geng, D. Kumar, and G. E. Christensen, "Transitive inverse-consistent manifold
registration." in Proc. Information Processing in Medical Imaging, 2005, pp.
468-479.

[50] A. Trouve, "Diffeomorphisms groups and pattern matching in image analysis," Int.
J Comput. Vision, vol. 28, no. 3, pp. 213-221, 1998.

[51] D. R. Forsey and R. H. Bartels, "Hierarchical b-spline refinement," Computer
Graphics, vol. 22, no. 4, pp. 205-212, 1988.










[52] C. Cocosco, V. Kollokian, R.-S. Kwan, and A. Evans, "Brainweb: online interface to
a 3-d mri simulated brain database," 1997, last accessed: July 2005. [Online].
Available: http://www.bic.mni.mcgill.ca/brainweb/

[53] P. Thevenaz and M. Unser, "Optimization of mutual information for multiresolution
image registration," IEEE Transactions on Image Processing, vol. 9, no. 12, pp.
2083-2099, December 2000.

[54] T. Chan and L. Vesse, "An active contour model without edges," in Intl. Conf on
Scale-space Theories in Computer Vision, 1999, pp. 266-277.

[55] N. Duta, A. K. Jain, and M.-P. Dubuisson-Jolly, "Automatic construction of 2d shape
models," IEEE Transactions Pattern Anal. Mach. Intell., vol. 23, no. 5, pp. 433-446,
2001.

[56] E. Klassen, A. Srivastava, W. Mio, and S. H. Joshi, "Analysis of planar shapes using
geodesic paths on shape spaces." IEEE Transactions Pattern Anal. Mach. Intell.,
vol. 26, no. 3, pp. 372-383, 2003.

[57] T. B. Sebastian, P. N. Klein, B. B. Kimia, and J. J. Crisco, "Constructing 2d curve
atlases," in IEEE Workshop on Mathematical Methods in Biomedical Image
Analysis, Washington, DC, USA, 2000, pp. 70-77.

[58] H. Tagare, "Shape-based nonrigid correspondence with application to heart motion
analysis." IEEE Transactions on Medical Imaging, vol. 18, no. 7, pp. 570-579, 1999.

[59] F. L. Bookstein, "Principal warps: Thin-plate splines and the decomposition of
deformations." IEEE Transactions Pattern Anal. Mach. Intell., vol. 11, no. 6, pp.
567-585, 1989.

[60] H. Chui, A. Rangarajan, J. Zhang, and C. M. Leonard, "Unsupervised learning of an
atlas from unlabeled point-sets." IEEE Transactions Pattern Anal. Mach. Intell.,
vol. 26, no. 2, pp. 160-172, 2004.

[61] S. Belongie, J. Malik, and J. Puzicha, "Shape matching and object recognition using
shape contexts," IEEE Transactions Pattern Anal. Mach. Intell., vol. 24, no. 4, pp.
509-522, 2002.

[62] H. Chui and A. Rangarajan, "A new algorithm for non-rigid point matching." in
IEEE Computer Vision and Pattern Recognition, 2000, pp. 2044-2051.

[63] H. Guo, A. Rangarajan, S. Joshi, and L. Younes, "Non-rigid registration of shapes
via diffeomorphic point matching." in International Symposium on Biomedical
Imaging, 2004, pp. 924-927.

[64] T. F. Cootes, C. J. Taylor, D. H. Cooper, and J. Graham, "Active shape models: their
training and application," Comput. Vis. Image Underst., vol. 61, no. 1, pp. 38-59,
1995.










[65] Y Wang and L. H. Staib, "Boundary finding with prior shape and smoothness
models," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 22,
no. 7, pp. 738-743, 2000.

[66] A. Hill, C. J. Taylor, and A. D. Brett, "A framework for automatic landmark
identification using a new method of nonrigid correspondence." IEEE Transactions
Pattern Anal. Mach. Intell., vol. 22, no. 3, pp. 241-251, 2000.

[67] Y Tsin and T. Kanade, "A correlation-based approach to robust point set
registration." in European Conference on Computer Vision, 2004, pp. 558-569.

[68] J. Glaunes, A. Trouve, and L. Younes, "Diffeomorphic matching of distributions: A
new approach for unlabelled point-sets and sub-manifolds matching." in IEEE
Computer Vision and Pattern Recognition, 2004, pp. 712-718.

[69] J. Lin, "Divergence measures based on the shannon entropy," IEEE Transactions
Information Theory, vol. 37, pp. 145-151, 1991.

[70] A. Hero, O. M. B. Ma, and J. Gorman, "Applications of entropic spanning graphs,"
IEEE Transactions Signal Processing, vol. 19, pp. 85-95, 2002.

[71] Y He, A. Ben-Hamza, and H. Krim, "A generalized divergence measure for robust
image registration," IEEE Transactions Signal Processing, vol. 51, pp. 1211-1220,
2003.

[72] D. M. Endres and J. E. Schindelin, "A new metric for probability distributions,"
IEEE Transactions Information Theory, vol. 49, pp. 1858-60, 2003.

[73] H. Chui and A. Rangarajan, "A new point matching algorithm for non-rigid
registration," Computer Vision and Image Understanding (CVIU), vol. 89, pp.
114-141, 2003.

[74] G. McLachlan and K. Basford, Mixture Model:Inference and Applications to
Clustering. New York: Marcel Dekker, 1988.

[75] A. L. Yuille, P. Stolorz, and J. Utans, "Statistical physics, mixtures of distributions,
and the em algorithm," Neural Comput., vol. 6, no. 2, pp. 334-340, 1994.

[76] R. Bansal, L. Staib, Z. Chen, A. Rangarajan, J. Knisely, R. Nath, and J. Duncan.,
"Entropy-based, multiple-portal-to-3d ct registration for prostate radiotherapy using
iteratively estimated segmentation," in Medical Image Computing and
Computer-Assisted Intervention, 1999, pp. 567-578.

[77] A. Yezzi, L. Zollei, and T. Kapur, "A variational framework for joint segmentation
and registration," in IEEE Workshop on Mathematical Methods in Biomedical Image
Analysis, 2001, pp. 388-400.

[78] P. Wyatt and J. Noble, "Mrf-map joint segmentation and registration," in Medical
Image Computing and Computer-Assisted Intervention, 2002, pp. 580-587.










[79] N. Paragios, M. Rousson, and V. Ramesh, "Knowledge-based registration &
segmentation of the left ventricle: A level set approach." in WACV, 2002, pp. 37-42.

[80] B. Fischl, D. Salat, E. Buena, and M. A. et.al., "Whole brain sementation:
Automated labeling of the neuroanatomical structures in the human brain," in
Neuron, vol. 33, 2002, pp. 341-355.

[81] S. Soatto and A. J. Yezzi, "Deformotion: Deforming motion, shape average and the
joint registration and segmentation of images," in European Conference on
Computer Vision, 2002, pp. 32-57.

[82] B. C. Vemuri, Y Chen, and Z. Wang, "Registration assisted image smoothing and
segmentation," in European Conference on Computer Vision, 2002, pp. 546-559.

[83] T. Zhang and D. Freedman, "Tracking objects using density matching and shape
priors." in IEEE International Conference on Computer Vision, 2003, pp.
1056-1062.

[84] F. Wang, B. C. Vemuri, M. Rao, and Y Chen, "A new & robust information theoretic
measure and its application to image alignment." in Proc. Information Processing in
Medical Imaging, 2003, pp. 388-400.















BIOGRAPHICAL SKETCH

Fei Wang was born in Yan Cheng, JiangSu, P. R. China. He received his Bachelor of

Science degree from the University of Science and Technology of China, P. R. China, in

2001. He earned his Master of Science and Doctor of Philosophy degree from the

University of Florida in December 2002 and August 2006 respectively. His research

interests include medical imaging, computer vision, pattern recognition, computer

graphics and shape modeling.