|
A RADIATION DOSIMETRY MODEL FOR
RADIOLABELED MONOCLONAL ANTIBODIES:
INDIUM-111 LABELED B72.3-GYK-DTPA
FOR COLORECTAL CANCER
By
LATRESIA ANN WILSON
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
1990
Copyright 1990
by
Latresia A. Wilson
ACKNOWLEDGEMENTS
First and foremost, I would like to thank my family for
their love and support throughout the years. I would
especially like to thank my uncle, Manuel Lewis Jr., for his
wisdom and support; my grandmother, Lula Lewis, for her
patience and understanding; my mother, Geraldine Solomon, for
her encouragement and support; my aunts, Evelyn Scott and
Marilyn Johnson, for believing in me; and my cousins for their
encouragement.
I would especially like to thank Dr. Genevieve Roessler
for believing and supporting me during the rough years. I am
very thankful to my advisor, Dr. William Properzio, for his
superb guidance and support throughout this study. I am also
grateful to my advisory committee, Dr. Emmett Bolch, Dr.
Walter Drane, Dr. David Hintenlang, and Evelyn Watson, for
their guidance and help in this project.
I would like to thank Evelyn Watson and the staff (Mike
Stabin, Audrey Schlafke-Stelson, Fanny Smith, and Stan Walls)
of the Radiopharmaceutical Internal Dosimetry Center of Oak
Ridge Associated Universities. I would like to thank Oak Ridge
SAssociated Universities for giving me the opportunity and
resources to learn from the best in my field. I would like to
especially thank Martha Kahl and Rana Yalcintes in the medical
iii
library of Oak Ridge Associated Universities for their
excellent support. I would like to thank the staff of the
Nuclear Medicine Department at the University of Tennessee
Hospital for their guidance, patience, and use of their
equipment. I would like to thank Phyllis Cotten, Carmine
Plott, James Stubbs, Pat Harp Family, and the Rose Foster
Family for making my stay in Tennessee an enjoyable one.
I would like to thank Dr. Steven Harwood, Michelle
Morrissey, Linda Zangara, Dr. Will Webster, Dr. Carroll and
Dave Laven and staff of the Nuclear Medicine Department at Bay
Pines Veterans Administrative Medical Center in Bay Pines,
Florida for providing the patient data, use of their
facilities, and financial and expert support in this project.
I would like to thank the people of the Department of
Environmental Engineering Sciences, Dr. Charles Roessler and
fellow graduate students for their encouragement and support.
I would like to also thank Dr. Libby Brateman for being there
to answer all my seemingly endless number of questions and for
providing support.
I would like to thank Dean Rodrick McDavis for his
continued support. And last, I would like to thank the
McKnight Foundation and the Florida Endowment Fund for Higher
Education, Dr. Israel Tribble and staff, for their financial
support for without which, this degree would never have been
undertaken.
TABLE OF CONTENTS
page
ACKNOWLEDGMENTS ............... ......................... iii
LIST OF TABLES ........................ ................ vii
LIST OF FIGURES .......................................... ix
ABSTRACT .......... ................ ..................... xi
CHAPTERS
1 INTRODUCTION ........................................ 1
2 MONOCLONAL ANTIBODIES .............................. 6
Immunoglobulin Structure ..................... 9
Variables Associated with Radioimmunoimaging
and Radioimmunotherapy ................. 14
Tumor Localization ..................... 17
Choice of Radiolabel ............... 18
Tumor Size Effect ................. 24
Fragment vs Whole Antibody ........ 25
Dose Administered Effect ........... 27
Labeling Method Effect ............. 28
Dose Administration Route .......... 29
Tumor Biology .... .................. 30
Other Factors .... .................. 31
3 RADIATION DOSIMETRY .... ........................... 34
MIRD Approach .................................. 36
"Traditional" Point Kernal Method ............ 38
Microdosimetry ................................. 41
4 MATERIALS AND METHODS ........ .................... 47
SPECT Model .................................... 49
Monte Carlo Model ........................... 49
Dosimetry Model ................................ 51
Single-Photon Emission Computed Tomography ... 51
SPECT Quantitation ............................. 53
Photon Attenuation ........................... 54
Photon Scatter .................................. 56
SPECT Camera System .......................... 57
Image Segmentation ......... ................... 58
Program SPECTDOSE .............................. 60
Subroutine THOLD .......... ................ 63
Subroutine CONTOUR ..................... 63
Subroutine OBJSELECT ................... 63
Subroutine CORGAN ............
Subroutine VOXFIL ............
Program ALGAMP ....................
Pixel and Slice Size Determination
Phantom Studies ...................
Phantom Study One ............
Phantom Study Two ............
Phantom Study Three ..........
Thermoluminescent Devices ....
Clinical Studies ..................
Patients .....................
Monoclonal Antibody ..........
Monoclonal Antibody Procedure
Blood Analyses ...............
HPLC Procedure ..........................
Image Analysis ..........................
5 RESULTS AND DISCUSSION ............................
Pixel and Slice Size Determination ...........
Phantom Study One ............................
Phantom Study Two ............................
Phantom Study Three ..........................
Clinical Study ...............................
6 SUMMARY AND CONCLUSIONS ...........................
REFERENCES .............................................
APPENDICES
A SAMPLE CALCULATIONS................................
B TLD CALIBRATION................................... ..
C SPECTDOSE PROGRAM .................................
BIOGRAPHICAL SKETCH ....................................
...........
...........
...........
...........
...........
...........
...........
...........
...........
...........
...........
154
157
158
181
67
69
69
72
73
74
76
80
84
85
85
86
86
88
88
89
90
91
91
97
112
129
133
140
LIST OF TABLES
Table
Table
Table
2-1
2-2
2-3
Table 2-4
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
Table
3-1
4-1
4-2
4-3
5-1
5-2
5-3
5-4
5-5
5-6
5-7
5-8
5-9
5-10
5-11
5-12
Properties of Human Immunoglobulins ... 11
Properties of Human IgG Subclasses . . . 12
Selected Radionuclides for Radioimmunodetection
and Radioimmunotherapy . . . .... .. 19
Radionuclides for Radioimmunodetection
and Radioimmunotherapy . . . . . 21
Sublethal Radiation Doses. . . . . . 44
Phantom Study One Acquisition Parameters .. .75
Phantom Study Two Acquisition Parameters .. .77
Phantom Study Two Experiments . . . . 79
Phantom Study One Threshold Determination .. 92
Phantom Study One Results . . . ... 94
Phantom Study Two Threshold Determination .98
Phantom Study Two Experiment One Results 100
Phantom Study Two Experiment Two Results .102
Phantom Study Two Experiment Three Results 104
Gaussian Prefilter Comparison: Actual versus
SPECT Measured Volume . . .. . . 107
Phantom Study Two Absorbed Dose Results . 110
Phantom Study Three TLD Measurements. . . 114
Phantom Study Three Geometric Factor Method
and MIRD Pamphlet No. 3 Results . . . 117
Phantom Study Three Dosimetry Model Results 118
Phantom Study Three Results . . . ... 120
vii
Table 5-13 Phantom Study Three Error Anaylsis. ... .121
Table 5-14 TLD Calibration Study Results ..... . 123
Table 5-15 Phantom Study Four TLD Measurements .... .127
Table 5-16 Phantom Study Four Results. .. . .. . 128
Table 5-17 Phantom Study Four Error Analysis . . .. .130
Table 5-18 Clinical Study Results . . ....... 131
viii
LIST OF FIGURES
Figure 1-1
Figure 2-1
Figure 2-2
Figure 2-3
Figure 2-4
Figure 4-1
Figure 4-2
Figure 4-3
Figure 4-4
Figure 4-5
Figure 4-6
Figure 4-7
Figure 4-8
Figure 4-9
Figure 4-10
Figure 4-11
Figure 4-12
Figure 4-13
Antibody Carriers for Diagnosis
and Therapy . . . . . . . . .
Monoclonal Antibody Production . . . .
HAT Mediated Hybridoma Production . . .
IgG Molecule . . . . . . . .
Enzymatic Digestion of IgG Molecule
into Fragments . . . . . . . .
Research Methodology. . . . . . .
Single-Photon Emission Computed Tomography
SPECT Model Flow Chart. . . . . .
SPECTDOSE Program Subroutine Flow Chart .
Illustration of Subroutine THOLD Object
Segementation .. . . . . . . .
Subroutine CONTOUR Object Segmentation
Subroutine CONTOUR Object Assignment . .
Illustration of Subroutine OBJSELECT
Selected Object Comparison . . . . .
ALGAMP Flow Chart . . . . . . .
Phantom Study Two Torso Phantom and Organ
Inserts . . . . . . . . .
Phantom Study Three Experiment One TLD
Location . . . . . . . . .
Phantom Study Three Experiment Two TLD
Location .. . . . . . . . .
B72.3 Linker Complex . . . . .
Figure 5-1 Phantom Study One:Actual versus SPECT
Measured Volume . . . . . . . 95
Figure 5-2 Phantom Study One:SPECT Measured versus
Actual Activity Concentration . . . . 96
Figure 5-3 Phantom Study Three TLD Experimental
Locations. . . . . . . . . 113
Figure 5-4 Phantom Study Four TLD Chip Packaging . 124
Figure 5-5 Phantom Study Four TLD Locations. . .. 126
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
A COMPUTERIZED RADIATION DOSIMETRY MODEL FOR
RADIOLABELED MONOCLONAL ANTIBODIES:
INDIUM-111 LABELED B72.3-GYK-DTPA
FOR COLORECTAL CANCER
By
LATRESIA ANN WILSON
DECEMBER 1990
Chairperson: William S. Properzio
Major Department: Environmental Engineering Sciences
A foundation was developed for a dosimetry
methodology that could be used to calculate absorbed doses
in target and nontarget tissues using uniformly and
nonuniformly distributed activity. In this methodology, a
dosimetry model was developed which consisted of three
independent models: 1) the SPECT Model, 2) the Monte Carlo
Model, and 3) the Dosimetry Model. The SPECT Model uses
Single-Photon Emission Computed Tomography (SPECT) images to
determine the volume and radioactive uptake. A computer
program was written to automatically read and analyze SPECT
images. This program uses an edge detection method to
determine the volume. Voxel elements within the identified
volume are used to calculate the activity concentrations.
The Monte Carlo Model uses a monte carlo simulation method
xi
and results of the SPECT Model to calculate the fraction of
photon energy deposited in target and nontarget tissues. The
Dosimetry Model combines the results of the SPECT and Monte
Carlo Models to determine the absorbed dose in target and
nontarget tissues.
Several phantom studies were conducted to verify the
ability of the Dosimetry Model to evaluate organ and tumor
uptake, sizes, and to calculate absorbed doses. Comparisons
were made between the Dosimetry Model, other calculational
methods (MIRDOSE2, Geometric Factor Method, MIRD Pamphlet
No. 3), and TLD measurements.
For diagnostic activity doses, the SPECT Model was
found to calculate organ volumes of the order of 1000 ml to
within fifteen percent of the actual volumes but it failed
to accurately calculate organ volumes of 200 ml or less.
No meaningful relationship was found between the actual
and SPECT measured activity concentrations.
The Dosimetry Model agreed within 12% when compared
with the Geometric Factor Method and the MIRD Pamphlet No.3
results using homogeneously and heterogeneously distributed
"'In. The TLD measurements were within 30% at most of the
other methods.
Results of the several phantom studies indicated the
Dosimetry Model was an appropriate methodology for
calculating absorbed doses for homogeneously distributed
activity. Further investigation is needed to determine the
xii
accuracy of the Dosimetry Model in the heterogeneously
distributed activity case.
The addition of photon attenuation and scatter
correction and nonpenetrating radiation transport is
pertinent to the accuracy of the dosimetry methodology.
xiii
CHAPTER 1
INTRODUCTION
In the United States, cancer is the second leading
cause of death with the number of annual deaths fast
approaching 400,000 (1). This value represents a little over
20% of all deaths. Women are more susceptible to cancer than
men and except for accidents, cancer kills more children
than any other illness (1). In England, cancer is the
leading cause of death in children 1-14 years of age (1).
Utilization of antibodies to fight cancer started as
early as 1946 when Pressman theorized that polyclonal
antibodies directed against antigens expressed on tumor
cells could be used to localize radionuclides in the tumor.
He believed that once the antibodies were bound to the
antigen-rich tumor site, the radioactivity could be detected
with a gamma scanning device or if the radionuclide
concentration in the tumor was sufficient, serve as local
radiation therapy. So, after a series of ingenious
experiments, he successfully demonstrated that immune
proteins could be used to target radioactivity to tumors in
2
living animals (2). Unfortunately at that time, it was
difficult to produce antibodies that would survive in
cultured media, thus limiting the ability to produce
sufficient amounts with the specificity needed for clinical
studies. This ultimately limited the further use of this
technology for many years to come.
In 1975, Kohler and Milstein introduced a new technique
called hybridization, which would allow for the production
of large quantities of identical (monoclonal) antibodies
(3). This technique made it possible for the methodologies
proposed by Pressman to be applied clinically. Kohler and
Milstein later went on to receive the Nobel Prize for their
contribution.
With the advent of the hybridization technique, there
was renewed interest in the use of radiolabeled antibodies
for tumor therapy. It is generally believed that monoclonal
antibodies attached to radiolabels for therapy
(radioimmunotherapy) may be effective in treating metastases
and small tumors, where surgery may not be feasible. This
new technique offers some ray of hope in the fight against
cancer.
Recent advances in biotechnology have given new hope to
achieving the ultimate goal of using monoclonal antibodies
for targeting radioactivity for the dual purpose of cancer
diagnosis and therapy (Figure 1-1). This potential has
Antibody
to p97
Label With
Small Amount of
Radioactivity
Patient With
Undisclosed Tumor
Melanoma
Tumor Therapy
Attach Anti-Tumor
Drugs or High Dose
Radioactivity
Patient With Tumor
Tumor
Destruction
Figure 1-1. Antibody Carriers for Diagnosis and Therapy"
Adapted from Reference 4
4
generated a significant amount of interest and growth in the
field of nuclear medicine over the past few years. This
growth, in turn, has generated many new problems and
questions. One of these problems, the radiation dosimetry of
using radiolabeled monoclonal antibodies, is the focus of
this research.
Current radiation dosimetry methods, which allow for
the calculation of absorbed doses for both target and
nontarget tissues, assume that the radiolabel's energy is
distributed uniformly throughout the target and nontarget
organ. This assumption is not valid in the case of
radioimmunotherapy, since it has been shown that
radiolabeled monoclonal antibodies distribute
heterogeneously throughout a given organ and on the tumor
cell (5). It is, therefore the objective of this research to
develop a foundation for a radiation dosimetry methodology
that could be utilized for radiolabeled monoclonal
antibodies; i.e., a methodology which would allow for the
calculation of absorbed doses in tissues with a
heterogeneous or homogeneous radioactivity distribution. A
computerized dosimetry model, which allows for the
calculation of absorbed doses to both target and nontarget
tissues after intravenous (IV) injection of Indium-111
labeled B72.3-GYK-DTPA monoclonal antibody directed against
colorectal cancer, will be proposed in this research.
Clinical applications and ease-of-use of this dosimetry
5
model will be emphasized. A comparison of the results from
this model with that of current dosimetry methods will be
made.
This dissertation is divided into five basic sections.
First, an overview of monoclonal antibodies and the factors
that affect their localization are presented. Second, there
is a discussion of the current radiation dosimetry methods
and their inadequacies for use with radiolabeled monoclonal
antibodies. Third, a discussion of the experimental methods,
computer models, and imaging techniques used in this study
are presented. Next, the computational results are presented
and analyzed. Finally, the results are summarized and
suggestions for future applications of this method are made.
CHAPTER 2
MONOCLONAL ANTIBODIES
Antibodies, or immunoglobulins, are proteins made by
many animal species as part of their specific response to
foreign substances (antigens). When antibody-antigen binding
occurs, this immunologic response usually results in the
destruction or elimination of the antigen.
Immunoglobulins are produced by the activity of the B
lymphocytes and possess specific binding regions that
recognize the shape of particular sites or determinants on
the surface of the antigen. An antigen may have several
determinants, or epitopes, each of which is capable of
stimulating one or more B lymphocytes. For this reason, an
antigenic challenge results in the production of a variety
of antibodies (6).
Early antibody production techniques employed the use
of animals, usually a mouse or rabbit, immunized with an
antigenic substance, to obtain antibodies, which were found
in the serum of the immunized animal. These antibodies were
7
polyspecific because they reacted with a wide variety of
antigenic binding sites.
Highly specific antibodies can be developed by
extracting individual lymphocytes and cloning them in tissue
culture; each clone would have the potential to manufacture
a single antibody species, a monoclonal antibody (Figure 2-
1). Unfortunately, normal antibody-producing cells do not
survive in culture media. It took Nobel laureates Kohler and
Milstein (3) to recognize that myeloma cells, which are
cancer cells that produce large amounts of identical but
nonspecific immunoglobulins, and which can survive in
cultures indefinitely, might be altered by the new
techniques of recombinant genetics to construct immortal
clones that secrete immunoglobulins.
Kohler and Milstein developed a method of producing
such monoclonal antibody strains by fusing the lymphocytes
from the spleen of an immunized mouse with mouse myeloma
cells, thus forming clones of hybrid cell lines, called
hybridomas (4). These cells are usually fused in
polyethylene glycol and result in clones that have the
specific-antibody characteristics of the lymphocytes and the
longevity of the myeloma cells. Additionally, pure hybridoma
cells are selectively grown in hypoxanthene-aminopterin-
thymidine (HAT) media since it supports neither the unfused
lymphocytes nor the myeloma cells. Once these hybridomas are
produced, they can be assayed for antibody activity and for
Antigen
Antigenic
?Determinant
Lymphocytes Myeloma Cells
o Fuse 0
Spleen / \
\Hybrid
00 0 Myeloma
SLymphocytes Cells
1 2 3 4 Clot
Antibody Antigen
Antiserumlonl
Polyclonal Antibodies Monclonal Antibodies
Figure 2-1. Monoclonal Antibody Production'
Adapted from Reference 4
ne
9
further selective cultivation (Figure 2-2). The reader is
referred to Reference 6 for an excellent review of the
techniques involved in the production, purification,
analysis, quality control, radiolabeling, and storage of
monoclonal antibodies.
Immunoqlobulin Structure
Immunoglobulins (Ig) are divided into five classes:
IgG, IgA, IgM, IgD, and IgE and can further be subdivided
(isotypes) on the basis of internal attributes (see Table 2-
1 and 2-2). IgM antibodies are often the first to appear
during immunization and IgE antibodies mediate
hypersensitivity reactions (8).
Immunoglobulins of all classes are composed of two
heavy (H) chains and two light (L) chains in their simplest
form. All classes share the same light chains and differ
solely in the structure of the heavy chains. The heavy
chains are attached to one another by means of one or more
disulfide bonds, and a light chain is attached to each heavy
chain by a disulfide bond (Figure 2-3). Isotypes differ
structurally in the number of disulfide bonds linking the
two heavy chains together, and they differ functionally in
their ability to fix complement and to interact with
effector cells such as macrophages and mast cells (Table 2-
2).
Cell culture
Myeloma Line
Fuse In -n
Polyethylene
Glycol
-----*----.
Spleen Cells Myeloma Cells
Hat Medium feJ0' *0 Select Hybrid Cells
Assay For Antibody
T -- Freeze
Clone @0
Assay For Antibody
Freeze
Redone Q ( c
P Analyze to Select Variants
Propagate -- -- Freeze
Desired Clones V Thaw
Gr Induce
Grow in
Mass Culture/
Antibody Antibody
Figure 2-2. HAT Mediated Hybridoma Production"
SAdapted from Reference 7
11
o0 In
> ac
S0 0
0 r-M a- 4 0
0 >
0 4i 0 0
m I -O O O
SC
SO D 0 0
0* r_ c O 0 U o (n
S-- o < 14 ( N
03 rl0
3) O 0 0 0
C *C) O u 0 .
U CCV O 0 C
-O (V-0 0 0- H N 0
H S 0 4 0 4J 0 N rO 0S
S o L (0 N x-- i e (n C M-
0 N
00 C4 1 00 tTO 0
w o 0 2( V o
) > .09 f-
Oz C V 0V O O 000+
V0 O (M x 4 U in > rl
a0- 0 C 0 cc0 En -4 0
-30 Q 0 u -C In m ) -04
-, qO 4.H 0 3: :r V (V C) cM wf t>) 4--)
41 .q V -4 4J 41 :3 41 m mT
IO c- U UI (V (- 0( I (
41 I m q c C Ln 41i a c
> 33 0 0 m C X C f. C)
20 1 > C.l) C) E a 0 M a <
-( OO o Lj u < O o oin )O O
20'C 0) -M .4i in 0 U < 0a U
00 a 0 0 041 00 0 \ m
"- Ul z z 0 zO f l r U< CE
0
PC
+ 0
S-4
I + I U
<3
+
+ +
4-)
o3
o
(O OC
~NW V
+ +1
+ '
-4
0) 0
aZ "
41J
0
a)CO
0 C
O~O.
--) 0 i
4-1
a)
+ +
U]
U 0 o
)
C
**
S 3
*-. C 0
4 -'-4
a) .0
0 4-'
=>, C
c.~-' a)
OC *- E
O- c O
S3- a)
SC E'
; ~! U
0
O
t
0
S00
c0
C3
cO
0
CC
-'-4 TO
.)
a)
1-I
cr n
(- c< n r
n
co
p1-
0 0
+
0
-4-
0 .0
+
C
+
C
O
a)
+
c
0
0
u
0
C
4)
0
U
M .
0
Ul E
C l
-4 X 0)
.0 i
4
o <
>4
-41
-4i
.0.
--4
C4-i
'4 U)
TO *'-4
H2N NH2
H2N NH2
L L
sH CS
HOOC COOH
s s
s s
H H
HOOC COOH
Figure 2-3. IgG Molecule*
Adapted from Reference 8
14
The antibody-specific sites of the immunoglobulins are
situated near the amino-terminal (NH2) end of each of the
four chains (Figure 2-3), and it is in this region (variable
region) that the greatest variability in amino acid sequence
occurs from immunoglobulin to immunoglobulin (9). Constant
amino acid sequences are found in the carboxyterminal (COOH)
regions (constant region) of the immunoglobulin chains
(Figure 2-3). The two variable regions bind to specific
antigenic sites and the constant region (Fc) interacts with
the host immune system.
Since the Fc region of the antibody is most likely to
trigger allergic responses, fragmentation has been used to
remove this portion from the antibody molecule. Pepsin, a
proteolytic enzyme, cleaves off most of the Fc region, which
leaves two Fab fragments bound together in a divalent
structure known as the F(ab')2 fragment (Figure 2-4). The
enzyme papain breaks the immunoglobulin into two monovalent
Fab fragments and an intact Fc fragment.
Variables Associated with Radioimmunoimaging
and Radioimmunotherapy
A number of variables that must be considered before
diagnostic and therapeutic applications of monoclonal
antibodies can be utilized. Generally monoclonal antibodies
alone are not effective in tumor destruction (10). This has
been attributed in part to the heterogeneous distribution of
tumor-associated antigens on cell surfaces, which leads to
VARIABLE
\ REGION
LIGHT CHAIN
HEAVY CHAIN
PEPSIN/ IgG
F(ab')2
+ FRAGMENTS OF Fc
PAPAIN
Fab I.I Fab
Figure 2-4. Enzymatic Digestion of IgG Molecule into
Fragments*
Adapted from Reference 9
16
variable attachments of the antibodies to the different
tumor cells; more antibodies are attached to those cells
which have significant amounts of the antigen on their
surfaces but none to other tumor cells that are devoid of
the specific antigen and are therefore allowed to
proliferate (10). Since the monoclonal antibody alone is not
cytotoxic, it usually acts as a carrier of a more cytotoxic
radionuclide or toxin. This introduces a number of more
complicating factors and variables, which include the
combined physical, chemical, and biological properties of
the antibody and radiolabel. The following is a summary of
the variables directly linked with the production of the
radiolabeled tumor-associated antibody for imaging and
therapy (11):
1) Physical properties of radionuclides
a) Physical half-life
b) Gamma energies and abundances
c) Photon yield per absorbed radiation dose
d) Parent-daughter relationship-stable decay
products
e) Ratio of penetrating to nonpenetrating
components
f) Particle radiation (3-,9+,IC, and Auger
electrons)
g) Production mode (availability)
2) Chemical properties
a) Stability of radionuclide-protein bond
b) Specific activity-number of labels per molecule
obtainable
c) Retention of immunological activity versus
specific activity
d) Addition of nonradioactive carrier-metal ion
contamination
e) Sample pH
17
3) Biodistribution and biological half-life
a) Route of administration and activity of initial
dose
b) Vascularity: Blood flow and interstitial fluid
space
c) Uptake of protein-bound form of the isotope
d) Plasma and whole body clearance
e) Relative size of tumor model
f) Size of animal or human model
g) Cell proliferation
h) Capillary and cell permeability
i) Presence of inflammation
4) Target-nontarget time-dependent ratio: dose to
tumor, whole body, and other sensitive organs
5) Immunological purity of the antibody and its
relative specificity
6) Characteristics of imaging system with respect to
the radiolabel properties
7) Marketability, availability, convenience
This list, which is by no means all-inclusive, is
complicated by the fact that each variable seems to be
related to a number of the other variables.
Tumor Localization
The localization of radiolabeled antibodies at tumor
sites is dependent on a number of factors as reported by
several investigators (12-27). These include the tumor size,
radiolabeling method, choice of radiolabel, type of antibody
(whole vs fragment), route of administration, tumor biology
(blood flow, vascular permeability etc.), and the dose
administered. However, tumor uptake is ultimately dependent
upon its antigen content (13).
Choice of Radiolabel
The choice of a radiolabel is dependent upon its
intended use: diagnostic or therapeutic applications. For
diagnostic applications, one is more concerned with the
sensitivity and specificity of the test with the least
radiation dose. This is obtained by the use of radionuclides
with a low equilibrium absorbed dose constant. In therapy,
the objective is to attain the highest differential
radiation dose, which requires the use of radionuclides with
a high equilibrium absorbed dose constant. The goal of both
applications is to attain the highest radiation dose factor
for the target site in comparison to the normal tissue (28).
Ideally, radionuclides which are particularly suited
for imaging with radiolabeled antibodies should be
characterized by 1) physical half-life of 6 hr to 8 days, 2)
gamma energy range of 80-240 keV, 3) high single energy
gamma abundance per decay, 4) small abundances and low-
energy particulate radiation, and 5) reasonable
radiolabeling chemical properties and stability (11).
Similarly, radionuclides used for therapy should have
complementary properties to the antibody-bound radionuclides
used in imaging. However, their decay should be
characterized with a large component of particulate
radiation with little or no accompanying gamma radiation
such that a high localized dose may be delivered (11). Table
2-3 lists the various radionuclides that meet the required
Table 2-3. Selected Radionuclides for Radioimmunodetectiont
Nuclide Half-life Primary Decay Characteristics
9Tc 6 h IT (99%); 6 = 141 keV (89%)
123 13 h EC (100%); 6 = 159 keV (83%)
11In 68 h EC (100%); 6 = 171 keV (88%)
6 = 245 keV (94%)
131I1 193.2 h '" (100%); 6 = 364 keV
97Ru 69 h EC (100%); 6 = 216 keV (86%)
67Cu 62 h (100%); 6 = 91 keV (7%)
6 = 93 keV (17%)
6 = 184 keV (47%)
Selected Radionuclides for Radioimmunotherapyt
Nuclide Half-life Primary Decay Characteristics
131I1 193.2 h p'(100%); 0.608 MeV (86%)
6 = 364 keV (82%)
9Y 64 h P-(100%); 2.29 MeV (100%)
67Cu 62 h p'(100%); 6 = 91 keV (7%)
6 = 93 keV (17%)
6 = 184 keV (49%)
212Bi 1 h a (36%)
3' (64%) 212Po(0.3 g- sec T,,,
a = 8.78 MeV)
21At 7.2 h a (41%); 5.9 MeV (41%)
EC (59%)
125I 144.5 h EC (100%);6 = 35 keV
x-rays = 27 keV
t Adapted from Reference 29
Potentially useful in therapy as well
SPotentially useful in imaging also
20
specifications for diagnostic and therapeutic applications
of monoclonal antibodies. Table 2-4 lists the advantages and
disadvantages of the use of the radionuclides found in Table
2-3.
Technetium-99m, "'In, and 131I are examples of
radionuclides that are currently under extensive use in
medical imaging (29-34). They have the advantage of
availability, well-known chemistry, and optimal half-life
and gamma decay energy. Unfortunately, 131I suffers from
dehalogenation in vivo, which allows for nonspecific uptake
of free iodine in sites other than the tumor sites,
especially in the thyroid, liver and spleen (30,35). This
makes identifying tumors in these organs by imaging nearly
impossible. Iodine-131 also delivers a high radiation dose
to normal tissues due to its long half-life and medium gamma
energy (36). Indium-lll in vivo metabolism is relatively
unknown, although it has been shown to have good affinity
once in the tumor, but if it comes off the antibody, it will
relocate to the liver, spleen, and bone marrow (30).
Technetium-99m has a chemistry problem; i.e., it is
difficult to obtain a stable bond between it and the
antibody. Childs and Hnatowich (37) found increased
stability when "CTc was coupled directly to the chelate DTPA
(diethylenetriaminepentacetic acid). Rhodes et al. (38) used
a pretinning method to successfully label ""c directly to
antibody fragments, which showed increased stability against
21
Table 2-4. Advantages and Disadvantages of Selected
Radionuclides for Radioimmunodetectiont
Nuclide Advantages Disadvantages
"Tc Availability Short T.
Decay energy Chemistry problem
13I Decay energy Availability
Iodine chemistry Cost ($20/mCi)
Short T,
"'In Decay energy In vivo metabolism
Optimal T.
Chelation chemistry
Availability
Iodine chemistry
Optimal T2
Decay energy
In vivo de-iodination
97Ru Chelation chemistry In vivo metabolism
Availability
Decay Energy
67Cu Optimal T' Decay Energy
Advantages and Disadvantages of Selected Radionuclides for
Radioimmunotherapyt
Nuclide Advantages Disadvantages
131I Availability Long tissue path
Imaging
Cost
Y 90Sr-Generator Chemistry problems
Pure p decay In vivo metabolism?
67Cu Imaging In vivo metabolism?
212Bi High LET decay Short T,
High LET decay
High LET decay
Unknown chemistry
Short T
Unknown chemistry
Must be in nucleus to
kill tumor
Adapted from Reference 29
1311
1251
I Adapted from
Reference 29
22
transchelation. Recently, Goldenberg and associates (39)
have reported in vivo retentions of 98% immunoreactivity in
patients using anti-CEA murine monoclonal antibody (IMMU-4)
Fab' labeled directly with "9Tc. Chen and colleagues (40)
also have reported good results using 9"Tc labeled
antibodies in the confirmation of diagnosis of uveal
melanoma.
Several alpha- and beta-emitting radionuclides have
potential for radioimmunotherapy as seen in Table 2-3.
Iodine-131 has been most commonly used and is currently
being utilized in human clinical studies (35,41,42).
However, the choice of 1311 has not been because it is the
optimum for radioimmunotherapy; two-thirds of its absorbed
dose equivalent is due to penetrating radiation, which
usually escapes the primary tumors and their metastases (4).
The Auger electrons of iodine-125 may be effective for
therapy when used in conjunction with antibodies that are
internalized rather than remaining on the cell surface (9).
The appeal of alpha-particles for radioimmunotherapy is
their short range (-50-90 gm) and high linear energy
transfer (LET) (-80 keV/gm), which produces extreme
cytotoxicity. An alpha-particle traversing the diameter of a
10 im nucleus deposits an energy of 800 keV, equivalent to
an absorbed dose of approximately 0.25 Gy (4). Potential
alpha-emitting radionuclides for radioimmunotherapy are
astatine-211 and bismuth-212 (Table 2-3). Experimental
23
trials with 21At-conjugated antibodies on a murine lymphoma
system are in progress by Harrison (43) and Vaughan (44).
Perhaps the half-life of bismuth-212 is too short (60.6 min)
to fully capitalize on the longer antibody retention in the
tumor, although Macklis found it to be highly cytotoxic to
the murine Thy 1.2" EL-4 tumor cell line (45). More recently
Simonson et al. (46) showed 212Bi to be also cytotoxic to the
LS174T cell line. Few suitable alpha-sources are available
because most alpha-emitters are heavy elements (A > 82)
which decay to unstable daughters. The recoil alphas
produced in the decay of these daughters rupture the
radionuclide-antibody bond, which allows the daughter
product to diffuse away from the tumor (5).
Yttrium-90 offers another possibility for use in
radioimmunotherapy and has the advantage in that it is a
pure beta emitter and is easily available by production from
a strontium-90 generator (Table 2-4). Unfortunately, it has
no gamma emissions to allow for useful biokinetic studies in
the patient and, once detached from the antibody, it
deposits in the bone in sufficient quantities to give a high
radiation dose to the marrow. Yttrium-90 is currently under
investigation by several groups (24,47-52). Sally DeNardo
and colleagues (51,53) found copper-67 to be one of the most
promising radionuclides for radioimmunotherapy because of
its short half-life, abundance of beta particles, and the
presence of 93 and 184 keV gamma emissions.
24
A potential radionuclide for radioimmunotherapy,
palladium-109, a predominately beta-emitting radionuclide
(EMx= 1 MeV; half-life= 13.4 h) that is available carrier-
free, was investigated by Fawwaz et al. (54), who labeled
it to an antimelanoma monoclonal antibody. Unfortunately,
they found that at least 60% of the radiolabeled antibody
preparation failed to bind to melanoma cells. They believed
this was partly the result of inactivation of the antibody
during purification, storage, or radiolabeling and/or the
presence of carrier 1P08d in the '09Pd preparation. A new
radionuclide for radioimmunotherapy, rhenium-186 (1.07 MeV
maximum beta, 9% abundant 137 keV gamma), is being
extensively evaluated in patients by Schroff and associates
(55). Preliminary findings indicate it to have similar in
vivo properties to 9Tc and is very stable in vivo.
Tumor Size Effect
Several investigators have found that tumor uptake of
the radiolabeled monoclonal antibody is inversely related to
the tumor size; i.e., the per gram uptake of monoclonal
antibodies decreases as the tumor size increases (12-
15,56,57). Pimm and Baldwin (14) have found a multitude of
parameters that could potentially account for this
relationship. These include changes in blood flow, degree of
necrosis, levels of cellular and intratumor or extravascular
antigen, and the presence of circulating tumor-derived
25
antigen. However, this relationship could not be duplicated
by Cohen et al.(16). In fact, their findings contradicted
those by other investigators, in that they found the total
tumor uptake increased with increasing tumor size. No
satisfactory explanation has been offered to explain the
differences between the findings. Pedley et al. (57) found
that for tumor weights greater than 100 mg a strong positive
correlation exists between absolute uptake and tumor weight
but found a poor correlation for smaller tumors. They thus
concluded that specific uptake was inversely proportional to
tumor size regardless of the antibody.
At present, although still controversial, one may
conclude that the relationship between the tumor size and
the antibody uptake is an inverse one.
Fragment versus Whole Antibody
Antibody fragments (Figure 2-4) reach their maximum
accumulation faster and clear from the body faster than
whole antibodies (9,17-19,29,53). However, whole antibodies
remain in the tumor longer to achieve higher concentrations.
Thus, the choice of antibody type depends on the
application. Radioimmunoimaging would benefit most from the
use of fragments, because of their early maximum
accumulation and faster clearance, which results in a lower
background (nonspecific uptake) level. Radioimmunotherapy
would benefit most from the use of whole antibodies because
26
the cytotoxic effect could be delivered over a longer period
of time.
The difference in tumor localization between the
fragments and whole antibody has been attributed to the
smaller weight of the fragments (55,000 daltons and 110,000
daltons for IgG Fab and F(ab')2 respectively) compared to
that of the whole IgG antibody (160,000 daltons), which
allows them to transverse the intravascular and
extravascular space much more quickly (53). This effect also
may be the result of differences in the valency of the
antibodies (9). Since Fab fragments are monovalent, their
bonds to cell-bound antigens are weaker than those of the
divalent whole antibodies and because of this, they shed the
tumor and are rapidly cleared from the body via the kidneys
(9). F(ab')2 fragments, on the other hand, are divalent,
but demonstrate similar kinetics to the Fab fragments.
Ballou et al. (19) compared IgM F(ab')2, fragments to whole
IgM antibodies. The weight of the IgM F(ab')2 fragment was
130,000 daltons, which is not considerably less than that of
a whole IgG antibody (160,000 daltons). However, the IgM
F(ab')Z did weigh considerably less than the whole IgM
antibody, which weighed 900,000 daltons. The F(ab')2, showed
a 1.6-fold faster whole body clearance and reached its
maximum uptake earlier than that of the whole IgM antibody.
However, its total uptake was lower than the whole IgM
antibody. Ballou suggests that this may be caused in part by
27
differences in metabolism between the whole antibody and
fragment and also possible changes in the antigen-binding of
the fragments resulting from low pH digestion.
The choice of antibody type will depend upon its
application. For imaging, fragments will most likely be
used. The best choice seems to be F(ab')2 because it remains
in the blood longer than Fab fragments. This results from
its larger molecular weight, which reduces its loss through
the kidneys (58). For therapy, whole antibodies will
probably be used. Perhaps F(ab')2 fragments will prove
superior in all cases, because they offer the advantages of
fragments and the lack of immunogenicity of whole
antibodies.
Dose Administered Effect
Eger et al. (20) found dose-dependent kinetics in 12
human patients with melanoma. They found that as the amount
of injected antibody increased, the plasma half-life
increased, which eventually resulted in a higher tumor
uptake. They also found that the radioactivity levels in the
spleen and marrow decreased as the amount of antibody
increased. This dose dependent effect was also seen by
Hnatowich et al. (21). Pedley et al. (57) also studied the
effect of tumor weight on uptake with escalating amounts of
antibody. They found that there was decreased uptake with
escalating amounts of antibody in small tumors. This effect
28
was thought to be the result of steric hindrance in the
small tumors, even though the rate of diffusion into the
tumor may have increased.
Labeling Method Effect
The method used to attach the radionuclide to the
antibody will affect the antibody's localization. If the
method is inefficient, in that the radiolabel detaches from
the antibody in vivo or if the radiolabel's radiation
destroys or alters the properties of the antibody, all hope
of tumor localization is lost and radioimmunotherapy is
rendered useless.
Since most suitable radiolabels for therapy are metals
(Table 2-3), early methods of antibody labeling attempted to
attach them directly to the antibody. This proved to be
highly unstable and the radiolabel detached from the
antibody in vivo (9). However, nonmetals, such as iodine,
are currently being attached directly to the antibody by the
lodogen or Chloramine-T methods (22). These elements also
suffer from instabilities and tend to dehalogenate in vivo
(18). The latest methods employ the use of a coupling agent,
usually a chelate, to attach the metallic radiolabel to the
antibody (9,21,23,24). The most widely used chelate is
diethylenetriaminepentaacetic acid (DTPA). The antibody is
attached to the DTPA which, in turn, is attached to the
radiolabel. The bonds formed with the DTPA are much stronger
29
than those of the direct-attachment method (23); thus the
chelate-coupled antibodies are much more stable in vivo
(53). Another advantage to using a chelate such as DTPA is
that many different chelate substitution levels on the
antibody can be achieved by straightforward manipulation of
the relative amounts of reactants or time of reaction with
the antibody (53). Other chelates have also been used and
their effects on the antibody biodistribution are
continually being investigated (59-62).
Dose Administration Route
The site where the antibody is administered affects not
only how fast the antibody reaches the tumor, but also how
much eventually localizes in the tumor (17,25). Obviously,
if one is interested in localization in the lymphatic
system, intralymphatical administration will prove superior
to the other routes (25). If there are ascites in the
peritoneal cavity, intraperitoneal administration would
prove superior over the other routes. Hnatowich et al. (47)
concluded that the use of intraperitoneal rather than
intravenous administration may be important in the
application of yttrium-90 because it probably offers a means
of reducing radiation exposures to the bone marrow and the
critical organ without reducing exposure to the tumor within
the peritoneum. Larson (17) found that the concentration of
radiolabeled antibodies in human tumors is tenfold less
30
after administration via the intravenous route than after
injection either subcutaneously, intralymphatically, or
intraperitoneally.
Tumor Biology
Tumors grow radially from a central group of cells;
therefore as the tumor enlarges, the dividing cells form a
shell around a relatively hypoxic core. When these cells
outgrow their blood supply, they die and form a necrotic
central nest containing some viable cells that are highly
resistant to radiation (26). Blood flow in this situation is
low which makes delivery of the radiolabeled antibody to the
tumor very difficult. Studies by Gullino and Grantham found
that the average value of blood supply to tumors was 0.14
0.01 ml per hour per mg of nitrogen and the blood supply was
independent of the host (27). Solid tumors were also found
to be angiogenesis dependent by Folkman (63). The
radiolabeled antibody must reach the tumor through
circulation, crossing the capillary wall and diffusing
throughout the interstitial fluid to reach the tumor cells.
The rate of diffusion across these barriers is slowed by the
large size of the antibody molecule (64). This diffusion
rate has, according to Winchell (65), an 18 to 24 hour half-
life. Diffusion of the labeled antibody from the vascular
compartment into the tumor is caused by the concentration
gradient between the blood and the tumor (26). The higher
31
the concentration of radiolabeled antibody in the blood
compared with the tumor, the higher the diffusion rate will
be. Leichner et al.(66) found that external-beam
irradiation increased the permeability of tumor vascularity,
which resulted in increased tumor uptake of radiolabeled
antiferritin.
Other Factors
Other factors may influence the localization of
radiolabeled monoclonal antibodies in the tumor, such as the
amount of circulating antigens in the vascular system and
the metabolism and catabolism of the antibody in vivo.
Circulating antigens in the blood may combine with
circulating labeled antibodies. This complex could be
phagocytized by the reticuloendothelial system to reduce the
number of labeled antibodies that reach the tumor site
(67,68). Pimm and Baldwin (69) found that the average rate
of catabolism of 125I-labeled-IgG, anti-CEA monoclonal
antibody was 1.64% of the administered dose per gram per 24
hours and that this rate was higher for tumor bearing mice
as opposed to nontumor bearing mice. They also concluded
that tumor localization by the labeled antibody is a dynamic
process with simultaneous localization and degradation.
Gatenby et al. (70) have shown that the level of oxygen in
the tumor or tumor region also affects the antibody
localization. They found that tumors or tumor regions with a
32
mean oxygen pressure of 16 mm Hg or less had lower antibody
uptake, even when the presence of antigen was confirmed by
biopsy. This suggests that physiological factors other than
antigen expression may affect antibody uptake. In the past
few years a factor that has become increasingly important
because of the increase in the number of human studies is
the development of human anti-mouse antibodies (HAMAs). The
body, in response to the injection of murine antibodies,
produces antibodies (HAMAs) against the murine antibody
which it recognizes as being foreign. This response can be
detected within one week of exposure to the mouse protein
and is maximal within 2-3 weeks of exposure (71). The timing
and detection of the HAMAs are influenced by the dose of the
mouse antibody administered (71). HAMA clearly alters the
pharmacokinetics of subsequent murine antibody infusions
and, depending on the dose of the murine antibody and titer
of HAMA, can interfere with radioimaging and therapy and can
lead to toxicity because of the immune complexes and their
redistribution (71). Scannon (72) found a rapid clearance of
the infused murine antibodies from the blood which limited
further administration. It has been suggested that antibody
fragments be used instead of whole antibodies, because they
lack the Fc region (Figure 2-4), which most likely triggers
the allergic response. Other approaches to reducing HAMA
include the use of chimeric (human-mouse) monoclonal
antibodies, chemical alteration of the murine Fc portion,
33
ultrapheresis of human plasma to remove Ig, and chemical
suppression of the immune response.
From the above discussion, one may conclude that the
localization of radiolabeled monoclonal antibodies at the
tumor site is dependent upon a number of seemingly
interrelated variables which may vary from patient to
patient. Larson (25) also concluded that tumor localization
varied considerably from patient to patient.
CHAPTER 3
RADIATION DOSIMETRY
Before radioimmunotherapy can be implemented
successfully, it is necessary to know the amount of
radiation absorbed by the target and nontarget tissues. This
has proved to be difficult because of the lack of
appropriate methods to measure the amount of radiation
absorbed in the tissues; i.e., the absorbed dose, which was
deposited there by radiolabeled antibodies. The lack of an
appropriate method for correlating non-uniform dose with
effect has also hindered the efforts to assess the absorbed
dose. Assessment of the absorbed dose is complicated by the
large number of interrelated factors that affect the
localization of the radiolabeled antibodies in vivo (see
Chapter 2). These factors require that the calculated
absorbed dose be patient-specific. The current methods used
to calculate absorbed dose are based on assumptions that are
not valid when radiolabeled antibodies result in a
nonuniform distribution are used.
35
Current dosimetry methods can be divided on the basis
of the approach taken to calculate the absorbed-dose. There
are three basic approaches (73): (a) those that utilize the
conventional Medical Internal Radiation Dose Committee
(MIRD) formulation, a macroscopic approach which was
developed to cater mainly to diagnostic situations usually
involving gamma emitters and whole organs rather than
discrete targets (74); (b) those that utilize Berger's point
kernels, a semi-microdosimetry approach which considers
small size targets but not very low energy emissions at the
level of cell dimensions (75); and (c) those that take a
microdosimetric approach, which investigates doses from
short range emissions located near the cell surface or cell
nucleus (76).
Since absorbed dose is defined as the amount of energy
deposited per unit mass by ionizing radiation at the site of
interest (77), dosimetry calculations require a knowledge of
the physical properties of the radiolabel, length of time
the radioactivity remains in the various sites, and the
distribution of the radionuclide to the various sites in the
body (28,78,79). The physical properties of the radiolabel
are perhaps the easiest to determine accurately and will be
known in detail if conventional labels are used (80). The
residence time and spatial distribution of the radiolabeled
antibody in vivo are not usually known and must be
determined prior to radioimmmunotherapy. These parameters
36
are usually determined by sequential, timed quantitative
imaging. Several investigators (10,11,78) have suggested
that a diagnostic study, as such, be performed prior to
radioimmunotherapy. In this diagnostic study, the antibody
would be labeled with a small amount of the therapeutic
agent or a short-lived isotope of the therapeutic agent in
an effort to reduce the hazard to the patient (69).
Medical Internal Radiation Dose Committee (MIRD) Formulation
The MIRD Formula (74) is the most widely accepted
method for calculating radiation absorbed dose from
internally deposited radionuclides. This method was
recommended by the Medical Internal Radiation Committee of
the Society of Nuclear Medicine in 1968 and was later
adopted for standard use by the International Commission on
Radiation Units (ICRU) (81) in 1971. MIRD is based on the
dose rate equation developed by Loevinger et al. (82) in
1956 and is expressed as
Dose rate= K x activity in target x energy of x absorbed 1)
to target mass of target emission fraction
where K is a constant which depends on the units used.
Several assumptions are made in this approach, the most
important in the present context being that in applying this
method to humans, an anthropomorphic phantom is used, which
in calculating the absorbed dose, does not take into account
37
the nonuniformity of the activity distribution. Thus, source
homogeneity is assumed throughout the organs. Humm (4) gives
two reasons why this assumption may not necessarily be valid
in the case of radioimmunotherapy. First, the irregular
nature of the tumor vasculature will result in a complex
pattern of diffusion gradients guiding the antibodies
through the tumor. Second, immunohistochemical studies with
antibodies have shown that the tumor antigens may not be
expressed uniformly throughout the whole tumor cell
population.
For the application of radioimmunotherapy, and assuming
that the activity remaining in the body after organ uptake
is distributed uniformly, the mean dose to the target
(tumor) is the sum of three components: a) the dose from
nonpenetrating radiations (radiation pathlength is smaller
than the dimensions of the organ in which it resides)
emitted within the target organ, b) the dose from
penetrating radiations (radiation pathlength is greater than
the dimensions of the organ in which it resides) emitted
within the target organ, and c) the dose from penetrating
radiations emitted by the activity in the rest of the body
(41). The absorbed fraction for nonpenetrating radiations is
assumed to be unity; i.e., all the energy emitted by the
source organ is absorbed in the source organ. With this in
mind, one proceeds to calculate the various parameters of
the MIRD equation for each component. The effective half-
38
time can be calculated from exposure rate measurements and
activity measurements of the blood and urine as a function
of time. Decay constants are calculated from a least-squares
fit of the time-varied target organ count rates.
Compartmental modeling is often employed to calculate the
cumulated activity, decay constants, and the other
parameters needed for the MIRD equation. Tumor and critical
organ volumes are determined from Computed Tomography (CT)
or Single-Photon Emission Computed Tomography (SPECT)
images.
For conventionally employed radionuclides such as 131I,
32, or 9Y and for targets greater than a centimeter in
diameter, the MIRD method holds quite reasonably (73).
Berqer's Point Kernels
This method is based on Berger's Point Kernels for
calculating the absorbed dose from beta-rays (75). If the
medium is assumed to be uniform and unbounded, the beta-ray
dosimetry problem can be divided into two separate parts: a)
determination of the distribution of absorbed dose around a
point isotropic source, which is often referred to as a
point kernel, and b) appropriate integration over the point
kernel weighted by the source density to obtain absorbed-
dose distributions for extended sources (83). Part a)
contains all the physical aspects of the problem and part b)
is entirely geometric. Using the principles of
39
superposition, the absorbed dose from one source element can
be added independently to the contribution from another
source element. Thus, a distributed radionuclide source can
be considered as a collection of independently acting
isotropic sources (83).
The beta-ray dose rate is expressed in the form
Rp = 1.38E-05 Eg P A Gy d"' 2)
where Rg is the beta-ray dose rate in the tissue, E is the
average beta-ray energy per disintegration in Mev, is the
isotropic specific absorbed fraction, and A is the Activity
of the radionuclide in Bq.
Since the dose rate is proportional to the average
concentration, the total beta-particle dose is obtained by
integrating the concentration over the time the tissue is
exposed to the beta particles:
D (t) f RO(t)dt 1.38E-05 E t A(t)dt Gy
where E is in Mev, t is in days, and A is in Bq. Thus,
whenever the average activity A(t), is known as a function
of time, the absorbed dose can be computed by integration.
Loevinger et al. (82) states that for purposes of
dosimetry, the tissue distribution can be represented by a
stable system of separate compartments interconnected by
first-order reactions. First-order reactions imply that the
40
total amount of radioactivity leaving a given compartment
per unit time is proportional to the amount present. The
rate of change of the total radioactivity in the ith
compartment is described by the following differential
equation:
n
dgi = -pq ki0qo + Z(kjiqj kijqi) 4)
dt j=1
qi = total radioactivity in the ith compartment
kij = constant fraction of the radioactivity in the
ith compartment transferred to the jth
compartment, per unit time
ki = constant fraction of the radioactivity in the
ith compartment transferred to outside the
system (excretion)
Ip = radioactive decay constant
n = number of compartments
The first term on the right represents the loss due to
radioactive decay, the next term the loss from the system by
excretion or fixation, the first term inside the bracket
represents the contribution of the (n-1) other compartments
to the ith compartment, and the second term inside the
bracket represents the loss from the ith compartment to the
(n-1) other compartments. Integrating this equation for qi
(pCi) and then dividing by the mass (g) of compartment i
gives the average concentration of radioactivity in the ith
compartment:
C,(t) = 3.7 x 10' qi(t)/mi Bq g-1 5)
Thus, it is now possible, using Equation (5), to calculate
the total beta-particle dose from Equation (3). Spencer (84)
41
showed the applicability of this method in
radioimmunotherapy.
A whole range of electron energies from 10 keV to over
1000 keV, as well as tumor sizes from single cells to 107
cells (each having a millimeter diameter), can be
encompassed with this approach.
Microdosimetry
Microdosimetry is most applicable for evaluating dose-
effect relationships. It uses the microscopic distribution
of radiation interactions with biological systems to explain
the effects of radiation on the system (76). In some
instances, the distribution of specific energy in small
targets, individual tracks, or even individual energy
absorption events such as single ionizations may be needed
to obtain meaningful dose-effect relationships.
Microdosimetry takes into account the statistical aspects of
the particle tract structure, energy distribution patterns,
and radionuclide distribution within tissues and provides a
means for determining the number and frequency of cells
irradiated, the probability densities in specific energy,
and the average dose delivered to cells of interest (85).
Charged-particle radiation interacts with atomic electrons
of the matter through which it passes, and ionization and/or
excitation energy is imparted with each interaction. The
charge and mass of the particle, its initial energy, and the
42
matter through which it travels determine the pattern of
energy loss, the distance traveled, and the direction taken
by the particle. Ionizations and excitations are produced
when the energy is transferred from the particle to the
medium.
The basic quantity that describes the energy imparted
to matter is the absorbed dose, which actually is a mean
value. By definition, the absorbed dose D is the quotient of
de by dm, where de is the mean energy imparted by ionizing
radiation to matter of mass dm (86):
D = de/dm 6)
The specific energy, z, a stochastic quantity with units
similar to absorbed dose, is defined as the quotient of e by
m, where e is the energy imparted by ionizing radiation to
matter of mass m (86):
z = e/m 7)
The mean absorbed dose in a volume is equal to the mean
specific energy z, in the volume:
D= 8)
The ratio e/m is highly dependent upon target size. As the
target size gets smaller and smaller, the variations in the
local dose becomes increasingly greater, and the average
dose value becomes less and less indicative of the complete
43
dose distribution (85). Thus, for very small target sites,
the concept of absorbed dose becomes increasingly abstract,
and the dose is better represented by a distribution of
doses in "specific energy". For a given value of target size
mass, this distribution is called the "probability density
in specific energy" and is denoted by f(z). The probability
that the specific energy received by a target site lies in
the infinitesimal range dz containing the value z is f(z)dz.
Methods for calculating the probability densities in
specific energy can be divided into four steps (85). The
first step involves characterizing the geometrical
relationship between the radioactive source distribution and
the target sites. Second, the density in specific energy
must be determined for a target at any distance from the
radioactive source and with all possible angles of
intersection considered. Third, the probability that a point
source exists at any given distance from the target must be
determined from the spatial distribution of sources. And
fourth, the densities from all point sources are convolved
using Fourier transforms to construct a new specific energy
density for the target population.
The product of a microdosimetry calculation is a
statistical distribution of doses to small sites from which
an average dose could be determined. The precise
relationship between the specific energy density (average
dose) and the resulting biological effects is not known;
44
therefore, the results from this approach are not directly
applicable to the rather different conditions found in
radioimmunotherapy. However, several investigators (85,87-
91) have proceeded to utilize this method for radiolabeled
antibody dosimetric calculations.
Critical Organs
The maximum radiation doses that radiosensitive organs
can tolerate and still continue to function adequately to
support life are listed in Table 3-1 (10). With current
systemic approaches to therapy, bone marrow toxicity has
been the dose-limiting side-effect (92). However, as shown
by in vivo radiolabeled antibody biodistribution studies,
the dose-limiting organ is most likely to be the liver or
kidney.
Table 3-1. Sublethal Radiation Dosest
Organ System Dose (Gy)
Bone Marrow < 2
Intestinal Mucosa < 7
Kidney < 15
Liver < 25
t Adapted from Reference 10
Leichner et al. (41), using the MIRD methodology, found that
the 131I radiation dose for four patients ranged from four to
10 Gy for the liver and from 1.1 to 2.2 Gy for total-body
45
irradiation. Vaughan et al. (93), using Berger's Point
Kernels, found that a tumor dose of two Gy in one week with
131I was associated with a whole-body dose of 17 Gy. Bigler
et al. (87) utilized a microdosimetric approach to calculate
the mean dose to the red marrow for a number of different
radiolabels. The mean dose ranged from 1.6 Gy with "As to
17 Gy with 131I for a large cell size.
In recent years with the advent of new materials
technology, Griffith et al.(94-95) and Wessels (96) have
developed a method for the direct measurement of absorbed
radiation dose through the use of teflon-imbedded, CaSO4:Dy
thermoluminescent dosimeters (TLD)*, which have been
modified to fit inside a 20-gauge needle. The TLDs are
directly implanted into the tissue of interest and are
subsequently recovered for read-out. They measured an
absorbed dose of 8.1 Gy for the 1311 labeled B72.3 colorectal
carcinoma mouse system and 17.4 Gy for the 131I labeled LYM-1
Raji B-cell lymphoma mouse system, which correlated well
with autoradiography measurements (95). This method is not
appropriate for human dosimetry studies because of patient
discomfort and tissue trauma.
In order for radioimmunotherapy to be successful, the
radiation dose deposited in the tumor and other critical
organs must be known accurately. Current dosimetric methods
do not adequately address the unique features proposed by
Teledyne, Inc., NJ.
46
the use of radiolabeled antibodies for the calculation of
absorbed dose. Therefore, new methods must be created. As
more clinical information using radiolabeled antibodies
becomes available, a better method may be defined, which can
be compared to direct measurements.
CHAPTER 4
MATERIALS AND METHODS
In this research, a foundation for a dosimetry
methodology to determine the absorbed dose in both target
and nontarget tissues using uniformly and nonuniformly
distributed activity has been developed. The calculation of
absorbed dose can be divided into two parts: 1) the
determination of the radionuclide concentration, and 2) the
determination of the amount of energy deposited in the
tissues of interest. This new dosimetry methodology uses
Single-Photon Emission Computed Tomography (SPECT) to
determine the radioactive uptake in the tissues and a Monte
Carlo method to determine the amount of energy deposited in
the tissues.
The research method utilized in this research is shown
in Figure 4-1. In this figure, the research method is
divided into three models: 1) the SPECT Model, 2) the Monte
Carlo Model, and 3) the Dosimetry Model. Results from the
SPECT and Monte Carlo Models are utilized in the Dosimetry
Model.
Figure 4-1. Research Methodology
49
SPECT Model
The SPECT Model employs the use of Single-Photon Computed
Tomography, a diagnostic imaging technique, to determine the
volume and radioactive uptake in the target and nontarget
tissues following injection of radiolabeled monoclonal
antibodies. A computer program, SPECTDOSE, was written to
calculate both target and nontarget tissue volumes and
radioactive uptake. SPECTDOSE uses edge detection and
contour tracing algorithms to determine the volume of the
various organs and tissues of interest. The SPECT image is
divided into several three-dimensional arrays of a
preselected size and number. Sixty-four arrays composed of
64 x 64 elements (pixels) are utilized in this research.
Each element of the array represents an image volume (voxel)
at a specified location. Each voxel contains an integer
value derived from the measured activity in the imaged
object. The total number of voxels and their location, image
intensity per voxel, and organ volume (total number of
voxels at a specified location) are determined in this
model. Results of this model are used in the Monte Carlo
Model.
Monte Carlo Model
This model uses a monte carlo method to calculate the
fraction of photon energy deposited per unit mass of target
and nontarget tissues (specific absorbed fraction). A monte
50
carlo computer program was obtained from Oak Ridge National
Laboratory, Oak Ridge, Tennessee (97). This program, called
ALGAMP, is a photon transport code which accurately
simulates the physical phenomena of the photon by the use of
the statistical nature of radioactivity. In ALGAMP, the
human body and organs are represented by a set of
mathematical equations known collectively as the Cristy
Parametized Phantom (98). The radioactive distribution
within each organ is assumed to be homogeneous in the Cristy
Parametized Phantom. ALGAMP was modified for use in this
research by the deletion of the Cristy Parametized Phantom
and the addition of a method which permits the direct use of
the voxel information created by the SPECT Model. The voxel
information generated by the SPECT Model defines the organ
volumes and locations of interest. In the dose calculation
each voxel value represents the heterogeneous radioactivity
distribution found in the organs following the use of
radiolabeled monoclonal antibodies. By use of the SPECT
image voxel information and the monte carlo simulation
method, the amount of photon energy deposited per tissue
mass, specific absorbed fractions, can be determined for
each organ volume and voxel. The specific absorbed fractions
are utilized in the Dosimetry Model.
51
Dosimetry Model
The Dosimetry Model uses the results of the SPECT and Monte
Carlo Models to determine the absorbed dose to both the
target and nontarget tissues. Voxel matrix values of the
tissue volumes determined in the SPECT Model are utilized in
the Monte Carlo Model to determine the specific absorbed
fractions in the tissues of interest. The Dosimetry Model
combines the specific absorbed fractions with the organs'
radioactive uptake determined in the SPECT Model to
calculate the absorbed dose. The absorbed dose is determined
for both the organ and organ voxels; i.e, the absorbed dose
can be calculated for each organ voxel also. The Dosimetry
Model retains the concepts of the MIRD Method in addition to
accounting for the heterogeneous distribution of
radioactivity exhibited in the organs and organ voxels
following the injection of radiolabeled monoclonal
antibodies into humans.
Single-Photon Emission Computed Tomography
Single-Photon Emission Computed Tomography (SPECT) is a
diagnostic imaging technique utilized in nuclear medicine,
in which, the differences in radioactive distribution of
internally administered radionuclides are exploited (99). In
SPECT, the detector, a gamma camera, rotates around the
patient while acquiring data (photon detection) (Figure 4-
2). With the use of a computer and several complicated
Single-Head SPECT
Gamma Camera
Dual-Head SPECT
Gamma Camera
Single-Head SPECT Unit
Multi-Detector Head
SPECT Unit
Figure 4-2. Single-Photon Emission Computed Tomography
Adapted from Reference 100
53
algorithms, the data is reprojected (reconstructed) into a
transverse section image (slice) of the activity
distribution. Basically, SPECT maps the three dimensional
concentration of a radionuclide by measuring the angular
distributions, or projections, of gamma ray intensities
emitted within the body. SPECT is also capable of
eliminating overlying and underlying source activities and
offers the potential for quantitating the radioactive uptake
in the patient (100)
SPECT Quantitation
SPECT quantitation of radionuclide activities in the
human body is affected by several physical and instrumental
factors including absorption attenuation of photons in the
patient, Compton scattered events, the system's finite
spatial resolution, and object size, finite number of
detected events, partial volume effects, the
radiopharmaceutical biokinetics, and patient and/or organ
motion. Other instrumentation factors such as calibration of
the center-of-rotation, sampling, and detector
nonuniformities will affect the SPECT measurement process
(100,101,102). Several of the major factors that affect
quantitation with SPECT systems are as follows (100):
1) Physical Factors:
a) Characteristic energy of the emitted photons
b) Radiation decay as a function of time
54
c) Attenuation of gamma photons within the
patient
d) Inclusion of scattered photons within pulse
height window
2) Anatomical/Physiological Factors:
a) Source size and location within the body
b) Patient and/or organ motion
c) Biokinetical behavior of radiopharmaceutical
within the body
3) SPECT System Factors:
a) Camera/collimator energy and spatial
resolutions
b) Detection efficiency
c) Changes in collimator geometric response with
distance from the collimator surface
d) Sensitivity variations across the camera
surface
e) Camera electronic variations, ADC errors, and
gantry mechanical variations with time and/or
position
f) Characteristics of reconstruction process
such as shape of filter function, linear and
angular sampling interval values, accuracy of
attenuation, nonuniformities, and scatter
compensation methods and accuracy of edge-
detection methods
Their relative importance depends on the type of
quantitative information desired and the biokinetic
properties of the radiopharmaceutical. The determination of
radionuclide concentration as a function of time for small
volume elements (voxels) within the body is affected most by
the factors listed above.
Photon Attenuation
The determination of the radionuclide concentration as
a function of time in the voxel elements is affected by the
55
absorption attenuation and scattering of photons. The effect
of attenuation results in a decrease in the measured gamma
ray intensity. There is self-attenuation in the source organ
and also attenuation in the surrounding body tissues. Most
attenuation compensation methods assume that the attenuation
coefficient, the fraction of the gamma-ray beam attenuated
per unit thickness of absorber (103), is constant. Although
this will provide a less accurate compensation within
regions where the value of the attenuation coefficient is
variable, it is the method utilized in this research. Other
attenuation compensation methods, those which do not assume
a constant attenuation coefficient, can be divided into
three classes: 1) Preprocessing Methods, 2) Intrinsic
Compensation Methods, and 3) Postprocessing Methods (100).
Preprocessing methods attempt to correct the projection
data prior to image reconstruction. These methods are
relatively easy to implement, however, they tend to generate
streak artifacts in the presence of noise. This method was
not used in this research because the antibody SPECT images
were very noisy.
Intrinsic compensation methods integrate attenuation
correction directly in the reconstruction algorithm. An
attenuation map is measured (by using a transmission source)
or assumed as part of the reconstruction algorithm. These
methods require the use of large computers and are time
56
consuming, thus preventing the use of this method in this
dissertation.
Postprocessing methods apply attenuation correction
after the image reconstruction has completed. This approach
is used most often in commercial SPECT systems and requires
the measurement or estimation of the patient's body contour.
In the human studies undertaken in this research, the
patient's body contour was not retained, which precluded the
use of this method in this research. The area of attenuation
compensation in SPECT is currently undergoing extensive
analysis and the reader is further directed to a number of
reports on this subject (104-119).
Photon Scatter
Compton scattering events degrade the image contrast
resulting in a major source of error in the quantification
of radionuclide concentrations. Scattered photons can
contribute as much as 50% of the total collected events in
SPECT (120). The use of a sodium iodine-thallium doped
detector in SPECT systems results in the inclusion of both
scattered and nonscattered photons in the photopeak energy
window. Several approaches have been attempted to compensate
for the scattered radiation, but none at this point have
proven to be of substantial value (100,101,102,121-130). No
scatter correction method was utilized in this research.
57
SPECT Camera System
SPECT was performed with a digital rotating gamma
camera* with a medium energy collimator and a 20% peak
energy window. A rotation of 360', 128 projections (2.81
apart), and a study of 26 minutes (12 s view'1) was used.
Data was acquired on a computer" in the 64 x 64 x 16 bit
mode (131). After acquisition, the raw image data is reduced
from 16 bits to 8 bits using the system software**. After
which, the data is prefiltered using a Gaussian filter of
the 24th order and a frequency cutoff of 0.20 (131). This
data is reconstructed using the high resolution
reconstruction algorithm, which is an iterative
reconstruction method (132,133), with one iteration and a
dampening factor of 0.5 (134). The dampening factor
indicates the level of contribution by the error image to
the production of the iterative transverse slices (135).
Sixty-four transaxial slices, one pixel thick, are created.
The size of the elemental voxel is one pixel in the x and y
axis (transaxial plane) and in the z direction (parallel to
the axis of rotation). One pixel was determined to be equal
to 6.9 mm in the patient studies.
Technicare Omega 500, Technicare Corporation,
Cleveland, OH 44139
ADAC DPS-3300, ADAC Laboratories, San Jose, CA 95138
ADAC Laboratories Version 4 System Software, San
Jose, CA 95138
58
Image Segmentation
Prior to determining the organ volumes, the SPECT image
must be segmented into the respective organs. Image
segmentation is the process of subdividing an image into its
constituent parts or objects. Segmentation algorithms are
generally based on two properties of the image gray-level
values: discontinuity and similarity. The gray-level value
is an integer that represents the image intensity. In the
discontinuity category, the image is partitioned on the
basis of abrupt changes in the gray level. Detection of
points, lines, and edges are of principle interest in this
category. In the similarity category, the image is divided
on the basis gray level similarities. Approaches to the
similarity category include thresholding and region growing
(136). Several segmentation or edge detection methods were
attempted prior to the selection of the Threshold
Segmentation Method in this research.
The Gradient Method, an approach that looks for
discontinuity, was attempted first. It is assumed in this
method that the regions of interest are homogeneous so that
the transition between two regions can be determined on the
basis of gray-level discontinuities alone. A local
derivative operator is determined, whereby the magnitude of
the first derivative indicates the presence of an edge and
the sign of the second derivative determines where the edge
pixel lies; the background or object side (137). Since the
59
first and second derivatives must be determined for each
image pixel, this method is computationally intensive.
A second approach, Histogram Segmentation Method, was
also attempted. This technique creates a histogram of the
gray-level contents of an image. The image is subdivided
into its constituents by use of the peaks and valleys in the
histogram, which represent the image object and background
regions respectively. Division between objects is difficult
when a deep valley or steep peak is not present (138). In
the presence of image noise, differentiation between peaks
and valleys is futile. Because of noisy SPECT images this
method was not used in this research.
The last approach attempted and used in this research
is the Threshold Segmentation Method. This technique
segments on the basis of gray-level similarity. A threshold
value is applied to the image, whereby any image pixel's
gray-level value that is greater than the threshold value is
considered to be a part of the object and any pixel with a
gray-level less than the threshold value is apart of the
background. Since the threshold depends only on each
pixel's gray-level, it is called global (139). This method
was selected because of its easy implementation, small
computation requirements, and excellent results when used
with noisy images.
60
Program SPECTDOSE
The Program SPECTDOSE was developed in this research to
calculate the necessary parameters proposed by the SPECT
Model; i.e., organ volumes and radioactive concentrations
(Figure 4-3). This program is written in Fortran-77 for a
VAX/VMS Operating System. The Program SPECTDOSE is divided
into a number subroutines (Figure 4-4). Before the SPECT
image could be utilized, its data format or the way the
image data was written to the file had to be determined. The
data format for the SPECT images was obtained with a promise
of confidentiality from the ADAC Corporation (140). The
reconstructed SPECT image data is stored in each voxel as
hexadecimal (base-16) numbers. The main program reads the
hexadecimal numbers into a logical array, where the values
(image count) are scaled between 0-255 intensity levels
(gray-levels) and read into an integer array. The resulting
reconstructed image data is represented as an interger which
has a value between 1 and 256. The image count can be
corrected for attenuation and radioactive decay at this
point by entering the appropriate linear attenuation
coefficient, radionulide half-life, and time of decay values
into the program. The image threshold value is entered and
the subroutine THOLD is called to segment the image into its
constituent objects. This process is repeated for each image
slice.
Figure 4-3. SPECT Model Flow Chart
Figure 4-4. SPECTDOSE Program Subroutine Flow Chart
Subroutine THOLD
The subroutine THOLD segments the image into various objects
using the Threshold Segmentation method (139) (Figure 4-5).
The objects are separated from the background pixels by
comparing their intensity values with a global threshold
value; all pixels with an intensity value higher than the
threshold belong to the object. The subroutine CONTOUR is
called to extract the objects from the segmented image.
Subroutine CONTOUR
The subroutine CONTOUR extracts the objects from the
segmented image (Figure 4-6). The extracted object's
boundary is traced and the resulting object is stored in a
binary file called OBJECT#.DAT. This process is repeated for
all objects in the segmented image. Each object file is
assigned a consecutive identification number; i.e.,
Objectl.dat, Object2.dat etc. (Figure 4-7). The extracted
object's characteristics, which include the number of
voxels, total count, maximum and minimum indices, volume,
and area, are written to the file, OBJVAL.DAT.
Subroutine OBJSELECT
The subroutine OBJSELECT integrates the extracted objects of
each slice into a single object; i.e., organ. The extracted
object that best represents the shape of the organ of
interest is determined. This object's, the selected object,
Objects are separated from the background in the
SPECT image using the Threshold Segementation
Method, which is implemented in Subroutine THOLD.
Figure 4-5. Illustration of Subroutine THOLD Object
Segmentation
w I
3
1 2
3
Segemented objects are extracted and separated into
separate files by Subroutine CONTOUR
Figure 4-6. Subroutine CONTOUR Object Segmentation
1 2
3
SLICE 1
OBJECT
1
2
3
4
5
SLICE 2
SLICE N
- K
Segmented objects from each image slice is extracted and
separated into separate files and assigned file names
in consecutive order by Subroutine CONTOUR
Figure 4-7. Subroutine CONTOUR Object Assignment
[7:[ I
-aa
67
identification number is entered into the program. The
selected object is then compared with the rest of the
objects (Figure 4-8). If ninety percent of the object's
voxels are the same for each slice as the selected object's
voxels, the object is considered to be apart of the organ.
This process is repeated for all organs. The organ's voxel
indices, count, name, identification number, and volume are
stored in the file, VOXEL.DAT.
Subroutine CORGAN
The subroutine CORGAN creates an organ given its voxel
indices and identification parameters. Each pixel is
assigned an integer value which will represent the organ
volume desired. The number of image slices included in the
organ is also assigned. The organ created is used as a
photon reflector or sink; i.e., it either scatters or
absorbs the incident photons, and represents the areas of
the body not included in the SPECT image. If the whole body
is to be included in the SPECTDOSE Program, it is necessary
to create those areas of the body not seen in the SPECT
images due to the limited field-of-view of the SPECT camera
and the lack of availability of whole body SPECT images.
These areas are created using this subroutine.
Select Object 2
for Comparison:
OBJECT
SELECTED
OBJECT
K-1
Selected objects are compared to the remaining objects
for possible inclusion into one larger object (organ)
Figure 4-8. Illustration of Subroutine OBJSELECT Selected
Object Comparison
Subroutine VOXFIL
Once the organ files have been created; i.e., each organ's
voxel information has been stored in a VOXEL.DAT file, the
subroutine VOXFIL assembles the organs (each VOXEL.DAT file)
into one larger file called VOXPHAN.DAT. The number of
organs; i.e., the number of VOXEL.DAT files, is entered into
the program. This file contains the voxel indices, image
counts, weighting factors, identification numbers, and the
total number of voxels for all organs. The VOXPHAN.DAT file
is read directly by the program ALGAMP.
Program ALGAMP
The program ALGAMP is a point energy gamma-ray monte
carlo radiation transport code for calculating specific
absorbed fractions of energy and absorbed dose data from
internal and external sources (97). This program is written
in Fortran and was developed at Oak Ridge National
Laboratory in Oak Ridge, Tennessee (97). A flow chart of
this program can be seen in Figure 4-9. This program is
composed of 30 or more subroutines and initially utilized
the organs of the parametized phantom model from the Cristy
Phantom Series (98). In this Series, the organs were
represented by mathematical equations of various geometrical
shapes, such as, spheres and cylinders. The equations were
confined to a small number of ALGAMP subroutines (GEOM,SUM1,
SUM2, and RESULT), which hasten the modification process.
Figure 4-9. ALGAMP Flow Chart
71
The Cristy Phantom Series was not used in this research
because it assumes a homogeneous source distribution, which
is not valid in the case when radiolabeled monoclonal
antibodies are used; the organs were not patient specific;
and its inability to represent diseased organs, which are
most often found in nuclear medicine patients.
Since each human is uniquely different, it was the
desire of this research to make this new dosimetry method
patient specific; i.e., the organs of the imaged subject are
utilized in the calculation. This can be achieved by using
the actual SPECT image to define the organ volumes and
radioactive uptake. Each organ's voxel information was
determined and compiled into the file VOXPHAN.DAT by the
SPECT Model. The VOXPHAN.DAT file is read by ALGAMP. Each
voxel inherently, at the level of the camera system's
resolution, accounts for the heterogeneous source
distribution exhibited in the organs at that level following
the uptake of the radiolabled monoclonal antibodies. If the
activity is not distributed heterogeneously at the level of
the camera system's resolution, which is approximately one
half centimeter in this research, the voxels will reflect
this and the activity will be assumed to be homogeneously
distributed. For this homogeneous case, no additional
modifications to ALGAMP would be needed. Several ALGAMP
subroutines (INPUT, SOURCE, SEARCH, GEOM, SUM1, SUM2,
72
RESULT, and RANPOS) were modified to accommodate the
inclusion of the voxel information.
For sources distributed in energy, the cumulative
distribution function (cdf) for the source energy spectrum
is used. A detailed cross-section table is generated for the
source energies of interest. Each photon is weighted by a
weighting factor which describes the probability of photons
existing in a given voxel. The photon weighting factor is
computed for each voxel by dividing the voxel image count by
the average voxel count for a given organ. The source photon
location is chosen by randomly sampling the voxel locations
in the VOXPHAN.DAT file. Photon collisions are scored by
determining the voxel location given the photon direction
coordinates. Scoring is tallied for each voxel and organ.
Pixel and Slice Size Determination
In order to determine the organ volumes, it is
necessary to determine the size of each image pixel and
slice in physical dimensions. Since these values are
dependent upon the camera system's electronics, they must be
determined after each camera adjustment or change. A pixel
and slice size determination study was conducted prior to
the patient and phantom studies and in the case when the
camera system's electronics were changed. Two line sources
(small tubes containing ~"Tc) of known length and distance
apart are imaged in the planar (static) mode. The line
73
sources are imaged in both the parallel and perpendicular
positions relative to the camera system's axis-of-rotation
(AOR) to detect changes in the x and y planes, which would
indicate the camera system was working improperly. The
system software was used to return the number of pixels in a
line drawn between the two line source centers in the planar
imaged. The image pixel size in centimeters per pixel equals
the distance between the two line sources in centimeters
divided by the number of pixels in the line drawn between
the line source centers. The slice size is determined by
dividing the length of the line sources in centimeters by
the number of transverse slices that is required to
transverse the length of the line sources. The result is
reported in centimeters per slice.
Phantom Studies
Since the Threshold Segmentation Method is used in this
research, a threshold value which best relates the actual
objects of interest to the resulting SPECT image objects
must be determined. Three phantom studies were conducted to
determine the best threshold value for a given volume and
condition and to verify that the SPECT, Monte Carlo, and
Dosimetry models were working properly. The first phantom
study consisted of several cylinders of different volumes
filled with homogeneously distributed activity ("'In) imaged
in air. The second phantom study tested a torso phantom
74
containing three organ inserts under three experimental
conditions. First, the organ inserts were filled with
homogeneously distributed activity and placed in the cold
(no activity present) water filling the torso phantom.
Second, the organ inserts were filled with heterogeneously
distributed activity and placed in the cold water filling
the torso phantom. And last, the heterogeneously distributed
organ inserts were placed in the hot (activity present)
water filling the torso phantom. The last phantom study
consisted of a single cylindrical volume filled with
activity homogeneously and heterogeneously distributed and
thermoluminescent dosimetry devices for measuring absorbed
dose.
Phantom Study One
It is necessary to determine the best threshold value
which will result in the SPECTDOSE program calculating the
most accurate organ volume. A phantom study using objects of
a known volume can be conducted to determine the threshold
value which results in the SPECTDOSE program calculating a
volume which is closest to the known volume. In this study,
five cylinders of different sizes were SPECT imaged using
the same setup parameters as in the Clinical Studies (see
the previous section, SPECT Camera System) (Table 4-1). The
resulting images were read by the program SPECTDOSE to
calculate the phantom volume and activity concentration
Table 4-1. Phantom Study One Acquisition Parameters
FIVE CYLINDERS OF TISSUE EQUIVALENT MATERIAL
SOURCE: INDIUM-111
VOLUME (ml)
30.56
438.71
496.17
616.39
6032.50
ACTIVITY (GBq)
1.13
16.28
18.36
22.86
229.33
ACQUISITION PARAMETERS:
360 degree rotation
128 views at 12 s view'
20% window over each peak
medium energy collimator
64 x 64 Matrix
PIXEL SIZE:
SLICE SIZE:
0.69 cm
0.71 cm
PHANTOM
76
using different threshold values. The results were analyzed
by linear regression to determine the correlation between
the threshold value, phantom volume, and activity
concentration. The threshold value that yielded the best
correlation between the actual phantom volume and SPECTDOSE
measured volume was used for that range of volumes and set
of conditions.
Phantom Study Two
This study was conducted to simulate the conditions in
which a patient has been injected with a radiolabeled
substance. A tissue-equivalent torso phantom with a liver,
spleen, and tumor insert was tested under several
experimental conditions. The study set up parameters can be
seen in Table 4-2. The tumor insert was placed 12 cm below
the liver insert and the spleen insert was placed right of
the tumor insert 2.5 cm below the liver insert (Figure 4-
10). In the first experiment, the organ inserts were filled
with homogeneously distributed "'In activity and placed
inside the torso phantom, which is filled with water with no
radioactivity in it. The amount of activity added to the
inserts and the acquisition parameters can be seen in tables
4-2 and 4-3. In the second experiment, the liver and spleen
inserts were filled with small glass beads (5 mm diameter)
and 11In. The beads were used to distribute the
radioactivity heterogeneously within those organs. This
Table 4-2. Phantom Study Two Acquisition Parameters
TORSO BODY PHANTOM WITH ORGAN INSERTS
SOURCE: INDIUM-111
PHANTOM VOLUME (ml)
LIVER 1200.00
SPLEEN 166.31
TUMOR 0.26
BODY PHANTOM 13854.42
ACQUISITION PARAMETERS:
360 degree rotation
128 views at 12 s view"
20% window over each peak
medium energy collimator
64 x 64 matrix
PIXEL SIZE: 0.80 cm
SLICE SIZE: 0.82 cm
-- 30cm -
Liver Insert
Volume = 1200 cm 3
Spleen Insert
Volume = 166.31 cm3
/-0.25 cm
h-i
1.3 cm
Tumor Insert
Volume = 0.26 cm3
Figure 4-10. Phantom Study Two Torso Phantom and Organ Inserts
Table 4-3. Phantom Study Two Experiments
I. HOMOGENEOUS DISTRIBUTION/COLD BACKGROUND
Activity is uniformly distributed within organ
inserts with no activity in body phantom
(background).
VOLUME (ml)
1200.00
166.31
0.26
ACTIVITY (MBq)
38.48
5.74
2.37
II. HETEROGENEOUS DISTRIBUTION/COLD
BACKGROUND
Glass beads were added to the liver and spleen
inserts displacing 150 and 40 ml of liquid,
respectively. No activity in body phantom.
VOLUME (ml)
1200.00
166.31
0.26
ACTIVITY (MBq)
32.45
4.22
2.29
III. HETEROGENEOUS DISTRIBUTION/HOT
BACKGROUND
Glass beads were added to the liver and spleen
inserts displacing 150 and 40 ml of liquid,
respectively. Activity was added to the
body phantom.
PHANTOM
LIVER
SPLEEN
TUMOR
BACKGROUND
VOLUME (ml)
1200.00
166.31
0.26
12307.89
ACTIVITY (MBq)
31.45
4.14
2.26
5.81
PHANTOM
LIVER
SPLEEN
TUMOR
PHANTOM
LIVER
SPLEEN
TUMOR
80
experiment simulates the condition in which the patient's
organ(s) has a cold nonradioactivee) tumor within it. The
heterogeneously distributed radioactive organs and the
homogeneously distributed tumor insert were placed in the
cold nonradioactivee) water of the torso phantom and SPECT
imaged. The third experiment consisted of the
heterogeneously distributed radioactive liver and spleen
inserts and the homogeneously distributed tumor insert from
experiment two placed in the torso phantom which is filled
with radioactive water and SPECT imaged. Indium-lll (5.8
MBq) was added to the water of the torso phantom (Table 4-
3). Each experiment was SPECT imaged three times to detect
camera fluctuations. If the results varied from image to
image, this would indicate that the camera system was not
working properly and that the variation in results was due
to an improperly working camera system.
Phantom Study Three
This study was undertaken to verify the results of the
Dosimetry Model by comparing its results to direct measuring
devices placed in the phantom and to results calculated by
other methods. Indium-lll was placed homogeneously and
heterogeneously in the Jaszczak Phantom**** along with
several thermoluminescent dosimetry devices. The Jaszczak
Phantom is a cylinder made of tissue-equivalent material
Data Spectrum Corporation, High Point, NC 27514
81
with a volume of 6032.50 milliliters. In the first
experiment, 73 megabecquerels of "'In was added to the water
of the Jaszczak Phantom and six TLD cards of two chips each
were placed on the walls in various locations in the phantom
(Figure 4-11). The phantom was SPECT imaged using the same
camera setup parameters reported in the SPECT Camera System
section of this chapter. The TLDs were exposed to 11In for a
half an hour in the phantom. In the second experiment, a
cubed insert (7 cm x 6 cm x 7 cm) filled with air was added
to the radioactive water of the phantom to distribute the
activity heterogeneously. Six new TLD cards were added to
the walls of the phantom and the cubed insert (Figure 4-12).
The phantom was SPECT imaged using the same camera setup
parameters noted above. The TLD cards, once exposed, were
read by an automatic TLD reader *. The TLD reader was
calibrated using a Cesium-137 needle source; whereby, two
TLD cards were exposed to the 137Cs source for each of the
four exposure times (Appendix B). Once calibrated, the
experimental TLD readings can be converted to exposure and
absorbed dose. The TLD results calculated at time infinity
were compared to the results of the Dosimetry Model,
Geometric Factor Method (142), and the results calculated by
use of data in MIRD Pamphlet No. 3 (143) at time infinity.
It is assumed in each of these calculational methods
TLD System 4000, Harshaw/Filtrol Partnership, Solon,
OH 44139
h-- 21.6 cm-
I TOP I
BOTTOM
n-- BACK PANEL
RIGHT
SIDE
PANEL
TOP
Jaszczak Phantom
LEFT
PANEL
E
CARD 6
BACK
PANEL
RIGHT
PANEL
BOTTOM
Figure 4-11. Phantom Study Three Experiment One TLD Location
LEFT
SIDE
PANEL
I I It
I I I
I I I
II I I
II I I
I I II
I I II
21.6 cm--
TOP I
E
Organ Insert 0
I,
00'
BOTTOM
Jaszczak Phantom
CARD
Organ Insert
TOP
Figure 4-12. Phantom Study Three Experiment Two TLD Location
BOTTOM
84
(Dosimetry Model, Geometric Factor Method, and MIRD Pamphlet
No. 3) that the activity in the source organs are removed
only by physical decay, the effective half-life is equal to
the physical half-life, and all non-penetrating radiation is
absorbed in the source organ.
Thermoluminescent Dosimeters (TLDs)
Ionizing radiation incident on a thermoluminescent
crystal elevates an electron from the valence band to the
conduction band to leave a hole in the valence band. The
electron and hole pair migrate throughout the crystal until
they are trapped at impurity sites. When the chip is heated,
energy is imparted to the electron which causes it to move
and to eventually recombine with its counterpart hole (or
electron). The recombination energy is released in the form
of visible light, which can be detected by a phototube. The
thermoluminescent crystal used in this research is lithium
fluoride, which is the most common thermoluminescent crystal
used today (142). The lithium fluoride chip has a useful
range that extends beyond 103 Sv and good linearity
response, which extends below 0.1 mSv. The dynamic range of
TLDs, in general, is large, with doses from a few mSv to 10
Sv. Two TLD-100 chips (0.318 cm x 0.318 cm x 0.009 cm)
placed in the Type G-l gamma card configuration *.... were
used in this research. Six Type G-l cards were used in each
Harshaw/Filtrol Partnership, Solon, OH 44139
85
experiment and were secured in place in the phantom with a
hot glue gun.
Clinical Studies
The results of Phase One, Two, and Three clinical
studies being conducted at Bay Pines Veterans Administration
Medical Center (VAMC) in Bay Pines, Florida, using indium-
111 labeled B72.3-GYK-DTPA directed against colorectal
cancer, will be utilized in this research. The goals of
Phase One, Two, and Three studies are similar to those of
this research, in that, they both seek to determine the
radiation absorbed dose. Phase One, Two, and Three studies
also seek to establish the radionuclide-antibody
biodistribution, dosage range, clearance half-life, critical
organs, and optimal imaging and sampling times in diseased
patients (143). The research at Bay Pines VAMC is being
sponsored by Cytogen Corporation of Princeton, New Jersey,
and is ongoing. The experimental protocol for these studies
is given below.
Patients
Sixteen male patients participating in a phase I-III study
using indium-lll labeled B72.3-GYK-DTPA were screened prior
to antibody infusion. The age of the patients ranged from
49-89 years with a mean age of 67 years. All subjects had
86
proven primary or were suspected of having recurrent
colorectal cancer.
Monoclonal Antibody
The antibody, B72.3, is a murine monoclonal antibody of the
IgGi subclass which detects a 200K-400K molecular weight
tumor-associated glycoprotein called Tag-72. The Tag-72
antigen has been found to be expressed on certain human
colon and human breast carcinoma cell lines. B72.3 is
coupled to In-ill by oxidation of the oligosaccharide
moieties on the constant region of the antibody molecule.
This provides a site for specific attachment of
radionuclides and other ligands, while retaining the
homogeneous antigen affinity and binding characteristics of
the antibody. B72.3 is conjugated with the linker complex,
glycyl-tyrosyl-(N-e-diethylenetriaminepentaacetic acid)-
lysine (GYK-DTPA) to produce B72.3-GYK-DTPA-In-111 (Figure
4-13).
Monoclonal Antibody Procedure
Each patient was given a pre-infusion diagnostic blood
screening workup prior to the antibody infusion. The
analyses included routine blood chemistries, hematology,
electrolyte, and urinalysis, serum TAG-72, HAMA and CEA
levels. Each patient was then given B72.3-GYK-DTPA in
randomly assigned doses of 0.5, 1 or 2 milligrams of
|
|
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 EMU5XHN22_CQX282 INGEST_TIME 2011-07-15T21:03:58Z PACKAGE UF00097390_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
PAGE 1
A RADIATION DOSIMETRY MODEL FOR RADIOLABELED MONOCLONAL ANTIBODIES; INDIUM-111 LABELED B72 . 3-GYK-DTPA FOR COLORECTAL CANCER By LATRESIA ANN WILSON 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 1990
PAGE 2
,iiiiii|«
PAGE 3
Copyright 1990 by Latresia A. Wilson
PAGE 4
ACKNOWLEDGEMENTS First and foremost, I would like to thank my family for their love and support throughout the years. I would especially like to thank my uncle, Manuel Lewis Jr., for his wisdom and support; my grandmother, Lula Lewis, for her patience and understanding; my mother, Geraldine Solomon, for her encouragement and support; my aunts, Evelyn Scott and Marilyn Johnson, for believing in me; and my cousins for their encouragement . I would especially like to thank Dr. Genevieve Roessler for believing and supporting me during the rough years. I am very thankful to my advisor. Dr. William Properzio, for his superb guidance and support throughout this study. I am also grateful to my advisory committee, Dr. Emmett Bolch, Dr. Walter Drane, Dr. David Hintenlang, and Evelyn Watson, for their guidance and help in this project. I would like to thank Evelyn Watson and the staff (Mike Stabin, Audrey Schlafke-Stelson, Fanny Smith, and Stan Walls) of the Radiopharmaceutical Internal Dosimetry Center of Oak Ridge Associated Universities. I would like to thank Oak Ridge >c Associated Universities for giving me the opportunity and resources to learn from the best in my field. I would like to especially thank Martha Kahl and Rana Yalcintes in the medical iii
PAGE 5
library of Oak Ridge Associated Universities for their excellent support. I would like to thank the staff of the Nuclear Medicine Department at the University of Tennessee Hospital for their guidance, patience, and use of their equipment. I would like to thank Phyllis Gotten, Carmine Plott, James Stubbs, Pat Harp Family, and the Rose Foster Family for making my stay in Tennessee an enjoyable one. I would like to thank Dr. Steven Harwood, Michelle fs Morrissey, Linda Zangara, Dr. Will Webster, Dr. Carroll and Dave Laven and staff of the Nuclear Medicine Department at Bay Pines Veterans Administrative Medical Center in Bay Pines, Florida for providing the patient data, use of their facilities, and financial and expert support in this project. I would like to thank the people of the Department of -\ Environmental Engineering Sciences, Dr. Charles Roessler and fellow graduate students for their encouragement and support. I would like to also thank Dr. Libby Brateman for being there to answer all my seemingly endless number of questions and for providing support. I would like to thank Dean Rodrick McDavis for his continued support. And last, I would like to thank the McKnight Foundation and the Florida Endowment Fund for Higher Education, Dr. Israel Tribble and staff, for their financial support for without which, this degree would never have been undertaken. IV
PAGE 6
TABLE OF CONTENTS page ACKNOWLEDGMENTS ^^^ LIST OF TABLES vii LIST OF FIGURES ix ABSTRACT xi CHAPTERS 1 INTRODUCTION 1 2 MONOCLONAL ANTIBODIES 6 Immunoglobulin Structure 9 Variables Associated with Radioimmunoimaging and Radioimmunotherapy 14 Tumor Localization 17 Choice of Radiolabel 18 Tumor Size Effect 24 Fragment vs Whole Antibody 25 Dose Administered Effect 27 Labeling Method Effect 28 Dose Administration Route 29 Tumor Biology 3 Other Factors 31 3 RADIATION DOSIMETRY 34 MIRD Approach 3 6 "Traditional" Point Kernal Method 38 Microdosimetry 41 4 MATERIALS AND METHODS 4 7 SPECT Model 4 9 Monte Carlo Model 4 9 Dosimetry Model 51 Single-Photon Emission Computed Tomography ... 51 SPECT Quantitation 53 Photon Attenuation 54 Photon Scatter 5 6 SPECT Camera System 57 Image Segmentation 58 Program SPECTDOSE 60 Subroutine THOLD 63 Subroutine CONTOUR 63 Subroutine OBJSELECT 63
PAGE 7
Subroutine CORGAN 57 Subroutine VOXFIL 69 Program ALGAMP 69 Pixel and Slice Size Determination 72 Phantom Studies 73 Phantom Study One 74 Phantom Study Two 7 6 Phantom Study Three 8 Thermoluminescent Devices 84 Clinical Studies 85 Patients 85 Monoclonal Antibody 86 Monoclonal Antibody Procedure 86 Blood Analyses 88 HPLC Procedure 88 Image Analysis 89 5 RESULTS AND DISCUSSION 90 Pixel and Slice Size Determination 91 Phantom Study One 91 Phantom Study Two 97 Phantom Study Three 112 Clinical Study 129 6 SUMMARY AND CONCLUSIONS 133 REFERENCES 14 APPENDICES A SAMPLE CALCULATIONS 154 B TLD CALIBRATION 157 C SPECTDOSE PROGRAM 158 BIOGRAPHICAL SKETCH 181 VI
PAGE 8
LIST OF TABLES Table 2-1 Properties of Human Immunoglobulins 11 Table 2-2 Properties of Human IgG Subclasses 12 Table 2-3 Selected Radionuclides for Radioimmunodetection and Radioimmunotherapy 19 Table 2-4 Radionuclides for Radioimmunodetection and Radioimmunotherapy 21 Table 3-1 Sublethal Radiation Doses 44 Table 4-1 Phantom Study One Acquisition Parameters ... 75 Table 4-2 Phantom Study Two Acquisition Parameters ... 77 Table 4-3 Phantom Study Two Experiments 79 Table 5-1 Phantom Study One Threshold Determination . . 92 Table 5-2 Phantom Study One Results 94 Table 5-3 Phantom Study Two Threshold Determination . . 98 Table 5-4 Phantom Study Two Experiment One Results . . 100 Table 5-5 Phantom Study Two Experiment Two Results . . 102 Table 5-6 Phantom Study Two Experiment Three Results . 104 Table 5-7 Gaussian Prefilter Comparison: Actual versus SPECT Measured Volume 107 Table 5-8 Phantom Study Two Absorbed Dose Results . . . 110 Table 5-9 Phantom Study Three TLD Measurements 114 Table 5-10 Phantom Study Three Geometric Factor Method and MIRD Pamphlet No. 3 Results 117 Table 5-11 Phantom Study Three Dosimetry Model Results . 118 Table 5-12 Phantom Study Three Results 120 vii
PAGE 9
Table 5-13 Phantom Study Three Error Anaylsis 121 Table 5-14 TLD Calibration Study Results 123 Table 5-15 Phantom Study Four TLD Measurements 127 Table 5-16 Phantom Study Four Results 128 Table 5-17 Phantom Study Four Error Analysis 130 Table 5-18 Clinical Study Results 131 viix
PAGE 10
LIST OF FIGURES Figure 1-1 Antibody Carriers for Diagnosis and Therapy 3 Figure 2-1 Monoclonal Antibody Production 8 Figure 2-2 HAT Mediated Hybridoma Production 10 Figure 2-3 IgG Molecule 13 Figure 2-4 Enzymatic Digestion of IgG Molecule into Fragments 15 Figure 4-1 Research Methodology 48 Figure 4-2 Single-Photon Emission Computed Tomography . 52 Figure 4-3 SPECT Model Flow Chart 61 Figure 4-4 SPECTDOSE Program Subroutine Flow Chart ... 62 Figure 4-5 Illustration of Subroutine THOLD Object Segementation 64 Figure 4-6 Subroutine CONTOUR Object Segmentation ... 65 Figure 4-7 Subroutine CONTOUR Object Assignment .... 66 Figure 4-8 Illustration of Subroutine OBJSELECT Selected Object Comparison 68 Figure 4-9 ALGAMP Flow Chart 70 Figure 4-10 Phantom Study Two Torso Phantom and Organ Inserts "78 Figure 4-11 Phantom Study Three Experiment One TLD Location 82 Figure 4-12 Phantom Study Three Experiment Two TLD Location 83 Figure 4-13 B72.3 Linker Complex 87 IX
PAGE 11
Figure 5-1 Phantom Study One: Actual versus SPECT Measured Volume 95 Figure 5-2 Phantom Study One: SPECT Measured versus Actual Activity Concentration 96 Figure 5-3 Phantom Study Three TLD Experimental Locations 113 Figure 5-4 Phantom Study Four TLD Chip Packaging . . . 124 Figure 5-5 Phantom Study Four TLD Locations 12 6
PAGE 12
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 A COMPUTERIZED RADIATION DOSIMETRY MODEL FOR RADIOLABELED MONOCLONAL ANTIBODIES: INDIUM-111 LABELED B72 . 3-GYK-DTPA FOR COLORECTAL CANCER By LATRESIA ANN WILSON DECEMBER 1990 Chairperson: William S. Properzio Major Department: Environmental Engineering Sciences A foundation was developed for a dosimetry methodology that could be used to calculate absorbed doses in target and nontarget tissues using uniformly and nonuniformly distributed activity. In this methodology, a dosimetry model was developed which consisted of three independent models: 1) the SPECT Model, 2) the Monte Carlo Model, and 3) the Dosimetry Model. The SPECT Model uses Single-Photon Emission Computed Tomography (SPECT) images to detennine the volume and radioactive uptake. A computer \ program was written to automatically read and analyze SPECT images. This program uses an edge detection method to detennine the volume. Voxel elements within the identified volume are used to calculate the activity concentrations. The Monte Carlo Model uses a monte carlo simulation method xi
PAGE 13
and results of the SPECT Model to calculate the fraction of photon energy deposited in target and nontarget tissues. The Dosimetry Model combines the results of the SPECT and Monte Carlo Models to determine the absorbed dose in target and nontarget tissues. Several phantom studies were conducted to verify the ability of the Dosimetry Model to evaluate organ and tumor uptake, sizes, and to calculate absorbed doses. Comparisons were made between the Dosimetry Model, other calculational methods (MIRDOSE2, Geometric Factor Method, MIRD Pamphlet No. 3) , and TLD measurements. For diagnostic activity doses, the SPECT Model was found to calculate organ volumes of the order of 1000 ml to within fifteen percent of the actual volumes but it failed to accurately calculate organ volumes of 200 ml or less. No meaningful relationship was found between the actual and SPECT measured activity concentrations. The Dosimetry Model agreed within 12% when compared with the Geometric Factor Method and the MIRD Pamphlet No. 3 results using homogeneously and heterogeneously distributed ^^^In. The TLD measurements were within 30% at most of the other methods. Results of the several phantom studies indicated the Dosimetry Model was an appropriate methodology for calculating absorbed doses for homogeneously distributed activity. Further investigation is needed to determine the Xll
PAGE 14
accuracy of the Dosimetry Model in the heterogeneously distributed activity case. The addition of photon attenuation and scatter correction and nonpenetrating radiation transport is pertinent to the accuracy of the dosimetry methodology Xlll
PAGE 15
CHAPTER 1 INTRODUCTION In the United States, cancer is the second leading cause of death with the number of annual deaths fast approaching 400,000 (1). This value represents a little over 20% of all deaths. Women are more susceptible to cancer than men and except for accidents, cancer kills more children than any other illness (1). In England, cancer is the leading cause of death in children 1-14 years of age (1) . Utilization of antibodies to fight cancer started as early as 1946 when Pressman theorized that polyclonal antibodies directed against antigens expressed on tumor cells could be used to localize radionuclides in the tumor. He believed that once the antibodies were bound to the antigen-rich tumor site, the radioactivity could be detected with a gamma scanning device or if the radionuclide concentration in the tumor was sufficient, serve as local radiation therapy. So, after a series of ingenious experiments, he successfully demonstrated that immune proteins could be used to target radioactivity to tumors in
PAGE 16
2 living animals (2) . Unfortunately at that time, it was difficult to produce antibodies that would survive in cultured media, thus limiting the ability to produce sufficient amounts with the specificity needed for clinical studies. This ultimately limited the further use of this technology for many years to come. In 1975, Kohler and Milstein introduced a new technique called hybridization, which would allow for the production of large quantities of identical (monoclonal) antibodies (3). This technique made it possible for the methodologies proposed by Pressman to be applied clinically. Kohler and Milstein later went on to receive the Nobel Prize for their contribution. With the advent of the hybridization technique, there was renewed interest in the use of radiolabeled antibodies for tumor therapy. It is generally believed that monoclonal antibodies attached to radiolabels for therapy (radioimmunotherapy) may be effective in treating metastases and small tumors, where surgery may not be feasible. This new technique offers some ray of hope in the fight against cancer. Recent advances in biotechnology have given new hope to achieving the ultimate goal of using monoclonal antibodies for targeting radioactivity for the dual purpose of cancer diagnosis and therapy (Figure l-l) . This potential has
PAGE 17
Label With Small Amount of Radioactivity Patient With Undisclosed Tumor Nuclear Image to Localize Tumor Antibody to p97 Melanoma Tumor Therapy Attach Anti-Tumor Drugs or High Dose Radioactivity Patient With Tumor Tumor Destruction Figure 1-1. Antibody Carriers for Diagnosis and Therapy Adapted from Reference 4
PAGE 18
4 generated a significant amount of interest and growth in the field of nuclear medicine over the past few years. This growth, in turn, has generated many new problems and questions. One of these problems, the radiation dosimetry of using radiolabeled monoclonal antibodies, is the focus of this research. Current radiation dosimetry methods, which allow for the calculation of absorbed doses for both target and nontarget tissues, assume that the radiolabel's energy is distributed uniformly throughout the target and nontarget organ. This assumption is not valid in the case of radioimmunotherapy , since it has been shown that radiolabeled monoclonal antibodies distribute heterogeneously throughout a given organ and on the tumor cell (5) . It is, therefore the objective of this research to develop a foundation for a radiation dosimetry methodology that could be utilized for radiolabeled monoclonal antibodies; i.e., a methodology which would allow for the calculation of absorbed doses in tissues with a heterogeneous or homogeneous radioactivity distribution. A computerized dosimetry model, which allows for the calculation of absorbed doses to both target and nontarget tissues after intravenous (IV) injection of Indium-111 labeled B72 . 3-GYK-DTPA monoclonal antibody directed against colorectal cancer, will be proposed in this research. Clinical applications and ease-of-use of this dosimetry
PAGE 19
5 model will be emphasized. A comparison of the results from this model with that of current dosimetry methods will be made. This dissertation is divided into five basic sections. First, an overview of monoclonal antibodies and the factors that affect their localization are presented. Second, there is a discussion of the current radiation dosimetry methods and their inadequacies for use with radiolabeled monoclonal antibodies. Third, a discussion of the experimental methods, computer models, and imaging techniques used in this study are presented. Next, the computational results are presented and analyzed. Finally, the results are summarized and suggestions for future applications of this method are made.
PAGE 20
CHAPTER 2 MONOCLONAL ANTIBODIES Antibodies, or immunoglobulins, are proteins made by many animal species as part of their specific response to foreign substances (antigens) . When antibody-antigen binding occurs, this immunologic response usually results in the destruction or elimination of the antigen. Immunoglobulins are produced by the activity of the B lymphocytes and possess specific binding regions that recognize the shape of particular sites or determinants on the surface of the antigen. An antigen may have several determinants, or epitopes, each of which is capable of stimulating one or more B lymphocytes. For this reason, an antigenic challenge results in the production of a variety of antibodies (6) . Early antibody production techniques employed the use of animals, usually a mouse or rabbit, immunized with an antigenic substance, to obtain antibodies, which were found in the serum of the immunized animal. These antibodies were
PAGE 21
7 polyspecific because they reacted with a wide variety of antigenic binding sites. Highly specific antibodies can be developed by extracting individual lymphocytes and cloning them in tissue culture; each clone would have the potential to manufacture a single antibody species, a monoclonal antibody (Figure 21) . Unfortunately, normal antibody-producing cells do not survive in culture media. It took Nobel laureates Kohler and Milstein (3) to recognize that myeloma cells, which are cancer cells that produce large amounts of identical but nonspecific immunoglobulins, and which can survive in cultures indefinitely, might be altered by the new techniques of recombinant genetics to construct immortal clones that secrete immunoglobulins. Kohler and Milstein developed a method of producing such monoclonal antibody strains by fusing the lymphocytes from the spleen of an immunized mouse with mouse myeloma cells, thus forming clones of hybrid cell lines, called hybridomas (4). These cells are usually fused in polyethylene glycol and result in clones that have the specific-antibody characteristics of the lymphocytes and the longevity of the myeloma cells. Additionally, pure hybridoma cells are selectively grown in hypoxanthene-aminopterinthymidine (HAT) media since it supports neither the unfused lymphocytes nor the myeloma cells. Once these hybridomas are produced, they can be assayed for antibody activity and for
PAGE 22
Antigen -Z'^"^Antigenic "^ — I — 'Determinant T /^ Lymphocytes Myeloma Ceils Spleen /^ M\ Q o o o Lymphocytes ipnocytes i Antibody/ A"<'9en Antiserum Fuse "^c; "^ / N Hybrid Myeloma Cells 4 Clone Polyclonal Antibodies Monclonal Antibodies Figure 2-1. Monoclonal Antibody Production* Adapted from Reference 4
PAGE 23
9 further selective cultivation (Figure 2-2) . The reader is referred to Reference 6 for an excellent review of the techniques involved in the production, purification, analysis, quality control, radiolabeling, and storage of monoclonal antibodies. Immunoglobulin Structure Immunoglobulins (Ig) are divided into five classes: IgG, IgA, IgM, IgD, and IgE and can further be subdivided (isotypes) on the basis of internal attributes (see Table 21 and 2-2) . IgM antibodies are often the first to appear during immunization and IgE antibodies mediate hypersensitivity reactions (8) . Immunoglobulins of all classes are composed of two heavy (H) chains and two light (L) chains in their simplest form. All classes share the same light chains and differ solely in the structure of the heavy chains. The heavy chains are attached to one another by means of one or more disulfide bonds, and a light chain is attached to each heavy chain by a disulfide bond (Figure 2-3). Isotypes differ structurally in the number of disulfide bonds linking the two heavy chains together, and they differ functionally in their ability to fix complement and to interact with effector cells such as macrophages and mast cells (Table 22) .
PAGE 24
10 ^ Cell culture Myeloma Line Fuse In ^Hi^-A. Polyethylene I'^O Glycol
PAGE 25
11 O C E e c e D 0) a o I i2 < O I— I
PAGE 26
12 (-1 O 0) ^
PAGE 27
1 3 HOOC COOH Figure 2-3. IgG Molecule ' Adapted from Reference 8
PAGE 28
14 The antibody-specific sites of the immunoglobulins are situated near the amino-terminal (NHj) end of each of the four chains (Figure 2-3) , and it is in this region (variable region) that the greatest variability in amino acid sequence occurs from immunoglobulin to immunoglobulin (9) . Constant amino acid sequences are found in the carboxyterminal (COOH) regions (constant region) of the immunoglobulin chains (Figure 2-3). The two variable regions bind to specific antigenic sites and the constant region (Fc) interacts with the host immune system. Since the Fc region of the antibody is most likely to trigger allergic responses, fragmentation has been used to remove this portion from the antibody molecule. Pepsin, a proteolytic enzyme, cleaves off most of the Fc region, which leaves two Fab fragments bound together in a divalent structure known as the F(ab')2 fragment (Figure 2-4). The enzyme papain breaks the immunoglobulin into two monovalent Fab fragments and an intact Fc fragment. Variables Associated with Radioimmunoimaginq and Radioimmunotherapv A number of variables that must be considered before diagnostic and therapeutic applications of monoclonal antibodies can be utilized. Generally monoclonal antibodies alone are not effective in tumor destruction (10). This has been attributed in part to the heterogeneous distribution of tumor-associated antigens on cell surfaces, which leads to
PAGE 29
15 LIGHT CHAIN HEAVY CHAIN ^/y^ VARIABLE "X REGION PEPSIN F(ab')2 + FRAGMENTS OF Fc s-s CONSTANT REGION '^° ^^\pAPAIN Figure 2-4. Enzymatic^ Digestion of IgG Molecule into Fragments* Adapted from Reference 9
PAGE 30
16 variable attachments of the antibodies to the different tumor cells; more antibodies are attached to those cells which have significant amounts of the antigen on their surfaces but none to other tumor cells that are devoid of the specific antigen and are therefore allowed to proliferate (10). Since the monoclonal antibody alone is not cytotoxic, it usually acts as a carrier of a more cytotoxic radionuclide or toxin. This introduces a number of more complicating factors and variables, which include the combined physical, chemical, and biological properties of the antibody and radiolabel. The following is a summary of the variables directly linked with the production of the radiolabeled tumor-associated antibody for imaging and therapy (11) : 1) Physical properties of radionuclides a) Physical half-life b) Gamma energies and abundances c) Photon yield per absorbed radiation dose d) Parent-daughter relationship-stable decay products e) Ratio of penetrating to nonpenetrating components f) Particle radiation (/3-,^+,IC, and Auger electrons) g) Production mode (availability) 2) Chemical properties a) Stability of radionuclide-protein bond b) Specific activity-number of labels per molecule obtainable c) Retention of immunological activity versus specific activity d) Addition of nonradioactive carrier-metal ion contamination e) Sample pH
PAGE 31
17 3) Biodistribution and biological half-life a) Route of administration and activity of initial dose b) Vascularity: Blood flow and interstitial fluid space c) Uptake of protein-bound form of the isotope d) Plasma and whole body clearance e) Relative size of tumor model f) Size of animal or human model g) Cell proliferation h) Capillary and cell permeability i) Presence of inflammation 4) Target-nontarget time-dependent ratio: dose to tumor, whole body, and other sensitive organs 5) Immunological purity of the antibody and its relative specificity 6) Characteristics of imaging system with respect to the radiolabel properties 7) Marketability, availability, convenience This list, which is by no means all-inclusive, is complicated by the fact that each variable seems to be related to a number of the other variables. Tumor Localization The localization of radiolabeled antibodies at tumor sites is dependent on a number of factors as reported by several investigators (12-27). These include the tumor size, radiolabeling method, choice of radiolabel, type of antibody (whole vs fragment) , route of administration, tumor biology (blood flow, vascular permeability etc.), and the dose administered. However, tumor uptake is ultimately dependent upon its antigen content (13).
PAGE 32
18 Choice of Radiolabel The choice of a radiolabel is dependent upon its intended use: diagnostic or therapeutic applications. For diagnostic applications, one is more concerned with the sensitivity and specificity of the test with the least radiation dose. This is obtained by the use of radionuclides with a low equilibrium absorbed dose constant. In therapy, the objective is to attain the highest differential radiation dose, which requires the use of radionuclides with a high equilibrium absorbed dose constant. The goal of both applications is to attain the highest radiation dose factor for the target site in comparison to the normal tissue (28) . Ideally, radionuclides which are particularly suited for imaging with radiolabeled antibodies should be characterized by 1) physical half-life of 6 hr to 8 days, 2) gamma energy range of 80-240 keV, 3) high single energy gamma abundance per decay, 4) small abundances and lowenergy particulate radiation, and 5) reasonable radiolabeling chemical properties and stability (11). Similarly, radionuclides used for therapy should have complementary properties to the antibody-bound radionuclides used in imaging. However, their decay should be characterized with a large component of particulate radiation with little or no accompanying gamma radiation such that a high localized dose may be delivered (11) . Table 2-3 lists the various radionuclides that meet the required
PAGE 33
19 Table 2-3. Selected Radionuclides for Radioimmunodetection"'' Nuclide Half-life Primary Decay Characteristics 99iftp^ 123111 In 13197 RU 67 CU* 6 h 13 h 68 h 193.2 h 69 h 62 h IT (99%) ; EC (100%) ; EC (100%) ; ^' (100%) ; EC (100%) ; P' (100%) ; 5 = 141 keV (89%) S = 159 keV (83%) 5 = 171 keV (88%) 7n^^t Cu* 212 Bi 211 At 125i 193.2 h 64 h 62 h 1 h 7.2 h 144.5 h /?'(100%) /3'(100%) ^(100%) a (36%) P' (64%) a (41%) EC (59%) 0.608 MeV (86%) 6 = 364 keV (82%) 2.29 MeV (100%)
PAGE 34
20 specifications for diagnostic and therapeutic applications of monoclonal antibodies. Table 2-4 lists the advantages and disadvantages of the use of the radionuclides found in Table 2-3. Technetiuin-99in, ^^^In, and ^'^I are examples of radionuclides that are currently under extensive use in medical imaging (29-34). They have the advantage of availability, well-known chemistry, and optimal half-life and gamma decay energy. Unfortunately, ^^^I suffers from dehalogenation in vivo, which allows for nonspecific uptake of free iodine in sites other than the tumor sites, especially in the thyroid, liver and spleen (30,35). This makes identifying tumors in these organs by imaging nearly impossible. Iodine-131 also delivers a high radiation dose to normal tissues due to its long half-life and medium gamma energy (36). Indium-Ill in vivo metabolism is relatively unknown, although it has been shown to have good affinity once in the tumor, but if it comes off the antibody, it will relocate to the liver, spleen, and bone marrow (30). Technetium-99m has a chemistry problem; i.e., it is difficult to obtain a stable bond between it and the antibody. Childs and Hnatowich (37) found increased stability when '*'*^c was coupled directly to the chelate DTPA (diethylenetriaminepentacetic acid). Rhodes et al. (38) used a pretinning method to successfully label '^'^c directly to antibody fragments, which showed increased stability against
PAGE 35
21 Table 2-4. Advantages and Disadvantages of Selected Radionuclides for Radioinununodetection'*' Nuclide Advantages Disadvantages 99nvp, 123n 111 In 13197 Ru Availability Decay energy Decay energy Iodine chemistry Decay energy Optimal T^^ Chelation chemistry Availability Iodine chemistry Optimal T^ Chelation chemistry 67, Cu Optimal T^ Short T^^ Chemistry problem Availability Cost ($20/mCi) Short T^ In vivo metabolism Decay energy In vivo de-iodination In vivo metabolism Availability Decay Energy Decay Energy Advantages and Disadvantages of Selected Radionuclides for Radio immunotherapy"*" Nuclide Advantages Disadvantages 13190, Availability Imaging Cost 67, Cu 212 Bi 211 At 125n '°Sr-Generator Pure /3' decay Imaging High LET decay High LET decay High LET decay Long tissue path Chemistry problems In vivo metabolism? In vivo metabolism? Short T^ Unknown chemistry Short T,^ Unknown chemistry Must be in nucleus to kill tumor "^ Adapted from Reference 29
PAGE 36
22 transchelation. Recently, Goldenberg and associates (39) have reported in vivo retentions of 98% immunoreactivity in patients using anti-CEA murine monoclonal antibody (IMMU-4) Fab' labeled directly with ''^c. Chen and colleagues (40) also have reported good results using '^^c labeled antibodies in the confirmation of diagnosis of uveal melanoma. Several alphaand beta-emitting radionuclides have potential for radioimmunotherapy as seen in Table 2-3. Iodine-131 has been most commonly used and is currently being utilized in human clinical studies (35,41,42). However, the choice of ^'^I has not been because it is the optimum for radioimmunotherapy; two-thirds of its absorbed dose equivalent is due to penetrating radiation, which usually escapes the primary tumors and their metastases (4). The Auger electrons of iodine-125 may be effective for therapy when used in conjunction with antibodies that are internalized rather than remaining on the cell surface (9) . The appeal of alpha-particles for radioimmunotherapy is their short range (-50-90 ^m) and high linear energy transfer (LET) (-80 keV//im) , which produces extreme cytotoxicity. An alpha-particle traversing the diameter of a 10 /im nucleus deposits an energy of 800 keV, equivalent to an absorbed dose of approximately 0.25 Gy (4). Potential alpha-emitting radionuclides for radioimmunotherapy are astatine-211 and bismuth-212 (Table 2-3) , Experimental
PAGE 37
23 trials with ^^^At-conjugated antibodies on a murine lymphoma system are in progress by Harrison (43) and Vaughan (44) . Perhaps the half-life of bismuth-212 is too short (60.6 min) to fully capitalize on the longer antibody retention in the tumor, although Macklis found it to be highly cytotoxic to the murine Thy 1.2* EL-4 tumor cell line (45). More recently Simonson et al. (46) showed ^^^Bi to be also cytotoxic to the LS174T cell line. Few suitable alpha-sources are available because most alpha-emitters are heavy elements (A > 82) which decay to unstable daughters. The recoil alphas produced in the decay of these daughters rupture the radionuclide-antibody bond, which allows the daughter product to diffuse away from the tumor (5) . Yttrium-90 offers another possibility for use in radioimmunotherapy and has the advantage in that it is a pure beta emitter and is easily available by production from a strontium-90 generator (Table 2-4). Unfortunately, it has no gamma emissions to allow for useful biokinetic studies in the patient and, once detached from the antibody, it deposits in the bone in sufficient quantities to give a high radiation dose to the marrow. Yttrium-90 is currently under investigation by several groups (24,47-52). Sally DeNardo and colleagues (51,53) found copper-67 to be one of the most promising radionuclides for radioimmunotherapy because of its short half-life, abundance of beta particles, and the presence of 93 and 184 keV gamma emissions.
PAGE 38
24 A potential radionuclide for radio immunotherapy , palladium-109, a predominately beta-emitting radionuclide (^(nax= ^ W®^' half-life= 13.4 h) that is available carrierfree, was investigated by Favn^az et al. (54), who labeled it to an antimelanoma monoclonal antibody. Unfortunately, they found that at least 60% of the radiolabeled antibody preparation failed to bind to melanoma cells. They believed this was partly the result of inactivation of the antibody during purification, storage, or radiolabeling and/or the presence of carrier ^°^Pd in the ^"'Pd preparation. A new radionuclide for radioimmunotherapy , rhenium-186 (1.07 MeV maximum beta, 9% abundant 137 keV gamma) , is being extensively evaluated in patients by Schroff and associates (55) . Preliminary findings indicate it to have similar in vivo properties to ''^c and is very stable in vivo. Tumor Size Effect Several investigators have found that tumor uptake of the radiolabeled monoclonal antibody is inversely related to the tumor size; i.e., the per gram uptake of monoclonal antibodies decreases as the tumor size increases (1215,56,57). Pimm and Baldwin (14) have found a multitude of parameters that could potentially account for this relationship. These include changes in blood flow, degree of necrosis, levels of cellular and intratumor or extravascular antigen, and the presence of circulating tumor-derived
PAGE 39
25 antigen. However, this relationship could not be duplicated by Cohen et al.(16). In fact, their findings contradicted those by other investigators, in that they found the total tumor uptake increased with increasing tumor size. No satisfactory explanation has been offered to explain the differences between the findings. Pedley et al. (57) found that for tumor weights greater than 100 mg a strong positive correlation exists between absolute uptake and tumor weight but found a poor correlation for smaller tumors. They thus concluded that specific uptake was inversely proportional to tumor size regardless of the antibody. At present, although still controversial, one may conclude that the relationship between the tumor size and the antibody uptake is an inverse one. Fragment versus Whole Antibody Antibody fragments (Figure 2-4) reach their maximum accumulation faster and clear from the body faster than whole antibodies (9,17-19,29,53). However, whole antibodies remain in the tumor longer to achieve higher concentrations. Thus, the choice of antibody type depends on the application. Radioimmunoimaging would benefit most from the use of fragments, because of their early maximum accumulation and faster clearance, which results in a lower background (nonspecific uptake) level. Radioimmunotherapy would benefit most from the use of whole antibodies because
PAGE 40
26 the cytotoxic effect could be delivered over a longer period of time. The difference in tumor localization between the fragments and whole antibody has been attributed to the smaller weight of the fragments (55,000 daltons and 110,000 daltons for IgG Fab and F(ab')2 respectively) compared to that of the whole IgG antibody (160,000 daltons), which allows them to transverse the intravascular and extravascular space much more quickly (53). This effect also may be the result of differences in the valency of the antibodies (9) . Since Fab fragments are monovalent, their bonds to cell-bound antigens are weaker than those of the divalent whole antibodies and because of this, they shed the tumor and are rapidly cleared from the body via the kidneys (9). F(ab')2 fragments, on the other hand, are divalent, but demonstrate similar kinetics to the Fab fragments. Ballou et al. (19) compared IgM F(ab')2^ fragments to whole IgM antibodies. The weight of the IgM F(ab')2^ fragment was 130,000 daltons, which is not considerably less than that of a whole IgG antibody (160,000 daltons). However, the IgM F(ab')2^ did weigh considerably less than the whole IgM antibody, which weighed 900,000 daltons. The F(ab')2^ showed a 1.6-fold faster whole body clearance and reached its maximum uptake earlier than that of the whole IgM antibody. However, its total uptake was lower than the whole IgM antibody. Ballou suggests that this may be caused in part by
PAGE 41
27 differences in metabolism between the whole antibody and fragment and also possible changes in the antigen-binding of the fragments resulting from low pH digestion. The choice of antibody type will depend upon its application. For imaging, fragments will most likely be used. The best choice seems to be F(ab')2 because it remains in the blood longer than Fab fragments. This results from its larger molecular weight, which reduces its loss through the kidneys (58) . For therapy, whole antibodies will probably be used. Perhaps F(ab')2 fragments will prove superior in all cases, because they offer the advantages of fragments and the lack of immunogenicity of whole antibodies. Dose Administered Effect Eger et al. (20) found dose-dependent kinetics in 12 human patients with melanoma. They found that as the amount of injected antibody increased, the plasma half-life increased, which eventually resulted in a higher tumor uptake. They also found that the radioactivity levels in the spleen and marrow decreased as the amount of antibody increased. This dose dependent effect was also seen by Hnatowich et al. (21). Pedley et al. (57) also studied the effect of tumor weight on uptake with escalating amounts of antibody. They found that there was decreased uptake with escalating amounts of antibody in small tumors. This effect
PAGE 42
28 was thought to be the result of steric hindrance in the small tumors, even though the rate of diffusion into the tumor may have increased. Labeling Method Effect The method used to attach the radionuclide to the antibody will affect the antibody's localization. If the method is inefficient, in that the radiolabel detaches from the antibody in vivo or if the radiolabel 's radiation destroys or alters the properties of the antibody, all hope of tumor localization is lost and radioimmunotherapy is rendered useless. Since most suitable radiolabels for therapy are metals (Table 2-3) , early methods of antibody labeling attempted to attach them directly to the antibody. This proved to be highly unstable and the radiolabel detached from the antibody in vivo (9) . However, nonmetals, such as iodine, are currently being attached directly to the antibody by the lodogen or Chloramine-T methods (22). These elements also suffer from instabilities and tend to dehalogenate in vivo (18) . The latest methods employ the use of a coupling agent, usually a chelate, to attach the metallic radiolabel to the antibody (9,21,23,24). The most widely used chelate is diethylenetriaminepentaacetic acid (DTPA) . The antibody is attached to the DTPA which, in turn, is attached to the radiolabel. The bonds formed with the DTPA are much stronger
PAGE 43
29 than those of the direct-attachment method (23); thus the chelate-coupled antibodies are much more stable in vivo (53) . Another advantage to using a chelate such as DTPA is that many different chelate substitution levels on the antibody can be achieved by straightforward manipulation of the relative amounts of reactants or time of reaction with the antibody (53) . Other chelates have also been used and their effects on the antibody biodistribution are continually being investigated (59-62) . Dose Administration Route The site where the antibody is administered affects not only how fast the antibody reaches the tumor, but also how much eventually localizes in the tumor (17,25). Obviously, if one is interested in localization in the lymphatic system, intralymphatical administration will prove superior to the other routes (25) . If there are ascites in the peritoneal cavity, intraperitoneal administration would prove superior over the other routes. Hnatowich et al. (47) concluded that the use of intraperitoneal rather than intravenous administration may be important in the application of yttrium-90 because it probably offers a means of reducing radiation exposures to the bone marrow and the critical organ without reducing exposure to the tumor within the peritoneum. Larson (17) found that the concentration of radiolabeled antibodies in human tumors is tenfold less
PAGE 44
30 after administration via the intravenous route than after injection either subcutaneously, intralymphatically , or intraperitoneal ly. Tumor Biology Tumors grow radially from a central group of cells; therefore as the tumor enlarges, the dividing cells form a shell around a relatively hypoxic core. When these cells outgrow their blood supply, they die and form a necrotic central nest containing some viable cells that are highly resistant to radiation (26) . Blood flow in this situation is low which makes delivery of the radiolabeled antibody to the tumor very difficult. Studies by Gullino and Grantham found that the average value of blood supply to tumors was 0.14 ± 0.01 ml per hour per mg of nitrogen and the blood supply was independent of the host (27) . Solid tumors were also found to be angiogenesis dependent by Folkman (63) . The radiolabeled antibody must reach the tumor through circulation, crossing the capillary wall and diffusing throughout the interstitial fluid to reach the tumor cells. The rate of diffusion across these barriers is slowed by the large size of the antibody molecule (64) . This diffusion rate has, according to Winchell (65), an 18 to 24 hour halflife. Diffusion of the labeled antibody from the vascular compartment into the tumor is caused by the concentration gradient between the blood and the tumor (26) . The higher
PAGE 45
31 the concentration of radiolabeled antibody in the blood compared with the tumor, the higher the diffusion rate will be. Leichner et al.(66) found that external-beam irradiation increased the permeability of tumor vascularity, which resulted in increased tumor uptake of radiolabeled antiferritin. Other Factors Other factors may influence the localization of radiolabeled monoclonal antibodies in the tumor, such as the amount of circulating antigens in the vascular system and the metabolism and catabolism of the antibody in vivo. Circulating antigens in the blood may combine with circulating labeled antibodies. This complex could be phagocytized by the reticuloendothelial system to reduce the number of labeled antibodies that reach the tumor site (67,68). Pimm and Baldwin (69) found that the average rate of catabolism of ^"i-iabeled-IgG, anti-CEA monoclonal antibody was 1.64% of the administered dose per gram per 24 hours and that this rate was higher for tumor bearing mice as opposed to nontumor bearing mice. They also concluded that tumor localization by the labeled antibody is a dynamic process with simultaneous localization and degradation. Gatenby et al. (70) have shown that the level of oxygen in the tumor or tumor region also affects the antibody localization. They found that tumors or tumor regions with a
PAGE 46
32 mean oxygen pressure of 16 nun Hg or less had lower antibody uptake, even when the presence of antigen was confirmed by biopsy. This suggests that physiological factors other than antigen expression may affect antibody uptake. In the past few years a factor that has become increasingly important because of the increase in the number of human studies is the development of human anti-mouse antibodies (HAMAs) . The body, in response to the injection of murine antibodies, produces antibodies (HAMAs) against the murine antibody which it recognizes as being foreign. This response can be detected within one week of exposure to the mouse protein and is maximal within 2-3 weeks of exposure (71) . The timing and detection of the HAMAs are influenced by the dose of the mouse antibody administered (71) . HAMA clearly alters the pharmacokinetics of subsequent murine antibody infusions and, depending on the dose of the murine antibody and titer of HAMA, can interfere with radioimaging and therapy and can lead to toxicity because of the immune complexes and their redistribution (71) . Scannon (72) found a rapid clearance of the infused murine antibodies from the blood which limited further administration. It has been suggested that antibody fragments be used instead of whole antibodies, because they lack the Fc region (Figure 2-4) , which most likely triggers the allergic response. Other approaches to reducing HAMA include the use of chimeric (human-mouse) monoclonal antibodies, chemical alteration of the murine Fc portion.
PAGE 47
33 ultrapheresis of human plasma to remove Ig, and chemical suppression of the immune response. From the above discussion, one may conclude that the localization of radiolabeled monoclonal antibodies at the tumor site is dependent upon a number of seemingly interrelated variables which may vary from patient to patient. Larson (25) also concluded that tumor localization varied considerably from patient to patient.
PAGE 48
CHAPTER 3 RADIATION DOSIMETRY Before radioiitimunotherapy can be implemented successfully, it is necessary to know the amount of radiation absorbed by the target and nontarget tissues. This has proved to be difficult because of the lack of appropriate methods to measure the amount of radiation absorbed in the tissues; i.e., the absorbed dose, which was deposited there by radiolabeled antibodies. The lack of an appropriate method for correlating non-uniform dose with effect has also hindered the efforts to assess the absorbed dose. Assessment of the absorbed dose is complicated by the large number of interrelated factors that affect the # ... W localization of the radiolabeled antibodies in vivo (see '^{•' Chapter 2). These factors require that the calculated V absorbed dose be patient-specific. The current methods used >>to calculate absorbed dose are based on assumptions that are iN ';^ not valid when radiolabeled antibodies result in a vV'Ar nonuniform distribution are used. V 34
PAGE 49
35 Current dosimetry methods can be divided on the basis of the approach taken to calculate the absorbed-dose. There are three basic approaches (73) : (a) those that utilize the conventional Medical Internal Radiation Dose Committee (MIRD) formulation, a macroscopic approach which was developed to cater mainly to diagnostic situations usually involving gamma emitters and whole organs rather than discrete targets (74); (b) those that utilize Berger's point kernels, a semi-microdosimetry approach which considers small size targets but not very low energy emissions at the level of cell dimensions (75) ; and (c) those that take a microdosimetric approach, which investigates doses from short range emissions located near the cell surface or cell nucleus (76) . Since absorbed dose is defined as the amount of energy deposited per unit mass by ionizing radiation at the site of interest (77) , dosimetry calculations require a knowledge of the physical properties of the radiolabel, length of time the radioactivity remains in the various sites, and the distribution of the radionuclide to the various sites in the body (28,78,79). The physical properties of the radiolabel are perhaps the easiest to determine accurately and will be known in detail if conventional labels are used (80) . The residence time and spatial distribution of the radiolabeled antibody in vivo are not usually known and must be determined prior to radioimmmunotherapy. These parameters
PAGE 50
36 are usually determined by sequential, timed quantitative imaging. Several investigators (10,11,78) have suggested that a diagnostic study, as such, be performed prior to radioimmunotherapy. In this diagnostic study, the antibody would be labeled with a small amount of the therapeutic agent or a short-lived isotope of the therapeutic agent in an effort to reduce the hazard to the patient (69) . Medical Internal Radiation Dose Committee (MIRD) Formulation The MIRD Formula (74) is the most widely accepted method for calculating radiation absorbed dose from internally deposited radionuclides. This method was recommended by the Medical Internal Radiation Committee of the Society of Nuclear Medicine in 1968 and was later adopted for standard use by the International Commission on Radiation Units (ICRU) (81) in 1971. MIRD is based on the dose rate equation developed by Loevinger et al. (82) in 1956 and is expressed as Dose rate= K x activity in target x energy of x absorbed 1) to target mass of target emission fraction where K is a constant which depends on the units used. Several assumptions are made in this approach, the most important in the present context being that in applying this method to humans, an anthropomorphic phantom is used, which in calculating the absorbed dose, does not take into account
PAGE 51
37 >. the nonuniformity of the activity distribution. Thus, source homogeneity is assumed throughout the organs. Humm (4) gives ^ two reasons why this assumption may not necessarily be valid in the case of rad io immunotherapy . First, the irregular i nature of the tumor vasculature will result in a complex pattern of diffusion gradients guiding the antibodies : through the tumor. Second, immunohistochemical studies with antibodies have shown that the tumor antigens may not be expressed uniformly throughout the whole tumor cell population. For the application of radioimmunotherapy , and assuming that the activity remaining in the body after organ uptake is distributed uniformly, the mean dose to the target (tumor) is the sum of three components: a) the dose from nonpenetrating radiations (radiation pathlength is smaller than the dimensions of the organ in which it resides) emitted within the target organ, b) the dose from penetrating radiations (radiation pathlength is greater than the dimensions of the organ in which it resides) emitted within the target organ, and c) the dose from penetrating radiations emitted by the activity in the rest of the body (41) . The absorbed fraction for nonpenetrating radiations is assumed to be unity; i.e., all the energy emitted by the • source organ is absorbed in the source organ. With this in mind, one proceeds to calculate the various parameters of the MIRD equation for each component. The effective half-
PAGE 52
^ 38 time can be calculated from exposure rate measurements and activity measurements of the blood and urine as a function of time. Decay constants are calculated from a least-squares fit of the time-varied target organ count rates. Compartmental modeling is often employed to calculate the cumulated activity, decay constants, and the other parameters needed for the MIRD equation. Tumor and critical ) organ volumes are determined from Computed Tomography (CT) or Single-Photon Emission Computed Tomography (SPECT) images, r" For conventionally employed radionuclides such as ^'^I, / '^P, or '°Y and for targets greater than a centimeter in diameter, the MIRD method holds quite reasonably (73) . Berqer's Point Kernels This method is based on Berger's Point Kernels for calculating the absorbed dose from beta-rays (75) . If the medium is assumed to be uniform and unbounded, the beta-ray dosimetry problem can be divided into two separate parts: a) determination of the distribution of absorbed dose around a point isotropic source, which is often referred to as a point kernel, and b) appropriate integration over the point kernel weighted by the source density to obtain absorbeddose distributions for extended sources (83). Part a) contains all the physical aspects of the problem and part b) is entirely geometric. Using the principles of
PAGE 53
39 superposition, the absorbed dose from one source element can be added independently to the contribution from another source element. Thus, a distributed radionuclide source can be considered as a collection of independently acting isotropic sources (83) . The beta-ray dose rate is expressed in the form R^ = 1.38E-05 E^ * A Gy d''' 2) where R^ is the beta-ray dose rate in the tissue, E^ is the average beta-ray energy per disintegration in Mev, * is the isotropic specific absorbed fraction, and A is the Activity of the radionuclide in Bq. Since the dose rate is proportional to the average concentration, the total beta-particle dose is obtained by integrating the concentration over the time the tissue is exposed to the beta particles: D^(t) r R^(t)dt 1.38E-05 E^ A(t)dt Gy where E^ is in Mev, t is in days, and A is in Bq. Thus, whenever the average activity A(t) , is known as a function of time, the absorbed dose can be computed by integration. Loevinger et al . (82) states that for purposes of dosimetry, the tissue distribution can be represented by a stable system of separate compartments interconnected by first-order reactions. First-order reactions imply that the
PAGE 54
40 total amount of radioactivity leaving a given compartment per unit time is proportional to the amount present. The rate of change of the total radioactivity in the ith compartment is described by the following differential equation: n ^i = -^p^i ^^,0^1 + 2:(k..qj k.^q.) 4) dt j = l q. = total radioactivity in the ith compartment kj. = constant fraction of the radioactivity in the ith compartment transferred to the jth compartment, per unit time kjQ = constant fraction of the radioactivity in the ith compartment transferred to outside the system (excretion) k = radioactive decay constant n = number of compartments The first term on the right represents the loss due to radioactive decay, the next term the loss from the system by excretion or fixation, the first term inside the bracket represents the contribution of the (n-1) other compartments to the ith compartment, and the second term inside the bracket represents the loss from the ith compartment to the (n-1) other compartments. Integrating this equation for q. (/iCi) and then dividing by the mass (g) of compartment i gives the average concentration of radioactivity in the ith compartment: C.(t) = 3.7 X 10^ q, (t)/m. Bq g"^ 5) Thus, it is now possible, using Equation (5) , to calculate the total beta-particle dose from Equation (3). Spencer (84)
PAGE 55
<^ 41 showed the applicability of this method in radioimmunotherapy . A whole range of electron energies from 10 keV to over 1000 keV, as well as tumor sizes from single cells to 10 cells (each having a millimeter diameter) , can be encompassed with this approach. Microdos imetrv Microdosimetry is most applicable for evaluating doseeffect relationships. It uses the microscopic distribution of radiation interactions with biological systems to explain the effects of radiation on the system (76) . In some instances, the distribution of specific energy in small targets, individual tracks, or even individual energy absorption events such as single ionizations may be needed to obtain meaningful dose-effect relationships. Microdosimetry takes into account the statistical aspects of the particle tract structure, energy distribution patterns, and radionuclide distribution within tissues and provides a means for determining the number and frequency of cells irradiated, the probability densities in specific energy, and the average dose delivered to cells of interest (85) . Charged-particle radiation interacts with atomic electrons of the matter through which it passes, and ionization and/or excitation energy is imparted with each interaction. The charge and mass of the particle, its initial energy, and the
PAGE 56
42 matter through which it travels determine the pattern of energy loss, the distance traveled, and the direction taken by the particle. Ionizations and excitations are produced when the energy is transferred from the particle to the medium. The basic quantity that describes the energy imparted to matter is the absorbed dose, which actually is a mean value. By definition, the absorbed dose D is the quotient of de by dm, where de is the mean energy imparted by ionizing radiation to matter of mass dm (86) : D = de/dm 6) The specific energy, z, a stochastic quantity with units similar to absorbed dose, is defined as the quotient of e by m, where e is the energy imparted by ionizing radiation to matter of mass m (86) : z = e/m 7) The mean absorbed dose in a volume is equal to the mean specific energy z, in the volume: D = z 8) The ratio e/m is highly dependent upon target size. As the target size gets smaller and smaller, the variations in the local dose becomes increasingly greater, and the average dose value becomes less and less indicative of the complete
PAGE 57
43 dose distribution (85). Thus, for very small target sites, the concept of absorbed dose becomes increasingly abstract, and the dose is better represented by a distribution of doses in "specific energy". For a given value of target size mass, this distribution is called the "probability density in specific energy" and is denoted by f(z). The probability that the specific energy received by a target site lies in the infinitesimal range dz containing the value z is f(z)dz. Methods for calculating the probability densities in specific energy can be divided into four steps (85) . The first step involves characterizing the geometrical relationship between the radioactive source distribution and the target sites. Second, the density in specific energy must be determined for a target at any distance from the radioactive source and with all possible angles of intersection considered. Third, the probability that a point source exists at any given distance from the target must be determined from the spatial distribution of sources. And fourth, the densities from all point sources are convolved using Fourier transforms to construct a new specific energy density for the target population. The product of a microdosimetry calculation is a statistical distribution of doses to small sites from which an average dose could be determined. The precise relationship between the specific energy density (average dose) and the resulting biological effects is not known;
PAGE 58
44 therefore, the results from this approach are not directly applicable to the rather different conditions found in radioimmunotherapy. However, several investigators (85,8791) have proceeded to utilize this method for radiolabeled antibody dosimetric calculations. Critical Organs The maximum radiation doses that radiosensitive organs can tolerate and still continue to function adequately to support life are listed in Table 3-1 (10). With current systemic approaches to therapy, bone marrow toxicity has been the dose-limiting side-effect (92). However, as shown by in vivo radiolabeled antibody biodistribution studies, the dose-limiting organ is most likely to be the liver or kidney. Table 3-1. Sublethal Radiation Doses''' Organ System Dose (Gy) Bone Marrow < 2 Intestinal Mucosa < 7 Kidney < 15 Liver < 2 5 ''' Adapted from Reference 10 Leichner et al. (41), using the MIRD methodology, found that the ^'^I radiation dose for four patients ranged from four to 10 Gy for the liver and from 1.1 to 2.2 Gy for total-body
PAGE 59
45 irradiation. Vaughan et al. (93), using Berger's Point Kernels, found that a tumor dose of two Gy in one week with 15^1 was associated with a whole-body dose of 17 Gy. Bigler et al. (87) utilized a itiicrodosimetric approach to calculate the mean dose to the red marrow for a number of different radiolabels. The mean dose ranged from 1.6 Gy with ^As to 17 Gy with ^^^I for a large cell size. In recent years with the advent of new materials technology, Griffith et al. (94-95) and Wessels (96) have developed a method for the direct measurement of absorbed radiation dose through the use of teflon-imbedded, CaSO,:Dy thermoluminescent dosimeters (TLD)*, which have been modified to fit inside a 20-gauge needle. The TLDs are directly implanted into the tissue of interest and are subsequently recovered for read-out. They measured an absorbed dose of 8.1 Gy for the ^^^I labeled B72.3 colorectal carcinoma mouse system and 17.4 Gy for the ^^^I labeled LYM-1 Raji B-cell lymphoma mouse system, which correlated well with autoradiography measurements (95). This method is not appropriate for human dosimetry studies because of patient discomfort and tissue trauma. In order for radioimmunotherapy to be successful, the radiation dose deposited in the tumor and other critical organs must be known accurately. Current dosimetric methods do not adequately address the unique features proposed by * Teledyne, Inc., NJ.
PAGE 60
46 the use of radiolabeled antibodies for the calculation of absorbed dose. Therefore, new methods must be created. As more clinical information using radiolabeled antibodies becomes available, a better method may be defined, which can be compared to direct measurements.
PAGE 61
CHAPTER 4 MATERIALS AND METHODS In this research, a foundation for a dosimetry methodology to determine the absorbed dose in both target and nontarget tissues using uniformly and nonuniformly distributed activity has been developed. The calculation of absorbed dose can be divided into two parts: 1) the determination of the radionuclide concentration, and 2) the determination of the amount of energy deposited in the tissues of interest. This new dosimetry methodology uses Single-Photon Emission Computed Tomography (SPECT) to determine the radioactive uptake in the tissues and a Monte Carlo method to determine the amount of energy deposited in the tissues. The research method utilized in this research is shown in Figure 4-1. In this figure, the research method is divided into three models: 1) the SPECT Model, 2) the Monte Carlo Model, and 3) the Dosimetry Model. Results from the SPECT and Monte Carlo Models are utilized in the Dosimetry Model. 47
PAGE 62
48 Organ/Tumor Biodistribution SPECT Model Volume Radiolabel Physics Radioactive Uptake Monte Carlo Model Absorbed Fraction Dosimetry Model Figure 4-1. Research Methodology
PAGE 63
49 SPECT Model The SPECT Model employs the use of Single-Photon Computed Tomography, a diagnostic imaging technique, to determine the volume and radioactive uptake in the target and nontarget tissues following injection of radiolabeled monoclonal antibodies. A computer program, SPECTI)OSE, was written to calculate both target and nontarget tissue volumes and radioactive uptake. SPECTDOSE uses edge detection and contour tracing algorithms to determine the volume of the various organs and tissues of interest. The SPECT image is divided into several three-dimensional arrays of a preselected size and number. Sixty-four arrays composed of 64 X 64 elements (pixels) are utilized in this research. Each element of the array represents an image volume (voxel) at a specified location. Each voxel contains an integer value derived from the measured activity in the imaged object. The total number of voxels and their location, image intensity per voxel, and organ volume (total number of voxels at a specified location) are determined in this model. Results of this model are used in the Monte Carlo Model . Monte Carlo Model This model uses a monte carlo method to calculate the fraction of photon energy deposited per unit mass of target and nontarget tissues (specific absorbed fraction) . A monte
PAGE 64
50 carlo computer program was obtained from Oak Ridge National Laboratory, Oak Ridge, Tennessee (97) . This program, called ALGAMP, is a photon transport code which accurately simulates the physical phenomena of the photon by the use of the statistical nature of radioactivity. In ALGAMP, the human body and organs are represented by a set of mathematical equations known collectively as the Cristy Parametized Phantom (98) . The radioactive distribution within each organ is assumed to be homogeneous in the Cristy Parametized Phantom. ALGAMP was modified for use in this research by the deletion of the Cristy Parametized Phantom and the addition of a method which permits the direct use of the voxel information created by the SPECT Model. The voxel information generated by the SPECT Model defines the organ volumes and locations of interest. In the dose calculation each voxel value represents the heterogeneous radioactivity distribution found in the organs following the use of radiolabeled monoclonal antibodies. By use of the SPECT image voxel information and the monte carlo simulation method, the amount of photon energy deposited per tissue mass, specific absorbed fractions, can be determined for each organ volume and voxel. The specific absorbed fractions are utilized in the Dosimetry Model.
PAGE 65
51 Dosimetry Model The Dosimetry Model uses the results of the SPECT and Monte Carlo Models to determine the absorbed dose to both the target and nontarget tissues. Voxel matrix values of the tissue volumes determined in the SPECT Model are utilized in the Monte Carlo Model to determine the specific absorbed fractions in the tissues of interest. The Dosimetry Model combines the specific absorbed fractions with the organs' radioactive uptake determined in the SPECT Model to calculate the absorbed dose. The absorbed dose is determined for both the organ and organ voxels; i.e, the absorbed dose can be calculated for each organ voxel also. The Dosimetry Model retains the concepts of the MIRD Method in addition to accounting for the heterogeneous distribution of radioactivity exhibited in the organs and organ voxels following the injection of radiolabeled monoclonal antibodies into humans. Single-Photon Emission Computed Tomography Single-Photon Emission Computed Tomography (SPECT) is a diagnostic imaging technique utilized in nuclear medicine, in which, the differences in radioactive distribution of internally administered radionuclides are exploited (99) . In SPECT, the detector, a gamma camera, rotates around the patient while acquiring data (photon detection) (Figure 42) . With the use of a computer and several complicated
PAGE 66
52 Single-Head SPECT Gamma Camera Dual-Head SPECT Gamma Camera F=^^ iiiliiiiiiiiiiniii iii|iii i iii|iii | iii Single-Head SPECT Unit Multi-Detector Head SPECT Unit Figure 4-2. Single-Photon Emission Computed Tomography* * Adapted from Reference 100
PAGE 67
53 algorithms, the data is reprojected (reconstructed) into a transverse section image (slice) of the activity distribution. Basically, SPECT maps the three dimensional concentration of a radionuclide by measuring the angular distributions, or projections, of gamma ray intensities emitted within the body. SPECT is also capable of eliminating overlying and underlying source activities and offers the potential for quantitating the radioactive uptake in the patient (100) SPECT Quantitation SPECT quantitation of radionuclide activities in the human body is affected by several physical and instrumental factors including absorption attenuation of photons in the patient, Compton scattered events, the system's finite spatial resolution, and object size, finite number of detected events, partial volume effects, the radiopharmaceutical biokinetics, and patient and/or organ motion. Other instrumentation factors such as calibration of the center-of-rotation, sampling, and detector nonuniformities will affect the SPECT measurement process (100,101,102). Several of the major factors that affect quantitation with SPECT systems are as follows (100) : 1) Physical Factors: a) Characteristic energy of the emitted photons b) Radiation decay as a function of time
PAGE 68
54 c) Attenuation of gamma photons within the patient d) Inclusion of scattered photons within pulse height window 2) Anatomical/Physiological Factors: a) Source size and location within the body b) Patient and/or organ motion c) Biokinetical behavior of radiopharmaceutical within the body 3) SPECT System Factors: a) Camera/collimator energy and spatial resolutions b) Detection efficiency c) Changes in collimator geometric response with distance from the collimator surface d) Sensitivity variations across the camera surface e) Camera electronic variations, ADC errors, and gantry mechanical variations with time and/or position f) Characteristics of reconstruction process such as shape of filter function, linear and angular sampling interval values, accuracy of attenuation, nonuniformities, and scatter compensation methods and accuracy of edgedetection methods Their relative importance depends on the type of quantitative information desired and the biokinetic properties of the radiopharmaceutical. The determination of radionuclide concentration as a function of time for small volume elements (voxels) within the body is affected most by the factors listed above. Photon Attenuation The determination of the radionuclide concentration as a function of time in the voxel elements is affected by the
PAGE 69
55 absorption attenuation and scattering of photons. The effect of attenuation results in a decrease in the measured gainina ray intensity. There is self-attenuation in the source organ and also attenuation in the surrounding body tissues. Most attenuation compensation methods assume that the attenuation coefficient, the fraction of the gamma-ray beam attenuated per unit thickness of absorber (103), is constant. Although this will provide a less accurate compensation within regions where the value of the attenuation coefficient is variable, it is the method utilized in this research. Other attenuation compensation methods, those which do not assume a constant attenuation coefficient, can be divided into three classes: 1) Preprocessing Methods, 2) Intrinsic Compensation Methods, and 3) Postprocessing Methods (100) . Preprocessing methods attempt to correct the projection data prior to image reconstruction. These methods are relatively easy to implement, however, they tend to generate streak artifacts in the presence of noise. This method was not used in this research because the antibody SPECT images were very noisy. Intrinsic compensation methods integrate attenuation correction directly in the reconstruction algorithm. An attenuation map is measured (by using a transmission source) or assumed as part of the reconstruction algorithm. These methods require the use of large computers and are time
PAGE 70
56 consuming, thus preventing the use of this method in this dissertation. Postprocessing methods apply attenuation correction after the image reconstruction has completed. This approach is used most often in commercial SPECT systems and reguires the measurement or estimation of the patient's body contour. In the human studies undertaken in this research, the patient's body contour was not retained, which precluded the use of this method in this research. The area of attenuation compensation in SPECT is currently undergoing extensive analysis and the reader is further directed to a number of reports on this subject (104-119). Photon Scatter Compton scattering events degrade the image contrast resulting in a major source of error in the quantification of radionuclide concentrations. Scattered photons can contribute as much as 50% of the total collected events in SPECT (120) . The use of a sodium iodine-thallium doped detector in SPECT systems results in the inclusion of both scattered and nonscattered photons in the photopeak energy window. Several approaches have been attempted to compensate for the scattered radiation, but none at this point have proven to be of substantial value (100,101,102,121-130). No scatter correction method was utilized in this research.
PAGE 71
57 SPECT Camera System SPECT was performed with a digital rotating gamma camera* with a medium energy collimator and a 20% peak energy window. A rotation of 360°, 128 projections (2.81° apart), and a study of 26 minutes (12 s view'^) was used. Data was acquired on a computer** in the 64 x 64 x 16 bit mode (131) . After acquisition, the raw image data is reduced from 16 bits to 8 bits using the system software***. After which, the data is prefiltered using a Gaussian filter of the 24th order and a frequency cutoff of 0.20 (131). This data is reconstructed using the high resolution reconstruction algorithm, which is an iterative reconstruction method (132,133), with one iteration and a dampening factor of 0.5 (134). The dampening factor indicates the level of contribution by the error image to the production of the iterative transverse slices (135) . Sixty-four transaxial slices, one pixel thick, are created. The size of the elemental voxel is one pixel in the x and y axis (transaxial plane) and in the z direction (parallel to the axis of rotation) . One pixel was determined to be equal to 6.9 mm in the patient studies. Technicare Omega 500, Technicare Corporation, Cleveland, OH 44139 ADAC DPS-3300, ADAC Laboratories, San Jose, CA 95138 ADAC Laboratories Version 4 System Software, San Jose, CA 95138
PAGE 72
58 Image Secnnentation Prior to determining the organ volumes, the SPECT image must be segmented into the respective organs. Image segmentation is the process of subdividing an image into its constituent parts or objects. Segmentation algorithms are generally based on two properties of the image gray-level values: discontinuity and similarity. The gray-level value is an integer that represents the image intensity. In the discontinuity category, the image is partitioned on the basis of abrupt changes in the gray level. Detection of points, lines, and edges are of principle interest in this category. In the similarity category, the image is divided on the basis gray level similarities. Approaches to the similarity category include thresholding and region growing (136) . Several segmentation or edge detection methods were attempted prior to the selection of the Threshold Segmentation Method in this research. The Gradient Method, an approach that looks for discontinuity, was attempted first. It is assumed in this method that the regions of interest are homogeneous so that the transition between two regions can be determined on the basis of gray-level discontinuities alone. A local derivative operator is determined, whereby the magnitude of the first derivative indicates the presence of an edge and the sign of the second derivative determines where the edge pixel lies; the background or object side (137). Since the
PAGE 73
59 first and second derivatives must be determined for each image pixel, this method is computationally intensive. A second approach. Histogram Segmentation Method, was also attempted. This technique creates a histogram of the gray-level contents of an image. The image is subdivided into its constituents by use of the peaks and valleys in the histogram, which represent the image object and background regions respectively. Division between objects is difficult when a deep valley or steep peak is not present (138) . In the presence of image noise, differentiation between peaks and valleys is futile. Because of noisy SPECT images this method was not used in this research. The last approach attempted and used in this research is the Threshold Segmentation Method. This technique segments on the basis of gray-level similarity. A threshold value is applied to the image, whereby any image pixel's gray-level value that is greater than the threshold value is considered to be a part of the object and any pixel with a gray-level less than the threshold value is apart of the background. Since the threshold depends only on each pixel's gray-level, it is called global (139). This method was selected because of its easy implementation, small computation requirements, and excellent results when used with noisy images.
PAGE 74
60 Program SPECTDOSE The Program SPECTDOSE was developed in this research to calculate the necessary parameters proposed by the SPECT Model; i.e., organ volumes and radioactive concentrations (Figure 4-3) . This program is written in Fortran-77 for a VAX/VMS Operating System. The Program SPECTDOSE is divided into a number subroutines (Figure 4-4). Before the SPECT image could be utilized, its data format or the way the image data was written to the file had to be determined. The data format for the SPECT images was obtained with a promise of confidentiality from the ADAC Corporation (140) . The reconstructed SPECT image data is stored in each voxel as hexadecimal (base-16) numbers. The main program reads the hexadecimal numbers into a logical array, where the values (image count) are scaled between 0-255 intensity levels (gray-levels) and read into an integer array. The resulting reconstructed image data is represented as an interger which has a value between 1 and 256. The image count can be corrected for attenuation and radioactive decay at this point by entering the appropriate linear attenuation coefficient, radionulide half-life, and time of decay values into the program. The image threshold value is entered and the subroutine THOLD is called to segment the image into its constituent objects. This process is repeated for each image slice.
PAGE 75
61
PAGE 76
62 £S Subroutine THOLD n Subroutine CONTOUR Jti Program SPECTDOSE Z3 I Subroutine OBJSELECT Subroutine FOPEN ^H Subroutine MENU Subroutine SOBJCT Subroutine VOXFIL Subroutine CORGAN Yi Subroutine VOXMAX Figure 4-4. SPECTDOSE Program Subroutine Flow Chart
PAGE 77
63 Subroutine THOLD The subroutine THOLD segments the image into various objects using the Threshold Segmentation method (139) (Figure 4-5) . The objects are separated from the background pixels by comparing their intensity values with a global threshold value; all pixels with an intensity value higher than the threshold belong to the object. The subroutine CONTOUR is called to extract the objects from the segmented image. Subroutine CONTOUR The subroutine CONTOUR extracts the objects from the segmented image (Figure 4-6). The extracted object's boundary is traced and the resulting object is stored in a binary file called OBJECT*. DAT. This process is repeated for all objects in the segmented image. Each object file is assigned a consecutive identification number; i.e., Objectl.dat, Object2.dat etc. (Figure 4-7). The extracted object's characteristics, which include the number of voxels, total count, maximum and minimum indices, volume, and area, are written to the file, OBJVAL.DAT. Subroutine OBJSELECT The subroutine OBJSELECT integrates the extracted objects of each slice into a single object; i.e., organ. The extracted object that best represents the shape of the organ of interest is determined. This object's, the selected object,
PAGE 78
64 r-
PAGE 79
65
PAGE 80
66 OBJECT SLICE 1 SLICE 2 SLICE N K Segmented objects from each image slice is extracted and separated into separate files and assigned file names in consecutive order by Subroutine CONTOUR Figure 4-7. Subroutine CONTOUR Object Assignment
PAGE 81
67 identification number is entered into the program. The selected object is then compared with the rest of the objects (Figure 4-8). If ninety percent of the object's voxels are the same for each slice as the selected object's voxels, the object is considered to be apart of the organ. This process is repeated for all organs. The organ's voxel indices, count, name, identification number, and volume are stored in the file, VOXEL.DAT. Subroutine CORGAN The subroutine CORGAN creates an organ given its voxel indices and identification parameters. Each pixel is assigned an integer value which will represent the organ volume desired. The number of image slices included in the organ is also assigned. The organ created is used as a photon reflector or sink; i.e., it either scatters or absorbs the incident photons, and represents the areas of the body not included in the SPECT image. If the whole body is to be included in the SPECTDOSE Program, it is necessary to create those areas of the body not seen in the SPECT images due to the limited f ield-of-view of the SPECT camera and the lack of availability of whole body SPECT images. These areas are created using this subroutine.
PAGE 82
68 Select Object 2 for Comparison: OBJECT SELECTED OBJECT 2 1
PAGE 83
69 Subroutine VOXFIL Once the organ files have been created; i.e., each organ's voxel information has been stored in a VOXEL.DAT file, the subroutine VOXFIL assembles the organs (each VOXEL.DAT file) into one larger file called VOXPHAN.DAT. The number of organs; i.e., the number of VOXEL.DAT files, is entered into the program. This file contains the voxel indices, image counts, weighting factors, identification numbers, and the total number of voxels for all organs. The VOXPHAN.DAT file is read directly by the program ALGAMP. Program ALGAMP The program ALGAMP is a point energy gamma-ray monte carlo radiation transport code for calculating specific absorbed fractions of energy and absorbed dose data from internal and external sources (97) . This program is written in Fortran and was developed at Oak Ridge National Laboratory in Oak Ridge, Tennessee (97) . A flow chart of this program can be seen in Figure 4-9. This program is composed of 30 or more subroutines and initially utilized the organs of the parametized phantom model from the Cristy Phantom Series (98). In this Series, the organs were represented by mathematical equations of various geometrical shapes, such as, spheres and cylinders. The equations were confined to a small number of ALGAMP subroutines (GE0M,SUM1, SUM2 , and RESULT), which hasten the modification process.
PAGE 84
70 jEI Get Input Data n Get Media Data Get Voxel Data ALGAMP i£ start Photon Histories Calculate Cross Sections Generate Random Numbers Print Results Score Collision Score Media and Voxel Hits T Figure 4-9. ALGAMP Flow Chart
PAGE 85
71 The Cristy Phantom Series was not used in this research because it assumes a homogeneous source distribution, which is not valid in the case when radiolabeled monoclonal antibodies are used; the organs were not patient specific; and its inability to represent diseased organs, which are most often found in nuclear medicine patients. Since each human is uniquely different, it was the desire of this research to make this new dosimetry method patient specific; i.e., the organs of the imaged subject are utilized in the calculation. This can be achieved by using the actual SPECT image to define the organ volumes and radioactive uptake. Each organ's voxel information was determined and compiled into the file VOXPHAN.DAT by the SPECT Model. The VOXPHAN.DAT file is read by ALGAMP. Each voxel inherently, at the level of the camera system's resolution, accounts for the heterogeneous source distribution exhibited in the organs at that level following the uptake of the radiolabled monoclonal antibodies. If the activity is not distributed heterogeneously at the level of the camera system's resolution, which is approximately one half centimeter in this research, the voxels will reflect this and the activity will be assumed to be homogeneously distributed. For this homogeneous case, no additional modifications to ALGAMP would be needed. Several ALGAMP subroutines (INPUT, SOURCE, SEARCH, GEOM, SUMl, SUM2 ,
PAGE 86
72 RESULT, and RANPOS) were modified to accoitimodate the inclusion of the voxel information. For sources distributed in energy, the cumulative distribution function (cdf) for the source energy spectrum is used. A detailed cross-section table is generated for the source energies of interest. Each photon is weighted by a weighting factor which describes the probability of photons existing in a given voxel. The photon weighting factor is computed for each voxel by dividing the voxel image count by the average voxel count for a given organ. The source photon location is chosen by randomly sampling the voxel locations in the VOXPHAN.DAT file. Photon collisions are scored by determining the voxel location given the photon direction coordinates. Scoring is tallied for each voxel and organ. Pixel and Slice Size Determination In order to determine the organ volumes, it is necessary to determine the size of each image pixel and slice in physical dimensions. Since these values are dependent upon the camera system's electronics, they must be determined after each camera adjustment or change. A pixel and slice size determination study was conducted prior to the patient and phantom studies and in the case when the camera system's electronics were changed. Two line sources (small tubes containing '^''^c) of known length and distance apart are imaged in the planar (static) mode. The line
PAGE 87
73 sources are imaged in both the parallel and perpendicular positions relative to the camera system's axis-of-rotation (AOR) to detect changes in the x and y planes, which would indicate the camera system was working improperly. The system software was used to return the number of pixels in a line drawn between the two line source centers in the planar imaged. The image pixel size in centimeters per pixel equals the distance between the two line sources in centimeters divided by the number of pixels in the line drawn between the line source centers. The slice size is determined by dividing the length of the line sources in centimeters by the number of transverse slices that is required to transverse the length of the line sources. The result is reported in centimeters per slice. Phantom Studies Since the Threshold Segmentation Method is used in this research, a threshold value which best relates the actual objects of interest to the resulting SPECT image objects must be determined. Three phantom studies were conducted to determine the best threshold value for a given volume and condition and to verify that the SPECT, Monte Carlo, and Dosimetry models were working properly. The first phantom study consisted of several cylinders of different volumes filled with homogeneously distributed activity (^^^In) imaged in air. The second phantom study tested a torso phantom
PAGE 88
74 containing three organ inserts under three experimental conditions. First, the organ inserts were filled with homogeneously distributed activity and placed in the cold (no activity present) water filling the torso phantom. Second, the organ inserts were filled with heterogeneously distributed activity and placed in the cold water filling the torso phantom. And last, the heterogeneously distributed organ inserts were placed in the hot (activity present) water filling the torso phantom. The last phantom study consisted of a single cylindrical volume filled with activity homogeneously and heterogeneously distributed and thermoluminescent dosimetry devices for measuring absorbed dose. Phantom Study One It is necessary to determine the best threshold value which will result in the SPECTDOSE program calculating the most accurate organ volume. A phantom study using objects of a known volume can be conducted to determine the threshold value which results in the SPECTDOSE program calculating a volume which is closest to the known volume. In this study, five cylinders of different sizes were SPECT imaged using the same setup parameters as in the Clinical Studies (see the previous section, SPECT Camera System) (Table 4-1) . The resulting images were read by the program SPECTDOSE to calculate the phantom volume and activity concentration
PAGE 89
75 Table 4-1. Phantom Study One Acquisition Parameters FIVE CYLINDERS OF TISSUE EQUIVALENT MATERIAL SOURCE: INDIUM-111 PHANTOM VOLUME (ml^ ACTIVITY (GBa) 1 30.56 1.13 2 438.71 16,28 3 496.17 18.36 4 616.39 22.86 5 6032.50 229.33 ACQUISITION PARAMETERS: 360 degree rotation 128 views at 12 s view'^ 20% window over each peak medium energy collimator 64 X 64 Matrix PIXEL SIZE: 0.69 cm SLICE SIZE: 0.71 cm
PAGE 90
76 using different threshold values. The results were analyzed by linear regression to determine the correlation between the threshold value, phantom volume, and activity concentration. The threshold value that yielded the best correlation between the actual phantom volume and SPECTDOSE measured volume was used for that range of volumes and set of conditions. Phantom Study Two This study was conducted to simulate the conditions in which a patient has been injected with a radiolabeled substance. A tissue-equivalent torso phantom with a liver, spleen, and tumor insert was tested under several experimental conditions. The study set up parameters can be seen in Table 4-2. The tumor insert was placed 12 cm below the liver insert and the spleen insert was placed right of the tumor insert 2 . 5 cm below the liver insert (Figure 410) . In the first experiment, the organ inserts were filled with homogeneously distributed "^In activity and placed inside the torso phantom, which is filled with water with no radioactivity in it. The amount of activity added to the inserts and the acquisition parameters can be seen in tables 4-2 and 4-3. In the second experiment, the liver and spleen inserts were filled with small glass beads (5 mm diameter) and "^In. The beads were used to distribute the radioactivity heterogeneously within those organs. This
PAGE 91
77 Table 4-2. Phantoin Study Two Acquisition Parameters TORSO BODY PHANTOM WITH ORGAN INSERTS SOURCE: INDIUM-111 PHANTOM VOLUME fml^ LIVER 1200.00 SPLEEN 166.31 TUMOR 0.26 BODY PHANTOM 13854.42 ACQUISITION PARAMETERS: 360 degree rotation 128 views at 12 s view''' 20% window over each peak medium energy collimator 64 X 64 matrix PIXEL SIZE: 0.80 cm SLICE SIZE: 0.82 cm
PAGE 92
Liver Insert Volume = 1200 cm 3 |— 5.5cm-^ 3^ o Spleen Insert Volume = 166.31 cm^ r 0.25 cm 1.3 cm Tumor Insert Volume = 0.26 cm^ Figure 4-10. Phantom Study Two Torso Phantom and Organ Inserts
PAGE 93
79 Table 4-3. Phantom Study Two Experiments I. HOMOGENEOUS DISTRIBUTION/ COLD BACKGROUND Activity is uniformly distributed within organ inserts with no activity in body phantom (background) . PHANTOM LIVER SPLEEN TUMOR VOLUME rml^ 1200.00 166.31 0.26 ACTIVITY (MBq) 38.48 5.74 2.37 II. HETEROGENEOUS DISTRIBUTION/COLD BACKGROUND Glass beads were added to the liver and spleen inserts displacing 150 and 40 ml of liquid, respectively. No activity in body phantom. PHANTOM LIVER SPLEEN TUMOR VOLUME rml) 1200.00 166. 31 0.26 ACTIVITY rMBa^ 32.45 4.22 2.29 III. HETEROGENEOUS DISTRIBUTION/HOT BACKGROUND Glass beads were added to the liver and spleen inserts displacing 150 and 40 ml of liquid, respectively. Activity was added to the body phantom. PHANTOM LIVER SPLEEN TUMOR BACKGROUND VOLUME (ml^ 1200.00 166.31 0.26 12307.89 ACTIVITY (MBql 31.45 4.14 2.26 5.81
PAGE 94
80 experiment simulates the condition in which the patient's organ(s) has a cold (nonradioactive) tumor within it. The heterogeneously distributed radioactive organs and the homogeneously distributed tumor insert were placed in the cold (nonradioactive) water of the torso phantom and SPECT imaged. The third experiment consisted of the heterogeneously distributed radioactive liver and spleen inserts and the homogeneously distributed tumor insert from experiment two placed in the torso phantom which is filled with radioactive water and SPECT imaged. Indium-Ill (5.8 MBq) was added to the water of the torso phantom (Table 43). Each experiment was SPECT imaged three times to detect camera fluctuations. If the results varied from image to image, this would indicate that the camera system was not working properly and that the variation in results was due to an improperly working camera system. Phantom Study Three This study was undertaken to verify the results of the Dosimetry Model by comparing its results to direct measuring devices placed in the phantom and to results calculated by other methods. Indium-111 was placed homogeneously and heterogeneously in the Jaszczak Phantom**** along with several thermoluminescent dosimetry devices. The Jaszczak Phantom is a cylinder made of tissue-equivalent material Data Spectrum Corporation, High Point, NC 27514
PAGE 95
81 with a volume of 6032.50 milliliters. In the first experiment, 73 megabecquerels of ^"in was added to the water of the Jaszczak Phantom and six TLD cards of two chips each were placed on the walls in various locations in the phantom (Figure 4-11) . The phantom was SPECT imaged using the same camera setup parameters reported in the SPECT Camera System section of this chapter. The TLDs were exposed to "^In for a half an hour in the phantom. In the second experiment, a cubed insert (7 cm x 6 cm x 7 cm) filled with air was added to the radioactive water of the phantom to distribute the activity heterogeneously. Six new TLD cards were added to the walls of the phantom and the cubed insert (Figure 4-12) . The phantom was SPECT imaged using the same camera setup parameters noted above. The TLD cards, once exposed, were read by an automatic TLD reader"***. The TLD reader was calibrated using a Cesium-137 needle source; whereby, two TLD cards were exposed to the ^^'cs source for each of the four exposure times (Appendix B) . Once calibrated, the experimental TLD readings can be converted to exposure and absorbed dose. The TLD results calculated at time infinity were compared to the results of the Dosimetry Model, Geometric Factor Method (142), and the results calculated by use of data in MIRD Pamphlet No. 3 (143) at time infinity. It is assumed in each of these calculational methods TLD System 4000, Harshaw/Filtrol Partnership, Solon, OH 44139
PAGE 96
82 BACK PANEL E RIGHT ^ SIDE « PANEL BOTTOM Jaszczak Phantom TOP E u 00 E o j CARD 4 f— ? ~~-l E u I CARD 6 LEFT PANEL BACK PANEL E u
PAGE 97
83 CARD -21.6 cmTOP Organ Insert E o <£> CO BOTTOM Jaszczak Phantom CARD 7CARD8 CARD 11 Organ Insert TOP BOTTOM Figure 4-12. Phantom Study Three Experiment Two TLD Location
PAGE 98
84 (Dosimetry Model, Geometric Factor Method, and MIRD Pamphlet No. 3) that the activity in the source organs are removed only by physical decay, the effective half-life is equal to the physical half-life, and all non-penetrating radiation is absorbed in the source organ. Thermoluminescent Dosimeters (TLDs) Ionizing radiation incident on a thermoluminescent crystal elevates an electron from the valence band to the conduction band to leave a hole in the valence band. The electron and hole pair migrate throughout the crystal until they are trapped at impurity sites. When the chip is heated, energy is imparted to the electron which causes it to move and to eventually recombine with its counterpart hole (or electron) . The recombination energy is released in the form of visible light, which can be detected by a phototube. The thermoluminescent crystal used in this research is lithium fluoride, which is the most common thermoluminescent crystal used today (142) . The lithium fluoride chip has a useful range that extends beyond 10^ Sv and good linearity response, which extends below 0.1 mSv. The dynamic range of TLDs, in general, is large, with doses from a few mSv to 10 Sv. Two TLD-100 chips (0.318 cm x 0.318 cm x 0.009 cm) placed in the Type G-1 gamma card configuration""" were used in this research. Six Type G-1 cards were used in each Harshaw/Filtrol Partnership, Solon, OH 44139
PAGE 99
85 experiment and were secured in place in the phantom with a hot glue gun. Clinical Studies The results of Phase One, Two, and Three clinical studies being conducted at Bay Pines Veterans Administration Medical Center (VAMC) in Bay Pines, Florida, using indiumIll labeled B72 . 3-GYK-DTPA directed against colorectal cancer, will be utilized in this research. The goals of Phase One, Two, and Three studies are similar to those of this research, in that, they both seek to determine the radiation absorbed dose. Phase One, Two, and Three studies also seek to establish the radionuclide-antibody biodistribution, dosage range, clearance half-life, critical organs, and optimal imaging and sampling times in diseased patients (14 3) . The research at Bay Pines VAMC is being sponsored by Cytogen Corporation of Princeton, New Jersey, and is ongoing. The experimental protocol for these studies is given below. Patients Sixteen male patients participating in a phase I-III study using indium-Ill labeled B72 . 3-GYK-DTPA were screened prior to antibody infusion. The age of the patients ranged from 49-89 years with a mean age of 67 years. All subjects had
PAGE 100
86 proven primairy or were suspected of having recurrent colorectal cancer. Monoclonal Antibody The antibody, B72.3, is a murine monoclonal antibody of the IgG, subclass which detects a 200K-400K molecular weight tumor-associated glycoprotein called Tag-72. The Tag-72 antigen has been found to be expressed on certain human colon and human breast carcinoma cell lines. B72.3 is coupled to In-Ill by oxidation of the oligosaccharide moieties on the constant region of the antibody molecule. This provides a site for specific attachment of radionuclides and other ligands, while retaining the homogeneous antigen affinity and binding characteristics of the antibody. B72.3 is conjugated with the linker complex, glycyl-tyrosyl(N-e-diethylenetriaminepentaacetic acid) lysine (GYK-DTPA) to produce B72 . 3-GYK-DTPA-In-lll (Figure 4-13) . Monoclonal Antibody Procedure Each patient was given a pre-infusion diagnostic blood screening workup prior to the antibody infusion. The analyses included routine blood chemistries, hematology, electrolyte, and urinalysis, serum TAG-72, HAMA and CEA levels. Each patient was then given B72 . 3-GYK-DTPA in randomly assigned doses of 0.5, 1 or 2 milligrams of
PAGE 101
87 -s-sGYK•DTPA Figure 4-13. B72.3 Linker Complex
PAGE 102
88 antibody radiolabeled with 129.5 185.0 MBq of indium-Ill through an in-dwelling intravenous catheter (Heparin lock) by slow intravenous push over five minutes. The subject's vital signs and blood pressure was monitored prior to, and for two hours after antibody infusion. Blood Analyses Blood samples were taken via a Heparin lock in the arm opposite the infusion site at 1, 15, 30, 60, 90, 120, and 240 minutes after antibody injection. Additional blood samples were taken at 8 , 24, and 30 hours and daily for seven days. All samples were then centrifuged and counted in an automated well counter. An aliquot of representative blood samples were analyzed by size exclusion HPLC chromatography . HPLC Procedure Representative blood and urine samples were analyzed by HPLC gel filtration chromatography with a ultraviolet radioisotope detector. Samples were diluted with a running buffer (PBS) prior to injection. Five-component standards and the data were analyzed using a computer, whereby separation was based on the component's molecular weight.
PAGE 103
89 Image Analysis Each patient was scanned by an external gamma camera**"*** at two hours, and on days one, three, seven post-infusion. Planar images (10 minutes/view) were acquired in both the anterior and posterior projection over the chest, abdomen, and pelvis. Standards of known activity were imaged along with the patient. Single Photon Emission Computed Tomography (SPECT) imaging was also performed on days three and seven. ******* Technicare Omega 500, Technicare Corporation, Cleveland, OH 44139
PAGE 104
CHAPTER 5 RESULTS AND DISCUSSION The objective of this research was to develop a foundation for a dosimetry methodology which could be used to calculate absorbed doses in target and nontarget tissues using uniformly and nonuniformly distributed activity. In this proposed methodology, a computer program, SPECTDOSE, was developed to calculate the target and nontarget tissue volumes and activity concentrations. ALGAMP, a monte carlo program, was modified to determine the specific absorbed fractions. The results of the SPECT and Monte Carlo models were combined in the Dosimetry Model to determine the absorbed dose. The accuracy of the SPECTDOSE program was accessed in several phantom studies; the results of which are presented in this chapter. Validation of the results from the modified ALGAMP program was made by comparing them to the results of the MIRD0SE2 program (141), the Geometric Factor Method (144,145), MIRD Pamphlet No. 3 results (146), and TLD measurements. The results of the pixel and slice size determination studies, the three phantom studies and the clinical study are presented in this chapter. 90
PAGE 105
91 Pixel and Slice Size Determination Studies Three pixel and slice size determination studies were performed: one prior to the Clinical Study, Phantom Study Two, and Phantom Study Three. The pixel size was determined to be 0.69 cm by 0.69 cm and the slice thickness was 0.71 cm for a 64 X 64 image in the Clinical Study. For Phantom Study Two, the pixel size was determined to be 0.80 cm and the slice size was 0.82 cm for a 64 x 64 image. The pixel size was 0.96 cm and the slice size was 0.98 cm in Phantom Study Three. The pixel and slice sizes had to be determined after the camera crystal was replaced prior to Phantom Study Two and after the camera system's electronics were adjusted prior to Phantom Study Three. Phantom Study One Phantom Study One was conducted to determine the best threshold value for a given range of volumes and to determine a relationship between the SPECT measured activity concentrations and the actual activity concentrations. Several threshold values were tested to determine the one which best related the actual phantom volumes to the SPECT measured volumes (Table 5-1) . The standard error of the estimate was used to determine the best threshold value for the range of volumes in this study. For the five different cylindrical volumes in this study, a threshold value of 0.52 was determined to have the lowest standard error of the
PAGE 106
92 Table 5-1. Phantom Study One Threshold Determination Threshold Standard Error of Estimate 0.20 87.4 0.25 76.0 0.30 76.1 0.35 74.6 0.45 69.1 0.50 63.5 0.52 63.4 0.53 117.0 0.55 92.8 0.60 88.8
PAGE 107
93 estimate (63.4 ml) (Table 5-1). The SPECT measured volumes and concentrations using the threshold value of 0.52 is shown in Table 5-2. Using the threshold value 0.52, a positive correlation can be seen between the actual and SPECT measured volumes in Figure 5-1. A positive correlation is also seen between the actual activity concentration and the SPECT measured activity concentration using the threshold value of 0.52 (Figure 5-2). The correlation coefficient between the actual activity concentration and the SPECT measured activity concentration was 0.75. The increase in photon scatter within the phantoms and the enhancement of the object's boundaries by the Gaussian prefilter, which made segmentation of the SPECT image easier, may account for the excellent correlation between the actual and SPECT measured volumes. The poor correlation between the actual and SPECT measured activity concentrations is due to photon absorption attenuation and scatter which reduces the measured gamma ray intensity and thus, the SPECT measured activity concentration. The poor correlation between the actual and SPECT measured activity concentration is also attributed to the fact that the activity concentrations were essentially the same for each volume, which would produce a horizontal line with zero slope and no correlation. Thus, the low correlation coefficient is reflecting this point. Another study in which the phantom activity concentrations are varied more is
PAGE 108
94 Table 5-2. Phantom Study One Results Threshold Value
PAGE 109
95 10000 E LlI O > o < 1000 — 100 -r 10 -r^ 10 100 1000 10000 SPECT MEASURED VOLUME (ml) Figure 5-1. Phantom Study One: Actual versus SPECT Measure Volume
PAGE 110
96 cr m O < o O O _l < I— O < 0.039 0.038 — 0.037 — 0.036 — 0.035 — 0.034 100 120 140 160 180 -1 SPECT MEASURED CONCENTRATION (Cts Vox ) Figure 5-2. Phantom Study One: SPECT Measured versus Actual Activity Concentration
PAGE 111
97 recommended to adequately assess the relationship between the actual and SPECT measured activity concentrations. Phantom Study Two Phantom Study Two consisted of three experiments in which three SPECT images were acquired for each experiment. The results presented for each experiment are the average of the results determined from each of the three images. In Experiment One, the organ inserts were filled with homogeneously distributed ^^^In and placed in the nonradioactive water of the torso phantom. The phantom was SPECT imaged sequentially three times. In each of the resulting images, the tumor insert was not seen. It was determined that the Gaussian prefilter in the process of smoothing the SPECT image (removing noise) had removed the tumor from the image. The amount of activity in the tumor was very small (2.37 MBq) and the tumor image had similar image intensity values as the image background which aided the Gaussian prefilter in assuming it was image noise. The SPECT Model determined the volume and activity concentration for the remaining organs (liver and spleen) . An average threshold value of 0.53 was used for this experiment (Table 5-3) . The average difference between the actual and SPECT measured volume for the liver was 180.95 ml and 843.18 for the spleen. The SPECT Model overestimated the spleen volume and underestimated the liver volume in each image, however,
PAGE 112
98 Table 5-3. Phantom Study Two Threshold Determination Experiment Threshold Value Average Threshold Value Correlation of Coef f icent COLD HOMO-1
PAGE 113
99 the liver estimations were closer to the actual values (Table 5-4). It can be inferred from this limited observation (two data points) that there is a positive correlation between the actual and SPECT measured volumes. This is perhaps due to the fact that image segmentation is easier for larger volumes (see Image Segmentation section in Chapter 4) . A negative correlation can also be inferred between the actual and SPECT measured activity concentrations. The conclusion that there is a positive correlation between the actual and SPECT measured volumes and a negative correlation between the actual and SPECT measured activity concentrations may be invalid if more data is acquired; however, the standard deviations support these observed trends. Since there are only two data points (liver and spleen) for this experiment, the number of inferences that can be made about the experiment is limited. In Experiment Two, the results were similar to those of Experiment One, but requiring a higher average threshold value of 0.56 (Table 5-3). The increase in the average threshold value from Experiment One to Experiment Two is due to the decrease in activity in the liver and spleen inserts and the increase of scatter photons within the phantoms, which manifests itself as noise in the image. When a large amount of noise is present in an image, a higher threshold value is required to segment objects from the noisy
PAGE 114
100 Table 5-4. Phantom Study Two Experiment One Results Liver Volume Image Actual (ml) SPECT (ml) Spleen Volume Actual (ml) SPECT (ml) 1
PAGE 115
101 background (Chapter 4, Image Segmentation section). The average SPECT measured volume was 1110,7 ml for the liver insert and 1091,6 ml for the spleen insert (Table 5-5). The tumor insert was not seen in the images of this experiment because it was filtered out of the images by the Gaussian prefilter. The SPECT Model underestimated the liver volume by an average of 89.3 ml, which represents a standard error of 7.4%, and overestimated the spleen volume by an average of 925.3 ml, which is a factor of six or more higher (Table 5-5) . Thus, the actual volume correlates positively with the SPECT measured volume in this experiment. The actual activity concentration correlated positively with the SPECT measured activity concentration; however, this observation is limited by the two data points. Again, due to the limited amount of data in this experiment, any other conclusions drawn from this data would be highly speculative. Experiment Three's results were consistent with the previous experiments. This experiment is a representation of the conditions most often found in nuclear medicine patients. The activity in the background of the torso phantom represents the blood pool of the patient and the organ inserts with heterogeneously distributed activity (glass beads placed in organ inserts to distribute activity heterogeneously) represents the patient's organ following injection of radiolabeled monoclonal antibodies. The tumor insert was seen in the images of this experiment after the
PAGE 116
102 Table 5-5. Phantom Study Two Experiment Two Results Liver Volume Image Actual (ml) SPECT (ml) Spleen Volume Actual (ml) SPECT (ml) 1
PAGE 117
103 use of the Gaussian prefilter. The average threshold value for this experiment was 0.56 (Table 5-3). The threshold value of Image One was 0.67, which was higher than Images Two and Three. The threshold value of Image One was higher than expected and was due in part to the movement (settling) of the glass beads within the organ inserts while the image was being acquired. The average threshold value in this experiment would increase in comparison to the two other experiments because there was an increase in the number of scattered photons within the torso phantom. The Gaussian prefilter would smooth out the objects' edges which would make it difficult to distinguish one object's edge from another in the presence of the image noise, which would make it difficult to see the tumor insert in the SPECT image. On the contrary, the tumor insert was seen in this experiment. No explanation is offered for this contradiction. The SPECT Model overestimated the volumes for all three inserts in this experiment with the smallest overestimation occurring for the liver volume (Table 5-6). The correlation between the Actual and SPECT measured volume for this experiment was positive. The large standard deviations for the mean of the volumes was an indication of the usefulness of this result and also demonstrated a positive trend in the
PAGE 118
104 Table 5-6. Phantom Study Two Experiment Three Results Image
PAGE 119
105 data (Table 5-6) . The correlation between the actual and SPECT measured activity concentration was negative. The standard deviations for this data indicate a negative trend (Table 5-6) . As the actual activity concentration increases, the SPECT measured activity concentration decreases. The results of this experiment demonstrate that the SPECT Model overestimates the volume and underestimates the activity concentration. This result can be remedied by the use of a higher threshold value, which would allow the SPECT Model to better predict the volumes and activity concentrations . To determine what effect the Gaussian prefilter had on the images, the results of Experiment One and Two were compared to the case in which no Gaussian prefilter was used. The raw image data for both experiments was reconstructed without the use of the Gaussian prefilter. Since the Gaussian prefilter was not used, the tumor insert was seen in the images for this test. In Experiment One, without the use of the Gaussian prefilter (Cold HomogeneousNo Gaussian prefilter), a threshold value of 0.62 was determined to produce a positive correlation between the actual and SPECT measured volumes. Using this threshold value, there was a negative correlation between the actual activity concentration and the SPECT measured activity concentration. The best threshold value for Experiment Two without the use of the Gaussian prefilter (Cold
PAGE 120
106 Heterogeneous-No Gaussian prefilter) was determined to be 0.60. The correlation was positive between the actual and SPECT measured volume and negative between the actual and SPECT measured activity concentration in this case. Comparisons can now be made between the images with and without the Gaussian prefilter to assess its effect. Table 5-7 shows the results of comparing the threshold values, correlation slope signs and magnitudes, and correlation coefficients for Image Two of Experiment One and Two with and without the Gaussian prefilter. From this test it was concluded that the absence of the Gaussian prefilter necessitated the use of a higher threshold value (Table 57) . This result was expected since there is more noise present in the image when the Gaussian prefilter is not used, which makes image segmentation difficult and requires a higher threshold value to segment the objects from the image background noise. One may conclude from this Study that the Gaussian prefilter smooths the image, which reduces the amount of noise present in the image. This effect acts to reduce the threshold value required to segment the image. The results of Phantom Study Two showed that the SPECT Model overestimates the activity concentration in the organs. This overestimation is greatest for the smaller volumes. Upon further investigation of this result, it was found that the calculation of the SPECT measured activity concentration in the program SPECTDOSE had been performed
PAGE 121
107 Table 5-7. Gaussian Prefilter Comparison: Actual Versus SPECT Measured Volume COLD HOMO-2 COLD HOMO COLD HETER-2 COLD HETER Parameter w/Gauss."' w/o Gauss. '•' w/Gauss. w/o Gauss, Threshold Value 0.52 0.62 0.56 0.60 Correlation Slope Sign n/a^ (+) n/a (-) Correlation Slope Magnitude n/a 1.7 n/a 0.63 Correlation Coefficient n/a 1.00 n/a 1.00 Gaussian Prefilter Comparison: Actual Versus SPECT Measured Activity Concentration COLD HOMO-2* COLD HOMO COLD HETER-2" COLD HETER PARAMETER w/Gauss.*" w/o Gauss. ''^ w/Gauss. w/o Gauss. Threshold Value 0.52 0.62 0.56 0.60 Correlation Slope Sign n/a (-) n/a (+) Correlation Slope Magnitude n/a 0.69 n/a 0.76 Correlation Coefficients n/a 1.00 n/a 0.83 Cold Homogeneous-Image 2 Cold Heterogeneous-Image 2 With Gaussian prefilter "*' Without Gaussian prefilter ^ n/a Not Applicable (only two data points)
PAGE 122
108 incorrectly. In the Subroutine THOLD of SPECTDOSE, the userentered threshold value is multiplied by the maximum image count and any pixels having values greater than this result, are considered to be apart of the object. This method is correct for volume calculations, but not for activity concentration calculations. For activity concentration calculations, the resulting value of the threshold value times the maximum image count is subtracted from each pixel's value and any pixel with a value greater than this result is considered to be apart of the object. A modification was made to the program SPECTDOSE, but unfortunately, the results were the same, i.e. the SPECT Model overestimates the activity concentration in the organs. Since the activity concentrations utilized in this study were determined using the relation between the actual and SPECT measured activity concentrations determined in Phantom Study One, the conclusions stated above may or may not be valid. Since the relationship developed between the actual and SPECT measured activity concentrations in Phantom Study One was only accurate for the limited activity concentrations, it may not be appropriate for the wider range of activity concentrations utilized in this study. However, this can be easily remedied by conducting another phantom study as in Phantom Study One in which the range of activity concentrations of interest are used.
PAGE 123
109 The accuracy of the Dosimetry Model developed in this research was tested by comparing the results of this study using the Dosimetry Model to those using the MIRD0SE2 program (141). In MIRD0SE2 , the liver and spleen are used both as organs of source activity and targets. The liver and spleen residence times, which are required for the MIRD0SE2 calculation, were calculated by using the organ anterior and posterior counts determined from Planar imaging taken at specific times and then fitting this data after decay correction to an exponential function by using a subroutine supplied with the MIRD0SE2 program. The Dosimetry Model requires the use of the pixel and slice sizes determined in the Pixel and Slice Size Determination Study. The results of both programs are presented in Table 5-8 for this study. In Experiment One, the percent difference between the Dosimetry Model and MIRD0SE2 absorbed dose result was ten percent for the liver and four percent for the spleen. The average absorbed dose in the liver was 141 mGy and 13 mGy in the spleen as calculated by the Dosimetry Model (Table 58). In Experiment Two, the average dose in the liver was 111 mGy and 109 mGy in the spleen as calculated by the Dosimetry Model (Table 5-8) . In this experiment, the percent difference between the Dosimetry Model and MIRD0SE2 absorbed dose was three percent for the liver and nine percent for the spleen (Table 5-8) .
PAGE 124
110 Table 5-8. Phantom Study Two Absorbed Dose Results EXPERIMENT 1:
PAGE 125
Ill In Experiment Three, the Dosimetry Model had a percent difference of 36% for the liver absorbed dose to that of MIRDOSE2. The Dosimetry Model and MIRDOSE2 differed by a factor of three in the calculation of absorbed dose for the spleen (Table 5-8) . The Dosimetry Model also calculated the absorbed dose for the tumor insert in this experiment, which was 1208 mGy (Table 5-8) . The MIRD0SE2 program does not calculate tumor absorbed doses. The Dosimetry Model absorbed dose estimate for the liver differed by as much as 36% to that of the MIRD0SE2 absorbed dose estimate and by as much as a factor of two for the spleen. This difference may be explained by the use of different organ volumes in the calculation of the absorbed doses by each method. The MIRD0SE2 program uses the organs of Reference Man in which the liver is 1800 ml and the spleen is 150 ml (103) . The Dosimetry Model used a smaller liver volume of 1200 ml and a larger spleen volume of 166.31 ml. A volume difference of 50% is found between each method's liver volume and a volume difference of 11% is found between the spleen volumes. The absorbed dose for the Dosimetry Model was higher than the MIRD0SE2 absorbed dose because of its higher nonpenetrating dose contribution due to the smaller liver volume used (specific absorbed fraction equals 1/m) . And accordingly, the MIRD0SE2 spleen absorbed dose was higher in Experiment One and Two. The volume differences do not completely account for the difference in
PAGE 126
112 absorbed doses for the liver and spleen by the two methods, therefore, other causes will be examined in the next phantom study. Phantom Study Three Phantom Study Three was conducted to test the accuracy of the Dosimetry Model. The Jaszczak Phantom was filled with homogeneously and heterogeneously distributed activity and thermoluminescent devices (TLDs) . A threshold value of 0.47 was found to give the best estimate of the Jaszczak Phantom volume (6103.8 ml) in Experiment One, which is the homogeneously distributed activity case. Experiment Two's, (heterogeneously distributed activity) , threshold value was 0.45 with a calculated volume of 6024.2 ml. The volume percent difference was less than 1.2% in Experiment One and 0.1% in Experiment Two. In Experiment One, the TLDs were placed in the phantom (cards one through six) and the phantom was filled with homogeneously distributed "^In in water (Figure 5-3) . The measured TLD absorbed doses and rates for this experiment are shown in Table 5-9. The average absorbed dose rate for Experiment One was 2,84 ± 0.41 mGy h''' (278 mGy) . In Experiment Two, the TLDS were placed in the phantom and on the organ insert (cards six through twelve) , and the phantom was filled with heterogeneously distributed "^In in water (Figure 5-3). The average absorbed dose rate for Experiment Two was 1.84 ±
PAGE 127
113 BOTTOM TOP r E o 00 E u CARD 4 } E u T CARD 6 LEFT PANEL BACK Jaszczak Phantom CARD7—
PAGE 128
114
PAGE 129
115 0.56 mGy h"^ (181 mGy) (Table 5-9). This result is consistent, in that, the average absorbed dose was lower because of the smaller amount of activity in the phantom, which was displaced by the organ insert. The average TLD absorbed dose rates were integrated out to infinity for each experiment to derive the total absorbed dose. It is assumed in this integration that the activity is removed only by decay and the effective half-life is equal to the physical half-life. The TLD results of this study were compared to the results calculated for this study by the use of the Geometric Factor Method (144,145) and the results found using data in MIRD Pamphlet No. 3 (146) for a similar phantom. In the Geometric Factor Method, the integral equation that represents the volume of interest, a cylinder in this study, located at a distance r from the point where the absorbed dose is desired, is approximated by an average geometric factor. The average geometric factor has been calculated and tabulated for cylinders of different heights and diameters (144). The average absorbed dose for gamma photons in the cylinder is calculated by multiplying the activity concentration of the cylinder by the specific gamma-ray emission and by the average geometric factor (145) and is tabulated in the reference for various radioactive sources. The specific gamma-ray emission is the gamma
PAGE 130
116 radiation exposure rate from a point source of unit activity at a unit distance (145) . It is assumed in this method that the source activity is uniformly distributed (144,145). Tabulated in MIRD Pamphlet No. 3 are absorbed fractions for various sized spheres which contain uniformly distributed gamma emitting activity. Absorbed fractions for a sphere of six kilograms approximated the cylinder used in this study. The chosen absorbed fractions are put into the MIRD Formula (74) to determine the average absorbed doses. The results of the Geometric Factor and MIRD Pamphlet No. 3 methods are seen in Table 5-10 and are reported as average absorbed doses at infinity. The total absorbed doses include the assumptions that the activity is removed only by physical decay, the effective half-life is equal to the physical half-life, and all non-penetrating radiations are absorbed in the target organ. Phantom Study Three was also analyzed by the Dosimetry Model. The resulting absorbed dose for Experiment One was 58 mGy and for Experiment Two, 51 mGy after 10000 histories (Table 5-11) . The coefficient of variation for the specific absorbed fractions was one percent for both Experiment One and Two. In the Dosimetry Model, the absorbed doses were calculated using the MIRD Formula (74) after the specific absorbed fractions were determined using the Monte Carlo Model (Appendix A) .
PAGE 131
117 Table 5-10. Phantom Study Three Geometric Factor Method and MIRD Pamphlet No. 3 Results Experiment
PAGE 132
118 Table 5-11, Phantom Study Three Dosimetry Model Results Experiment 1 Experiment 2 Activity Ag (MBq) 43.40 41.44 Photon Histories 10000 10000 Specific Absorbed Fraction (kg'^) 0.03 0,03 Coefficient of Variation (%) 1.1 1.2 Absorbed Dose (mGy) 58.3 51.1
PAGE 133
119 A comparison can now be made between the four methods for determining absorbed dose; whereby, each method's total absorbed dose values are shown in Table 5-12. The absorbed doses calculated using the calculational methods compared well with each other; however, the TLD measurements were a factor of two or more greater than these other methods. If the average absorbed doses calculated by the use of the data in MIRD Pamphlet No. 3 are considered the standard by which all calculations will be compared, the standard error between the results of using MIRD Pamphlet No. 3 and the other methods and measurements are presented in Table 5-13. The Geometric Factor Method underestimated the absorbed dose in Experiment One by one percent and overestimated the absorbed dose in Experiment Two by six percent. The Dosimetry Model underestimated the absorbed dose by 10% and 11% in Experiment One and Two respectively. The TLD measurements overestimated the absorbed dose by a factor of four and three for experiments One and Two respectively. A complete error analysis was initiated to determine the cause and contribution of errors in this study, especially in the TLD measurements. It was discovered that the error of greatest significance was found in the TLD measurements and was associated with the TLD reader and its initial calibration. A calibration study; whereby, several TLDs were exposed to known amounts of activity were read by the same TLD reader as used in Phantom Study Three, was conducted to
PAGE 134
120 Table 5-12. Phantom Study Three Results Experiment
PAGE 135
121 Table 5-13. Phantom Study Three Error Analysis Standard Error between Standard Error* TLD Average Experiment (%)
PAGE 136
122 estimate the amount error associated with the TLD reader. The results of this study, which was conducted as a blind study; i.e., the TLD exposure was unknown to the TLD reader operator prior to reading the exposed chips, are presented in Table 5-14. At the lower exposures, the TLD reader overestimated the exposures by as much as 52% and at the higher exposures, it overestimated the exposures by 39%. This accounts for most of the error found between the TLD measurements and the MIRD Pamphlet No. 3 calculation; however, the TLD measurements were still one order of magnitude above the calculated results using MIRD Pamphlet No. 3. The remaining error was attributed to the dose calibrator (~3%) , the monte carlo calculation in MIRD Pamphlet No. 3 (-2%) , and to the TLD exposure time (-20%) ; however, since the exact cause of the TLD measurement error could not be identified, the results of this experiment will not be utilized in this research. A second phantom study using the same setup as in Phantom Study Three was conducted, but with TLD chips which were read by a more reliable TLD chip reader and in which all parameters, such as, the exposure time and packet drying were monitored more closely. Two TLD chips (lithium fluoride) were wrapped in black paper to limit exposure to light and sealed in polyethylene plastic (Figure 5-4). Seventeen chip packets were adhered to various locations throughout the inside of the Jaszczak
PAGE 137
123 Table 5-14. TLD Calibration Study Actual Measured Standard Exposure Exposure Error (mR) (mR) (%) 55.0 83.6 52.0 55.0 76.4 38.9 55.0 74.2 34.9 Bkg* Bkg . 99.0 121.6 22.8 99.0 122.0 23.2 99.0 145.9 47.3 Bkg Bkg . 148.0 191.3 29.2 148.0 193.7 30.9 148.0 215.0 45.3 Bkg Bkg . 198.0 247.5 38.7 198.0 275.0 38.9 198.0 252.1 27.3 Bkg Background Exposure
PAGE 138
124 Polyethelene Plastic Black Paper Polyethelene Plastic Black Paper Sealed Chip Packet Sealed Polyethelene Packet Chip 1 Chip 2 Figure 5-4. Phantom Study Four TLD Chip Packagi ng
PAGE 139
125 Phantom by a hot glue gun (Figure 5-5) . One millicurie of ''^^In was homogeneously distributed in water which was then added to the phantom. The phantom was SPECT imaged using the same setup parameters as used in Phantom Study Three. The chips were exposed to the uniformly distributed activity for an hour during the phantom preparation and SPECT imaging. Each chip was read for 3 seconds using a TLD reader* The resulting exposure, exposure rate, absorbed dose rate, and total absorbed dose at time infinity for each chip packet is shown in Table 5-15. The average absorbed dose for this study was 70 ± 12 mGy (Table 5-15) . The centerline chip packets had the highest average absorbed dose rate (0.84 mGy h"^) and the right panel chip packets had the lowest absorbed dose rate (0.53 mGy h'^) . The phantom was imaged on its side (axis parallel to the camera's AOR) , with the right panel facing upward and the left panel on the bottom. The activity settled to the bottom of the phantom to expose the left panel more than the right panel. The TLD results were compared to the calculated results using the Geometric Factor Method, MIRD Pamphlet No. 3, and the Dosimetry Model (Table 5-16) . Using the MIRD Pamphlet No. 3 results as the standard, the TLDs overestimated the total absorbed dose by 30%; however, this is accounted for in the standard deviation of the mean TLD absorbed dose Harshaw 2000 A/B, Hawshaw/Filtrol Partnership, Solon, OH 44139
PAGE 140
T E u
PAGE 141
127 Table 5-15. Phantom Study Four TLD Results Location Experiment 1 Absorbed Dose Rate Total (mGy h"^) (n»Gy) Back Panel
PAGE 142
128 Table 5-16. Phantom Study Four Results
PAGE 143
129 reading, which was 12 itiGy (Table 5-17) . The Geometric Factor Method underestimated the total absorbed dose by one percent and the Dosimetry Model underestimated the absorbed dose by three-tenths of a percent (Table 5-17) . Again, the TLD measurements were higher, but not by a factor of two or more. The higher TLD results are attributed to exposure times which were longer than those accounted for. The MIRD Pamphlet No. 3 results were expected to be higher than the Geometric Factor Method and the Dosimetry Model because it used spherical geometry, which was more efficient at absorbing dose than cylinders. There was good agreement among all of the absorbed dose method for this experiment. This experiment clearly showed the excellent agreement between the Dosimetry Model results and the other method results and proved that the Dosimetry Model is an accurate methodology for calculating absorbed doses in tissues whose activity is distributed homogeneously. Clinical Study The proposed dosimetry methodology in this research, the Dosimetry Model, was used to estimate the absorbed dose in the organs of one patient. Data from one patient participating in the clinical study being conducted at Bay Pines Veterans Medical Center was available for analysis. The absorbed dose results of patient ABG using MIRD0SE2 and the Dosimetry Model can be seen in Table 5-18. The patient
PAGE 144
130 Table 5-17. Phantom Study Four Error Analysis Standard Error between MIRD Pamphlet No. 3 Results : Standard Standard Standard Error* Error Error TLD Geometric Factor Dosimetry Average Method Model Experiment (%) (%) (%) 30.5 1.4 0.32 Standard Error = 100 x (X MIRD Pamphlet No. 3 value) MIRD Pamphlet No. 3 value X = TLD, Geometric Factor, or Dosimetry Model value
PAGE 145
131 Table 5-18. Clinical Study Results Patient: ABG MIRD0SE2 MIRD0SE2 Organ Absorbed Dose Volume (mGy) (ml) Liver 63.7 Spleen 113.1 Tumor n/a* Dosimetry Model Absorbed Dose (mGy) 1800 547.6 150 732.5 n/a 2421.5 Dosimetry Model Volume (ml) 301.9 103.2 8.9 n/a Not applicable (MIRD0SE2 does not calculate tumor absorbeds doses)
PAGE 146
132 was injected with 5.49 itiCi of indium-Ill labeled B72.3-GYKDTPA. As seen in Table 5-18, the MIRDOSE2 program underestimated the absorbed doses for both the liver (factor of nine) and spleen (factor of 7) and did not calculate the absorbed dose for the tumor. The underestimation of the absorbed doses for the liver and spleen by the MIRD0SE2 program was due to the use of larger liver and spleen volumes (Table 5-18) . The actual smaller volume resulting from tumor specific concentrations by the monoclonal antibody results in an expected higher absorbed dose due to the higher nonpenetrating absorbed dose component. The Dosimetry Model is felt to give a better representation of the absorbed doses in these organs.
PAGE 147
CHAPTER 6 SUMMARY AND CONCLUSIONS The objective of this research was to develop a foundation for a dosimetry methodology that could be used to calculate absorbed doses in target and nontarget tissues using uniformly and nonuniformly distributed activity. In this proposed methodology, a dosimetry model was developed which used Single-Photon Emission Computed Tomography (SPECT) to determine the volume and radioactive uptake in the tissues and a Monte Carlo method to determine the amount of energy deposited in the tissues. The dosimetry model was divided into three independent models (the SPECT Model, the Monte Carlo Model, and the Dosimetry Model); whereby, each model completed a specific task, in the SPECT Model, a computer program, SPECTDOSE, was developed to calculate the target and nontarget tissue volumes and activity concentrations using a edge detection method and contour tracing algorithm. The edge detection method utilized in this research (Threshold Segmentation Method) required the use of a threshold value, a percentage of the SPECT image's maximum intensity value, to segment and extract the tissue volumes of interest from the SPECT image 133
PAGE 148
134 for volume and activity quantitation. In the Monte Carlo Model, a monte carlo program, ALGAMP, was modified to determine the specific absorbed fractions. The results from the SPECT and Monte Carlo models were combined in the Dosimetry Model to determine the target and nontarget tissues absorbed dose. Several phantom studies were conducted to determine the accuracy of the Dosimetry Model and to verify the ability of the Dosimetry Model to evaluate organ and tumor uptake volumes, sizes, and to calculate absorbed doses. Comparisons were made between the Dosimetry Model, other calculational methods (MIRD0SE2, Geometric Factor Method, MIRD Pamphlet No. 3), and TLD measurements. The Standard Error of the Estimate was used to determine the best image threshold value, which was found to be affected by image noise. A higher threshold value was required to segment the image as the background image noise was increased. The image threshold value was determined for each experimental condition and range of volumes of interest. For organ doses consistent with a diagnostic administration, the SPECT Model was found to calculate organ volumes, which were of the order of 1000 ml, to within fifteen percent of the actual volumes. The SPECT Model failed to accurately calculate organ volumes of approximately 200 ml or less with the use of the diagnostic
PAGE 149
135 administration dose levels. This was due in part to the difficulty of discerning the smaller volumes from the image background noise using the Threshold Segmentation Method. Since the organ or tissue volumes were discerned on the basis of the amount of activity they contained (image intensity values) , it is felt that at therapeutic dose administration levels, the SPECT Model would more accurately calculate organ size for the smaller tissue volumes. The higher tissue activities found with the use of therapeutic dose administration levels would create a large edge gradient between the tissue volumes and the image background noise, which would facilitate image segmentation by the Threshold Segmentation Method. Other image segmentation methods, such as the Gradient and Histogram methods, might be used to give a better estimate of the organ and tumor volumes when computer time and speed is not a concern. A combination of the Threshold Segmentation and Histogram Methods will be attempted in the future of this research. The activity concentrations were not varied enough to establish a meaningful relationship between the actual and SPECT measured activity concentrations. The radiation absorbed dose determined with the Dosimetry Model agreed within 12% to that determined by , other calculational methods (Geometric Factor Method, MIRD Pamphlet No. 3 results) using homogeneously and heterogeneously distributed ^"in. The TLD measurements were
PAGE 150
136 within 3 0% at most of the other methods (Dosimetry Model, Geometric Factor Method, MIRD Pamphlet No. 3) . The results of the several phantom studies indicated the Dosimetry Model was working properly and that it is an appropriate methodology for calculating absorbed doses for homogeneously distributed activity. Further investigation is needed to determine the accuracy of the Dosimetry Model when heterogeneously distributed activity is used. The MIRD0SE2 Program has been widely used for absorbed dose evaluation in nuclear medicine. In MIRD0SE2, the activity is assumed to be uniformly distributed in the source organ, which is not the case when radiolabeled monoclonal antibodies are used. This assumption, thus makes it inappropriate for use with heterogeneously distributed activity. The methodology proposed in this research provides a means of calculating absorbed doses for heterogeneously distributed activity. In comparison to MIRD0SE2 , which uses the fixed organ sizes of Reference Man, the research Dosimetry Model is patient specific. The Dosimetry Model uses the organ tumor volumes of the patient of interest and calculates the specific absorbed fractions for those specific tissue volumes. MIRD0SE2 has no mechanism for calculating absorbed doses for tumors or nonstandard organs. The Dosimetry Model only requires one patient SPECT image at each time point; whereby, the MIRD0SE2 program requires several planar images to be taken at each time point to
PAGE 151
137 assess the organ uptake. Thus, the Dosimetry Model reduces patient discomfort, which accompanies long image times and reduces the patient load, because the patient does not have to return for additional images. It is clear that the Dosimetry Model has several advantages over the MIRD0SE2 approach. Corrections for photon absorption attenuation (nonlinear) and scatter and the inclusion of electron transport mechanisms were not accounted for in the development of this Dosimetry Model. Their inclusion should be considered in any further refinement of the methodology. Although not investigated in this study, new methods for determining photon absorption attenuation and scatter have been proposed. These methods include the use of simultaneous transmission and SPECT imaging and photon absorption attenuation and scattering computer simulations. The transmission image maps the object's attenuation and the buildup factors necessary for photon scatter correction (147,148). A monte carlo program has been developed which simulates photon absorption attenuation and scattering in a patient (148) . The results of this program are convolved with the SPECT image raw data to remove the effects of photon absorption attenuation and scatter. It was assumed in all the calculational methods utilized in this research that all nonpenetrating radiation was absorbed in the source organ. This assumption is
PAGE 152
138 significant, in that, it overestimates the nonpenetrating absorbed dose contribution for small volumes in both the homogeneous and heterogeneous activity distribution case. Thus, it is pertinent that nonpenetrating radiation transport be included in the dosimetry methodology to accurately assess the absorbed dose in target and nontarget tissues using uniformly and nonuniformly distributed activity. An electron transport code developed at the Stanford Linear Acceleration Center (149) is being considered for future incorporation into the Dosimetry Model developed in this research. A significant problem facing this research was the assessment of the accuracy of the absorbed dose results calculated by the research model. Phantoms were employed to assess the accuracy of the absorbed doses determined by the Dosimetry Model by comparing the results measured with TLDs and that calculated by other calculational methods (MIRDOSE2, Geometric Factor Method, MIRD Pamphlet No. 3). The phantom absorbed dose assessment for a homogeneous activity distribution of ^"in showed close agreement between the four methods. No comparisons were made for the heterogeneously distributed activity case. Only one patient was analyzed using the Dosimetry Model developed in this research; the results of which, differed greatly from those found using MIRD0SE2 . No direct or indirect method of measurement in a living patient was
PAGE 153
139 available for comparison with the calculational methods. Since the long term goal of the dosimetry methodology proposed in this research is to use it in human clinical studies, verification of the Dosimetry Model's results in humans will be difficult. In this research, a foundation was developed for a dosimetry methodology that could be used to calculate absorbed doses in target and nontarget tissues using uniformly and nonuniformly distributed activity. This proposed methodology showed good agreement with the current dosimetry methods for uniformly distributed activity. But before this methodology could be used in clinical studies, several shortcomings must be resolved. This can be done by by improving the organ and tumor volume estimates by the use of other image segmentation methods or a combination of methods. The addition of photon absorbtion attenuation and scatter correction and nonpenetrating radiation transport should be considered. Phantom studies in which the range of volumes and activity concentrations simulate the range of clinical interest should also be considered.
PAGE 154
140 REFERENCES 1. Rubin, P:Statement of the clinical oncologic problem. In: Clinical OncolooY . Rubin P, ed. , Atlanta, Ga., American Cancer Society, 1987, pp2-19. 2. Pressman D: Radiolabeled antibodies. Ann NY Acad Sci 69:644-650, 1957. 3. Kohler G,Milstein C:Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256:495-497, 1975. 4. Milstein C:Monoclonal antibodies. Sci Am 243 (4) : 66-74 , 1980. 5. Humm JL: Dosimetric aspects of radiolabeled antibodies for tumor therapy. J Nucl Med 27:1490-1497, 1986. 6. Zola H: Monoclonal Antibodies; A Manual of Techniques . Boca Raton, Florida, CRC Press, Inc., 1987. 7. Keenan AM,Harbert JC, Larson SM:Monoclonal antibodies in nuclear medicine. J Nucl Med 26:531-537, 1985. 8. Ross NR,Milgrom F,Van Oss CJ; Principles of Immunology . New York, Macmillan Publishing, 1979. 9. Weir DM: Immunology . New York, Churchill Livingstone, 1977. 10. Fawwaz RA,Wang TS , Srivastava SC, Hardy MA:The use of radionuclides for tumor therapy. Nucl Med Biol 13(4) :429-436, 1986. 11. Wessels BW,Rogus RD: Radionuclide selection and model absorbed dose calculations for radilabeled tumor associated antibodies. Med Phys 11 (5) : 638-645, 1984. 12. Williams LE,Duda RB,Proffitt RT,Beatty BG,Beatty JD, Wong JYC, Shively JE, Paxton RJ: Tumor uptake as a function of tumor mass:A mathematic model. J Nucl Med 29:103-109, 1988.
PAGE 155
141 13. Hagan PL,Halpern SE,Dillman RO,Shawler DL, Johnson DE, Chen A,Krishnan L,Frinc}ce J, Bartholomew RM, David GS, Carlo D: : Tumor size: Effect of monoclonal antibody uptake in tumor models. J Nucl Med 27:422-427, 1986. 14. Pimm MV, Baldwin RW: Effect of tumor size on monoclonal antibody uptake in tumor models. J Nucl Med 27:17881789, 1986. 15. Macey DJ,DeNardo SJ,DeNardo GL,DeNardo GL, Goodnight JK, Unger MW:Uptake of indium-lll-labeled monoclonal antibody ZME-018 as a function of tumor size in a patient with melanoma. Amer J Phvsl Imaq 3:1-6, 1988. 16. Cohen MB,Saxton RE, Mann B: Effect of tumor size on monoclonal antibody uptake in tumor models. J Nucl Med 27:1789-1790, 1986. 17. Larson SM: Lymphoma, melanoma, colon cancer: Diagnosis and treatment with radiolabeled monoclonal antibodies. Radio 165:297-304, 1987. 18. Douillard J-Y,Chatal J-F, Saccavini JC,Curtet C,Kremer M,Peuvrel P,Koprowski H: Pharmacokinetic study of radiolabeled anti-colorectal carcinoma monoclonal antibodies in tumor-bearing nude mice. Eur J Nucl Med 11:107-113, 1985. 19. Ballou B,Reiland J,Levine G:Tumor location using F(ab')2 from a monoclonal IgM antibody : Pharmacokinetics. J Nucl Med 26:283-292, 1985. 20. Eger RR,Covell DG, Carrasquillo JA,Abrams PG,Foon KA, Reynolds JC,Schroff RW, Morgan AC, Larson SM,Weinstein JN:Kinetic model for the biodistribution of an In-111labled monoclonal antibody in humans. Cancer Res 47:33283336, 1987. 21. Hnatowich DJ, Griffin TW,Kosciuczyk C,Rusckowski M,Childs RL,Mattis JA,Shealy D,Doherty OW: Pharmacokinetics of In111-labeled monoclonal antibody in cancer patients. J Nucl Med 26:849-858, 1985. 22. Fraker PJ , Speck JC: Protein and cell membrane iodinations with a sparingly soluble chloramide 1, 3,4, 6-tetrachloro 3a-6a-diphenyl-glycouril . Biochem Biophvs Res Commun 80:849-857, 1978. 23. Gobuty AH, Kim EE,Weiner RE:Radiolabeled monoclonal antibodies: Radiochemical pharmacokinetic and clinical challenges. J Nucl Med 26:546-548, 1985.
PAGE 156
142 24. Hnatowich DJ,Virzi F,Doherty PW: DTPA-coupled antibodies labeled with yttriuin-90. J Nucl Med 26:503-509, 1985. 25. Larson SMiRadiolabeled monoclonal anti-tumor antibodies in diagnosis and therapy. J Nucl Med 26:538-545, 1985. 26. Goodwin DA: Pharmacokinetics and antibodies. J Nucl Med 28:1358-1362, 1987. 27. Gullino PM, Grantham FH: Studies on the exchange of fluids between host and tumor. II. The blood flow of hepatomas and other tumors in rats and mice. J Natl Cancer Inst 27:1465-1491, 1961. 28. Hosain F,Hosain P:Selection of radionuclides for therapy. In: Therapy in Nuclear Medicine . Spencer RP, ed., New York, Grune & Stratton, 1978, pp33-43, 29. Strudler PK, Larson SM:Radiolabeled monoclonal antibodies: A "decisive " technology. J Nucl Med 13: 46-52, 1985. 30. d'Agincourt L:New radiopharmaceuticals opening doors for SPECT. Diaq Imaq June:94-100, 1986. 31. Engelstad BL, Ramos EC, Stoudemire J,0'Connell JW, Villanueve J, Faulkner DB,Hattner RS,Spitler LE,Scannon P: Improved immune-specificity in monoclonal radioimmunoiraaging using dual radionuclide color functional maps. Invest Radiol 21:917-921, 1986. 32. Keenan AM,Weinstein JN,Mulshine L, Carrasguillo JA,Bunn PA, Reynolds JC, Larson SM: Immunolymphoscintigraphy in patients with lymphoma after subcutaneous injection of indium-lll-labeled TlOl monoclonal antibody. J Nucl Med 28:42-46, 1987. 33. Nelp WB,Eary JF, Jones RF,Hellstrom KE,Beaumier PL,Krohn KA: Preliminary studies of monoclonal antibody lymphoscintigraphy in malignant melanoma. J Nucl Med 28:34-41, 1987. 34. Perkins AC, Pimm MV: Differences in tumour and normal tissue concentrations of iodineand indium-labelled monoclonal antibody: The effect on image contrast in clinical studies. Eur J Nucl Med 11:295-299, 1985. 35. Order SE, Stillwagon GB, Klein JL,Leichner PK,Siegelman SS,Fishman EK,Ettinger DS,Haulk T,Koper K, Finney K, Surdyke M,Self S,Leibel S: Iodine 131 antif erritin, a new treatment modality in hepatoma: A radiation therapy oncology group study. J Clin Oncol 3:1573-1582, 1985.
PAGE 157
143 36. Washburn LC,Hwa Sun TT, Crook JE,Byrd BL, Carlton JE Hung Y-W,Steplewski ZS r'^Y-labeled monoclonal antibodies for cancer therapy. Nucl Med Biol 13:453-456,1986. 37. Childs RL,Hnatowich DJrOptimum conditions for labeling of DTPA-coupled antibodies with Technetium-99in. J Nucl Med 26:293-299, 1985. 38. Rhodes BA,Zamora PO, Newell KD,Valdez EF:Technetiuin-99in labeling of murine monoclonal antibody fragments. J Mucl Med 27:685-693, 1986. 39. Goldenberg DM, Ford EH, Lee RE, Hall TC, Horowitz JA, Sharkey RM, Hansen HJ:A new 99m-Tc labeling system for cancer radioimmunodetection. Fourth International Conference on Monoclonal Antibody Immunoconiuqates for Cancer . San Diego, March 1989, p24. 40. Chen DCP, Pascal SG,Siegel ME, Liggett PE,Kempf RA, Payne D, Smith R: Radioimmunodetection of uveal and cutaneous melanomas with Tc99m labeled monoclonal antibody. Fourth International Conference on Monoclonal Antibody Twnniinn coniuqates for Cancer . San Diego, March 1989, pl26. 41. Leichner PK, Klein JL, Garrison JB, Jenkins RE,Nickoloff EL, Ettinger DS, Order SE: Dosimetry of ^^^-I-labeled antiferritin in hepatoma: A model for radioimmunoglobulin dosimetry. Intl J Oncol Biol Phvs 7:323-333, 1981. 42. Freiherr G:Monoclonal antibodies against cancer:Clinical application. Res Resources Rpt 8:1-8, 1984. 43. Harrison A,Royle L: Preparation of a ^"At-IgG conjugate which is stable in vivo. Int J Appl Radiat Isot 35:1005-1008, 1984. 44. Vaughan ATM,Bateman WJ, Fisher DR:The in vivo fate of ^^^At labeled monoclonal antibody with known specificity in a murine system. Int J Radiat One Biol Phvs 8:19431946, 1982. 45. Macklis RM, Kinsey BM,Kassis AI,Ferrara JLM,Atcher RW, nines JJ, Coleman CN,Adelstein SJ,Burakoff S J : Radioimmunotherapy with alpha-particle-emitting immunoconjugates. Science 240:1024-1026, 1988.
PAGE 158
144 46. Simonson RB,Ultee ME, Hauler JA, Alvarez VL: Radio immunotherapy of peritoneal human colon cancer xenografts with site-specifically 212Bi-labeled antibody. Fourth International Conference on Monoclonal Antibod y TTnTnn noconiuaatea for Cancer . San Diego, March 1989, pl42. 47. Hnatowich DJ,Chinal M,Siebecker DA,Gionet M, Griffin T, Doherty PW, Hunter R,Kase KR: Patient biodistribution of intraperitoneally administered Yttrium-90-labeled antibody. J Nucl Med 29:1428-1434,1988. 48. Royston I, Parker BA,Vassos A,Hupf H,Amox D, Green MR, Miller R, Halpern S:'°Yttrium anti-idiotype monoclonal antibody therapy of Non-Hodgkin' s lymphoma . Fourth International Conference on Monoclonal Antibody Immunoconjuqates for Cancer . San Diego, March 1989, pl42. 49. Vriesendorp HM,Stinson R,Onyekwere 0,Leichner P, Klein J, Wessesls B, Williams J: Bone marrow toxicity of polyclonal antiferritin in rat, dog and man. Fourth International Conference on Monoclonal Antibody Immunoconjuqates for Cancer . San Diego, March 1989, pl42. 50. Crook JE, Washburn LC,Lee Y-CC,Byrd BL,Sun TTH, Steplewski Z:Tissue distribution of '°Y-labeled monoclonal antibody C017-1A in young and old athymic mice. Fourth International Conference on Monoclonal Antibody Immunoconiuqates for Cancer . San Diego, March 1989, pl42. 51. DeNardo SJ,Meares CF,Deshpande SV,Moi MK,McCall MJ:Remarkably stable metal chelates for radioimmunotherapy with Cu-67 and Y-90. Fourth International Conference on Monoclonal Antibody Immunoconiuqates for Cancer . San Diego, March 1989, pl42. 52. Lee Y-CC, Washburn LC,Sun TTH,Byrd BL, Crook JE, Steplewski Z : Radiotherapy of human colorectal carcinoma xenografts using Y-90-labeled C017-1A. Fourth International Conference on Monoclonal Antibody Immunoconiuqates for Cancer. San Diego, March 1989, pl42. 53. DeNardo SJ,Jungerman JA, DeNardo GL, Laguanas-Solar MC, Cole WC,Meares CF:The choice of radionuclides for radioimmunotherapy. Proceedings of the International Symposium of the Developinq Role of Short-lived Radionuclides in Nuclear Medicine Practice . Davis, CA. , DOE Publication, pp401-414.
PAGE 159
145 54. Fawwaz RA,Wang TST,Srivastava AC, Rosen JM,Ferrone S, Hardy MA,Alderson PO: Potential of Palladium-109-labeled antimelanoma monoclonal antibody for tumor therapy. J Nucl Med 25:796-799, 1984. 55. Schroff R,Hanelin L, Vanderheyden J-L,Fer M,Weiden P, Breitz H, Fisher D,Abrams P,Ratliff B,Appelbaum J, Morgan CFritzberg A: Preliminary clinical evaluation of a Re-186 labeled anti-CEA F(ab')2 antibody fragment as a potential radioimmunotherapy agent. Fourth International Conference on Monoclonal Antibody im mnwftgiQ niuqates for Cancer . San Diego, March 1989, pl42. 56. Khaw BA,Bailes JS, Schneider SL, Lancaster J, Powers J, Strauss HW,Laster JC,McGuire WL: Human breast tumor imaging using ^^^In labeled monoclonal antibody lAthymic mouse model. Eur J Nucl Med 14:362-366, 1988. 57. Pedley RB,Boden J , Keep PA,Harwood PJ, Green AJ, Rogers GT: Relationship between tumour size and uptake of radiolabeled anti-CEA in a colon tumour xenograft. Eur J Nucl Med 13:197-202,1987. 58. Davis KH:Use of Gallium-67 citrate and Indium-Ill monoclonal antibodies in the detection of metastatic melanoma. J Nucl Med Tech 15:25-29, 1987. 59. Matzku S , Schuhmacher J, Kirchgebner H,Brugger J: Labeling of monoclonal antibodies with a '^^Ga-phenolic aminocarboxylic acid chelate. Part II: Comparison of immunoreactivity and biodistribution of monoclonal antibodies labeled with "^^Ga-chelate or with ^^^I. Eur J Nucl Med 12:405-412, 1986. 60. Ward MC, Roberts KR,Westwood JH,Coombes RCC,McCready VR: The effect of chelating agents on the distribution of monoclonal antibodies in mice. J Nucl Med 27:1746-1750, 1986. 61. Goodwin DA,Meares CF,McCall MJ,Haseman MK,McTigue M, Diamond CI , Chaovapong W: Chelate conjugates of monoclonal antibodies for imaging lymphoid structures in the mouse. J Nucl Med 26:493-502, 1985. 62. Esteban JM,Schlom J,Gansow OA,Atcher RW,Brechbeil MW, Simpson DE,Colcher D:New method for the chelation of Indium-Ill to monoclonal antibodies: Biodistribution and imaging of athymic mice bearing human colon carcinoma xenografts. J Nucl Med 28:861-870, 1987.
PAGE 160
146 63. Folkman J: How is blood vessel growth regulated in normal neoplastic tissue?-GHA Clowes Memorial Award Lecture. Cancer Res 46:467-473, 1986. 64. Anghileri LJ, Crone-Escanye MC,Thouvenot P, Brunette F, Robert J: Mechanisms of gallium-67 accumulation by tumors: Role of cell membrane permeability. J Nucl Med 29:663-668, 1988. 65. Winchell HS :Mechanisms for localization of radiopharmaceuticals in neoplasms. Semin Nucl Med 6:371-378, 1976. 66. Leichner PK, Klein JL, Order SE: Dosimetric considerations in radioimmunotherapy of patients with hepatoma. Proceedings of the Fourth International Radiopharmaceutical DosimetrY Symposivun , SchlafkeStelson AT, Watson EE, eds. , Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, pp477-492. 67. Belliveau RE,Witek JT: Possible therapeutic use of radiolabeled antibodies: A review. Therapy in Nuclear Medicine . Spencer RP, ed. , New York, Grune & Stratton, Inc., 1978, pp295-303. 68. Hagan PL,Halpern SE,Chen A,Krishnan L,Frincke J, Bartholomew RM, David GS, Carlo D:In vivo kinetics of radiolabeled monoclonal anti-CEA antibodies in animal models. J Nucl Med 26:1418-1423, 1985. 69. Pimm MV, Baldwin RW: Comparative biodistributions and rates of catabolism of radiolabelled anti-CEA monoclonal antibody and control immunoglobulin in nude mice with human tumour xenigrafts showing specific antibody localization. Eur J Nucl Med 13:258-263, 1987. 70. Gatenby RA,Moldofsky PJ,Weiner LM:Metastatic colon cancer: Correlation of oxygen levels with 1-131 F(ab')2 uptake. Radio 166:757-759, 1988. 71. Dillman RO:Overview of the clinical problem of human antimouse antibodies. Fourth International Conference on Monoclonal Antibody Immunoconiuqates for Cancer . San Diego, March 1989, p61. 72. Scannon PJ: Suppression of the human immune response to immunotoxins with drugs. Fourth International Conference on Monoclonal Antibody Immunoconjuqates for Cancer . San Diego, March 1989, p62.
PAGE 161
147 73. Myers MJ: Dosimetry for radiolabelled antibodies-macro or micro? Int J Cancer Supplement 2:71-73, 1988. 74. Loevinger R, Herman M:A schema for absorbed dose calculations for biologically-distributed radionuclides. J Nucl Med 9:9-14, 1968. 75. Berger MJ: Distribution of absorbed dose around point sources of electrons and beta particles in water and other media. J Nucl Med 12 , Supplement 5:5, 1971. 76. Diel JH:Microdosimetery of internally deposited radionuclides. Int J AdpI Radiat Isot 33:967-979, 1982 77. Loevinger R,Budinger TF, Watson EE; MIRD Primer for Absorbed Dose Calculations . New York, The Society of Nuclear Medicine, 1988, pp 1-21. 78. Myers MJ,Epenetos AA, Hooker G: Practical assessment of radiation doses using labelled antibodies for therapy. Nucl Med Biol 13:437-446, 1986. 79. Myers MJ, Hooker GR,Epenetos AA: Dosimetry of radiolabeled monoclonal antibodies used for therapy. Proceedings of the Fourth International Radiopharmaceutical Dosimetry Symposium . SchlafkeStelson AT, Watson EE, eds.. Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, PP477-492. 80. Dillman LT,Von der Lage FC: Radionuclide decay schemes and nuclear parameters for use in radiation-dose estimation. Medical Internal Radiation Dose (MIRD) Pamplet 10, revised. New York, Society of Nuclear Medicine, September 1975. 81. ICRU Report 32:Methods of Assessment of Absorbed Dose in Clinical Use of Radionuclides. Washington D.C., International Commission on Radiation Units and Measurements . 1979. 82. Loevinger R,Holt JG,Hine GJ: Internally administered radioisotopes. Radiation Dosimetry . New York, Academic Press, 1956, p804-821. 83. Berger MJ: Beta-ray dosimetry calculations with the use of point kernels. Medical Radionuclides; Radiation Dose and Effects Proceedings . Oak Ridge Associated Universities, Oak Ridge, Tennessee, December 1969, AEC20, PP63-86.
PAGE 162
148 84. Spencer RP: Applied principles of radiopharmaceutical use in therapy. Nucl Med Biol 13:461-463, 1986. 85. Fisher DR:The microdosimetry of monoclonal antibodies labeled with alpha emitters. Proceedings of the Fourth International Radiophainnaceutical Dosimetry Syniposimn Schlafke-Stelson AT, Watson EE, eds . , Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, pp2636. 86. ICRU Report 33: Radiation quantities and units. Washington D.C., International Commission on Radiation Units and Measurements . 1980. 87. Bigler RE,Zanzonico PB,Cosma M: Adjuvant radioimmunotherapy for micrometastases: A strategy for cancer cure. Radiolabeled Monoclonal Antibodies for Imaging and Therapy . Srivastava SC, ed., New York, Plenum Publishing Corp., 1988, pp409-430. 88. Eckerman KF,Ryman JC,Taner AC, Kerr GD:Traversal of cells by radiation and absorbed fraction estimates for electrons and alpha particles. Proceedings of the Fourth International Radiopharmaceutical Dosimetry Symposium . Schlafke-Stelson AT, Watson EE, eds. , Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, pp6781. 89. Adelstein SJ,Kassis AI , Sastry KSR:Cellular vs organ approaches to dose estimates. Proceedings of the Fourth International Radiopharmaceutical Dosimetry Symposium . Schlafke-Stelson AT, Watson EE , eds. , Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, ppl390. Rao DV,Govelitz GF, Sastry KSR, Howell RW: Spermatogonia! cell killing by radiolabeled methionine:A comparative study of the effects of Se-75, S-35, and H-3 . Proceedings of the Fourth International Radiopharmaceutical Dosimetry Symposium . Schlafke-Stelson AT, Watson EE, eds., Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, pp52-66. 91. Langmuir VK: Dosimetry models for radioimmunotherapy. Med Phys 15(6) :867-873, 1988. 92. O'Donoghue JA,Wheldon TE:Predicted allowable doses to normal organs for biologically targeted radiotherapy. Br J Radiol 61:264-266, 1988. 93. Vaughan ATM, Anderson P, Dykes PW, Chapman CE,Bradwell AR: Limitations to the killing of tumours using radiolabelled antibodies. Br J Radio 60:567-578, 1987.
PAGE 163
149 94. Griffith MH,Wijewardene C,Wessels BW: Micro TLD and autoradiographic measurements for radiolabeled antibody uptake in tumor bearing mice. Med Phys 12(5): 872, 1985 95. Griffith MH,Yorke ED,Wessels BW,DeNardo GL,Neacy WP: Direct dose confirmation of quantitative autoradiography with micro-TLD measurements for radioimmunotherapy. J Nucl Med 29:1795-1809, 1988. 96. Wessels BW, Griffith MH:Miniature thermoluminescent dosimeter absorbed dose measurements in tumor phantom models. J Nucl Med 27:1308-1314, 1986. 97. Ryman JC, Warner GG,Eckerman KF:ALGAMP-a Monte Carlo radiation transport code for calculating specific absorbed fractions of energy from internal or external photon sources. ORNL/TM-8377 . 1986. 98. Cristy M:Mathematical phantoms representing children of various ages for use in estimates of internal dose. NUREG/CR-1159 . ORNL/TM-367, 1980. 99. Croft BY: Sinale-Photon Emission Computed T omography. Chicago, Year Book Medical Publishers, Inc., 1986, pi. 100. Jaszczak RJ, Greer KL, Coleman RE:SPECT quantification of regional radionuclide distributions. Proceed ings of the Fourth International Radiopharmaceutical Dos imetry Symposium . Schlafke-Stelson AT, Watson EE , eds . , Oak Ridge Associated Universities, Oak Ridge, Tennessee, November 1985, pp82-96. 101. losilevsky G, Israel 0,Frenkel A,Even-Sapir E,Ben-Haim S, Front A,Kolodny GM, Front D:A practical SPECT technique for quantification of drug delivery to human tumors and organ absorbed radiation dose. Semin Nucl Med XIX: 33-46, 1989. 102. Jaszczak RJ, Coleman RE, Whitehead FR:Physical factors affecting quantitative measurements using camera-based single photon emission computed tomography (SPECT) . IEEE Trans Nucl Sci NS-28:69-80, 1981. 103. Cember H: Introduction to Health Physics . New York, Pergamon Press, 1983, pill. 104. Axelsson B,Israelsson A,Larsson SA: Studies of a technique for attenuation correctionin single photon emission computed tomography. Phys Med Biol 32:737-749, 1987.
PAGE 164
150 105. Bailey DL,Hutton BF:The accuracy of SPECT attenuation correction using simultaneous transmission and emission tomography. J Nucl Med 28:677, 1987. 106. Chang, LT:A method for attenuation correction in radionuclide computed tomography. IEEE Trans Nucl Sci NS25:638-643, 1978. 107. Gait JR, Hamilton DR, Garcia EV:SPECT quantitation: Attenuation correction in inhomogeneous media utilizing a transmission scan and assigned attenuation coefficients. J Nucl Med 26:797-798, 1988. 108. Guzzardi R,Licitra G:Three dimensional attenuation coefficients distribution for SPECT of the chest. J Nucl Med 28:128-129, 1987. 109. Keller AM, Simon TR,Smitherman TC,Mellay CR,Dehmer GJ: Direct determination of the attenuation coefficient for radionuclide volume measurements. J Nucl Med 28:102107, 1987. 110. Lembcke TM,Ficken VJ, Allen EW: Optimum energies for dual energy attenuation correction in SPECT. J Nucl Med 30:5ab, 1989. 111. Malko JA,Van Heertum RL,Gullberg GT,Kowalski WP: SPECT liver imaging using an iterative attenuation correction algorithm and an external flood source. J Nucl Med 27:701-705, 1986. 112. Manglos SH,Jaszczak RJ, Floyd CE,Hahn LJ , Greer KL, Coleman RE:Nonisotropic attenuation in SPECT: Phantom tests of quantitative effects and compensation techniques. J Nucl Med 28:1584-1591, 1987. 113. Marshall CH: Absolute values of attenuation coefficient in computedtomography. Phys Med Biol 24:828, 1979. 114. Pellettier JL,Touzery C,Coitoux P:Relative efficiency of attenuation correction algorithms for emission tomography : Results from simulated data. Proceedings of the Third World Congress of Nuclear Medicine and BjoIocty 1:510-513, 1982. 115. Tsui BMW,Gullberg GT,Edgerton ER, Ballard JG, Perry JR, McCartney WH,Berg J: Correction of nonuniform attenuation in cardiac SPECT imaging. J Nucl Med 30:497-507, 1989. 116. Walters TE, Simon W,Chesler DA,Correia JA: Attenuation correction in gamma emission computed tomography. J Comp Ass Tomo 5:89-94, 1981.
PAGE 165
151 117. Webb S, Flower MA,Ott RJ, Leach MO: A comparison of attenuation correction methods for quantitative single photon emission computed tomography. Phys Med Biol 28:1045-1056, 1983. 118. Zhang CG,DeNardo GL,Macey DJ,DeNardo SJ, Custer T: Determination of the attenuation correction factor (ACF) from a transmission image for absolute quantitation of radionuclide. J Nucl Med 29:872, 1988. 119. Todd-Pokropek A:The elimination of non-stationary effects as part of attenuation scatter and PSF correction in tomography. J Nucl Med 28:660-661, 1987. 120. Heller SL, Goodwin PN:SPECT instrumentation: Performance, lesion detection, and recent innovations. Semin Nucl Med Vol XVII:184-199, 1987. 121. Beck Jw,Jaszczak RJ,Starmer CF:The effect compton scattering on quantitative SPECT imaging. Proceedings of the Third World Congress of Nuclea r Medicine and BioloqY 1:1042-1045, 1982. 122. Axelsson B,Msaki P,Israelsson A: Subtraction of comptonscattered photons in single-photon emission computerized tomography. J Nucl Med 25:490-494, 1984. 123. Bloch P, Sanders T:Reduction of the effects of scattered radiation on a sodium iodide imaging system. J Nucl Med 14:67-72, 1972. 124. Egbert SD,May RS:An integral-transport method for compton-scatter correction in emission computed tomography. IEEE Trans Nucl Sci NS-27 : 543-547 , 1980. 125. Floyd CE,Jaszczak RJ, Greer KL, Coleman RE: Deconvolution of compton scatter in SPECT. J Nucl Med 26:403-408, 1985. 126. Jaszczak RJ, Greer KL, Floyd CE: Improved SPECT quantification using compensation for scattered photons. J Nucl Med 25:893-900, 1984. 127. Koral KF,Swailem FM, Buchbinder S,Clinthorne NH, Rogers WL, Tsui BMW: In SPECT dual-energy-window compton correction, is the k value object invariant? J Nucl Med 29:797, 1988. 128. Koral KF, Buchbinder S,Clinthorne NH, Rogers WL,Tsui BMW, Edgerton ER: Compensation for attenuation and comptonscattering in absolute quantification of tumor activity. J Nucl Med 28:577, 1987.
PAGE 166
152 129. Koral KF,Wang X,Clinthorne NH, Rogers WL, Floyd CE,Jaszczak RJrCompton-scatter estimation by spectral fitting: Reproducibility comparison to monte carlo simulation. J Nucl Med 26:797, 1988. 130. Logan KW,McFarland WD: Direct scatter compensation by photopeak distribution analysis. J Nucl Med 29:797, 1988. 131. ADAC Laboratories Version 4 Operator's Manual, pE32-18, 1987. 132. Budinger TF,Gullberg GT:Three-Dimensional reconstruction in nuclear medicine emission imaging. IEEE Trans Nucl Sci NS-21:2-20, 1974. 133. Oppenheim BE:More Accurate algorithms for iterativedimensional reconstruction. IEEE Trans Nucl Sci NS91:72-77, 1972. 13 4. ADAC Laboratories Rotational Tomography Manual, pp4.124.13, 1987. 135. ADAC Laboratories Rotational Tomography Manual, p4.6, 1987. 136. Gonzalez RC,Wintz P: Digital Image Processing . Massachusetts, Addison-Wesley Publishing Company, Inc., 1987, p331. 137. Gonzalez RC,Wintz P: Digital Image Processing . Massachusetts, Addison-Wesley Publishing Company, Inc., 1987, p334-340. 138. Gonzalez RC,Wintz P: Digital Image Processing . Massachusetts, Addison-Wesley Publishing Company, Inc., 1987, pl44-158. 139. Gonzalez RC,Wintz P: Digital Image Processing . Massachusetts, Addison-Wesley Publishing Company, Inc., 1987, P354-368. 140. ADAC Laboratories Version 4 Operator's Manual, pE32-8, 1987. 141. Stabin M: MIRD08E2 Program . Oak Ridge, Tennessee, Oak Ridge Associated Universities, 1985. 14 2. Budnitz RJ,Nero AV, Murphy DT, Graven R: Instrumentation For Environmental Monitoring . Berkeley, California, John Wiley & Sons, Inc., 1983, pp285-288.
PAGE 167
153 143. Bennett JM: Basic Concepts in investigational therapeutics. Tn ; CA ini eal Oncoloav . Rubin P, ed. , New York, American Cancer Society, 1983, pp96-99. 14 4. Focht EF,Quiinpy EH,Gershowitz M: Revised Average Geometric Factors for Cylinders in Isotope Dosage. Part I. Radiology 85:151-152, 1965. 145. Cember H: Introduction to Health PhV3ic3. New York, Pergamon Press, 1983, ppl35-155. 146. Brownell GL,Ellett WH,Reddy AR:MIRD Pamplet 3: Absorbed Fractions for Photon Dosimetry. J Nucl Med Supplement No. 1:29-39, 1968. 147. Siegel JA,Lee RA,Steimle VS,Pawlyk DA,Khalvati S,Murthy S, Front D, Horowitz JA, Sharkey RM,Goldenberg DM: Monoclonal antibody (Mab) image quantitation using the buildup factor. J Nucl Med 31:783, 1990. 148 Ljungberg M: Developement and Evalua tion of Attenuation and Scatter Correction Techniques fo r SP ECT U sing the Monte Carlo Method . Sweden, Ljungbergs Tryckeri AB, Klippan, 1990. 149. Oak Ridge National Laboratory Radiation Shielding Information Center, Oak Ridge, Tennessee 37831.
PAGE 168
154 APPENDIX A SAMPLE CALCULATIONS Calculations made using the MIRD0SE2 Program, the Geometric Factor Method, MIRD Pamphlet No. 3, the Dosimetry Model, and in the phantom studies, utilized the procedures and formulas given in this appendix. Sample calculations are given below for each dosimetry method: MIRD0SE2 Program: Biodistribution and organ planar uptake data are fitted by an exponential curve using the program Plot, which is included along with the MIRDOSE2 program. The organ residence times are determined from the curve fits. The effective half-life is assumed to be infinite and equal to the physical half-life (T ) . The residence times are then inputed into MIRD0SE2 to determine the absorbed doses: Residence Time = 1.44 3 x T x Organ Uptake Absorbed Dose = A x MIRD0SE2 results (Gy/MBq injected) Geometric Factor Method: D = C X F X g F = Specific Gamma-Ray Emission = 3.65E-09 x Zn^En^^2 0.8959 E^^j = 0.172 Mev n247 = 0.9395 E^^j = 0.247 Mev F("^In) = 3.65E-09 X [ (0 . 8959) (0 . 172 )+( . 9395) ( . 247 ) ] = 1.409E-09 C Kg"^ m^ MBq"^ h'^ C = Activity Concentration = 1 . 25E-03/6032 . 5 = 7.667E-03 MBq m^ g = 89.6 cm = 0. 896 m D = 7.667E+03 X 1.409E-09 X 0.896 = 9.679E-06 C Kg''' h'^ D = (9.679E-06 C Kg'^ h"^) x (37 x ^^JpJH^P^) Gy h"^ ^n/Pm ~ Mass Energy-Absorption Coefficient of Tissue = 0.0299 cm^g'^
PAGE 169
155 /i^Pg = Mass Energy-Absorption Coefficient of Air = 0.0268 cm^ g'^ D = 9.679E-06 x 37 x 0.0299/0.0268 = 3.995E-04 Gy h'^ D = Penetrating Absorbed Dose at infinity = 1.443 X Tp X D Tp = Physical Half-Life = 67.92 h D = Absorbed Dose Rate (Gy h'^) D«p = 1.443 X 67.92 X 3.995E-04 = 3 9.17 mGy D = Nonpenetrating Absorbed Dose at Infinity = 1.443 X T X A X En^Ej A = Activity /iCi D = 1.443 X 67.92 h X 1250MCi X [ (0 . 1197 )( . 166) ] X (0.001 Kg g'^) = 2 4.34 mGy D. = D^ + D^ =63.5 mGy MIRD Pamphlet No. 3: *l^2 = 0.2365/6 = 0.0394 Kg"'' $247 = 0.2398/6 = 0.0400 Kg''' A^72 0.3282 g rad /iCi"^ h"^ A247 = 0.4942 g rad nCi''^ h"'' D„ = 1.443 X T X A X T.A.i. = 1.443 X 67.92 X 1250 X [( . 3282 )( . 0394 ) + (0.4942) (0.0400) + ( . 1197 )( . 166) ] X 0.001 Kg g"'' =64.4 mGy
PAGE 170
156 Dosimetry Model: Specific Absorbed Fractions (SAF) determined in the Monte Carlo Model are used along with the injected activity (A^) and radionulcide energy parameters to determine the absorbed doses: D, = 1.443 X T X Ag X [ (2 . 13Zn,.E,. x SAF ) + (2.13Zn^E. X SAF^p)] ^ = 2.13EnjEj = 0.8224 g rad ^Ci"'' h''' A -1 v,-1 = 2.13Zn.Ei = 0.1197 g rad ^iCi'^ h np 1 t D, = 1.443 X 67.92 X 1250 X [( . 8224 )( . 03366) + (0.1197) (0.166) ] X 0.001 = 58.3 mGy Standard Error; Standard Error = ( (X MIRD Pamphlet No. 3 result) /MIRD Pamphlet No. 3 result) x 100% = ((278.4 64.4)/64.4) X 100% = 332.3% X = TLD result, or Geometric Factor Method result, or Dosimetry Model result
PAGE 171
157 APPENDIX B TLD CALIBRATION The TLDs used in Phantom Study Three were calibrated using a radium needle source to relate the exposure reading to absorbed dose. The radium needle's exposure rate was 4.00E-07 C Kg'''. Two TLD cards (two TLD chips each) were exposed to the radium needle for a specified length of time for a total of four time intervals. The resulting TLD measurements (nC) are correlated with the radium needle exposures (mR) . The following correlation was found between the TLD readings and the radium needle exposure: Exposure (mR) = 67.5 x TLD measurement (nC) +20.9 The exposure results were converted to absorbed dose rate by using the mass energy-absorption coefficients and a constant. Absorbed Dose Rate = Exposure x 37 x \i-JvJ\^JV '^n/Pm ^ mass energy-absorption coefficient in tissue = 0.0299 cm^ g"^ Mg/Pg = mass energy-absorption coefficient in air = 0.0268 cm^ g"^ Absorbed Dose (at «) = 1.443 x Tp x Absorbed Dose Rate
PAGE 172
158 APPENDIX C SPECTDOSE PROGRAM The following pages of this appendix contain the computer code developed for the program SPECTDOSE. This program is written in Fortran-77' for a Vax/VMS Operating System". ' American National Standard X3. 9-1978 " Digital Equipment Corporation, Chelmsford, MA 01824
PAGE 173
159 Program SPECT_Dose c c This program uses SPECT images to calculatec object c volumes, store pixel locations, and pixel count. c*************** Written by Latresia A. Wilson ************* c Integer ia (262144) ,es,bs,ds,ac, rd,proj (64 , 64) ,ch, imm Integer sat,choice, tv, cum Integer*4 bslice, esl ice, m, slice, s Real*4 threshold, u, ad, atten, thalf, chalf, lambda Real*4 ct, decay Logical*! a(262144) Logical aa(262144) Character fname*40 Common pixel c c*********************************^*^^^^^^^^^^^^^^^^^^^^^^^^ c cum=0 5 Call menu read (5, *) choice if ((choice .gt. 5) . or . (choice .It. l))then write(6,*) 'Choice must be between 1 and 5, Try Again' • go to 5 end if if (choice .eg. 2) then if(sat .eg. l)go to 6 write(6,*) 'Enter the Number of Object Files*' read(5,*)imm 6 cum=cum+l if(cum .gt. l)go to 17 open (unit=7 , f ile= ' voxel ' , status= ' new ' ) go to 17 end if if(choice .eg. 3)go to 18 if(choice .eg. 4)go to 19 if(choice .eg. 5)go to 1501 1 write(6,*) 'The maximum image size is 64*64' write (6,*) 'The default directory is $Diskr account current] : ' write (6,*) • • write(6,*) 'Enter the Directory and Filename:' read(5,800)fname 800 format(a34) write (6,*) 'The maximum number of image slices is 64 • 11 write(6,*) 'Enter the beginning and ending slice number (1, 64=all) : ' read(5,*)bs,es ds=es-bs if(ds .GT. 64) then
PAGE 174
160 write(6,*) 'Maximum number of slices is 64, Try Again!' go to 11 end if bslice=( (bs-l)*4096)/512 + 1 eslice=(es*4 096)/512 write(6,*) 'Enter the Image Pixel Size (cm):' read(5, *)pixel write (6,*) 'Do you want to Correct for Attenuation (1-Yes,0-No)?' read (5, *) ac atten=l if(ac .eq. 1) then write(6,*) 'Enter the Attenuation Coefficient (cm-1) : ' read (5, *)u write (6,*) 'Enter the Attenuated Distance (Organ Depth) (cm) : ' read (5, *) ad atten=exp (u*ad) end if write (6,*) 'Do you want to Correct for Radioactive Decay (1-Yes, 0-No) ? ' read(5,*)rd decay=l if(rd .eq. 1) then write(6,*) 'Enter the Radioactive Halflife (days):' read ( 5,*) Thai f chalf=1440*thalf lambda=0. 693/chalf write(6,*) 'Enter Correction Time (min) : ' read(5,*)ct decay=exp ( 1 ambda *ct ) end if c c*********************************************************** open (unit=4 , f ile=f name , access= 'direct ' , status= ' old ' ) do 10 i=bslice,eslice k=(i-l)*512 read(4,rec=i) (a(k+j) , j=l,512) do 20 j=l,512 aa (k+j ) =a (k+j ) *atten*decay int=aa(k+j) if(int .It. 0) int=int+256 ia(k+j )=int 20 continue 10 continue c c For each slice, read values into matrix pro j ( ) s=(bs-l)*4096 Write(6,*) 'If Threshold Value the same for all slices, Enter 1: ' read(5,*)tv if(tv .eq. l)then
PAGE 175
161 write(6,*) 'Enter the Threshold Value for all Slices:' read ( 5 , * ) threshold end if c open objval.dat file open (unit=l , f ile= ' objval ' , status= ' new ' ) write(l,705) 705 fonnat(' z ' , 2x, ' Obj ' , Ix, ' 6ox ' , Ix, 'Totct ' , 3x, ' Avgct ' , 2x, 1 'Volume' ,2x, 'Width' ,lx, 'Length' ,lx, 'Area' ,lx, 2 'Center' , Ix, 'Minx' , Ix, 'Maxx' , Ix, 'Miny ' , Ix, 'Maxy ' ) do 30 slice=bs,es do 40 i=l,64 in=(i-l) *64+s do 50 j=l,64 proj (i, j)=ia(j+m) 50 continue 40 continue s=slice*4096 c c Call threshold method to segment image if(tv .eq. l)go to 15 write(6,*)' For Slice=' , slice, ' Enter its Threshold:' read ( 5 , * ) threshold 15 if (slice .eq. bs)imm=l call thold (proj , threshold, slice, imm) c 30 continue close (unit=l) imm=imm-l sat=l go to 5 17 call obj select (imm) go to 5 18 call corgan go to 5 19 call voxfil go to 5 1500 close(unit=7) 1501 stop end
PAGE 176
162 c c c. Subroutine Menu This subroutine displays a menu of choices wr ite( 6, *) 'Choose the Number of the Option of your write(6, *) write(6, *) write (6, *) write (6, *) write(6, *) write (6,*) write(6,*) return end Choice: • 1) Image Segmentation by 2) Object Extraction and 3) Create an Organ' 4) Create Voxel Phantom' 5) Quit Program' Threshold Method' Volume Calculation'
PAGE 177
c c c 163 Subroutine Thold(Data, tval , zslice, incr) This subroutine segments the image into objects with pixel values greater than the inputed threshold. c : Integer data (64,64), seg (64,64), tot , incr Integer*4 maxval , threshval , zslice Real*4 tval Common pixsize c ! • * ' c Find maximum count value in image maxval=0 do 60 i=l,64 do 70 j=l,64 if(data(i,j) .gt. maxval) then maxval=data ( i , j ) else maxval =maxval end if 70 continue 60 continue c c Determine threshold value threshval=nint(tval*maxval) c c Zero segmented image matrix do 80 i=l,64 do 90 j=l,64 seg(i, j)=0 90 continue 80 continue c Segement image do 100 i=l,64 do 110 j=l,64 if(data(i,j) .gt. threshval) then seg(i, j)=data(i, j) end if 110 continue 100 continue c c c call subroutine contour do 120 i=l,64 do 130 j=l,64 tot=0 if(seg(i,j) .eq. 0) go to 130 call contour ( i, j , seg, incr, tot, pixsize, zslice) 130 continue 120 continue close (unit=3) return end
PAGE 178
164 Subroutine Contour (1 ,in,bin, inx, totalct,pixs, z) c c This subroutine traces out objects in the segmented c image. Stores the objects indices, volume, voxel id, c and count, c Integer l,m, inx, bin (64, 64) ,minx,maxx,miny,maxy, width Integer length, totalct,numvox, area, xloc (256) ,yloc(256) Integer obj (64 , 64) , objct,p, r, q, zmin, zmax, o,c(256) ,w,t Integer chkx, chky , temp(64 , 64) ,pw, tx, ty , tc (256) Integer xs(4096) ,ys(4096) ,cs(4096) ,cont(64,64) Integer*4 z Real volume, center, avgct Real*4 pixs,voxvol Character oname*15 , inum*4 , objname*15 c voxvol=pixs**3 k=l c Zero indice matrices do 200 i=l,64 xloc(i)=0 yloc(i) =0 200 continue c c Setup temporary matrix to store ct array, bin () do 290 i=l,64 do 300 j=l,64 temp(i, j)=bin(i, j) 300 continue 290 continue c c Trace object contour xloc(k)=l yloc(k) =m obj (l,m) =1 9 if (bin(l-l,m+l) .ne. 0) then 1=1-1 m=m+l k=k+l xloc(k)=l yloc(k) =m obj (l,m) =1 go to 9 end if 31 if (bin(l,m+l) .ne. 0) then m=m+i k=k+l xloc(k)=l yloc(k) =m obj (l,m) =1 go to 9 end if
PAGE 179
165 41 if (bin(l+l,in+l) .ne. 0) then 1=1+1 in=m+l k=k+l xloc(k)=l yloc(k)=m obj {l,in)=l go to 9 end if 51 if (bin(l+l,in) .ne. 0) then 1=1+1 k=k+l xloc(k)=l yloc(k)=in obj (l,in)=l go to 41 end if 61 if (bin(l+l,in-l) .ne. 0) then 1=1+1 in=in-l k=k+l xloc(k)=l yloc(k)=in obj (l,m)=l go to 41 end if 71 if (bin(l,in-l) .ne. 0) then if((l .eq. xloc(l)) .and. (m .eq. yloc(l))) go to 401 k=k+l xloc(k)=l yloc(k)=in obj (l,in)=l go to 61 end if 81 if (bin(l-l,in-l) .ne. 0) then 1=1-1 if((l .eq. xloc(l)) .and. (m .eq. yloc(l))) go to 401 k=k+l xloc(k)=l yloc(k)=in obj (l,in)=l go to 71 end if 91 if (bin(l-l,in) .ne. 0) then if((l .eq. xloc(l)) .and. (m .eq. yloc(l))) go to 401 k=k+l xloc(k)=l yloc(k)=in obj (l,in)=l
PAGE 180
166 go to 81 end if 101 if (bin(l-l,m+l) .ne. 0) then 1=1-1 in=in+l if((l .eq. xloc(l) ) .and. (m .eq. yloc(l))) go to 401 k=k+l xloc(k)=l yloc(k) =in obj (1 ,in) =1 go to 81 end if 201 if (bin(l,m+l) .ne. 0) then m=m+l if((l .eq. xloc(l) ) .and. (m .eq. yloc(l))) go to 401 k=k+l xloc(k)=l yloc(k)=in obj (l,m) =1 go to 81 end if 301 if (bin(l+l,m+l) .ne. 0) then 1=1+1 in=in+l if((l .eq. xloc(l) ) .and. (m .eq. yloc(l))) go to 401 k=k+l xloc(k)=l yloc(k)=m obj (l,in)=l go to 101 end if c. . , c 401 C, c Reset bin pixels if isolated point if(k .eq. 1) then obj (xloc(l) ,yloc(l) )=0 go to 1000 end if 210 Calculate width, length, center, area ininx=xloc (1) inaxx=xloc(l) miny=yloc (1) inaxy=yloc(l) do 210 i=2,k if(xloc(i) .It. if(xloc(i) .gt. if(yloc(i) .It. if(yloc(i) .gt. continue width=inaxy-ininy length=niaxx-minx area=width* length minx) ininx=xloc ( i) inaxx)inaxx=xloc (i) miny ) niiny=yloc ( i) maxy ) maxy=yloc ( i)
PAGE 181
167 center=real ( (inaxx-ininx)/2 . 0) c ' ' * * ' j c Write object contour to file object. dat write (unit=inum,fint=' (14) ') inx if(inum(l:3) .eq. ' ' ) then onaine= ' OBJECT ' // inum (4:4) else if (inuin(l:2) .eq. ' ' ) then oname= ' OBJECT ' // inum (3:4) else if (inuin(l:l) .eq. ' ')then onaine= ' OBJECT ' // inum (2:4) else oname= ' OBJECT ' // inum (1:4) end if open (unit=2 , f ile=oname , status= ' new ' ) write (2,703)2, oname 703 format(2x, 'Slice=' ,I2,10x,al5) do 270 i=l,64 write (2, 704) (obj ( i , j ) , j=l , 64) 704 format(64I2) 270 continue close (unit=2) c c Zero contour matrix do 279 i=l,64 do 278 j=l,64 cont(i, j)=0 278 continue 279 continue c Count voxels p=l w=0 numvox=0 objct=0 do 220 i=l,k chkx=xloc(i) chky=yloc(i) if(i .eq. 1) go to 223 if(w .eq. 0) go to 222 do 230 t=l,w if(chkx .eq. xloc(c(t))) go to 220 230 continue 222 if(i .eq. k) go to 224 223 do 240 j=i+l,k if (xloc(j) .eq. chkx) then w=w+l c(w)=j end if 240 continue 224 if((w .eq. O).or.(i .eq. k) ) then numvox=numvox+ 1 obj ct=ob j ct+bin ( chkx , chky ) cont ( chkx , chky ) =b in ( chkx , chky ) bin (chkx, chky) =0
PAGE 182
168 Ob j ( chkx , chky ) =0 go to 220 end if zinin=yloc(c(p) ) zinax=yloc(c(p) ) do 250 o=p+l,w if (yloc(c(o) ) .It. zmin) zinin=yloc(c(o) ) if (yloc(c(o) ) .gt. zmax) zmax=yloc(c (o) ) 250 continue if (chky .gt. zmax) zinax=chky q=chkx if ((zmax .gt. 0). and. (zmin .eq. 0)) then numvox=numvox+ 1 objct=objct + bin (chkx, chky) cont ( chkx , chky ) =bin ( chkx , chky ) bin (chkx, chky) =0 obj (chkx, chky) =0 go to 220 end if do 260 r=zmin,zmax numvox=numvox+ 1 objct=objct+bin(q, r) cont (q, r)=bin(q,r) bin(q, r) =0 obj (q,r)=0 260 continue p=w+l 220 continue c c Calculate object parameters volume=numvox*voxvol totalct=totalct+ob j ct avgct=totalct/numvox c c Write object info to file objval.dat write (1,706) z, inx, numvox, total ct, avgct, volume, width, 1 length , area , center , minx , maxx , miny , maxy 7 06 format(lx,I2,lx,I3,2x,I4,lx,I6,lx,F7.2,lx,F7.2,lx,I4,2x, 1 I4,3x,I4,lx,F5.2,3x,I2,3x,I2,3x,I2,3x,I2) c. c Write object pixels to file objpix.dat write (unit=inum, fmt= ' (14) ' ) inx if(inum(l:3) .eq. ' ' ) then objname='0BJPIXV/inum(4 : 4) else if(inum(l:2) .eq. ' ' ) then obj name= ' OBJPIX ' // inum (3:4) else if(inum(l:l) .eq. ' ' ) then objname='0BJPIX'//inum(2 :4) else obj name= • OBJPIX ' //inum (1:4) end if open (unit=3 , f ile=objname, status= ' new' ) write(3,707) 707 formate Z • , 3x, 'X ' , 3x, ' Y ' , 4x, 'Ct ' )
PAGE 183
169 c Remove duplicate values from xloc and yloc pw=0 do 275 i=l,64 do 276 j=l,64 if(cont(i,j) .ne. 0)then pw=pw+l xs(pw)=i ys(pw)=j cs (pw) =cont ( i , j ) end if 276 continue 275 continue do 274 i=l,pw write(3,708)z,xs(i) ,ys(i) ,cs(i) 708 format (Ix, I2,2x,I2,2x,I2,2x,I4) 274 continue c inx=inx+l 1000 return end
PAGE 184
170 Subroutine Obj select (objnum) c c This subroutine sums the objects in each slice to form c a complete organ/object , whereby its volume is c calculated. The object is assigned an ID number , medium, c organ, which is written to a file, c Integer obj ,medno, idl, id2 , id3 , id4,x(4096) ,y(4096) Integer zs (4096) , oct (4096) counter , obj num, xi Integer sxloc(30000) , syloc (30000) , soctn ( 30000) Integer incl , octer, nw, pp, ss (30000) ,selob(64) Integer*4 totct , obj vox , otot , nvoxel Real psize,sw,ssr(30000) , sxrloc (30000) , syrloc (30000) Real zmin, zmax,yinin,ymax,xmin,xmax Real*4 organvol , select , avgct,wt ( 3 0000) Character organ*16 c write (6, *) 'Enter the Selected Object Contour Number:' read (5, *)obj write(6, *) 'Enter the Image Pixel Size (cm):' read(5, *) psize write ( 6, *) 'Enter the Image Slice Size (cm):' read (5, *) sw write(6, *) 'Enter Organ Name:' read(5,810) organ 810 format(al6) write(6, *) 'Enter Organ Medium Number (1-bone, 2-soft tiss, 3-lung, 4-void) ' read (5, *)medno write(6, *) 'Enter Whole Body Region Number:' read (5,*) idl write(6, *) 'Enter Organ ID Number:' read (5,*) id2 write(6, *) 'Enter Organ Subregion Number:' read (5,*) id3 write (6, *) 'Enter Organ Subregion Grouping Number:' read (5,*) id4 c c Open Selected Object File Call fopen(obj , counter, zs,x,y,oct) c c Calculate select object total count, number of voxels, c volume and pixels totct=0 objvox=0 xi=0 Call sobjct (counter, oct, totct, objvox,xi) c Store selected object pixels in separate matrix nw=xi do 370 i=l,nw ss(i) =zs( i) sxloc(i)=x(i)
PAGE 185
171 syloc(i)=y(i) soctn(i) =oct (i) 370 continue otot=totct nvoxel=objvox pp=l selob (pp) =obj c Open other object files do 380 j = l,objnuin do 381 i=l,pp if(j .eq. selob(i)) go to 380 381 continue call fopen( j , counter, zs,x,y, oct) call sob jet (counter, oct, totct, objvox, xi) c Check the number object pixels in select object, c include > 90% octer=0 incl=0 do 390 i=l,xi octer=octer+l do 400 k=l,nw if((x(i) .eq. sxloc(k) ) . and. (y (i) .eq. syloc(k) ) ) then incl=incl+l end if 400 continue 390 continue c Include if 90% of object pixels in selected object select=real ( incl/octer ) if(select .ge. 0.90)then do 410 i=l,xi k=nw+i ss(k)=zs(i) sxloc(k)=x(i) syloc(k) =y (i) soctn(k) =oct (i) 410 continue pp=pp+l selob (pp) =j otot=otot+totct nvoxel=nvoxel+objvox nw=nw+xi end if 380 continue c Calculate weighting factor avgct=real (otot/nvoxel) do 415 j=l,nw c Converts z,x,y into real coordinates (cm) ssr(j)=ss(j) *sw sxrloc(j)=(sxloc(j)-32) *psize syrloc( j)=(syloc(j)-3 2) *psize c Calculate voxel weights wt ( j ) =soctn ( j ) /avgct
PAGE 186
172 415 continue c Calculate organ volume organvol=nvoxel* (psize**3) c Find organ minimum and maximum coordinates call voxmax(ssr, sxrloc,syrloc,nw, zmax, xmax,ymax, zmin, 1 xmin,yTnin) c Write results to file write (7, 416)psi2e,sw 416 format(lx,F5.3,lx,F5.3) write (7, 417) nvoxel, zmax, xmax,ymax, zmin, xmin,yinin, 1 organvol , avgct 417 format (Ix, 15, Ix, F8 . 3 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, 1 F8.3,lx,F8.3, lx,F8.3,lx,F8.3) do 420 i=l,nw write(7,820) ssr (i) ,sxrloc(i) ,syrloc(i) ,soctn(i) , 1 wt ( i ) , organ , medno , idl , id2 , id3 , id4 82 format (lx,F7. 3 , Ix, F7 . 3 , Ix, F7 . 3 , Ix, 13 , Ix, F5 . 3 , Ix, alO , 1 lx,Il,2x,I3,lx,I3,lx,I3,lx,I3) 420 continue creturn end
PAGE 187
173 Subroutine Sobjct (ct,octs, stot, sobjvox, sxi) c c This subroutine determines the number of voxels and total c number of voxels. c Integer octs (4096) , sxi, ct Integer*4 stot, sobjvox c sxi=0 sobjvox=0 stot=0 do 310 i=l,ct sxi=sxi+l sobjvox=sobjvox+l stot=stot+octs ( i ) 310 continue c return end c************************************************ ***********
PAGE 188
174 Subroutine FOPEN(sobj , cter , szs , sx, sy , soct) c c This subroutine opens files based on sobj and returns c slice, x,y, and pixel ct c Integersobj , cter , szs (4096) , sx (4096) , sy (4096) , soct (4096) Character selob j * 15 , head* 2 , num* 3 c write (unit=num, f mt= ' (13) ' ) sobj if(nuin(l:l) .eq. ' ' ) then selob j = • OBJPIX • //num ( 3 : 3 ) else selob j =' OBJPIX • //num ( 2 : 3 ) end if selobj='0BJPIXV/num(l:3) open (unit=3 , f ile=selobj , status= ' old • ) read(3,305)head 305 format(a20) cter=l 307 read(3,900,end=306)szs(cter) ,sx(cter) ,sy(cter) ,soct(cter) 900 format(lx,I2,2x,I2,2x, I2,2x,I4) cter=cter+l go to 307 306 close(unit=3) cter=cter-l c return end
PAGE 189
175 C C c c. c, c 405 Subrout ineVoxmax ( z , x , y , n , zmax , xmax , ymax , zmin , xmin , ymin) This subroutine finds the maximum and minimum voxel location Integer n Real z(30000) ,x(30000) ,y(30000) , zmax, xmax, ymax Real zmin, xmin, ymin Find minimum and maximum z coordinate zmin=z (1) zmax=z (1) do 405 k=2,n if(z(k) .gt. if(z(k) .It. continue c c 407 c c zmax) zmax=z (k) zmin) zmin=z (k) 412 Find maximum and minimum x xmax=x(l) xmin=x(l) do 407 k=2,n if(x(k) .gt. xmax) xmax=x(k) IF(x(k) .It. xmin) xmin=x(k) continue Find maximum and minimum y ymax=y (1) ymin=y (1) do 412 k=2,n if(y(k) .gt. ymax)ymax=y (k) if(y(k) .It. ymin)ymin=y (k) continue return end
PAGE 190
176 Subroutine Corgan c c This subroutine allows for the creation of an organ c for the reflection properties. The user inputs the voxels c and their characteristics, c Integer bos, eos,ho, xvmin, yvmin, xvmax, yvmax, idO, idl, id2 Integer id3, id4 ,inedno,x(4096) ,y (4096) , z (4096) , set Integer*4 cttot , nvox, c, r Real xr(20000) ,yr(20000) ,zr(20000) , ops, oss, zmax, xmax Real ymax, zinin,XTnin,yTnin Real*4 vorgan,wt (20000) ,avgct Character organ*16 c c=l r=0 write(6,*)' Enter the Image Pixel Size (cm):' read (5 , *) ops write(6,*)' Enter the Image Slice Size (cm):' read (5, *) oss write (6,*)' Enter the Organ Name:' read (5, 902) organ 902 format(al6) write (6,*)' Enter the Organ Number:' read (5,*) idO write(6, *) 'Enter Organ Medium Number (1-bone, 2-soft 1 tiss, 3-lung, 4-void) ' read (5, *)medno write (6,*)' Enter Whole Body Region Number:' read (5,*) idl write(6,*)' Enter Organ ID Number:' read (5,*) id2 write (6,*)' Enter Organ Subregion Number:' read(5,*)id3 write (6,*)' Enter Organ Subregion Grouping Number:' read(5,*)id4 cttot=0 nvox=0 write(6,*)' The maximum image size is 64x64.' write (6,*)' The maximum number of image slices is 64.' write(6,*) ' ' 2 write(6,*)' Enter the beginning and ending slice 1 number ( 1, 64=all) : ' read ( 5 , * ) bos , eos if((eos .gt. 64). or. (bos .It. l))then write(6,*)' Maximum number of slices is 64, Try 1 Again! ' go to 2 end if write (6,*) 'Enter l,If all slices contain the same number 1 of voxels: ' read (5, *)ho
PAGE 191
177 if(ho .eq. l)go to 3 do 15 i=bos,eos write(6,905) i 905 formate' For Slice=',I2) write (6,*)' Enter minimum and maximum X voxel 1 coordinates: ' read ( 5 , * ) xvmin , xvmax write (6,*)' Enter minimum and maximum Y voxel 1 coordinates: ' read ( 5 , * ) yvmin , yvmax write(6,*) 'Enter the slice voxel count (0-255) read(5, *) set do 2 5 k=xvmin, xvmax do 3 5 j=yvmin, yvmax nvox=nvox+l cttot=cttot+sct x(nvox) =k y (nvox) =j z (nvox) =i 35 continue 25 continue go to 4 15 continue go to 7 3 write (6,*)' Enter minimum and maximum X voxel 1 coordinates: ' read ( 5 , * ) xvmin , xvmax write (6,*)' Enter minimum and maximum Y voxel 1 coordinates: ' read ( 5 , * ) yvmin , yvmax write(6,*) 'Enter the voxel count (0-255):' read (5, *) set do 45 i=bos,eos do 46 j=xvmin, xvmax do 47 k=yvmin, yvmax nvox=nvox+l cttot=cttot+sct x(nvox)=j y (nvox)=k z (nvox)=i 47 continue 46 continue 45 continue 4 avct=real (cttot/nvox) do 55 l=c,nvox+r wt (l)=sct/avct 55 continue c=nvox+l r=nvox if(ho .ne. l)go to 15 7 open (unit=7,file=' voxel ' , status='new' ) do 56 m=l,nvox
PAGE 192
178 xr(in) = (x(in)-32) *ops yr{in) = (y(m)-32) *ops zr (m)=z (m) *oss 56 continue c Calculate organ volume vorgan=nvox* (ops**3) c Find minimun and maximum organ coordinates callvoxmax (zr , xr,yr,nvox, zmax, xmax,ymax, zmin, xmin, ymin) c Write results to file write (7, 906) nvox, zmax , xmax , ymax , zmin, xmin, ymin, vorgan 906 format (Ix, 15, Ix, F8 . 3, Ix, F8. 3, 1X,F8. 3, lX,F8. 3, Ix, 1 F8.3,1X,F8.3,1X,F8.3) do 57 n=l,nvox write (7, 910)zr(n) ,xr(n) ,yr(n) ,sct,wt(n) , organ, medno, ido, 1 idl,id2,id3,id4 910 format(lx,F7.3,lx,F7.3,lx,F7.3,lx,I3, Ix, F5 . 3 , Ix, AlO , Ix, 1 Il,2x,I3,lx,I3,lx,I3,lx,I3,lx,I3) 57 continue close (unit=7) c return end
PAGE 193
179 Subroutine Voxfil c c This subroutine reads the voxel data files, "voxel .dat" , c for all organs to create one large file, "voxphan" . c Integer norg, fvox(20) ,inedno(100000) ,idl(100000) Integer id3 (100000) , id4 (100000) ,ct (100000) ,d, id2( 100000) Integer*4 tnvox Real 2(100000) ,x(100000) ,y(100000) ,voluine(20) Real zinax(20) ,xinax(20) ,psize,ssize Real zinin(20) ,xinin(20) ,yinin(20) ,yinax(20) Real*4 wt(lOOOOO) Character orgf il*15,nuin*2 Character*16 organ(lOOOOO) ,ninorg(20) c d=l tnvox=0 wr ite( 6, *) 'Maximum number of organs is 20!' write(6, *) 'Enter the number of organs (#voxel files):' read (5, *) norg do 21 i=l,norg write (unit=num, f mt= ' (12) ' ) i if(num(l:l) .eg. ' ' ) then orgf il= ' voxel . dat ; ' //num ( 2 ; 2 ) else orgf il=' voxel.dat; '//num(l:2) end if open(unit=ll, f ile=orgf il , status= ' old ' ) c Read pixel size and slice size read (ll,610)psi2e,ssize 610 format(lx,F5.3,lx,F5.3) c Read number of voxels in the file read (11, 62 0) fvox(i) , zmax(i) ,xmax( i) , ymax (i) , zmin(i) , 1 xmin (i) ,ymin(i) , volume (i) 62 format ( Ix, 15 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, 1 F8.3,lx,F8.3) tnvox=tnvox+fvox ( i) do 22 j=d, tnvox read(ll,625,end=23)z(j) ,x( j ) ,y ( j ) ,ct ( j ) , wt ( j ) ,organ(j) , 1 medno(j) ,idl(j) , id2 ( j ) , id3 ( j ) , id4 ( j ) 62 5 format (Ix, F7 . 3 , Ix, F7 . 3 , Ix, F7 . 3 , Ix, 13 , Ix, F5 . 3 , Ix, AlO, 1 lx,Il,2x,I3,lx,I3,lx,I3,lx,I3) 2 2 continue 2 3 nmorg(i)=organ (tnvox) close(unit=ll) d=tnvox+l 21 continue open (unit=12 , f ile= ' voxphan ' , status= ' new ' ) c Write results to file c Total number of voxels and organs, pixel and slice size write (12 , 627) tnvox, norg,psize, ssize 627 format(lx,I7,lx,I2,lx,F5.3,lx,F5.3)
PAGE 194
180 c Write #voxels, minimum and maximum coordinates, volume c of each organ do 25 j=l,norg write (12 ,628) nmorg( j ) , fvox( j ) , zmax( j ) ,xmax( j ) ,ymax( j ) , 1 zmin(j) ,xmin(j) ,ymin(j) ,volume(j) 628 format (lx,A10, Ix, 15 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, F8 . 3 , Ix, 1 F8.3,lx,F8.3,lX,F8.3,lx,F8.3) 25 continue do 24 k=l,tnvox write(12,629)z(k) ,x(k) ,y(k) ,ct(k) ,wt(k) ,organ(k) , 1 medno(k) ,idl(k) , id2(k) ,id3(k) ,id4(k) 629 format(lx,F7.3, Ix, F7 . 3 , Ix, F7 . 3 , Ix, 13 , Ix, F5 . 3 , lx,A10, 1 lx,Il,2x,I3,lx,I3,lx,I3,lx,I3) 24 continue close(unit=12) c return end
PAGE 195
181 BIOGRAPHICAL SKETCH Latresia Ann Wilson was born in 1963 in Miami, Florida. She grew up and attended high school in Ocala, Florida whereupon she graduated in 1981. She received a Bachelor of Science Degree in Nuclear Engineering in 1985 and a Master of Science Degree in Nuclear Engineering in 1986 from the University of Virginia. At present she is completing her Ph.D. degree in Medical Health Physics in the Department of Environmental Engineering Sciences at the University of Florida.
PAGE 196
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of P]j,iJ.osophy . Wilnam s. Proi Associate Professo^ Environmental Engineering Sciences I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. L '>zZ.^i-t^-;c,^ 1 (^-^-.L^t^ Genevieve S. Roessler Associate Professor of Nuclear Engineering Sciences I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor oj Philosophy. W. Emmett Bolch Professor of Environmental Engineering Sciences I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Waltdr E. Drane Assistant Professor of Radiology
PAGE 197
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. David E. Hinteni^ng Assistant Professor of Nuclear Engineering Sciences I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philos<^ Evelyj5^-' E. Watson Oak Ridge Associated Universities This dissertation was submitted to the Graduate Faculty of the College of Engineering and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. DecQDber 1990 / ^^^t-^KjQi • lS.^j^j^ Dean, College of Engineering yib
PAGE 198
/^ds'S
|
