<%BANNER%>

Detection and Characterization of Retinal Disruption in Mice Using High Angular Resolution Diffusion Microscopy (HARDM)

Permanent Link: http://ufdc.ufl.edu/UFE0043756/00001

Material Information

Title: Detection and Characterization of Retinal Disruption in Mice Using High Angular Resolution Diffusion Microscopy (HARDM)
Physical Description: 1 online resource (175 p.)
Language: english
Creator: Chandra, Saurav B
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: diabetes -- diffusion -- dti -- hardi -- hardm -- imaging -- microscopy -- mri -- retina -- retinopathy -- stem
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Diabetic retinopathy is the leading cause of blindness among diabetics. One of the primary concerns with this pathology is that it cannot be diagnosed in its early stages. This project aims to use Magnetic Resonance Imaging to investigate techniques for early detection of the disease. Primary protocols were developed on healthy control animals to establish the feasibility of the proposed methods, followed by experiments on diabetic mice. For these studies, we used High Angular Resolution Diffusion Microscopy (HARDM) as a non-invasive tool to detect this disease at an early stage. A spatial resolution up to 19µm was achieved to visualize the organized structure within the retinal layers in healthy mice. In contrast to this, in diabetic animals, the integrity of the retina was found to be compromised. These findings were arrived at using image analysis, calculation of fractional anisotropy (FA, a unit describing organized structure within a region of interest), and statistical analysis. FA in the retina was observed to be markedly higher in healthy controls as compared to diabetic mice. A novel technique was developed, which uses the direction of eigenvectors in a region of interest to perform a quantitative analysis, thereby characterizing the extent of retinal disruption due to diabetic retinopathy or other pathologies. These techniques have successfully been applied to track the repair of the retina in mice models of retinal disruption before and after stem cell treatment. Similar procedures along with novel cell tracking methods were applied to track the migration of endothelial progenitor cells (EPCs), which are essentially stem cells whose primary function is to repair damaged tissue. The tracking was accomplished by labeling the EPCs using Magnetic Iron Oxide Nanoparticles (MION) which is detectable in MRI. These studies aim to contribute towards early detection of retinal diseases using non-invasive diffusion MRI.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Saurav B Chandra.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Forder, John.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0043756:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043756/00001

Material Information

Title: Detection and Characterization of Retinal Disruption in Mice Using High Angular Resolution Diffusion Microscopy (HARDM)
Physical Description: 1 online resource (175 p.)
Language: english
Creator: Chandra, Saurav B
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: diabetes -- diffusion -- dti -- hardi -- hardm -- imaging -- microscopy -- mri -- retina -- retinopathy -- stem
Biomedical Engineering -- Dissertations, Academic -- UF
Genre: Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Diabetic retinopathy is the leading cause of blindness among diabetics. One of the primary concerns with this pathology is that it cannot be diagnosed in its early stages. This project aims to use Magnetic Resonance Imaging to investigate techniques for early detection of the disease. Primary protocols were developed on healthy control animals to establish the feasibility of the proposed methods, followed by experiments on diabetic mice. For these studies, we used High Angular Resolution Diffusion Microscopy (HARDM) as a non-invasive tool to detect this disease at an early stage. A spatial resolution up to 19µm was achieved to visualize the organized structure within the retinal layers in healthy mice. In contrast to this, in diabetic animals, the integrity of the retina was found to be compromised. These findings were arrived at using image analysis, calculation of fractional anisotropy (FA, a unit describing organized structure within a region of interest), and statistical analysis. FA in the retina was observed to be markedly higher in healthy controls as compared to diabetic mice. A novel technique was developed, which uses the direction of eigenvectors in a region of interest to perform a quantitative analysis, thereby characterizing the extent of retinal disruption due to diabetic retinopathy or other pathologies. These techniques have successfully been applied to track the repair of the retina in mice models of retinal disruption before and after stem cell treatment. Similar procedures along with novel cell tracking methods were applied to track the migration of endothelial progenitor cells (EPCs), which are essentially stem cells whose primary function is to repair damaged tissue. The tracking was accomplished by labeling the EPCs using Magnetic Iron Oxide Nanoparticles (MION) which is detectable in MRI. These studies aim to contribute towards early detection of retinal diseases using non-invasive diffusion MRI.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Saurav B Chandra.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Forder, John.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0043756:00001


This item has the following downloads:


Full Text

PAGE 1

1 DETECTION AND CHARACTERIZATION OF RETINAL DISRUPTION IN MICE USING HIGH ANGULAR RESOLUTION DIFFUSION MICROSCOPY (HARDM) By SAURAV BIMALKUMAR CHANDRA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORID A IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 2

PAGE 2

2 201 2 Saurav Bimalkumar Chandra

PAGE 3

3 To my parents and sister

PAGE 4

4 ACKNOWLEDGMENTS Before beginning my PhD degree I was advised by many that it was a daunting but learnt on the job is something that was written a long time ago in the Bible ve asked for advice and help and readily received it from many. This is my humble attempt at summarizing the vast amount of help I have received from many people over the years First and foremost, I would like to thank my adviser and mentor, Dr. John For der for his excellent advice and patience throughout my time in his l ab. I am particularly grateful to him for encouraging me to think independently and creatively, which is an essential quality all researchers need to learn as early as possible in their c areer He is also a real pleasure to be with, thanks to his cheerful disposition There has never been My heartfelt appreciation also goes to Dr. Stephen Blackba nd for h is splendid advice, interesting ideas and extremely entertaining conversations Though I always had an inclination towards MR Imaging, it was his course that I took in my early graduate studies that really fueled my interest further. I would also like to a ppreciate the helpful counsel I received from my committee members I would like to thank Dr. Thomas Mareci for providing the spherical coordinates for the 21 direction tensor protocol and for his extremely thought provoking ideas on MR Imaging methods. My project has largely encompassed ophthalmology, and Dr. Maria Grant being the expert in this field has been very supportive. Her lab was responsible for providing samples and personnel expertise for many of my studies. Dr. Christopher Batich has been a con stant support with his helpful comments especially on my col laborative Neuroscience project.

PAGE 5

5 My committee members were not only helpful with their advice, they also took a special interest in me in the final phase of my degree and made sure I understand t he fundamentals of MR Physics. They challenged me and pushed me to become a better scientist. I thank them all once again for making me the confident researcher that I can say I am today. In my graduate career I have had the good fortune of collaborating with many people from various departments. I am thankful for the collaboration with Dr. Nicholas Simpson at the College of Medicine. He was responsible for providing samples from mice models of diabetic retinopathy. I would like to thank Dr. Angelos Barmpo utis for his help with image processing and developing the diffusion MR image processing software, fanDTasia which I have used heavily for my research ; Dr. Sergio LiCalzi for his collaboration on some of my initial work and for training me for my in vivo MR imaging and Sergio Caballero also from the Grant lab. I would like to thank Dr. Joanna Collingwood for her immense contribution towards my development as a scientist She has been more of a friend than a superior and has truly been and continues to be an inspiration in my professional career. I would like to thank Dr. Mark Davidson who trained me and gave me the confidence to run experiments on my own on the X Ray Synchrotron at the Advanced Photon Source at the Argonne National Laboratory. My colleag ues and friends at the McKnight Brain Institute have truly been a pleasure to work with over the years I have learnt valuable lessons from each and every person mentioned here I would like to thank Dr. Vijay Antharam for his continuous help and support as a friend and colleague ; Dan Clark for his help on building an RF coil for my study and also for his interesting and funny anecdotes which

PAGE 6

6 always cheer ed up the mood in the lab ; Dr. Min Sig Hwang for his constant helpful tips on MRI and on life in genera l ; Dr. Nelson Klahr for his interesting company and Dan Plant for helping me with technical issues with imaging and shimming for so many years I would like to thank the Blackband lab members, Dr. Jeremy Flint, Choong Lee and Sharon Portnoy, who have alway s been more than happy to let me use their lab resources as required. Some people who have helped me learn about the various aspects of MR imaging especially in my initial years are Dr. Kyle Padgett, Dr. Samuel Grant, Dr. Peter Thelwall, Gary Blaskowski, Xeve Silver, James Rocca, Barbara Beck, Dr. David Peterson and Kelly Jenkins have been able to even begin my PhD, let alone complete it. I would like to thank and acknowledge my parents Mr. Bimal Kumar Chandra and Mrs. Leena Chandra for their love and support and my sister Dr. Shravani Gupta who was my main inspiration for aspiring to work in a field related to medicine. My family has been my strength and support when I most needed them. Each and every one mentioned in this section has been instrumental in my quest for acquiring this highest and terminal degree in academia.

PAGE 7

7 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIATIONS ................................ ................................ ........................... 14 ABSTRACT ................................ ................................ ................................ ................... 16 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 18 Basic Principles of MRI ................................ ................................ ........................... 18 Nuclear Induction ................................ ................................ ............................. 18 Relaxation ................................ ................................ ................................ ........ 20 T1 recovery ................................ ................................ ................................ 21 T2 decay ................................ ................................ ................................ .... 22 T2* decay ................................ ................................ ................................ ... 22 Basic Pulse Timing Parameters ................................ ................................ .............. 23 Repetition Time (TR) ................................ ................................ ........................ 23 Echo Time (TE) ................................ ................................ ................................ 23 Image Weighting and Contrast ................................ ................................ ............... 23 Contrast Mechanisms ................................ ................................ ....................... 23 T1 Contrast and Weighting ................................ ................................ ............... 24 T2 Contrast and Weighting ................................ ................................ ............... 25 Proton Density Contrast and Weighting ................................ ........................... 26 Image Quality and Trade offs ................................ ................................ ................. 26 Signal to No ise Ratio (SNR) ................................ ................................ ............. 26 Magnetic field strength ................................ ................................ ............... 27 Proton density ................................ ................................ ............................ 27 Voxel volume ................................ ................................ ............................. 27 TR, TE and flip angle ( ) ................................ ................................ ............ 27 Number of averages (NA) ................................ ................................ .......... 28 Receive bandwidth (BW) ................................ ................................ ........... 28 Coil type ................................ ................................ ................................ ..... 30 Spatial Resolution ................................ ................................ ............................ 30 Scan Time ................................ ................................ ................................ ........ 31 Trade offs ................................ ................................ ................................ ......... 32 Image Artifacts ................................ ................................ ................................ ........ 32 Ghosting ................................ ................................ ................................ ........... 32 Aliasing ................................ ................................ ................................ ............. 33 Frequency wrap ................................ ................................ ......................... 33

PAGE 8

8 Phase wrap ................................ ................................ ................................ 34 Gibbs Artifact ................................ ................................ ................................ .... 34 Moir Fringes ................................ ................................ ................................ ... 35 Chemical Shift Artifact ................................ ................................ ...................... 35 Zipper Artifact ................................ ................................ ................................ ... 36 Magnetic Susceptibility Artifact ................................ ................................ ......... 36 Basic Pulse Sequences ................................ ................................ .......................... 37 Spin Echo ................................ ................................ ................................ ......... 37 Gradient Echo ................................ ................................ ................................ .. 38 Diffusion MRI ................................ ................................ ................................ .......... 38 Diffusion ................................ ................................ ................................ ........... 38 Diffusion Weighted Imaging (DWI) ................................ ................................ ... 39 Diffusion Tensor Imaging (DTI) ................................ ................................ ........ 40 Tensor model ................................ ................................ ............................. 41 Bloch Torrey equations ................................ ................................ .............. 41 Fractional anisotropy (FA) ................................ ................................ .......... 42 Fiber tracking ................................ ................................ ............................. 43 High Angular Resolution Diffusion Imaging (HARDI) ................................ ........ 44 Diffusion MRI as a Diagnostic Tool ................................ ................................ .. 45 2 STUDY OF CONTROL RETINA USING HIGH ANGULAR RESOLUTION DIFFUSION MICROSCOPY (HARDM) ................................ ................................ ... 70 Background and Significance ................................ ................................ ................. 70 Materials and Methods ................................ ................................ ............................ 71 Animal Preparation and Sample Setup ................................ ............................ 71 MR Experiments ................................ ................................ ............................... 71 High resolution anatomical scans ................................ .............................. 72 High angular resolution diffusion microscopy (HARDM) ............................ 72 Data Analysis ................................ ................................ ................................ ... 72 Results ................................ ................................ ................................ .................... 75 High Resolution Anatomical Microscopy ................................ .......................... 75 High Angular Resolution Diffusion Microscopy (HARDM) ................................ 75 Discussion ................................ ................................ ................................ .............. 77 3 STUDY OF DIABETIC RETINOPATHY ................................ ................................ .. 90 Background and Significance ................................ ................................ ................. 90 Etymology of Diabetes Mellitus ................................ ................................ ........ 90 Balance of Glucose and Insulin ................................ ................................ ........ 90 Diabetic Retinopathy ................................ ................................ ........................ 92 Pathophysiology ................................ ................................ ......................... 92 Treatment methods ................................ ................................ .................... 93 Materials and Methods ................................ ................................ ............................ 94 Animal Preparations ................................ ................................ ......................... 94 MR Imaging ................................ ................................ ................................ ...... 95 Data Processing ................................ ................................ ............................... 95

PAGE 9

9 Results ................................ ................................ ................................ .................... 97 Retinal Disintegration in Diabetic sa mples ................................ ....................... 97 ...................... 97 Discussion ................................ ................................ ................................ .............. 98 4 REPAIR OF RETINAL PIGMENT EPITHELIUM (RPE) LAYER USING HEMATOPOIETIC STEM CELLS ................................ ................................ ......... 114 Background and Significance ................................ ................................ ............... 114 Methods ................................ ................................ ................................ ................ 116 Animal Preparation ................................ ................................ ......................... 116 MR Imaging ................................ ................................ ................................ .... 117 Res ults ................................ ................................ ................................ .................. 118 Discussion ................................ ................................ ................................ ............ 119 5 LABELLING OF STEM CELLS WITH MONOCRYSTALLINE IRON OXIDE NANOPARTICLES (MION) FOR TRACKING AND LOCALIZA TION BY MAGNETIC RESONANCE IMAGING ................................ ................................ ... 137 Background and Significance ................................ ................................ ............... 137 Methods ................................ ................................ ................................ ................ 140 MION Synthesis ................................ ................................ ............................. 140 Ex Vivo Imaging ................................ ................................ ............................. 140 Detection of MION labeled Stem Cells in Mouse Hind Limb .......................... 141 Murine Model of Choroidal Neovascularization (CNV) ................................ ... 141 Results ................................ ................................ ................................ .................. 142 Optimiz ation of Stem Cell Labeling ................................ ................................ 142 In Vivo Imaging ................................ ................................ ............................... 143 Discussion ................................ ................................ ................................ ............ 143 6 CONCLUSIONS AND FUTURE DIRECTIONS ................................ .................... 152 LIST OF REFERENCES ................................ ................................ ............................. 158 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 175

PAGE 10

10 LIST OF TABLES Table page 1 1 The results of optimizing image quality ................................ ............................... 68 1 2 Parameters and their associated tradeoffs. ................................ ........................ 69 2 1 Mean FA values in the retina versus the vitreous humor in individual control eyes. The difference in FA in the two groups was found to be statistically significant (p<0.0001). ................................ ................................ ........................ 89 3 1 Individual FA values in retinas and vitreous humor (used as an internal control) of healthy control eyes.. ................................ ................................ ....... 112 3 2 Individual FA valu es in retinas and vitreous humor (used as an internal control) of diabetic eyes.. ................................ ................................ .................. 113 4 1 Mean FA values in the retina of each sample (for all groups) along with the corresponding R 2 ................................ ................................ ............................. 136

PAGE 11

11 LIST OF FIGURES Figure page 1 1 Precession of hydrogen nucleus ................................ ................................ ......... 46 1 2 Behavior of hydr ogen proton spins. ................................ ................................ .... 47 1 3 T1 recovery curve ................................ ................................ ............................... 48 1 4 T2 decay curve ................................ ................................ ................................ ... 49 1 5 Free induction decay ................................ ................................ .......................... 50 1 6 The relationship between T1 or T2 and the molecular tumbling rate. L is the Larmor frequency. ................................ ................................ .............................. 51 1 7 T1 differences between fat and water ................................ ................................ 52 1 8 T2 differences between fat and water ................................ ................................ 53 1 9 Larger flip angle yield s a larger transverse magnetization, and hence a higher signal. ................................ ................................ ................................ ...... 54 1 10 Receive bandwidth vs SNR. Less noise can be sampled if the receive bandwidth is reduced. ................................ ................................ ......................... 55 1 11 Aliasing and undersampling ................................ ................................ ................ 56 1 12 Phase wrap artifact.. ................................ ................................ ........................... 57 1 13 Cause of Gibbs art ifact.. ................................ ................................ ..................... 58 1 14 Chemical shift and pixel shift. ................................ ................................ ............. 59 1 15 Mechanism of Spin Echo technique. ................................ ................................ .. 60 1 16 Pulse sequence diagram for a simple 2D spin echo sequence. ......................... 61 1 17 Isotropic versus anisotropic diffusion ................................ ................................ .. 62 1 18 .............................. 63 1 19 Effect of molecular motion in the presence of diffusion gradients. Protons that moved in between the t wo gradients are indicated by boxes. ............................ 64 1 20 Tensor model, eigenvalues and eigenvectors. ................................ ................... 65

PAGE 12

12 1 21 Range of FA. FA = 0 for complet e isotropy, 1 = 2 3 FA=1 for complete 1 2 3 =0 ................................ ................................ ............ 66 1 22 Schematic diagram for fiber tracking. ................................ ................................ 67 2 1 Ex vivo Sample setup. The samples were packed with gauze to prevent movement during imaging. ................................ ................................ ................. 80 2 2 Schematic of the o rganizational structure of the mouse retina .......................... 81 2 3 3D high resolution gradient echo image of an in vitro eye with clearly differentiated retinal layers. ................................ ................................ ................ 82 2 4 HARDM images of control eye in alterna te views. ................................ .............. 83 2 5 A 3D reconstruction of fiber tracts of a control eye ................................ ............. 84 2 6 Close up view of fiber tracts in the retina ................................ ............................ 85 2 7 Comparison of mean fractional anisotropy of the retina as compa red to the vitreous humor (N=4) ................................ ................................ .......................... 86 2 8 Illustration of the pr oposed focal point calculation method. ................................ 87 2 9 Evaluated mixture of Gaussian probabilities for healthy control eyes ................. 88 3 1 Compa rison between a control retina and a diabetic retina .............................. 103 3 2 A comparison of the fractional anisotropies in the retinas of c ontrol versus diabetic eyes ................................ ................................ ................................ ..... 104 3 3 A schematic illustrating the angular difference calculation method. The eigenvectors within a selected quadrant of the retina are displayed. ................ 105 3 4 Plot of points corresponding to angles ( ) made by eigenvectors with the reference line. ................................ ................................ ................................ ... 106 3 5 reference line is plotted for each eigenvector lying in the selec ted quadrant. ... 107 3 6 The distribution of points in individu al control samples is compared ................ 108 3 7 Diabetic eyes: The angles ( reference line is plotted for each eigenvector lying in the selected quadrant. ... 109 3 8 The distribution of points in individual diabetic samples is compared. .............. 110 3 9 For the diabetic group: plot of R 2 versus the number of days the blood glucose level of the mouse was above 300 mg/dl ................................ ............ 111

PAGE 13

13 4 1 Comparison of the RPE layer of mice from different groups ............................. 123 4 2 HARDM of eye from Group 1, which is a representative of a restored RPE layer. Yellow arrows indicate the RP E layer. ................................ .................... 124 4 3 HARDM of eye from Group 2, which is a representative of a restored RPE layer. Yellow arrows indicate the RPE layer. ................................ .................... 125 4 4 A retina from group 4 is shown. HARDM detects loss of structure due to a depleted RPE layer shown by yellow arrows. ................................ ................... 126 4 5 Comparison of FA in the retinas of repaired RPE laye r (Groups 1, 2), Untreated controls (Group 3) and ablated RPE layer (Group 4) ....................... 127 4 6 line is plotted for each eigenv ector lying in the selected quadrant. ................... 128 4 7 The group 1 distributions of points in individual samples are compared ........... 129 4 8 G line is plotted for each eigenvector lying in the selected quadrant. ................... 130 4 9 The group distributions of points in indi vidual samples are compared. ............. 131 4 10 line is plotted for each eigenvector lying in the selected quadrant. .................. 132 4 11 The group 3 distributions of points in i ndividual samples are compared. .......... 133 4 12 igenvectors to the reference line is plotted for each eigenvector lying in the selected quadrant. ................... 134 4 13 The group 4 distributions of points in i ndividual samples are compared ........... 135 5 1 In vitro MR Imaging of labeled and unlabeled stem cells. ................................ 148 5 2 Injection of MION labeled stem cells in a mouse hind leg ................................ 149 5 3 In vivo scan of a healthy control mouse eye. ................................ .................... 150 5 4 In vivo tracking of stem cells injected in murine eyes with different cond itions 151

PAGE 14

14 LIST OF ABBREVIATION S AD Average Diffusivity ADC Apparent Diffusion Coefficient bFGF Basic Fibroblast Growth Factor BM S C Bone Marrow Stem Cells BRB Blood Retinal Barrier DR Diabetic Retinopathy DTI Diffusi on Tensor Imaging DWI Diffusion Weighted Imaging EPC Endothelial Progenitor Cell FA Fractional Anisotropy FDT f anDTasia FID Free Induction Decay FOV Field of View Gd DTPA Gadolinium Diethylene Triamine Pentaacetic Acid HARDI High Angular Resolution Diffu sion Imaging HARDM High Angular Resolution Diffusion Microscopy HSC H ematopoietic S tem C ells IGF 1 Insulin like Growth Factor I LV Lentivirus MION M onocrystalline I ron O xide N anoparticles MRI Magnetic Resonance Imaging MR M Magnetic Resonance Microscopy MT X Matrix NA Number of Averages

PAGE 15

15 NEX Number of Experiments or Number of Averages (same as NA) NMR Nuclear Magnetic Resonance RA Relative Anisotropy RF Radio Frequency ROI Region of Interest RPE Retinal Pigment Ep ithelium SNR Signal to Noise Ratio SW Sweep Wi dth T Tesla TE Echo Time TEM T ransmission E lectron M icroscopy TR Repetition Time VEGF Vascular Endothelial Growth Factor VH Vitreous Humor

PAGE 16

16 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial F ulfillment of the Requirements for the Degree of Doctor of Philosophy DETECTION AND CHARACTERIZATION OF RETINAL DISRUPTION IN MICE USING HIGH ANGULAR RESOLUTION DIFFUSION MICROSCOPY (HARDM) By Saurav Bimalkumar Chandra August 201 2 Chair: John R. Forder Major: Biomedical Engineering Diabetic retinopathy is the leading cause of blindness among diabetics One of the primary concerns with this pathology is that it cannot be diagnosed in its early stages. This project aims to use m agnetic r esonance i maging to investigate techniques for early detection of the disease. Primary protocols were developed on healthy control animals to establish the feasibility of the proposed methods, followed by experiments on diabetic mice. For these studies, w e used h igh a ngul ar r esolution d iffusion m icroscopy (HARDM) as a non invasive tool to detect this disease at an early stage. A spatial resolution up to 19 m was achieved to visualize the organized struc ture within the retinal layers in healthy mice. In contrast to this, i n diabetic animals the integrity of the retina was found to be compromised. These findings were arrived at using image analysis, calculation of fractional anisotropy (FA ) which is a unit describing organized structure within a region of interest, and sta tistical analysis. FA in the retina was observed to be markedly higher in healthy controls as compared to diabetic mice. A novel technique was developed, which uses the direction of eigenvectors in a region of interest to perform a quantitative analysis, t hereby characterizing the extent of retinal disruption due to diabetic

PAGE 17

17 retinopathy or other pathologies. These techniques have successfully been applied to track the repair of the retina in mice models of retinal disruption before and after stem cell treat ment. Similar procedures along with novel cell tracking methods were applied to track the migration of endothelial progenitor cells (EPCs), which are essentially stem cells whose primary function is to repair damaged tissue. The tracking was accomplished b y labeling the EPCs using m agnetic i ron o xide n anoparticles (MION) which is detectable in MRI. These studies aim to contribute towards early detection of retinal diseases using non invasive diffusion MRI.

PAGE 18

18 CHAPTER 1 INTRODUCTION Basic Principles of MRI This chapter contains an overview of some of the basic concepts and principles of Magnetic Resonance Imaging (MRI). Imaging applications based on these concepts have been used to acquire data for this dissertation. The chapter begins with the discovery of the phenomenon known as Nuclear Magnetic Resonance (NMR) by Felix Bloch, followed by some ground breaking contributions made by some of the pioneers in this field. The early innovations in the field of NMR physics led to the development of the first two dimens ional image of biological tissue, thus introducing the term Magnetic Resonance Imaging. This was soon followed by the discovery that these concepts could be applied to differentiate between tissue types, as well as to detect differences between various pat hological states of tissue. This formed the basis of using MRI in medicine. Nuclear Induction known as Nuclear Magnetic Resonance (Bloch, 1946) This phenomenon was discovered independently at the same time by Edward Purcell (Purcell et al., 1946) The motion of the hydrogen proton in an external magnetic field (B 0 ) is like a spinning gyroscope (Haacke et al., 1999) field produces a torque, which causes it to precess about B 0 (Figure 1 1) In the absence of an external magnetic field, the nuclei are aligned in random directions (Figure 1 2), but after the application of B 0 the spins align either with or against B 0 Spins that align along B 0 are said to be in the parallel direction, and are in the low

PAGE 19

19 energy state. Spins aligned opposite to B 0 are known to be in the anti parallel direction and are in the higher energy state (Mitche ll and Cohen, 1999) The frequency with which the hydrogen nucleus oscillates in a magnetic field is its resonant frequency, known as the Larmor Frequency named after the British physicist, Sir Joseph Larmor (Oldendorf and Oldendorf, 1991) It is described by the Larmor Equation: 0 = B 0 (1 1 ) where 0 is the Larmor frequency, is a proportionality constant called the gyromagnetic ratio, equal to 42.58 MHz/T for the hydrogen proton ( 1 H), and B 0 is the magnitude of the external magnetic field. According to Bloch, the collection of spins in a medium could be represented as a net nuclear magnetiza tion (M). If the spins within M do not interact with each other, but only interact with the external magnetic field (B), then these spins will experience a torque given by (Elster and Burdette, 2001) He develop ed an equation which describes the relationship between these parameters. The resulting equation, known as the Bloch E quation is as follows: (1 2 ) where M is the net nuclear magnetization, B is the external magnetic field and when B = B 0 is constant, this equation predicts that the motion of M will be a precession around B 0 w ith frequency 0 (Figur e 1 1). Along with Edward Purcell, Felix Bloch was able to measure magnetic resonance in solids and liquids. Both Bloch and Purcell received the Nobel Prize in 1952 for their discoveries which led to the birth of MR Spectroscopy.

PAGE 20

20 tissue varies according to the type of tissue (Damadian, 1971) Soon after, in 1973, Paul Lauterbur and Pete r Mansfield produced the first 2 dimensional MR image (Lauterbur, 1973; Mansfield and Grannell, 1973) They received the Nobel Prize for this ground breaking discovery in 2003. Major developments in this field led to the term Magnetic Resonance Imaging (MRI) which is used widely today in disease diagnosis, drug discovery etc. Some of the important principles of MR Physics are discussed in this dissertation. Relaxation In the presence of an external magnetic field B0, the net magnetic moment or net magnetization vector (NMV ) is aligned along B 0 (M 0 = M z ) ( Westbrook, C et al., 2005) T his net magnetization of protons at equilibrium i s called longitudinal magnetization. At this point, a 90 frequency, flips the magnetization to the transverse plane (M 0 = M XY ). When the RF p ulse is switched off, the NMV is again influenced by B 0 and it tries to realign with B 0 In order to do so, the hydrogen nuclei must lose the energy given to them by the RF pulse. The process by whi ch hydrogen loses this energy is known as relaxation which results in two simultaneous, but independent processes the recovery of magnetization in the longitudinal plane and the decay of magnetization in the transverse plane. The recovery of longitudinal magnetization is caused by a process called T1 recovery. The decay of transverse magnetization is caused by a process termed as T2 decay. These form the basis for detecting contrast in tissue.

PAGE 21

21 T1 r ecovery T1 recovery is caused by the nuclei giving up their energy to the surrounding environment or lattice, and it is termed as spin lattice relaxation. Energy released to the surrounding lattice causes the magnetic moments of nuclei to recover their longitudinal magnetization. This rate of recovery is an exponential process, with a recovery time constant called th e T1 relaxation time. This is explained in detail in the following paragraph. The net magnetization of protons at equilibrium is called longitudinal magnetization, which is oriented in the same direction as the external magnetic field (M0= MZ). At this poi precessional frequency, flips the magnetization to the transverse plane (M0 = MXY). The time taken for the recovery of the longitudinal magnetization is known as longitudinal relaxation. The app roach to the equilibrium magnetization state is initially rapid but slows as the overall system comes close to its balanced state. The rate at which a medium approaches its equilibrium is determined by a time constant known as its T1 which is characteristi c of the tissue at the given field strength. The recovery is exponential and T1 is the time required for the z component of M to return to about 63% of its original value following an excitation pulse (Figure 1 3). The rate of recovery is the inverse of th e T1 time (1/T1). This concept can be described by the equation: ) (1 3) M 0 is the equilibrium magnetization and is fixed by the strength of the magnetic field. By convention, the direction of the external magnetic fi eld is along the Z axis. M z is equal to M 0 multiplied by a factor that depends on time (t) and the constant T1.

PAGE 22

22 T2 d ecay T2 decay is caused by nuclei exchanging energy with their neighboring nuclei. The energy exchange is caused by the magnetic fields of each nucleus interacting with its neighboring spins and is hence as termed spin spin relaxation. This results in the exponential decay or loss of transverse magnetization following the 90 RF pulse. The time taken for the transverse magnetization to decay to 37% of its original value is known as the T2 relaxation time of the tissue (Figure 1 4). The decay curve shown in the figure is described by the equation: (1 4) If T2 is short, the MR signal decays rapidly, and if T2 is long, the MR signal decays slowly. T2* d ecay As the magnitude of transverse magnetization decreases, so does the magnitude of the voltage induced in the receiver coil. The induction of the reduced signal is called the free induction decay (FID) signal. T2* decay is the decay of the FID following the RF excitation pulse. This decay is faster than the T2 decay since it is a combination of two effects the T2 decay itself as well as the dephasing of spins due to magnetic field inhomogeneities, which are areas within the magnetic field that do not exactly match the external field strength. The Larmor frequency of a nucleus is proportional to the magnetic field strength it experiences. So if a nucleus lies in an area of inhomogeneity with a higher field strength the precessional frequency of the nucleus increases. Conversely, if a nucleus experiences a lower field strength, the precessional frequency of the nucleus decreases. This relative acceleration and deceleration, as a result of

PAGE 23

23 magnetic field inhomogeneit ies and differences in the precessional frequency causes immediate dephasing of the NMV and produces a FID as shown in Figure 1 5. This dephasing is predominantly responsible for the exponential T2* decay. Basic Pulse Timing Parameters A pulse sequence co nsists of a combination of RF pulses, signals and periods of recovery. The main two timing parameters discussed here are repetition time (TR) and echo time (TE). Repetition Time (TR) This is the time from the application of one excitation RF pulse to the next RF pulse for each slice. The TR determines the amount of T1 relaxation that is allowed to occur from one RF pulse to the next. Echo Time (TE) This is the time from the application of the RF pulse to the peak of the signal induced in the coil. The TE determines how much decay of the transverse magnetization is allowed to occur. The TE thus controls the amount of T2 relaxation that has occurred when the signal is read. Image Weighting and Contrast Contrast Mechanisms Image contrast is mainly obtained through the mechanisms of T1 recovery, T2 decay and proton density. The proton density (PD) of a tissue is the number of protons per unit volume of the tissue. The higher the PD, the more signal is available from that tissue. For example, the PD of the lun gs is much lesser than the PD of the pelvic region. Therefore, the lungs appear dark as compared to the pelvic region since less signal is available.

PAGE 24

24 T1 and T2 relaxation depend on certain factors. The type of molecular packing of the tissue is one such fa ctor. Tissues with closely spaced molecules exchange energy more efficiently with hydrogen nuclei and vice versa. This is especially important in T2 relaxation processes. Another important factor is the match of the molecular tumbling rate with the Larmo r frequency. The molecular tumbling rate is the rate at which a molecule moves, and which governs the T1 and T2 relaxation times of a particular tissue. Tumbling rates close to the Larmor frequency will produce a short T1. Tumbling rates both longer and sh orter than this increases the T1 value. T2 is more or less directly proportional to the molecular tumbling rate (Liney, 2005). The relationship between T1 or T2 and the molecular tumbling rate is shown in Figure 1 6 (Elster and Burdette, 2001) The potentia l energy of the protons in the tissue also plays an important role. This potential energy can be in the form of thermal energy within the protons. Tissue with low energy is more capable of absorbing energy from hydrogen nuclei and vice versa. This is espec ially important in T1 relaxation processes. Different types of tissue generate different tissue contrast. In order to understand the mechanisms which govern contrast, the two extremes in contrast discussed here are fa t and water. T1 Contrast and Weighting Fat has low inherent energy and can therefore absorb energy more easily into its lattice. To be efficient at T1 relaxation, the molecule must be precessing near the Larmor frequency. Fat molecules are closely packed together and their molecular tumbling r ate matches the Larmor frequency. Therefore the energy exchange from hydrogen nuclei to the surrounding lattice is efficient. This means that the NMV of fat realigns quickly with B 0 and hence the T1 time of fat is short.

PAGE 25

25 In the case of water, the situatio n is quite the opposite. Water has high inherent energy and is therefore unable to absorb energy easily into its lattice. The mobility of water molecules is high and they are spaced apart. Their molecular tumbling rate is relatively fast and does not match the Larmor frequency. This leads to inefficient energy exchange from hydrogen nuclei to the surrounding lattice. This means that the NMV of water realigns slowly with B 0 and hence the T1 time of water is long. In a T1 weighted image, the contrast depends on the differences in the T1 times between fat and water (and all other tissues with intermediate signal). The TR must be short enough so that neither fat nor water has had sufficient time to fully return to B 0 Hence for T1 weighting, the TR must be shor t (Figure 1 7 ). T2 Contrast and Weighting T2 decay occurs due to the dephasing of spins caused by the interaction of neighboring spins. As stated earlier, fat molecules are closely packed together and their molecular tumbling rate matches the Larmor frequ ency. Hence the energy exchange is efficient, causing spins to dephase quickly and the loss of transverse magnetization is rapid. Therefore, the T2 time of fat is short. The energy exchange in water is less efficient because of the loosely spaced molecules along with the fact that the molecular tumbling rate does not match the Larmor frequency. Due to this, the loss of transverse magnetization is gradual and the T2 time of water is therefore long. As the transverse magnetization decays faster in fat compar ed to water, water appears bright on a T2 contrast image. In T2 weighted images, the difference in contrast depends on the differences between the T2 times of different tissue types. TE must be long enough to allow sufficient time to decay. If it is too sh ort, then there will be

PAGE 26

26 a very small difference in contrast. Therefore for T2 weighting, the TE must be long (Figure 1 8 ). Proton Density Contrast and Weighting A proton density image is one where the difference in the numbers of protons per unit volume is the main contributing factor in forming image contrast (Brown and Semelka, 2003) Hence to achieve proton density weighting, the effects of T1 and T2 contrast must be diminished. A long TR allows both fat and water to fully recover their longitudinal magn etization, thus diminishing T1 weighting. A short TE does not give enough time for the transverse magnetization to decay in either fat or water, thus diminishing T2 weighting. Therefore an image with a long TR and short TE produces a proton density image w here the inherent proton density is the main factor contributing to tissue contrast. Image Quality and Trade offs The quality of an image is determined by various factors, some of which are explained below. Altering one may affect one or more factors, ult imately affecting image quality. So in order to optimize image quality, it is crucial to be aware of the associated trade offs involved. The main considerations determining image quality are signal to noise ratio (SNR), spatial resolution and scan time. Si gnal to Noise Ratio (SNR) Noise is random and occurs at all frequencies. The signal is acquired at specific time points. The SNR depends on the following factors: Magnetic field strength Proton Density Voxel Volume TR, TE and flip angle Number of Averages

PAGE 27

27 Receive bandwidth Coil type Magnetic f ield s trength If the magnetic field strength is high, more number of protons align parallel to B 0 thereby increasing the energy gap between the high and low energy states of the protons. Due to this, the NMV increases in size and as a result, there is more magnetization available, which means that there is more signal. Hence SNR increases with magnetic field strength. Proton d ensity As the number of protons in an area of interest increases, so does the signal. This is why the SNR in the lungs for example (an area of low PD), is much lesser than the SNR of say, the brain. Voxel v olume A voxel is the product of pixel area and the slice thickness. A large voxel obviously contains more spins than a smaller voxel and hence has a higher SNR. For example, if the pixel dimensions cannot be changed in order to maintain 2D spatial resolution (to be discussed later), then a thicker slice can contribute more signal than a thinner slice. TR, TE and f lip a ngle ( ) A long TR allows more recovery of the longitudinal magnetization so that more spins are available to be flipped in the next excitation pulse. The TE controls the amount of transverse magnetization that is allowed to decay before an echo is collected. A longer TE ensures that less transverse magnetization is available to be rephased to produce an echo. Hence SNR and TE are inversely related.

PAGE 28

28 The flip angle controls the amount of transverse magnetization that is created to induce the signal. Figure 1 9 il lustrates how using a larger flip angle increases the transverse component of the signal, and consequently, the SNR. A 90 flip angle creates maximum signal. Number of a verages (NA) This is the number of times data is collected in each slice. Data contains both signal and noise, but noise being random in space and time, may be in a different position during each acquisition Signal, however, is not random and occurs at the same place when it is collected. Doubling the NA does not double the SNR because noise is also involved. Doubling the NA increases the SNR by a factor of 2 (or 1.4). A higher NA obviously results in longe r scan times. Receive b andwidth (BW) The range of frequencies sampled during the application of the readout gradient is known as the receive bandwidth (BW) or BWread. BWread determines the range or number of frequencies to digitize during readout. The fre quency matrix determines the number of data points (Nread) collected during readout. The echo is centered in the middle of the readout window. There is also a relationship between SNR, BW and spatial resolution ( spatial resolution will be discussed in det ail in a later section in this chapter) which can be described mathematically as follows: According to Nyquist the noise in a conductor o r biological tissue is given by, (1 5) where k is the Boltzmann constant, T is the absolute temper ature of the conductor, R is

PAGE 29

29 the resistance and BW is the frequency bandwidth in Hz. BW is the only parameter that can be controlled. The relationship between the BW, sampling time or acquisition time (AT) and number of data points is given as: (1 6 ) Also, the acquisition time step increment t = acquisition time (AT)/number of data points. So equation 1 6 can also be written as: (1 7 ) Noise is equally distributed over the bandwidth and is also equally divided among N read and N phase data points (for 2D images). (1 8) For these equations, we will denote FOV read and FOV phase as L read and L phase respectively and the resolutions in the corresponding directions as L read and L phase respectively. S ince the signal, the measured SNR in a voxel can be given as, (1 9) Therefore at a constant resolution the SNR is optimal when the BW is set to a When the signal is acquired N acq times, the signal is directly proportional to the number of averages, but the noise increases as the square root of the number of averages. Hence the resulting SNR is given by,

PAGE 30

30 (1 10) Figure 1 10 explains practically how less noise can be sampled if the receive bandwidth is reduced. Given all these factors, this action must be taken with caution because reducing the BW increases the minimum TE F or example, let us cons ider a BW read of 32 kHz with a sampling time of 8 ms used to acquire a frequency matrix of 256. If the BW read is halved to 16 kHz, then to maintain a frequency matrix of 256 (or in other words, to maintain spatial resolution ) the sampling time must be inc reased to 16 ms. Since the TE increases with the sampling time, the minimum TE thus increases if BW is reduced. Coil t ype The type of coil used affects the amount of signal received, and therefore the SNR. It is essential that the volume of tissue to be im aged fills the sensitive volume of the coil. For example, using a large coil may cause aliasing due to signal being produced from outside the field of view (FOV). A quadrature coil increases SNR because signal is received using two coils. A surface coil po sitioned correctly can maximize signal. Maximum signal from a surface coil can be obtained if the coil is positioned perpendicular to B 0 Spatial R esolution Spatial resolution is the ability to distinguish between two points in an image as separate and di stinct. This depends on voxel size which in turn depends on slice thickness, matrix size and FOV. A thinner slice thickness increases the ability to resolve small structures in the through plane or slice select plane and therefore increases spatial resolu tion. Thicker

PAGE 31

31 slices also cause undesirable partial voluming, in which case the voxel contains two different tissues and therefore possess es a signal average of both tissues. However, thinner slices reduce the SNR as discussed earlier. The matrix size dete rmines the number of pixels in the image. A larger matrix size equates to finer pixels which increase spatial resolution. The FOV along with the matrix size determines the pixel dimensions, which gives us the spatial resolution. Smaller pixel dimensions me an a better spatial resolution. The relationship is given as: (1 11 ) Scan Time The scan time is the time required to complete the data acquisition. It is mainly affected by the TR, phase matrix (or number of phase en codings ) and number of averages. The scan time can be calculated by the product of these three parameters (TR N phase NA) for 2D imaging. For 3D imaging, the number of encodings in the slice select gradient must also be included. Hence the total scan ti me can be calculated by TR N phase NA N slice TR being the time of each repetition, therefore increasing it, will increase the total scan time. The phase matrix is the number of phase encodings, which determines the number of lines of K space that ar e filled to complete a scan. For example, if the phase matrix is 128, it means that TR is applied 128 times to fill 128 lines in K space. By increasing the phase matrix to 256, the scan time is doubled because now TR must repeat 256 times.

PAGE 32

32 Trade offs Ideal ly an image must have high SNR, good spatial resolution and a short scan time. But it is obvious that there are many trade offs involved when optimizing image quality. Some parameters are often compromised in favor of another depending on the requirement o f the study. Some of these factors have been discussed in the preceding sections of this chapter. A summary of these factors may be found in Tables 1 1 and 1 2. Table 1 1 provides the result of optimizing image quality, while Table 1 2 lists an in depth su mmary of the parameters and their associated tradeoffs. Image Artifacts MR images often contain artifacts, many of which cannot be eliminated altogether. But they may be reduced by making certain adjustments. Some of the common image artifacts in MRI and their remedies are discussed below. Ghosting Ghosting or phase mismapping is caused by motion along the phase encoding direction. Common reasons for this motion are patient movement or sample vibration. The reason for the artifact being in this direction i s because the phase encoding gradient has a different amplitude in every TR while the frequency and slice select gradients have a constant amplitude. Therefore as the anatomy moves during the scan, it is misplaced in the phase encoding direction as the pha se gradient changes. For example, if the chest area is being imaged, the chest wall is located at a particular location during inspiration, but may have moved to another location during the next phase encoding at expiration Therefore the same part of the anatomy will have two different phase values, leading to mismapping.

PAGE 33

33 One of the methods to reduce this artifact is to swap the phase and frequency encoding directions so that the movement is in the frequency encoding direction. Gating techniques may also h elp reduce this artifact. In cardiac gating for example, the excitation pulse is coordinated with the R wave of the systole. Therefore each acquisition is acquired at the same phase of the cardiac cycle, thus reducing phase mism apping. Aliasing Aliasing or wrap around is produced when anatomy outside the FOV is mapped inside the FOV. Sometimes signal from nearby anatomy which is outside the FOV is picked up by the receiver coil. If the data is under sampled, the signal is mismapped inside the FOV instead of outside. This artifact may occur in both the frequency as well as the phase encoding directions. Frequency wrap Frequency wrap is aliasing along the frequency direction. It is caused by undersampling of frequencies. According to the Nyquist theorem, fr equencies must be mapped at least twice per cycle for correct mapping (Figure 1 11 A ). If frequencies are under sampled as shown in Figure 1 11 B and 1 11 C then the signal is misinterpreted as being at the wrong frequency. This may cause signals from outsid e the FOV to be mistaken for signals inside the FOV and consequently showing up in the image. One way to fix this is to increase the sampling rate, but this may increase noise. Frequency filters are often applied which filter out frequencies outside the FO V. Another remedy is to increase the FOV which will encompass the entire signal produced, but this results in loss of spatial resolution. Using surface coils instead of volume coils may help

PAGE 34

34 reduce aliasing as size of the coil and good placement can ensure that signal outside the FOV is not acquired. Phase wrap Aliasing along the phase direction of the image is known as phase wrap. Any signal is allocated a phase value depending upon its position along the phase axis (from 0 to 360 ). Figure 1 1 2 shows the different phases along the phase encode direction denoted by arrows. This phase curve is repeated on both sides of the FOV. Phase values outside the FOV with the corresponding phase values inside the FOV (shown by red arrows) are sometimes mapped i nto the FOV resulting in a phase wrap Anti aliasing along the phase axis is achieved by doubling the FOV in the phase encode direction, while also doubling the number of phase encoding steps (to maintain spatial resolution). However, this may result in increased scan times, and may be compensated for by reducing the number of averages. Gibbs Artifact Gibbs artifact is also known as truncation artifact or ringing artifact. It produces banding patt erns near the edge of an image (high contrast interfaces). Fourier transforms are used i n order to reconstruct MR signals into an image Theoretically, any signal can be represented as an infinite summation of sin e waves of different amplitudes, phases and frequencies (Elster and Burdette, 2001) In MR imaging, however, we are restricted to sampling a finite number of frequencies and must the refore approximate the image by using relatively few harmonics in its Fourier representation. The Fourier series then is cut short or truncated, hence the name. Figure 1 13 shows how using fewer harmonics produce s oscillations at the edge of a

PAGE 35

35 square wavef orm These oscillations are a result of truncating or under representing the Fourier series due to less number of samples being measured. Truncation artifacts may occur in both the frequency as well as the phase encoding direction, but since the phase mat rix is often reduced in order to reduce scan times, the truncation artifact is usually seen in this direction. In order to reduce this artifact, u ndersampling of data must be avoided by increasing the number of encoding steps. M oir Fringes This artifact i s characterized by a circular banding pattern superimposed on to the image. It is most commonly observed in gradient echo images. This is caused due to aliasing of signals with alternating phases. A common way to avoid this artifact is to use spin echo p ulse sequences. Chemical Shift Artifact Chemical shift artifact produces a dark edge at the interface between fat and water. It occurs along the frequency encoding axis. Fat precesses at a lower frequency than water. The difference in their precessional frequencies at 1.5 T is 220 Hz. This difference increases at higher field strengths and leads to the artifact known as chemical shift, expressed in parts per million (ppm). The value of the chemical shift is the same in any magnetic field strength The mag nitude of the fat water shift artifact can be determined by knowing the receiver bandwidth, the number of pixels in the frequency encoding direction and the inherent difference between fat and water resonance (220 Hz at 1.5 T). In Figure 1 14 consider a b andwidth of 32 kHZ and a matrix of 256. This means that each pixel has an individual frequency range of 125 Hz (32 kHz/256). Therefore fat and water protons

PAGE 36

36 adjacent to each other are mapped 1.76 pixels apart (220/125). Now if the bandwidth is reduced to 1 6 kHz, then each pixel can accommodate only 62.5 Hz (16 kHz/256). Therefore adjacent fat and water protons will be mapped 3.52 pixels apart (220/62.5) Alternately, if the BW is kept constant, but the magnetic field strength increases the difference in pr ecessional frequencies between fat and water also increases. This leads to fat and water being mapped further apart than in lower field strengths. Chemical shift artifact can be reduced by using a wide receive bandwidth and small FOV. The tradeoff on SNR i n this case must be considered. At higher field strengths where there are significant chemical shift artifacts, chemical saturation techniques are used to saturate out signal from fat or water. By doing so, either fat or water is nulled and chemical shift artifact is eliminated. Zipper Artifact A Zipper artifact appears as a dense line or a zip like artifact across the image in the frequency encoding direction caused by RF interference. This is caused by a leak in RF shielding. It can be reduced by electric screening to reduce leaks in the RF shield ing. Magnetic Susceptibility Artifact Magnetic susceptibility artifacts are distortions in the image or signal voids at certain areas generally caused by the presence of metal objects or magnetic particles in the imaging volume. It may also be caused by differences in susceptibility between certain tissues themselves, for example, air tissue interfaces. This artifact can be reduced by several methods depending on the cause of the artifact. All metal objects must be removed. Using spin echo sequences is an effective technique as the 180 refocusing pulse is effective at compensating for phase

PAGE 37

37 differences. Also, using a short TE reduces this artifact because it allows less time for dephasing between tissues with susceptibility differences. Basic Pulse Sequences Spin Echo In order to count er the issue of the decaying signal due to magnetic field inhomogeneities, the concept of pin e cho was developed by Erwin Hahn in 1950 (Hahn, 1950) Spin echoes are sometimes still The Hahn Spin Echo technique consists of a 90 flip angle followed by a 18 0 flip separated This sequence was used for the measurement of T 2 values. It is possible to measure T 2 since spin echo techniques compensate for local inhomogeneities o f the static magnetic field. The mechanism of this technique is illustrated in Figure 1 1 5 The 90 RF pulse flips the magnetization in the x y plane (Figure 1 15 1) when all the spins are in phase. Over time, the spins begin to dephase (Figure 1 15 2). At this point (Figure 1 15 3), a 18 0 refocusing pulse is applied, rotating all of the spins around the same axis. Due to this, the faster spins are now placed behind compared to the slower spins. The faster spins now continue to precess more rapidly than the slower spins and soon all the spins rephase to form the spin echo (Figure 1 15 4). A typical 2D spin echo pulse sequence with one refocusing pulse is illustrated in Figure 1 1 6 The first line (RF) describes the sequence of radio frequency pulses ( 9 0 and 18 0 ) G SS describes the timing of the slice select gradient. G PE indicates the multiple phases of varying strengths of the phase encoding gradient. G R final analog signal is converted to a digital value.

PAGE 38

38 Gradient Echo Gradient echo pulse sequences use a variable flip angle (usually less than 90 ). A small flip an gle partially flips the longitudinal magnetization into the transverse plane, and a smaller transverse magnetization is generated (less than what would have been created using a 90 flip angle). Hence shorter repetition times and, consequently, shorter sca n times can be used. After the RF pulse is withdrawn, the FID signal decays rapidly due to T2* dephasing. Instead of a 180 RF pulse as in the case of a spin echo sequence, a gradient is used to rephase the FID. The frequency encoding gradient is used for this purpose, and it is initially applied negatively to speed up the dephasing of the FID, and then its polarity is reversed, thereby rephasing the spins. Diffusion MRI Diffusion Diffusion occurs as a result of the constant movement of water molecules cau sed by collisions with surrounding molecules (Moritani et al, 2005). This motion, known as Brownian motion, was named after the nineteenth century botanist, Robert Brown, who had first observed this phenomenon in 1827 (Brown, 1828) For example, when a drop of ink falls into a still container of water, initially the ink will be concentrated in a small volume of water, but soon the ink mixes (diffuses) with the rest of the water. The molecules are free to diffuse in any direction as they are not restricted by boundaries. In this case, diffusion may be equal in all directions and is therefore known as isotropic diffusion (Figure 1 1 7 A ). However, in a biological medium like the human body for example, this motion is r estricted by boundaries such as membranes, macromolecules, ligaments etc. Sometimes this diffusion can be directional, depending on the tissue structure, or in other words, may have a preferential direction of diffusion, and is hence

PAGE 39

39 termed as anisotropic diffusion (Figure 1 1 7 B ). For example, diffusion in the central nervous system is preferential because diffusion tends to occur along the direction of the nerve fibers. The diffusion perpendicular to the axon is minimal. The net displacement of molecules i s known as the apparent diffusion coefficient (ADC). The measurement of diffusion can be performed using a pulse sequence containing two magnetic field gradients on either side of the 180 refocusing pulse. These magnetic field gradients of certain duratio n add a smaller magnetic field to spins located in different regions within the tissue. By applying another gradient after the 180 refocusing pulse, information is obtained about how much the spins have moved (diffused) during this time. Stationary spins will acquire no net phase change, while moving spins acquire a phase change which results in loss of signal. This is explained in detail in the following paragraphs. Diffusion Weighted Imaging (DWI) The Stejskal Tanner pulse sequence which was originally u sed for NMR spectroscopy, soon became the hallmark of diffusion imaging studies worldwide (Stejskal and Tanner, 1965) Figure 1 1 8 depicts this sequence where the diffusion gradients are a 1 1 9 (Mori, 2007) The first gradient dephases the spins, while the second gradient rephases these spins after a specific rephase completely, whereas spins that moved along the direction of the field gradient are rephased to different extents. In other words, spins which move (due to diffusion) along the direction of the applied diffusion gradients will acquire phase changes which are not reversed. As

PAGE 40

40 a result, intravoxel dephasing occurs, causing loss of signal. Thus, the NMR signal intensity is attenuated relative to molecular diffusion. The amount of signal attenuation field gradients The signal (S) in a voxel is determined using the equation: S = S 0 e bD ( 1 12 ) where S is the diffusion weighted signal, S 0 is the signal without diffusion effects b is the b value (a factor summar izing the gradient strength and timing parameters) and D is the diffusion coefficient. The relationship of the b value with the gradient strength and timing parameters involved in acquiring the diffusion weighted image is given as follows: 2 G 2 2 ) ( 1 13 ) where = standard gyromagnetic ratio, G = strength of the diffusion gradient = duration of each gradient, and = time interval separating the pair of diffusion gradients. [ is the effective diffusion weighti ng time. These parameters are illustrated in Figure 1 13. Varying the above mentioned parameters will change the b value. A higher b value results in higher diffusion weighting, which is often required depending on the application. However, the tradeoff is that as diffusion weighting increases, the signal decreases. Hence certain factors need to be considered while choosing the correct b value. Diffusion Tensor Imaging (DTI) Diffusion tensor imaging (DTI) is a technique that enables the measurement of the r estricted diffusion of water in tissue

PAGE 41

41 Tensor m odel Anisotropic diffusion is mathematically described using a geometric quantity known as 20 ). It is visualized as an ellipsoid which has three tensor eigenvectors V 1 V 2 and V 3 (LeBihan et al., 2001) The shape of the tensor ellipsoid is determined by the corresponding tensor eigenvalues 1 2 and 3 The major axis of this tensor is aligned along the primary eigenvalue ( 1 ) which is also the direction of fastest diffusion. 2 and 3 are the secondary and tertiary eigenvalues respectively. The tensor can be further explained by a 3x3 matr ix containing nine elements as follows: = (1 14 ) where = Diffusion Tensor Typically, the gradients are only applied in three directions yielding D xx D yy and D zz Hence D xy = D yx D xz = D zx and D yz = D zy Therefore only six unique scalar elements are required to completely describe the tensor. Bloch Torrey e quations Torrey modified the equation of motion for transverse magnetization to include diffusion in the presence o f a magnetic field gradient, This modified Block Torrey equation is as follows (Haacke et al., 1999; Torrey, 1956) : = i + (1 1 5 ) where

PAGE 42

42 For isotropic diffusion, the diffusion tensor (DT) can be written as, = = D (1 16 ) In this case, the Bloch Torrey equation simplifies to, = i + D (1 17 ) This equation has the solution, = A(t) ( 1 1 8 ) where = For a constant gradient the transverse magnet ization becomes, = A(t) (1 19 ) Fractional a nisotropy (FA) One of the most commonly used measures of anisotropy is fractional anisotropy (FA) (LeBihan et al., 2001) It is a scalar v alue which descri bes the fraction of diffusivity that is due to anisotropic diffusion (Basser and Pierpaoli, 1996) It is a value between 0 and 1, with 0 denoting isotropic diffusion, i.e. diffusion is unrestricted in all directions, while 1 denotes maximum anisotropy, i.e. diffus ion only occurs along one primary direction while it is completely restricted along a ll other directions (Figure 1 21 ). The value of FA is calculated from eigenvalues ( ) of the tensor model as follows: FA = (1 20 ) where =

PAGE 43

43 Another way of representing anisotropy is relative anisotropy (RA) and volume ratio (VR). RA is given by the formula as follows: R A = (1 21 ) where = Volume ratio is given by the formula, VR = (1 22 ) These indices are all scaled from zero to one. Fiber t racking Fiber tracking or fiber tractography provides an important kind of anatomical information connectivity information within tissue (Mori, 2007) It can be described as a tool which groups pixels (or voxels) based on the ir fiber orientation. Figure 1 22 is a schematic diagram of the linear line propagation approach for fiber tracking (Mori and Zijl, 2002) The shape of the tensor ellipsoid indicates the direction of greatest diffusion (fiber orientations) at each pixel. These are linearly connected to form 'fibers tracts '. Fiber tracking is increasingly used i n practice to understand underlying microanatomy, for example, neural pathways in the central nervous system. These techniques can be applied to various other areas of interest. Chapters 2 and 3 use this technique to investigate the fiber orientation in mu rine retinal layers.

PAGE 44

44 High Angular Resolution Diffusion Imaging (HARDI ) It has been established that diffusion tensor imaging is an exceptional tool in evaluating tissue microstructure and that it provides information about direction of diffusion within ti ssue. However, the fact remains that it is often unable to resolve fiber crossings to the fullest extent. For example, if within a voxel, if there are two fibers perpendicular to each other, the resulting primary eigenvector will not represent the underlyi ng fiber direction. This may lead to loss of information as conventional DTI does not compensate for this fact and the reported anisotropy may be incorrect. An advanced image acquisition technique known as high angular resolution diffusion imaging (HARDI) is used to address this particular issue (Frank, 2001; Tuch et al., 2002) The HARDI technique evaluates the apparent diffusion coefficients along many diffusion gradient orientations. It acquires images using a co nstant b value with the diffusion weighted gradient directions uniformly distributed around a particular point. Various tensor fitting models have been developed to extract information about underlying fiber orientations within complex tissue (zarslan and Mareci, 2003) An example of a diffusion displacement probability function is the diffusion orientation transformation (DOT) developed by Ozarslan, et al., 2006. They demonstra ted that higher order tensor models such as a rank 6 DT model resolved fiber crossings to a greater extent than a lower order tensor model such as a rank 2 DT model (zarslan et al., 2005) Progress made in this line of work aims to extract more information from underlying microa natomy (zarslan et al., 2006a; zarslan et al., 2006b) Established histology of the murine retina does not suggest fiber crossings within retinal layers, hence a rank 2 diffusion tensor model was used for all dif fusion data acquisitions. The

PAGE 45

45 spherical coordinates used for the acquisition of HARDI data have been previously described by zarslan et al. (zarslan and Mareci, 2003; zarslan et al., 2005) In this dissertation, HARDI methods have been employed to investigate reti nal microstructure with high resolutions up to 31 m. Since it has been established that imaging resolutions better than 100 m are classified as microscopy (Glover and Mansfield, 2002) the term high angular resolution diffusion microscopy (HARDM) is used in this dissertation. Diffusion MRI as a Diagnostic Tool Initial work in the early 90s by Moseley et al. (Moseley et al., 1990; Moseley et al., 1991) laid the foundation for the use of diffusion weighted MRI to investigate white matter. Diffusion weighted imaging became widely used clinically due to its ability to detect acute brain ischemia (Song et al., 2003; Sotak, 2002) Recently, DTI has been used to characterize brain development and aging, as well as structural chan ges induced by infectious, neurological, and neuropsychiatric disorders. Diffusion MRI quickly gained prominence due to its ability to map neural pathways (Basser et al., 2000; Inglis et al., 2001; Mori et al., 200 1) It soon became the modality of choice for imaging of the brain and central nervous system. It became widely used clinically to detect acute brain ischemia (Sotak, 2002; LeBihan, 200 3) Recently, DTI has been used to characterize brain development and aging, as well as structural changes induced by infectious, neurological, and neuropsychiatric disorders.

PAGE 46

46 Figure 1 1 Precession of hydrogen nucleus

PAGE 47

47 Figure 1 2. Beha vior of hydrogen proton spins. A) in the absence of an external magnetic field; B) in the presence of a magnetic field Bo (External Mag. Field) A B

PAGE 48

48 Figure 1 3. T1 recovery curve

PAGE 49

49 Figure 1 4. T2 decay curve

PAGE 50

50 Figure 1 5. Free induction decay Mxy MR signal Time FID Signal

PAGE 51

51 Figure 1 6. The relationship between T1 or T2 and the molecular tumbling rate. L is the Larmor frequency.

PAGE 52

52 Figure 1 7 T1 differences between fat and water

PAGE 53

53 Figure 1 8 T2 differences between fat and water

PAGE 54

54 Figure 1 9 Larger flip angle yields a larger transverse magnetization, and hence a higher signal.

PAGE 55

55 Figure 1 10 Receive bandwidth vs SNR. Less noise can be sampled if the receive bandwidth is reduced.

PAGE 56

56 Figure 1 11. Aliasing and undersampling ( Westbrook, C et al., 2005). A ) When the s ignal is sampled twice per cycle, the waveform is interpreted accurately. B ) Sampling once per cycle misinterprets the waveform as a straight line. C ) If sampled less than once per cycle, the waveform is misinterpreted as being at the wrong frequency (alia sed).

PAGE 57

57 Figure 1 12. Phase wrap artifact. Phase values outside the FOV with corresponding phase values inside the FOV (shown by red arrows) are sometimes mapped onto the FOV resulting in a phase wrap.

PAGE 58

58 A B Figure 1 13: Cause of Gibbs artifact. A ) Un dersampling of data incorrectly represents interfaces of high and low signal on the image. This gives rise to oscillations leading to banding patterns on the image. B ) Avoiding undersampling results in sharp edges reducing this artifact.

PAGE 59

59 Figure 1 14. C h emical shift and pixel shift. A) Each pixel has an individual frequency range of 125 Hz (32 kHz/256). Therefore fat and water protons adjacent to each other are mapped 1.76 pixels apart (220/125). B) When BW is reduced to 16 kHz, then each pixel can accom modate only 62.5 Hz (16 kHz/256). Therefore adjacent fat and water protons will be mapped 3.52 pixels apart (220/62.5).

PAGE 60

60 Figure 1 1 5 M echanism of Spin Echo technique 1) magnetization in the x y plane 2) S pins begin to dephase 3) applied, rotating all of the spins around the same axis. F aster spins are now placed behind the slower spins. The faster spins now continue to precess mo re rapidly than the slower spins 4) A ll the spins rephase to form the spin echo

PAGE 61

61 Figure 1 1 6 Pulse sequence diagram for a simple 2D spin echo sequence.

PAGE 62

62 Figure 1 1 7 Isotropic versus anisotropic diffusion. A) Isotropic Diffusion: Similar probability of molecular displacement in all directions due to absence of boundaries. B) Anisotropic Diffusion : Greater molecular diffusion in a preferred direction due to restricted boundaries.

PAGE 63

63 Figure 1 1 8 ce (Stejskal and Tanner, 1 965) with slice select and diffusion sensitive gradients applied along one physical dimension (shown shaded)

PAGE 64

64 Figure 1 1 9 Effect of molecular motion in the presence of diffusion gradients. Protons that moved in between the two gradients are indicated by boxes.

PAGE 65

65 Figure 1 20 Tensor model eigenvalues and eigenvectors. A) Primary eigenvalues 1 2 3 shown along the three axes. B) Corresponding eigenvectors V 1 V 2 and V 3 are shown.

PAGE 66

66 Figure 1 21 Range of FA. FA = 0 for complete isotropy, 1 = 2 = 3 FA=1 for complete 1 = 2 = 3 =0

PAGE 67

67 Figure 1 22 S chematic diagram fo r fiber tracking. It can be described as a tool which groups pixels (or voxels) based on their fiber orientation. The shape of the tensor ellipsoid indicates the direction of greatest diffusion (fiber orientations) at each pixel. These are linearly connect ed to form 'fibers tracts shown by the dotted red lines

PAGE 68

68 Table 1 1. The results of optimizing image quality To optimize image Adjusted Parameter Consequence Maximize SNR NEX matrix slice thickness BW read Maximize resolution (assuming a square FOV) Minimize scan time se matrix SNR movement artifact imaging

PAGE 69

69 Table 1 2. Parameters and their associated tradeoffs. Parameter Benefit Limitation Scan Time T1 weighting Number of Slices T2 weighting NA Scan Time Signal Averaging NA Scan T ime Slice Thickness Resolution Coverage Partial voluming Slice Thickness Resolution Partial voluming Coverage Resolution Coverage Aliasing Res olution Aliasing Coverage Resolution Scan Time SNR if pixel small Scan Time Resolution BW read Chemical Shift minimum TE BW read SNR Che mical Shift Large coil sensitive to artifacts aliasing with small FOV Small coil SNR less sensitive to artifacts less prone to aliasin g with small FOV

PAGE 70

70 CHAPTER 2 STUDY OF CONTROL RETINA USING HIGH ANGULAR RESOLUT ION DIFFUSION MICROSCOPY (HARDM) Background and Significance In recent years, diffusion tensor imaging (DTI) has found various applications, most of which hav e been to investig ate neuronal fiber orientation and muscle tissue microstructure ( Basser et al., 2000; Damon et al., 2002; Forder et al., 2001; Mori et al., 2002; Moseley et al., 1990; Scollan et al., 1998 ) The capability of DTI to assess skeletal muscle tissue damage has been established ( Zaraiskaya et al., 2006 ) As DTI can be used to study various microstructure and the lack thereof, t his study uses these capabilities to investigate the retina. Though the microstructure in the retina may not be as orga nized as that in myelinated neurons or muscle tissue, it is known to consist of layers with fairly distinct boundaries. Thus the primary motivation to use DTI is its ability to detect differences between normal and pathogenic tissue. The purpose of this s tudy is to determine the feasibility of HARDM in measuring anisotropic water diffusion in the normal control retina. This is the first phase of the project as achieving this will pave the way for non invasive early detection of various ocular diseases like diabetic retinopathy. As a first step, it was important to obtain anatomical images of the highest possible resolution of the mouse eye to distinguish individual cell layers within the retina. Some groups have been successful in observing multiple layers in the retina ( Cheng et al., 2006 ) However, our efforts have resulted in imaging further differentiation within these layers. We conducted high resolution anatomical MR microscopy of the murine retina followed by HARDM scans which enabled us to evaluate the anatomy and physiology of the retinal tissue. We also developed a novel technique where the primary orientation of the diffusion tensor

PAGE 71

71 ellipsoids in the photoreceptor cell layer of healthy retinas was studied. These were observed to point towards the center of the eye. Materials and Methods Animal Preparation and Sample Setup All animal procedures were performed in accordance with the ILAR Guide for Care and Use of Laboratory Animals. We used C57BL/6J mice obtained from the Jackson Laboratory (Bar Harb or, ME). This strain of mouse is usually used in diabetic research since diabetes can be fairly easily induced in them in a relatively short span of time ( Kaplan et al., 2006; Rees et al., 2005 ) All experiments were performed ex vivo The animals were sa crificed and the eyes were immediately excised and fixed in formalin solution (3% w/v). The eyes were maintained in the formalin solution for at least a period of 24 hours in order to permit complete fixation. Twenty four hours prior to imaging, they were transferred into cold phosphate buffered saline (PBS) solution to remove the fixative completely. This is a required procedure since aldehyde fixative solutions are known to cause a reduction in T1 and T2 of tissue as well as increase the extracellular app arent diffusion coefficient ( Shepherd et al., 2009 ) On the day of the experiment, the eyes were transferred to a 5 mm NMR tube and suspended in 43; 3M Corp.) to remove extraneous signals from the bathing medium. The sample was packed with gauze both above and below to prevent movement. This experimental setup is illustrated in Figure 2 1. MR Experiments All MR experiments were performed on a 17.6 T, 750 MHz (for 1 H) wide bore (89mm) NMR spectrometer (Bruker Instruments, Billerica, MA) at t he Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) Facility in the McKnight Brain

PAGE 72

72 Institute at the University of Florida, Gainesville, FL. The RF coil for the ex vivo experiments used was a birdcage coil (Doty Scientific) with diameter = 5 mm, length = 3.5 cm. High r esolution a natomical s cans A s chematic of the organizational structure of the mouse retina (Figure 2 2) shows the various layers that make up the neural retina (Ramalho et al., 2004) High resolution anatomical scans were obtain ed (Figure 2 3) which showed the retinal layers. A 3D gradient echo pulse sequence was used for this purpose using the following parameters: TR = 500 ms, TE = 7.5 ms, FOV = 0.38 cm x 0.38 cm x 0.38 cm, matrix size 200 x 180 x 80, NA=6. This yielded a final voxel resolution of 19 m x 21 m x 47.5 m. Total time for the acquisition was 12 hours. High a ngular r esolution d iffusion m icroscopy (HARDM) A 21 direction high angular resolution diffusion microscopy (HARDM ) sequence was used for the diffusion experi ments. A rank 2 tensor model was used as the fitting function for the HARDM data and the spherical coordinates used for the acquisition have been previously described by zarslan et al. ( zarslan and Mareci, 2003; zarslan et al., 2005 ) HARDM images of co ntrol eye were obtained (Figure 2 4) using the following parameters : TR=2500 ms, TE=1 5 96 ms, SW=50 kHz, gradient strength = 800 mT/m value = 1000 s/mm 2 n umber of a verages (NA) = 6, slice thickness = 0.25 mm matrix size = 128 x 128, FOV= 4 m m x 4 mm, spatial resolution = 31 m, total acquisition time = 12 hours. Data Analysis All the diffusion tensor images were processed using f anDTasia software developed at the Department of Computer and Information Sciences, University of

PAGE 73

73 Florida ( Barmpoutis 2009 ) This is a software application in the public domain for viewing multi layer image files in FDT format, which contains information about the magnetic gradient directions, b values and acquired DW MR images. It is particularly designed t o work with diffusion tensor MR images. The graphical user interface (GUI) provides the user with several options of viewing the data. Some of the features of the software include primary, secondary, and tertiary eigenvector visualization along with fract ional anisotropy (FA) maps, average diffusivity, and tractography visualization Each diffusion tensor is depicted by a colored ellipsoid, the color conveying the direction of the primary eigenvector. The raw data were converted using MATLAB into FDT (fan DTasia) format (Barmpoutis 2009) The software was used to obtain detailed maps of color coded diffusion tensor ellipsoids showing direction of diffusion in the retinal layers. Since the slices could be rotated, it enabled us to obtain a 3D visualization displaying the diffusion tensor ellipsoids in each slice providing deeper understanding of fiber orientation. The diffusion tensor field was estimated from diffusion weighted images using the constraint variational principle method described previously ( W ang et al., 2004 ) Consequently, a tensor spline was fitted to the diffusion tensor field in order to obtain a continuous and smoothly varying approximation of the tensor field (Barmpout is et al., 2006) This proc ess significantly reduces noise present in the diffusion tensor field, thus displaying a clearer picture of the fiber structure underneath. For the visualization of the tensor fields we represented the diffusion tensor at each v oxel by a Gaussian ellipsoid. The x, y, z components of the dominant eigenvectors of the diffusion tensors were assigned Red Green Blue (RGB) colors respectively for visualization purposes. The

PAGE 74

74 same color scheme was employed in the images showing the prima ry eigenvector field of a diffusion tensor dataset. A novel method of analyzing the water diffusion in the retina is presented. As can be seen from Figure 2 5 and 2 6 the fiber tracts corresponding to the rods and cones in the photoreceptor cell layer of the retina are aligned in such a way that they tend to point towards the center of the eye (Figure 2 8 A ). If the lines traced along these eigenvectors are extended, they intersect near the center of the eye (Figure 2 8 B ). Control eyes were subjected to t his technique; the extracted region shown in Figure 2 8 C corresponds to the retina T he isosurfaces of the Gaussian probabilities are plotted as thin cylindrical surfaces shown as lines. These lines were found to converge very close to the center of the ey e where the mixture of Gaussians has the global maximum (as shown by the red circle in Figure 2 8 C ). T he global maximum values near the center of the eyes (focal point) are shown as grey dots (Figure 2 9 C ,D). The maximum probability points of three contr ol eyes are projected on a probabilistic 3D eye atlas ( Figure 2 9 E). This calculation was performed by evaluating a quantitative measure using the following method. Each diffusion tensor of the retinal layer is represented by a multivariate Gaussian proba bility, whose isosurfaces have the same orientations as the primary eigenvector of the corresponding diffusion tensor. In the first step the region of the 3D HARDM dataset corresponding to the retinal area is extracted This can be done by thresholding th e diffusion weighted images, in which the retina has much bright er intensities compared to the rest of the regions of the eye. After having extracted the retinal regions we compute the 3x3 eigenvector matrix V i from the diffusion tensor

PAGE 75

75 matrix D i of each denotes the lattice index of the voxels in the retinal region. We assume that the first column of the matrices V i correspond to the principal eigenvector of D i isosurface of the probability takes an almost cylindrical shape parallel to the primary eigenvector and having radius equal to half pixel. Using these probabilities we formulate a mixture of Gaussians p(x)= 1/ n N ( x ; p i V i V i T ) and we evaluate it for every lattice point of the original 3D DTI volume. Results High R esolution A natomical M icroscopy Histochemical studies performed by various sources confirm that the mouse retina has an organized structure with several dist inct layers as is seen in Figure 2 2 (Ramalho et al., 2004) It is important to be able to visualize these layers using MRI for proceeding with the diffusion experiments. Using magnetic resonance microscopy (in plane resolution of 19 m) we were able to successfully identify these various layers in the retina without destruction of the tissue sample (Figure 2 3 ). Some ringing artifact (or Gibbs artifact) may be observed in the image. Gibbs art ifact is explained in Chapter 1. A discussion of the manifesta tion of this artifact in the image is included in the discussion section of this chapter. Apart from the retinal layers, other significant ocular anatomy like the vitreous humor and lens were observed We calculated the thickness of the neural retina to b e between 200 to 250 m. High Angular Resolution Diffusion Microscopy (HARDM) A sufficiently high spatial resolution (31 m) was achieved for the HARDM scans, which was essential to calculate rotationally invariant scalar measures of anisotropy

PAGE 76

76 such as fr actional anisotropy (FA) within the retina as well as the orientation and magnitude of restricted water diffusion. Figure 2 4 shows three dimensional reconstructions of HARDM scans of a control eye in alternate views. The direction of diffusion in the reti na shown by individual tensors is observed to be consistent when viewed from various angles (A and B) Figures 2 5 show s a three dimensional reconstruction of fiber tracts of a control eye. Figure 2 6 is a close up of the section highlighted by the yellow box in Figure 2 5 The retinal layers are marked by black arrows in both figures. The helical structure of the mouse lens can be seen by the fiber tracts (Fig ure 2 5 ). The lens which fills the central part of the murine eye is largely made of primary and s econdary lens fibers ( Chalupa and WIlliams, 2008 ) These fibers are oriented in a helical fashion with the older primary fibers at the center, filling the lumen of the lens vesicle. The secondary lens fibers are newer fibers towards the outer surface of t he lens and possess sufficient anisotropic diffusion to permit fiber tracking. Figure 2 7 is a comparison of FAs in the retina versus that of the vitreous humor (a region without restricted diffusion) in each sample (N =4). A paired student t test was perf ormed comparing the two regions of interest which were determined to be statistically significant (p<0.01). Figure 2 8 A shows the eigenvector field of the retina of a 2D slice from a control mouse eye Figure 2 8 B gives a s chematic showing how lines draw n through the eigenvectors can be extended towards the center. The central focal point is shown by the red dot The extracted region shown in Figure 2 8 C corresponds to the retina. Some of the isosurfaces of the Gaussian probabilities are plotted as thin cy lindrical surfaces

PAGE 77

77 These lines intersect near the center of the eye, where the mixture of Gaussians has the global maximum. Figure 2 9 C ,D shows the global maximum values near the c enter of the eye (focal point), shown as grey dots. Discussion The high ima ge resolution acquired in the anatomical images provides insight into the tissue microstructure. It may be observed that some ringing artifact (truncation or Gibbs artifact) exists in the anatomical image (Figure 2 3 ). This banding pattern is primarily see n in the phase encoding direction (up down direction in the image) and is a result of truncating or under representing the Fourier series due to less n umber of samples being measured, or in other words, due to a reduced matrix size in the phase encoding di rection. Sometimes, in an effort to reduce the total scan time, the phase encoding steps are reduced, which clearly is the case here. However, in this particular image, the presence of the banding pattern plays an important role as it prevents the correct visualization of the retinal layers. Therefore an effort must be made to reduce this artifact. This may be done by either interchanging the frequency and phase encoding directions or by increasing the phase encode matrix size from 180 to say, 256, which wo uld increase the total scan time from 12 hours to 17 hours. Considering that this is an ex vivo sample, this may be a possible option. Alternately, a matrix size between 180 and 256 may be optimized to reduce the Gibbs artifact in the image. Doing so will present a clearer picture of the existing layers. There are also many post processing methods performed by various groups which may also be considered (Liang et al., 1992; Smith and Nichols, 1990; Archibald and Gelb, 2002; Callaghan et al., 2005). Commonly employed techniques tend to reduce the oscillations caused due to the Gibbs artifact thereby eliminating the banding patterns which lead to a smoothing

PAGE 78

78 effect on the final processed image. However, it must be noted that this leads to some loss of data an d also reduced spatial resolution. There are also certain methods which attempt to avoid loss of data by extrapolating data in k space by making certain assumptions based on prior knowledge of the anatomy ( Block et al., 2008). However, given the current co ndition of the image, it may be observed that there are layers visible on certain areas of the retina which are not much affected by this artifact. For example, on the right side of the image, a part of the retina (highlighted by yellow arrows) shows at le ast seven different layers. This part of the image is not affected by the truncation artifact as the layers from outside to inside are in the frequency encode direction (right to left in the image). The non invasive nature of the experiments does not requi re tissue destruction or staining as required by the Berkowitz group where the sclera, aqueous humor, iris, and lens were separated for closer inspection (Berkowitz et al., 1999) A healthy retina has strong boundaries between its layers compared to the su rrounding anatomy. The significantly high FA values in the retina compared to the low FA values in the vitreous humor reflect the presence of an organized structure in the retina. The Duong group used fMRI to understand layer specific physiology ( Duong et al 2008 ) The Chen group has used DTI (6 directions) to detect basic differences between wild type and mouse models of retinal degeneration ( Chen et al., 2010 ) Both groups being limited by resolution were able to resolve the retina up to only 3 layer s, each of which actually comprise of sub layers. We were not only able to resolve the retina up to 9 layers, we also take a different approach by using a 21 direction HARDM protocol enabling effective fiber tracking so as to actually extend the primary ei genvectors towards the

PAGE 79

79 center of the eye. The novel analysis method may also pave the way for a radically different and decisive technique to analyze the onset of disease. It is hypothesized that retinas affected by diabetic retinopathy may have structural disorganization, due to the disruption of the blood retinal barrier. The high image resolution may be able to detect damaged areas with retinas Moreover, the FA in such retinas may be significantly lower than healthy retinas. T ensor ellipsoids in the im ages indicate the preferential direction of water diffusion in the retinal layers. The predominant direction of diffusion is observed to be perpendicular to the retinal layers rather than parallel to them. T his layer c orresponds to the photoreceptor segmen ts of the rods and cones (inner and outer segments) In future studies involving diabetic eyes, it is possible that this alignment of tensors is altered. The fiber tracking in the eye provides an in depth understanding of the underlying anatomy. It is inte resting to note the ability of this technique to visualize the helical structure of the mouse lens. We had hypothesized that if the primary eigenvectors of the normal retina were perpendicular to the retina, then it would be logical that the lines extendi ng along their direction would converge in the center of the eye, while it would not do so in degenerated retinas. Our focal point calculations perfectly agreed with this hypothesis. This may not necessarily be the case in diseased eyes. The capability to prove this and use these methods is an important step in determining various cases of retinal degeneration.

PAGE 80

80 Figure 2 1: Ex vivo Sample setup. The samples were packed with gauze to prevent movement during imaging.

PAGE 81

81 Figure 2 2 Sch ematic of the o rganizational structure of the mouse retina (Ramalho et al., 2004) S C, sclera; CH, choroids; RPE, retinal pigment epithelium; ROS,photoreceptor segments of rods and cones; ONL, nuclei of rods and cones; OSL, outer synaptic layer; INL, neu ron nuclear layer; ISL, inner synaptic layer; GCL, ganglion cell layer. ~ 250 m Outer Layer Inner Layer

PAGE 82

82 Figure 2 3. 3D high resolution gradient echo image of an in vitro eye with clearly differentiated retinal layers. The retinal layers visible are: Sclera (SC), choroid (CH), retinal pigment epithelium (RPE), photoreceptor segments of rods and cones (ROS), outer nuclear layer of rods and cones (ONL), outer synaptic layer (OSL), inner nuclear layer (INL); inner synaptic layer (ISL),ganglion cell layer (GCL). The yellow arrows indicate the retinal layers undistorted by truncation artifacts. SC CH RPE ROS ONL OSL INL ISL GCL V itreous H umor Lens

PAGE 83

83 A B Figure 2 4 HARDM i mages of control eye in alternate views. The direction of diffusion in the retina is observed to be consistent when viewed from various angles (A and B) Certain important anatomical parts are labeled. Retina Optic Nerv e Head Optic Nerve

PAGE 84

84 Figure 2 5 A 3D reconstruction of fiber tracts of a control eye The yellow box indicates a section of the retina which has the structured organization of the photoreceptor layer.

PAGE 85

85 Figure 2 6 Close up view of fiber tracts in the retina corresponding to the yellow box in Figure 2 4 The layer shown by black arrows have the organized structure of the photoreceptor layer

PAGE 86

86 Figure 2 7. Comparison of mean fractional anisotrop y of the r etina as compared to the vitreous h umor (N=4). The difference in FA in the two groups was found to be statistically significant (p<0.0001). 0.67 0.06 0.14 0.02 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Retina VH

PAGE 87

87 A Figure 2 8 Illustration of the proposed focal point calculation method. A ) The eigenvector field of the retina of a 2D slice from a control mouse eye. B) Schematic showing how the eigenvectors are extended towards the center. The central focal point is shown by the red dot. C ) The extracted region cor responding to the retina. The projected eigenvectors converge very close to the center of the eye (shown by red circle) B C C

PAGE 88

88 Fig ure 2 9 Evaluated mixture of Gaussian probabilities for healthy control eyes ( A,B ) obtained by drawing lines extending along the length of the eigenvectors in the retina (in 3D space). The global maximum values near the center of the eyes (focal point) ar e shown as grey dots ( C,D). The maximum probability points of three control eyes are projected on a probab ilistic 3D eye atlas ( E) A C D B E

PAGE 89

89 Table 2 1. Mean FA values in the retina versus the vitreous humor in individual control eyes. The difference in FA in the two groups was found to be statistically significant (p<0.0001). Sample # FA in Retina FA in Vitreous Humor 1 0.65 0.13 2 0 .58 0.16 3 0.71 0.12 4 0.69 0.12 5 0.73 0.15 Mean FA SD 0.67 0.06 0.14 0.0 1 8

PAGE 90

90 CHAPTER 3 STUDY OF DIABETIC RE TINOPATHY Background and Significance Etymology of Diabetes Mellitus The etymology of the name Diabetes Mellitus is attributed to the Romans and Greeks ( Williams and Pickup, 2004 ) The Romans identified this condition in which the mellitus the Latin word for sweet. This sweet, honey like taste was also reported by the Indian Hindu physician, Susruta in the 5 th century AD. He also mentions two forms of the disease which is classified today as Diabetes type 1 and type 2. The Greeks discovered that when people with this condition drank, the fluids came out in the urine as soon as they went in the mouth, like a s iphon. The Greek word for siphon is diabetes. This is the origin of the modern name for the disease, d iabetes m ellitus. Balance of Glucose and Insulin Diabetes mellitus is broadly defined as a disease characterized by chronically high blood glucose levels primarily caused by the inability of the human body to produce and/or use insulin (Williams and Pickup, 2004) Normally, glucose acts as an energy source for tissue cells and is absorbed in the intestines and channeled to the rest of the body via the bloo dstream. Since the need for glucose varies during the day, excess glucose is stored in the liver and muscles in the form of glycogen (chains of glucose molecules). When glucose is required, glucose is mobilized from the stored glycogen. This blood glucose level in the body is maintained in the body by two hormones having opposite actions: insulin and glucagon, which are both produced in the pancreas. Insulin

PAGE 91

9 1 a protein hormone that contains 51 amino acids. Insulin is required by almost all of the body's cells, but its major targets are liver cells, fat cells and muscle cells. While insulin lowers the blood glucose levels, glucagon raises blood glucose levels ( Kieffer et al., 1999 ) The pancreas releases glucagon when glucose levels fall too low. Glucagon causes the liver to convert stored glycogen into glucose, which is released into the bloodstream. High blood glucose levels stimulate the release of insulin. Insulin all ows glucose to be taken up and used by insulin dependent tissues. Thus, glucagon and insulin are part of a feedback system that keeps blood glucose levels at a stable level. A healthy pancreas adjusts the amount of insulin base d on the level of glucose. Bu t diabetes causes this process to break down and blood sugar levels become too high. Diabetes can be categorized into four types ( Saudek et al., 1997 ) : Type I Diabetes This is caused due to the complete or almost complete destruction of the pancreatic bet a cells (the cells that produce insulin). Type II Diabetes This is due to insulin resistance, or the body not responding to the insulin produced in the pancreas. Gestational diabetes mellitus (GDM) This occurs for the first time in pregnancy. Other types of diabetes There are caused by specific conditions such as endocrinopathies, genetic syndromes, etc In all the cases, glucose can no t move into the cells and blood glucose levels can become high. Over time, these high glucose levels can cause serious comp lications. Since p eople with t ype 1 diabetes are unable to produce insulin they require insulin right away whereas e xercise and weight control has been found to have a positive

PAGE 92

92 effect on type 2 diabetics. Other types of diabetes may be helped by treating their underlying cause. Gestational diabetes must be treated aggressively with proper diet and insulin Diabetic Retinopathy Diabetic retinopathy (DR) commonly affects both t ype 1 as well as type 2 diabetics ( Frank, 2004 ) In 2010, about 25.8 million peo ple had diabetes in the United States alone, out of which 4.2 million adults aged 40 years or older had diabetic retinopathy ( National Diabetes Statistics 2011 ) Diabetic retinopathy (DR) does not have any symptoms in its early stages, which is why in most people it remains undetected until it has progressed to a later stage. Pathophysiology Diabetic retinopathy is characterized by early re tinal microvascular dysfunction retinal blood vessels do not grow after birth. Diabetes causes retinal hypoxia ( Pierc e et al., 1995 ) an early marker of DR, which in turn results in neovascularization in the retina. The newly formed blood vessels do not have the properties of the original retinal blood vessels thus leading to hemorrhage and subsequent blurring of vision. There are three growth factors that are essentially signaling proteins affecting vascular growth, which contribute to retinal neovascularization. They are basic fibroblast growth factor (bFGF), insulin like growth factor I (IGF I) and VEGF. Out of these, V EGF is the chief contributing factor ( Ferrara et al., 1997; Frank et al., 1994; Pierce et al., 1995; Qaum et al., 2001 ) These damages change the anatomy of the blood retinal barrier (BRB) and subsequently make it more permeable. Diabetic retinopathy has fou r stages:

PAGE 93

93 Mild Nonproliferative Retinopathy This is the earliest stage where microaneurysms occur. They are small swellings in the retina's microvasculature. Moderate Nonproliferative Retinopathy This is a progressive stage where some of the blood vessels responsible for nourishing the retina get blocked. Severe Nonproliferative Retinopathy Many more blood vessels are blocked, depriving several areas of the retina with their blood supply. These areas of the retina send signals to the body to grow new blood vessels for nourishment. Proliferative Retinopathy At this advanced stage, the signals sent by the retina for nourishment trigger the growth of new blood vessels. This condition is called proliferative retinopathy. The new blood vessels formed in the sta ge of proliferative retinopathy are abnormal and fragile. They grow along the retina and along the surface of the clear, vitreous gel that fills the inside of the eye. By themselves, these blood vessels do not cause symptoms or vision loss. However, they h ave thin, fragile walls. If they leak blood, severe vision loss and even blindness can result. Treatment m ethods There are a number of basic eye exams used to detect later stages of this disease along with macular edema today. These include tests like visu al acuity tests, dilated eye exams and tonometry A visual acuity test is an eye exam checking for clarity of vision at various distances During a d ilated eye exam, the pupils are dilated using drops. Special magnifying lenses are used for closer examinat ion of the retina and optic nerve for signs of damage. Tonometry tests measure ocular pressure and look for abnormalities. The tests are performed to check for leaking blood vessels, retinal swelling (for macular edema), fatty deposits on the retina whic h are signs of leaking blood vessels and damaged nerve tissue Advanced eye exams include a fluorescein

PAGE 94

94 angiogram which involves the injection of dyes in order to identify leaking blood vessels. Current treatment methods usually do not include invasive pro cedures for the initial three stages of DR unless there are signs of macular edema. The progression of the retinopathy is prevented by controlling blood sugar, cholesterol and blood pressure. Proliferative retinopathy is generally treated using a laser sur gery technique known as scatter laser treatment which shrinks abnormal blood vessels. A disadvantage of this procedure is that it may affect adversely color vision and night vision. Another procedure called vitrectomy removes blood from the center of the e ye. All these procedures are non invasive in nature, hence the need arises for finding a way to detect and possibly characterize the early onset of the disease before the adverse effects or symptoms begin to surface. The experiments described in this chapt er attempt to address this very need. Materials and Methods Animal Preparations All animal protocols were in accordance to University of Florida Institutional Animal Care and Use Committee (IACUC) guidelines. C57 black mice were used for the studies. All i maging was performed ex vivo using a 17.6T, 750 MHz (for 1 H) wide bore (89mm) Bruker NMR spectrometer. Diabetic eyes came from mice with t ype I diabetes induced by a single dose of alloxan. For the injections, the m ice were heated (under a sun lamp) to di late the blood vessels. Alloxan was injected into the lateral tail vein in a small volume (~200 microliters). Doses were either 62.5 or 75 mg/kg body weight. Blood glucose was measured at regular intervals after a 4 hour fasting period by pricking the ta il vein. The instrument used to measure the droplet was a One Touch Ultra a common instrument used by diabetics. The mice were not maintained on

PAGE 95

95 supplemental insulin (blood glucose >200mg/dl). The mice were euthanized by cervical dislocation The excise d eyes were fixed in a 2% paraformaldehyde/1% glutaraldehyde solution in Tyrode's buffer. The eyes were maintained in the formalin solution for at least a period of 24 hours in order to permit complete fixation. Twenty four hours prior to imaging, they wer e transferred into cold phosphate buffered saline (PBS) solution to remove the fixative completely. This is a required procedure since aldehyde fixative solutions are known to alter the water relaxation and diffusion propertie s of tissue (Shepherd et al., 2009 ). On the day of the experiment, the eyes were transferred to a 5 43; 3M Corp.) to remove extraneous signals from the bathing medium. The sample setup was similar to the experiments in Chapter 2 and illustr ated in Figure 2 1. MR Imaging The diffusion images were obtained using a 21 direction high angular resolution diffusion imaging (HARDI) sequence. The parameters used were: TR=2500 ms, TE=16 b value = 1000 s/mm 2 NA=6, Matrix size =128 x 128 and FOV= 0.4 cm x 0.4 cm, Spatial Resolution = 31m x 31m Data Processing S tatistical comparisons of fractional anisotropy in the retina were made using a n un paired s tudent s t test, with significance established at the p<0.05 level. A novel analysis technique was optimized to l ook at yet another way of differentiating between diabetic and control eyes. The central focal point calculation method described in Chapter 2 was used to identify the eigenvectors in the retina which are projected inwards It is established in the previou s chapter that in control eyes, these eigenvectors intersect near the center of the eye Now since it is presumed that in

PAGE 96

96 diabetic eyes, the eigenvectors would be in disarray following retinal disruption, these eigenvectors would in fact not intersect in t he center of the eye. Hence the first step would be to check whether this intersection actually happens at the center or not. A second method was devised to measure the angular difference ( ) between consecutive eigenvectors in the retina. This method is illustrated in Figure 3 3 and Figure 3 4 In Figure 3 3 consider the eye with the eigenvectors aligned as shown in the retina. For consistency in each sample, the selected eigenvectors lay within a 2D quadrant of the eye, which included the retina Figure 3 4 shows that t he first eigenvector at the bottom of the quadrant would make a small angle ( 1 ) with the reference line The consecutive eigenvector would presumably make a bigger angle ( 2 ) with the same reference line and so on. Next, each of these angles are plotted on a graph as shown. Since these angles gradually and consistently increase in value, the curve made by these points can be assumed to be a linear curve as shown. The angular difference between consecutive eigenvectors is signified by the term in this chapter (for example, 2 1 ) For a control eye, this constant because the eigenvectors would gradually shift radially along the curve of the retina. However, this may not be the case for a diabetic retina These ang les made by the eigenvectors for each eye were plotted for each group The statistics associated with these plots included R 2 for each curve. These values provide the value of goodness of fit. An R 2 of 1 means a perfectly linear line, while lower values si gnify non linearity. A 95% confidence interval was also calculated for each curve. These values are listed in Tables 3 1 and 3 2 for each group respectively. All statistics involving

PAGE 97

97 goodness of fit and confidence intervals were performed using Microsoft Excel 2010 and Graphpad Prism 5 a software for performing statistical analysis. Results I n Chapter 2 we observe that the direction of diffusion in the control neural retina is perpendicular to the layers rather than parallel to them. This layer correspo nds to the photoreceptor s egments of the rods and cones (inner and outer segments). I n this study a systematic comparison has been made between control retinas and retinas from diabetic mice. Retinal D isintegration in Diabetic samples The integrity of the retinal layers is compromised in diabetic eyes with elevated glucose levels Results show significant differences between this group and control samples. Figure 3 1 illustrate s an example of direct comparison between a control retina (Figure 3 1 A) and a d iabetic retina (Figure 3 1 B) The diabetic retina clearly shows signs of dama ge due to diabetic retinopathy (section highlighte d by white circle in Figure 3 1 B). A point to be noted is that c ertain areas of the diabetic retina may be intact (similar to con trol retina) while certain areas have been compromised due to DR. Fractional anisotropy values were calculated in the retina of both groups (N=5 for each group). T he mean FA is significantly reduced in the diabetic retina s as compared to t he mean FA in co ntrol retinas (Figure 3 2 ). Using an unpaired student t test between the two groups the difference in FA is considered to be statistically significant (p<0.0001). Comparison of Curves between Groups The ang les made by the eigenvectors to the reference for each eye were plotted for control (Figure 3 5) as well as diabetic retina s (Figure 3 7 ) A total of 5 samples

PAGE 98

98 were analyzed for each group. A reference i s shown (in green) which would be the result of perfectly aligned eigenvectors pointing towards the center of the eye It is clearly revealed that in the case of control eyes, the curves follow a linear pattern close to the reference and consequently, hav e high R 2 values On the other hand, t he curves corresponding to the diabetic eyes do not follow any particular pattern at all. As a result, the R 2 value of the curves are significantly lower than their control counterparts. R 2 values for the control group as well as the diabetic group are listed in Tables 3 1 and 3 2 respectively. A 95% confidence interval for each individual sample is shown for both groups (Figure 3 6 and 3 8 ) The range of slopes for each confidence interval is listed in Tables 3 1 and 3 2. This range provides an estimate of the goodness of fit of the curves. As is expected, the control group has a consistently narrower range of slopes for each confidence interval. This is another straightforward method of differentiating between the two groups. A correlation was established between the R 2 of the angular difference curves from the diabetic samples and coinciding blood glucose levels. It was observed that the value of R 2 ( Figure 3 9 Table 3 2) Discussion Diabetic retinopathy (DR) does not have any symptoms in its early stages, which is why in most people it remains undetected until it has progressed to a later stage. Various treatment methods exist today fro m basic eye exams (visual acuity tests, dilated eye exams, tonometry etc) to advanced procedures (fluorescein angiograms, laser surgery, vitrectomy) to treat progressive stages of DR. These procedures are usually invasive in nature. Due to these reasons, t here is a dire need for detecting this

PAGE 99

99 disease at an early stage. By doing so, the condition can be curbed and will not require invasive treatment procedures (described above) at later stages. As BRB permeability has been established to be a direct indicat ion of the extent of retinopathy, various research groups have measured the loss of BRB integrity using tracers, contrast agents and histology tests. Early studies have used fluorescein as a marker, and its transport across the BRB has been characterized t o measure retinal permeability ( Tsuboi et al ., 1984 ) The Berkowitz group used a contrast agent to assess the breakdown of the BRB ( Berkowitz et al., 1999; Berkowitz et al., 1991; Berkowitz et al., 1992 ) By injecting Gd DTPA (gadolinium diethylene triami ne pentaacetic acid) in the disrupted BRB, the surrounding water proton relaxation rate was found to increase in direct proportion to the Gd DTPA concentration. The Duong group ( Duong et al ., 2002 ) observed increased fMRI activation in the retina choro id complex after visual stimulation. Chen at al. ( Chen et al., 2008 ) have provided insight to retinal tissue organization as well as carried out DTI studies to estimate differences between wild type and mouse models of retinal degeneration. These studies a re examples of using various techniques to study retinal function, but lack by way of spatial resolution needed to carry out effective fiber tracking within the retinal layers required to detect early changes in the retina due to pathologies. Various othe r techniques such as Doppler flowmetry, retinal photography, scanning laser ophthalmoscopy, and retinal oximetry measurements are used to detect DR ( Denninghoff et al ., 2000 ) but early detection of the disease through non invasive means has been a problem. Va rious histological studies have established that healthy retinas have a highly organized structure. Our earlier studies (Chapter 2) have shown that high field ex vivo

PAGE 100

100 MR imaging is capable of distinguishing individual retinal layers in C57BL/J mice. Since fractional anisotropy is a reliable marker for determining organized structure, we decided to use diffusion studies to investigate the presence or lack of organized structure in the retina. Our studies show that FA is consistently higher in the retina com pared to its surrounding anatomy Retinas obtained from healthy control animals exhibit a highly organized retinal structure. Diabetic retinas are subjected to various changes in anatomy and physiology due to the progression of diabetic retinopathy. The ne ovascularization around the retina caused by VEGF signaling and the increased permeability of the blood retinal barrier are some of the chief contributing factors These changes cause a marked disruption in the natural organizational structure of the retin al layers. The HARDM techniques established in this chapter are successful in detecting alterations in the eye following damage due to diabetic retinopathy. FA values calculated in this group are significantly lower than in their control counterparts. The FA in the vitreous humor for both groups did not show any significant change. The new technique developed to differentiate between control and diabetic eyes involving the eigenvector projections is found to be very effective The plot of angular difference s between consecutive eigenvectors provides a novel method to distinguish between the two groups. It was observed that the curves made by the eigenvectors to the reference line in the control group followed a linear regression with high values of R 2 Quite the opposite was observed to be true for d iabetic eyes with high blood glucose levels In the diabetic eyes, n ot only are the R 2 values significantly less, there are many more points lying outside the 95% confidence interval in this group. In the case of the control group, t he 95% confidence intervals include most of the points in the curve. Due to this, t he

PAGE 101

101 upper and lower limits of the confidence interval are also closer to each other in the control group. For each group, t he difference in the upper and lower limits of the 95% confidence interval bands along with their means are listed in Tables 3.1 and 3.2. The control group has a much lower mean than the diabetic group, which implies that there is excellent goodness of fit of the trend line for the samp les in the control group. Depending on the linearity of the curves a correlation between the shape of the curve and progression of disease may be characterized. G iven the findings in this study this possibly signifies that the value of the R 2 could be a direct indication of the extent of diabetic retinopathy. The plot of R 2 versus the number of days the blood glucose level of the mouse was above 300 mg/dl reaffirms this hypothesis, that the longer the mouse has been diabetic, the lower is the value of R 2 When the animal has diabetes for a prolonged period of time, due to the progression of the retinopathy, the retina is continually disrupted. This would cause loss of organization in the retina, consequently leading to lower value s of R 2 These techniques may aid future studies by providing early detection of pathogenic conditions in the retina. The key to making this possible in clinical practice is to first achieve the ability to observe these changes in the murine retina in vivo This would require the development of special RF coils capable of the high SNR and resolution requirement essential for these studies. Some progress has been made in this line of work and it will be discussed later in chapter 5 of this dissertation. The studies described in thi s chapter being ex vivo in nature, take the advantage of long scan times in order to get better SNR and spatial resolution For in vivo imaging, this aspect will need to be addressed by developing pulse sequences designed to reduce

PAGE 102

102 scan times. Once these p reclinical studies are successful, steps may be ta ken to translate this research i nto a clinical setting.

PAGE 103

103 A B Figure 3 1. Comparison between a control retina ( A) and a diabetic retina (B). The control retina shows an organized structure whereas i n the diabetic eye, t he retina is damaged (shown by white circle ).

PAGE 104

104 Figure 3 2 A comparison of the fractional anisotropies in the retinas of c ontrol versus d iabetic eyes. Using an unpaired student t test, th is difference was found to be statistically significant (p<0.0001). FA in the vitreous humor (used as an internal control) in both groups did not show any significant difference. 0.71 0.04 0.13 0.02 0.51 0.04 0.14 0.02 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Control Retina Control VH Diabetic Retina Diabetic VH FA

PAGE 105

105 Figure 3 3 A schematic illustrating the angular difference calculation method. The eige nvectors within a selected quadrant of the retina are displayed. Retina Optic Nerve Eigenvectors 90 Selected Quadrant

PAGE 106

106 A B Figure 3 4 Plot of points corresponding to angles ( ) made by eigenvectors with the reference line. A ) In the selected quadrant, lines are extended along the eigenvectors converging at the center of the eye. B ) A plot of the angles made by the extensions with the reference line

PAGE 107

107 Figure 3 5. Control eyes : reference line is plotted for each eigenvector lying in the selected quadrant. The process is repeated for 5 control eyes (labeled as Ctrl for Control) The green curve serves as a reference curve wh ich would be the result of perfectly aligned eigenvectors.

PAGE 108

108 Figure 3 6. The distribution of points in individual control samples is compared. 95% confidence intervals for each trend line are also displayed to show goodness of fit

PAGE 109

109 Figure 3 7 Diabetic eyes: reference line is plotted for each eigenvector lying in the selected quadrant. The process is repeated for 5 diabetic eyes (labeled as DB for Diabetic) The green curve serves as a reference curve whic h would be the result of perfectly aligned eigenvectors.

PAGE 110

110 Figure 3 8 The distribution of points in individual diabetic samples is compared. 95% confidence intervals for each trend line are also displayed.

PAGE 111

111 Figure 3 9 For the diabetic group: plot of R 2 versus the number of days the blood glucose level of the mouse was above 300 mg/dl. According to the graph a trend may be observed suggesting that the longer the mouse has been diabetic, the lower is the value of R 2

PAGE 112

112 Ta ble 3 1. Individual FA values in retinas and vitreous humor (used as an internal control) of healthy control eyes. The R 2 values of each curve (Figure 3 5) are listed. The difference between the upper and lower slopes of the 95% confidence interval (CI) fo r each sample is listed. Sample # FA in Retina FA in VH R 2 Diff bet wee n upper and lower slopes of 95% CI band 1 0.71 0.14 0.98 0.704 2 0.67 0.12 0.97 0.352 3 0.77 0.14 0.88 0.544 4 0.69 0.11 0.64 0.959 5 0.73 0.15 0.96 1.255 Mean FA SD 0.71 0.04 0.13 0.02

PAGE 113

113 Table 3 2 Individual FA values in retinas and vitreous humor (used as an internal control) of diabetic eyes. The R 2 values of each curve (from Figure 3 7 ) are listed. The difference between the upper and lower slopes of th e 95% confidence interval (CI) for each sample is listed. The number of days that each animal had a blood glucose level of 300 mg/dl and above is also listed. Sample # FA in Retina FA in VH R 2 Diff bet wee n upper and lower slopes of 95% CI band # days bl ood glucose level > 300 mg/dl 1 0.55 0.14 0 .65 2.87 13 2 0.54 0.16 0 .29 2.38 26 3 0.47 0.14 0 .21 2.74 31 4 0.53 0.13 0 .26 3.13 44 5 0.46 0.11 0 .7 0 2.46 12 Mean FA SD 0.51 0.04 0.14 0.02

PAGE 114

114 CHAPTER 4 REPAIR OF RETINAL PI GMENT EPITHELIUM (RP E ) LAYER USING HEMATOPOIETIC STEM C ELLS Background and Significance Found in the bone marrow, hematopoietic stem cells (HSCs) give rise to all blood cell types ( Mller et al., 2002; Mller Sieburg et al., 1992 ) Adult HSCs are restricted in their capacity to differentiate unlike embryonic stem cells which have the ability to differentiate into all lineages ( Sengupta et al., 2003; Sengupta et al., 2009 ) HSCs can be programmed to differentiate into specific tissue by transducing them with vectors expressing spe cific genes. The ability to control the differentiation of adult hematopoietic stem cells (HSCs) would promote development of new cell based therapies to treat multiple degenerative diseases. Degenerative diseases of the eye, particularly those involving damage to or loss of the retinal pigment epithelium (RPE) or the retina, are a major health complication associated with aging and diabetes. The RPE layer is a thin lining at the posterior section of the eye. The retina is supported and nourished by this cell layer. Being an important part of the ocular anatomy, t here is a dire need for methods of repairing damaged RPE. Fortunately, mouse models exist for RPE damage that can be exploited for new experimental therapeutic cell based strategies. We hypothesiz ed that in hematopoietic stem cells (HSCs), the expression of a gene that is both unique to a terminally differentiated cell type and a transcriptional modulator would shift the balance toward differentiation of HSCs into that unique cell type. Thus, by pr omoting HSC differentiation more readily into the obligatory cell type, the repair process could be enhanced. We selected the eye not only because of the increasing number of individuals with vision threatening conditions, but also because the eye serves a s an

PAGE 115

115 ideal model system; the retina is highly specialized, allowing precise identification and localization of stem cell derived tissue. The retina is critically dependent on RPE cells, which help to maintain neural retinal structure and function, and if d amaged lead to retinal degeneration and vision loss. One candidate for directing HSC differentiation into RPE is the RPE specific protein RPE65. epithelium The RPE65 protein is responsible fo r converting light entering the eye into electrical signals that are transmitted to the brain. The light sensitive pigments in the retina convert a molecule known as 11 cis retinal to another molecule called all trans retinal. This conversion triggers a ch emical reaction that creates the electrical signals detected by the brain. Hence RPE65 is critical for the normal formation of 11 cis retinal, and thus for photoreceptor function. In addition, RPE65 modulates the availability of retinoic acid, a known tra nscriptional regulator and differentiation inducer ( Gollapalli et al., 2004 ) The Grant group at University of Florida has used HSCs to provide repair functions induced injury in a model of choroidal neovascularizati on ( Chan Ling et al., 2006 ) They have further been able to restore visual function in mice impaired by retinal degeneration using HSCs to repair the retinal pigment epithelium (RPE) layer ( Sengupta et al., 2009 ) Their previous studies used systemic injecti on of sodium iodate (NaIO3) to ablate the retinal pigment epithelial (RPE) layer in C57BL6 mice and initiate neural retinal degeneration. The HSCs were infected ex vivo with lentiviral vectors expressing the RPE specific gene RPE65, leading to their differ entiation into RPE cells thereby repairing this layer of the retina

PAGE 116

116 The aim of this study is to investigate the capability of HARDM techniques to observe the repair of the RPE layer using a transgenic murine model of retinal degeneration (rd11). The key f inding in this study is that the differences in fraction al anisotropy values across various groups provide an indication of the condition of the RPE layer. HARDM is successful in tracking the repair of the RPE layer. Further research in this line aims to e stablish the use of HSCs in restoring visual function and using DTI as a non invasive tool to track the progress of stem cells in vivo Methods All animal protocols were in accordance to the University of Florida Institutional Animal Care and Use Committee (IACUC) guidelines. All image processing was done using fanDTasia software described in Chapter 2. Animal Preparation The mouse model used was a retinal degeneration model (rd11) described by the Davisson group at Jackson laboratories ( Hawes et al., 2002 ) The HSC infection technique used was the same as described in Sengupta et al. (Sengupta et al., 2009). Ex vivo HSC gene transfer with either retrovirus or lentivirus (LV) is typically followed by drug resistant selection in order to ensure that 100% of the infected HSCs are expressing the transgene and to expand the cell population prior to administration to ensure sufficient numbers of cells to repair the damaged tissue. This approach requires cell proliferation without induction of cell differentiation. T he LV infection protocol minimized exposure of the cells to ex vivo culturing conditions. Specifically, the cells were centrifuged at 300g for 5 minutes, the supernatant removed, and the cells ), polybreen (10

PAGE 117

117 The study samples were divided into 4 group s, each with N=3. Group 1 spontaneously developed retinal degeneration (rd11) that received by intravitreous injection, 5,000 purified c kit+/Sca 1+ bone marrow derived HSCs that were transduced with lentiviral vector expressing the RPE65 gene. Group 2 were rd/rd mice that received by systemic (intravenous) injection of 15,000 purified c kit+/Sca 1+ bone marrow derived HSCs transduced with lentiviral vector expressing the RPE65 gene. This was to test if the systemic route differed in its effect from local injection (i.e. at the site of injury). Group 3 were untreated controls with healthy retinas. Group 4 were rd/rd animals that received by intravitreous injection, 10,000 purified c kit+/Sca 1+ bone marrow derived HSCs transduced with lentiviral vector expressing the LacZ gene (as a control for transduction) MR Imaging The sample preparation for the excised eye s prior to MR imaging is the same as described in Chapter 2. All samples were scanned using a 17.6T, 750 MHz Bruker spectrometer at the Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) Facility in the McKnight Brain Institute at the University of Florida, Gainesville, FL. A 21 direction high angular resolution diffusion microscopy (HARDM) sequence was used with the following t ypical parameters: TR=2500 ms, TE=1 5 96 ms, SW=50 kHz, gradient strength = 800 mT/m mm 2 NA=6, slice thickness = 0.25 mm matrix size=128 x 128, FOV= 4 m m x 4 mm, spatial resolution = 31 m, total acquisition time = 12 hours. A rank 2 tensor model was used as the fitting function for the HARDM data and the spherical coordinates used for t he acquisition have been previously described by zarslan et al. (zarslan and Mareci, 2003; zarslan et al., 2005). All data processing was performed using f anDTasia. Image p rocessing and analysis techniques were the same as that used in Chapters 2 and 3.

PAGE 118

118 Results F igure 4 1 is a c omparison of the RPE layer of mice from different groups studied by Sengupta et al ( Sengupta et al., 2009 ) Panel A shows a n ormal uninjured eye stained with hematoxylin and eosin depicting integrity of the RPE layer and photorece ptors. Figure 4 1 B is a n eye with a damaged RPE layer where it is injected with LacZ infected HSCs. LacZ was generated as a control or companion recombinant virus. Figure 4 1 C is a repaired RPE layer after being injected with RPE65 infected HSCs. Figure 4 2 and Figure 4 3 are HARDM image s of eyes from group 1 and group 2 respectively clearly showing repaired RPE layer s No significant differences were observed in FA means or in the organizational structure of the retina between these groups HARDM data as well as t he mean FA in group 3 indicate good retinal integrity as is expected since the eye samples were obtained from untreated control animals HARDM of eyes from group 3 are not shown in this chapter since they are similar to control eyes from chapter 3 In group 4, the retinas exhibit loss of organizational structure due to the disruption in the RPE layer as can be seen from Figure 4 4. The mean FA of the retinas in g roup 4 was significantly lower than each of the previous groups (g roups 1, 2 and 3). In this case, the depleted RPE layer (denoted by the red circle) is clearly detected Fig ure 4 5 is a comparison of the FA in the retinas of Groups 1, 2, 3 and 4. The angles between consecutive eigenvectors in the retina (in selected quadrants of each sample ) were plotted using the same method described in chapter 3. We had observed in our earlier studies that the points obtained from control eyes followed a linear regression which was close to the reference line. The reference line is a result of

PAGE 119

119 perfectly a ligned vectors. The graphs obtained by similar methods for each group are shown in Figures 4 6, 4 7, 4 8 and 4 9 respectively. A 95% confidence interval for each individual sample is shown. Discussion HSCs have been used for their ability to replenish an d restore function in various cell types (Mller Sieburg et al., 1 992) They are also known for their ability to multiply into daughter cells to a very large extent ( Grant et al., 2002 ) The Grant group has used HSCs to repair the B s previously ruptured using a laser injury ( Chan Ling et al., 2006 ) They have further been able to restore visual function in mice impaired by retinal degeneration (Sengupta et al., 2009) As previously shown, treatment of mice with a single dose of 100 m g/kg of sodium iodate resulted in rounding up, detachment, and loss of RPE cells by 7 days after treatment (Mizota et al., 1997) This neural retinal degeneration was initiated by systemic injection of sodium iodate (NaIO3) to ablate the RPE which was in turn repaired using HSCs. In this study, we, in collaboration with the Grant group, used rd11 mice which are genetically altered to have ablated RPE layers. To enhance the reparative capacity of adult HSCs, the HSCs were programmed for cellular fate requir ed to rep air the specific injured tissue in this case, the cells of the depleted RPE layer. This process requires not only precise differentiation into RPE cells but also sufficient HSC recruitment and proliferation at the site of injury to restore prope r tissue function. RPE65 is critical for the normal formation of 11 cis retinal, and thus for photoreceptor function RPE65 has secondary functions and important roles in development that enable it to serve outside the visual cycle. The RPE specific gene ( RPE65 ) contains the RPE65 protein which is responsible for converting light entering the eye into electrical signals that are transmitted to the brain.

PAGE 120

120 T he HSCs were infected ex vivo with lentiviral vectors expressing RPE65, leading to their differentiatio n into RPE cells thereby repairing this layer of the retina. Histological validation of the repair process has been documented ( Sengupta et al., 2009 ) Recent studies have used diffusion techniques to study the repair of the spinal cord (Kamble et al., 201 1), and other tissue but they have not been applied to track stem cell therapy before. HARDM techniques established in chapter 2 of this dissertation were successfully used to investigate the capability of HSCs to repair the RPE layer In our previous stu dies (Chapter 2 and 3), we have shown that HARDM techniques are capable of detecting irregularities in retinal integrity Fractional anisotropy, which is a tested parameter for detecting structure, is used in this study as well. The key finding is that the differences in FA values across various groups provide an indication of the condition of the RPE layer. The retinas in groups 1 and 2 were found to be comparable to group 3 (untreated controls). This is apparent visually from the images as well as from th e mean FA in each group. This is an indication that the genetically modified HSC infected with the RPE65 gene contribute positively to the repair process of the RPE layer. The FA in group 4 was found to be significantly less than each of the other three gr oups. This is not unexpected, as the LacZ gene is not known to have any effect on the reparation process of the RPE layer. Our HARDM techniques are successful in detecting differences between the repaired versus unrepaired RPE layer s. G roups 1 and 2 do not show any significant difference in their organizati onal Given these results, it may be assumed that the repair mechanism or rate of repair is independent of the route of administration of stem cells. As is expected, in the sa mples

PAGE 121

121 in group 4, t he LacZ gene is not capable of restoring the RPE layer and this is readily apparent due to the lack of organizational structure in the RPE ( Figure 4 4 ) The mean FA in this group is significantly lower than the FA in groups 1, 2 and 3 A s a high FA value is attributed to the presence of organized structure, this is a direct indication of loss of organized structure in the RPE layer. This finding corroborates with the findings propos ed by Sengupta et al. (Sengupta et al., 2009). As we have previously observed in our diabetic retinopathy studies (Chapter 3), t he linear regression curves are a very good indication of the condition of the retina, and in this case, the RPE layer. It was observed that the curves in groups 1, 2 and 3 followed a l inear regression with high values of R 2 Group 4 on the other hand, demonstrated curves which did not follow a linear regression, and consequently had much lower R 2 values as compared to the other groups. Moreover, i n the case of group 1 through 3, the 95% confidence intervals includ e most of the data points This shows that there is excellent goodness of fit for th ese group s In the case of group 4, it is quite the opposite. The R 2 values are also significantly less for this groups indicating loss of struc ture in the retina. Further research in this line of work would aim to use non invasive in vivo MRI to track the restoration process initiated by the HSCs. The key to making this possible in clinical practice is to first achieve the ability to observe the se changes in the murine retina in vivo The studies described in this chapter being ex vivo in nature, use long scan times. For in vivo imaging, many aspects will need to be addressed including developing pulse sequences to reduce scan times while maintai ning image quality and developing special RF coils capable of the high sensitivity essential for these studies.

PAGE 122

122 Once these preclinical studies are successful, steps may be taken to translate this research onto a clinical setting. This would not only establ ish the use of HSCs in restoring visual function but also demonstrate the use of diffusion MRI in stem cell therapy.

PAGE 123

123 A B C Figure 4 1. Comparison of the RPE layer of mice from different groups studied by Sengupta et al. (Sengupta et al ., 2009) A) N ormal uninjured eye stained with hematoxylin and eosin showing integrity of the RPE layer and photoreceptors. B) An eye with a damaged RPE layer where it is injected with LacZ infected HSCs. C) A repaired RPE l ayer after being injected with RPE65 infected HSCs.

PAGE 124

124 Figure 4 2 HARDM of eye from Group 1 which is a representative of a restored RPE layer. Yellow arrows indicate the RPE layer.

PAGE 125

125 Figure 4 3 HARDM of eye from Group 2, which is a represen tative of a restored RPE layer Yellow arrows indicate the RPE layer.

PAGE 126

126 Figure 4 4 A retina from group 4 is shown. HARDM detects loss of structure due to a depleted RPE layer shown by yellow arrows.

PAGE 127

127 Figure 4 5. Comparison of FA in the retinas of repaired RPE laye r (Groups 1, 2), Untreated controls (Group 3) and ablated RPE layer (Group 4) 0.69 0.05 0.77 0.03 0.71 0.02 0.47 0.15 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Group 1 Group 2 Group 3 Group 4 FA

PAGE 128

128 Figure 4 6 Group 1 : line is plotted for each eigenvector lying in the selected quadrant. The dotted red line is used as a reference as it would be the result of perfectly aligned eigenvectors. 0 10 20 30 40 50 60 70 80 90 100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 # Points Group 1

PAGE 129

129 Figure 4 7 The group 1 distributions of points in individual samples are compared. 95% confidence intervals for each trend line are also displayed. s)

PAGE 130

130 Figure 4 8 Group 2 : l ine is plotted for each eigenvector lying in the selected quadrant. The dotted red line is used as a reference as it would be the result of perfectly aligned eigenvectors. 0 10 20 30 40 50 60 70 80 90 100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 # Points Group 2

PAGE 131

131 Figure 4 9. The group distributions of points in indi vidual samples are compared. 95% confidence intervals for each trend line are also displayed.

PAGE 132

132 Figure 4 10 Group 3 : reference line is plotted for each eigenvector lying in the selected quadrant. The dotted red line is used as a reference as it would be the result of perfectly aligned eigenvectors. 0 10 20 30 40 50 60 70 80 90 100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 # Points Group 3

PAGE 133

133 Figure 4 11. The group 3 distributions of points in individual samples are compared. 95% confidence intervals for each tre nd line are also displayed.

PAGE 134

134 Figure 4 12 Group 4 : reference line is plotted for each eigenvector lying in the selected quadrant. The dotted red line is used as a reference as it would be the result of perfectly aligned eigenvectors. 0 10 20 30 40 50 60 70 80 90 100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 # Points Group 4

PAGE 135

135 Figure 4 13. The group 4 distributions of points in individual samples are compared. 95% confidence intervals for each trend line are also displayed.

PAGE 136

136 Table 4 1. Mean FA values in the retina of each sample (for all groups) along with the co rresponding R 2 Sample # FA in Retina R 2 Group 1 1 0.70 0.77 2 0.74 0.81 3 0. 63 0.82 Mean FA SD 0.69 0.05 Mean R 2 0.80 0.03 Group 2 1 0.79 0.87 2 0.74 0.85 3 0.8 0.79 Mean FA SD 0.77 0.03 Mean R 2 0.83 0.04 Group 3 1 0.69 0.89 2 0.73 0.87 3 0.71 0.94 Mean FA SD 0.71 0.02 Mean R 2 0.90 0.04 Group 4 1 0.29 0.05 2 0.54 0.018 3 0.58 0. 35 Mean FA SD 0.47 0.15 Mean R 2 0.1 4 0.18

PAGE 137

137 CHAPTER 5 LABELLING OF STEM CE LLS WITH MONOCRYSTAL LINE IRON OXIDE NANOPARTICLES (MION) FOR TRACKING AND LOC ALIZATION BY MAGNETI C RESONANCE IMAGING Background and Significance The project described in Chapter 4 successfully establishes the fact that hematopoietic stem cells are able to repair the retinal p igment epithelium layer in a murine model of retinal degeneration. Another approach to tracking the migration of epithelial stem cells was taken by way of labeling or tagging the stem cells with nanoparticles which are detectable in MRI. In this study, end othelial progenitor cells (EPCs) were labeled in vitro and injected into the target area. These then migrated to the specific site of the damaged tissue in order to repair the retinal pigment epithelium. While stem cell therapy has succeeded in facilit ating tissue regeneration and promoting healing, reproducible clinical benefits have remained elusive for reasons that have not yet been defined ( Deans et al., 2000; Pittenger et al., 1999 ) Studies by the Grant group and others have demonstrated that very f ew stem cells localized within a precise target area are sufficient to accelerate tissue repair ( Espinosa Heidmann et al., 2003; Espinosa Heidmann et al., 2005 ) On the other hand, other reports demonstrate that aberrant localization of active stem cells ca n reduce t herapeutic utility ( Barbash et al., 2003 ) Given the additional, well established fact that hematopoietic stem cells (HSC) must migrate to areas of injury to initiate the repair process, the ability to track the migration of stem cells would grea tly enhance their therapeutic utility. Moreover, because HSC and most other stem cell populations possess vast potential for self replication, it is critical to be able to document the precise number of stem cell present, localize these cells and establish their viability after i nfusion. Therapeutic

PAGE 138

138 success will increase as methods are developed to optimize localization of infused cells and to direct infused stem cells to specific targets. One stem cell population with marked clinical utility is bone marrow derived stem cells (BMDC). These cells contribute to revascularization of wounds and damaged tissues. Numerous animal models have demonstrated the ability of transplanted marrow derived progenitors to accelerate revascularization and promote healing after they home to sites of damage. Homing of these cells is influenced by numerous factors, including alterations in the extracellular environment and the presence of physical barriers such as disrupted tissues, neoplastic cells, scarring, and altered oxygen c ontent. Moreover, loss of blood flow results in areas of ischemia which compromise the delivery of BMDC to sites of injury. As few as 5 st em 2 r eadily accelerate tissue repair and provide as many as half of the cells that contribute to new vessel development ( Chan Ling et al., 2006; Harris et al., 2006 ) To date, only a few methods ( Shapiro et al.,20 06; Heyn et a l., 2006; Stroh et al., 2005 ) have been developed to localize very low numbers of engrafted stem cells. Thus novel strategies to detect small numbers of transplanted stem cell populations such as BMDC are greatly needed. Stem cells are often confined within a small but critical area before they avidly proliferate, differentiate, and finally assemble into functional tissues. The development of methods to localize these cells would help identify aberrant l ocalization of transplanted stem cells. Here, we present a possible solution to this problem that employs high field strength magnetic resonance imaging of stem cells prelabeled with monocrystalline iron oxide.

PAGE 139

139 Magnetic resonance imaging (MRI) is widely used in clinical settings due to its excellent soft tissue contrast, high anatomical detail and minimal invasiveness. MRI sensitivity is improving by increasing signal to noise ratio (S N R ) through use of higher magnetic field strengths, high performance gradients, and better radio frequency (RF) coils ( Beck et a l., 2002 ) all of which enable better resolution in a shorter time. High field monitoring of transplanted cells with a resolution of 25 to Transplanted cells must be labeled with a MR contrast agent in order to be visualized. Among the several MR contrast agents available, monocrystalline iron oxide nanoparticles (MION) have been successfully used for cell tracking in vivo (for review, see references Bjornerud and Joh ansson, 2004; Bulte and Kraitchman, 2004;Wu et al., 2004). MIONs are an attractive contrast agent because of their small size (5 30 nm) and because their cellular uptake occurs by adsorptive pinocytosis (Allen and Meade, 2003; Bulte et al., 2002; Dodd et al., 2001; Frank et al., 2003; Sundstrom et al., 2004; Weissleder, 1999; Weissleder and Mahmood, 2001) minimizing cell alterations. I n this report, we evaluated the utility of MION labeling for assessing the d istribution of a well characterized human CD13 3+, population by MRI. We optimized the clinical applicability of our approach by not utilizing any agents to specifically enhance MION uptake by these cells. We assessed the effects of MION labeling on various aspects of cell functi on and used MRI to evaluate cell numbers and distribution.

PAGE 140

140 Methods MION Synthesis MIONs were prepared as previously described ( Kang,Y.S. 1996 ) by mixing FeCl 2 and FeCl 3 solutions with NaOH solution yielding a precipitate of Fe 3 O 4 (FeO/Fe 2 O 3 ). The precipit ate was collected via centrifugation and neutralized with HCl. Cationic colloidal nanoparticles were separated by centrifugation and stabilized with 1 M sodium citrate. Resultant MIONs were dialyzed against 0.1 M sodium citrate (pH 7.0) resulting in a brow n transparent solution. The particle size of the MION preparation was determined by transmission electron microscopy (TEM), and was estimated to be approximately 4 nm. The iron concentration of MION solutions was determined using a Cary 100 UV Vis spectrop hotometer. Solutions were acidified with HCl, buffered with sodium citrate (pH 4.0) and the iron was fully reduced to Fe 2+ with 0.5 M hydroquinone. o phenanthroline dye (20 mM) was used to form a colored complex with Fe 2+ complex, and the iron concentratio n was calculated based on the absorbance measurement at 580 nm ( Oca Cossio,J. 2004 ) Ex V ivo I maging Magnetic resonance images were acquired using a wide bore (89 mm) 750 MHz a 30 tip angle. The field of view was set at 19 mm19 mm, and the matrix size was 512400, to yield an in ave rages was 6. Total time for imaging was 1 h and 9 min, unless otherwise specified. To test the feasibility of imaging single MION labeled cells, stem cells were labeled as described above and then dispersed at an extremely low density of approximately 200

PAGE 141

141 cells/ml in 2% agarose gel in a 5 mm NMR glass tube. The individual 5 mm vials were placed inside a 20 mm NMR tube along with two 5mm marker vials. One marker vial used to e nsure proper identification of the vials in post processing. The agar samples were imaged using T2 weighted gradient echo imaging techniques. In order to provide direct evidence that the areas of decreased image intensity corresponded to MION located withi n labeled stem cells, cells were plated on fibronectin coated coverslips and examined using MRI and light microscopy. Detection of MION labeled Stem Cells in Mouse Hind Limb C57BL/6J mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Hind l imbs were injected intramuscularly at three sites with specific numbers of MION labeled c irculating CD34 + cells from pooled healthy human donors (Lonza, Rockland, ME) as indicated, suspended in 50 l of saline. Imaging was performed on three mice ( Figur e 5 2) approximately 1 hour post injection, using a gradient echo technique to enhance detection perturbations in T2. The field of view was 2 cm x 1.2 cm, and a 256 x 128 matrix was employed, yielding an in plane resolution of 78 m x 94 m. The tissue slice thickness was 700 m and the repetition time was 300 ms. Echo time was varied from 5 ms to 12 ms in 1 ms increments. Two days post imaging, mice were sacrificed and hind limbs fixed in 4% paraformaldehyde for histological analisys as described below Murine Model of Choroidal N eovascularization (CNV) C57BL/6J mice ( n =3/condition) (wild type, WT) (Jackson Laboratory) 6 8 weeks of age, were treated with laser photocoagulation in a manner similar to that used by ( Ryan,

PAGE 142

142 SJ 1979 ) and previously described by ( Sengupta et al., 2003 ) Brie fl y, each eye received three laser burns to rupture Bruch's membrane, resulting in a bubble over the ruptured area. The right eye of either laser injured or non injured animals were intravitreally injected with 50,000 huma n CD133 + CD14 and CD34 + MION labeled stem cells resuspended in 2 l saline while corresponding left eyes were injected with unlabeled stem cells or saline. All animals were imaged at 24 and 48 h after laser injury and stem cell injection. The overall in vivo setup included respiration monitoring, anesthesia and heating pads for the mice. The parameters used were TR=5000 ms, TE=10 ms, matrix size=512256, FOV=1.2 cm 0.9 cm and slice thickness=0.25 mm. At the time of euthanasia, the eyes were enucleated and incubated in 4% paraformaldehyde for 1 h and rinsed in PBS for 30 min before paraf fi n embedding. All statistics involving the plotting of linear regression curves and confidence intervals was performed using Microsoft Excel 2010 and Graphpad Prism 5 a software for performing statistical analysis. Results Optimization of Stem Cell Labeling Figure 5 1 illustrates how MION labeling allows imaging of a single stem cell in vitro A) Stem cells were labeled with increasing concentrations of MION (as indica ted), dispersed in 2% agar, and placed in NMR tubes for imaging. Blank: agar alone. Control 1: unlabeled cells. Fluorinert and H2O were used as controls for MRI. Shown is a representative slice through the tubes. B: Close up of tube # 5 in panel A. MION la beled cells are indicated (arrows). The larger dark spots represent air bubbles that formed during agar solidification. C and D: Stem cells cultured on fibronectin coated coverslips were incubated in medium without (C) or in the presence of (D) MION overni ght. After

PAGE 143

143 extensive washes to remove non internalized MION, samples were fixed in 4% paraformaldehyde and then imaged as described. Small areas within the coverslip were characterized by a decrease in image intensity, consistent with the presence of MION (D). E and F: The same samples shown in panels C and D, where stained for iron Data in the hind limb model reported herein support our contention that a very small number of human stem cells, perhaps as few as 5 cells, can be detected in vivo by this appr oach ( Figure 5 2). In V ivo Imaging Our initial scans ( Figure 5 3 ) were successful in achieving the resolution and depth penetration required to visualize the anatomy of the mouse eye. These imaging parameters were in turn employed to track the migration o f the stem cells towards the site of injury ( Figure 5 4 ). Panel A shows a m ouse eye intravitreally injected with unlabeled stem cells and imaged by MRI at 24 h ours following injection. B: The MION labeled cells are localized to the edge of the lens mor e an terior in the uninjured eye C: Intravitreal injection of MION labeled stem cells into mouse eyes that underwent rupture of Bruch's Membrane (3 burns) and imaged at MRI 24 h ours following lasering. The MION labeled cells are localized to the edge of the le ns in the posterior region of the eye beginning to incorporate into the retina and traversing the retina toward the area of injury. D: Intravitreal injection of unlabeled CD34 + cells into a mouse eye 24 h ours following injection. Dotted circles enclose MIO N labeled cells. Discussion Stem and progenitor cell therapies have shown success in regeneration and repair of damaged tissue, but one difficulty encountered in optimizing these techniques is the inability to follow the fate of the transplanted stem cells in the recipients ( Daldrup Link et

PAGE 144

144 al., 2005 ) To date, a wide spectrum of in vivo cell tracking techniques are available. However, all have limitations usually associated with inadequate contrast and potential radiation damage. While the use of scintigr aphic techniques to track labeled cells provides increased sensitivity in isolated tissue ( Kiessling et al., 2008; Thompson et al., 2005 ) these techniques do not provide the spatial resolution that is possible with high field MR, ( Badea et al., 2008; He rschman et al., 2003 ) Reporter genes, while providing much greater sensitivity for identifying labeled cells, generally lack spatial specificity for in vivo detection. MR imaging is well suited for this task, because it can enable both whole body examinat ions and subsequent detailed depictions of host organs with near microscopic anatomic resolution and excellent soft tissue contrast. In addition, MR imaging allows repetitive investigations without known side effects and without risking radiotoxic damage t o the transplanted cells. However, to visualize and track stem and progenitor cells, they must be labeled with MR imaging contrast agents. The d evelopment of dedicated labeling techniques has been investigated ( Bulte et al., 2 004; Crich et al., 20 04; Winna rd et al., 2003 ) Cell labeling techniques are restricted by the concentration of internalized contrast agent which limits sensitivity for detection by MRI. This constraint can be overcome by using very high magnetic field strengths ( Lewin et al., 2000 ) I n both experimental and clinical settings, cellular MRI offers the advantage of non invasive monitoring of the distribution of transplanted cells in real time at a single cell. Among the several MRI contrast agents available, MIONs have been successfully used to magnetically label cells ( Bjornerud et al., 2004; Bulte et al., 2004; Wu et al., 2004 ) They are our contrast agent of choice because of their small size (5 to

PAGE 145

145 30 nm) and because, in stem cells, MION are internalized without use of transfection agents or other agents that disrupt cell membranes. Both particle size and coating (citrate, carboxydextran, or others) impact the route (pinocytotic, phagocytotic, etc .) and efficacy of uptake in different cell lines ( Rogers et al., 2006 ) The optimal choice for labeling of a particular type of cell will be dependent upon these factors and the requisite sensitivity that is needed. The particles we used were comprised of a 4 nm iron oxide core with an anionic citrate coating, resulting in a hydrodynamic diameter of approximately 8 nm. Paramagnetic iron oxide particles in use for magnetic resonance contrast range from ultrasmall superparamagnetic iron oxide (USPIO; b40 nm) to superparamagnetic iron oxide (SPIO; usually characterized as 100 300 nm) to evaluate the effects on cell viability and sensitivity for detection of one species of iron oxi de contrast agent one which has been identified in previous studies to permit detection of small numbers of cells ( Heyn et al., 2006; Stroh et al., 2005 ) Susceptibility based detection methods, especially with paramagnetic agents, result in local field perturbations that are reflected by regions of decreased image density that may extend far beyond the image location of the agent itself. In the case of the in vitro experiments reported here, we sought to minimize this problem by serial dilution of the ce ll suspensions to yield very low density in both the agar and coverslip preparations. Moreover, the choice of echo time (15 ms) at this field (17.6 T) was chosen to maintain spatial information (i.e., longer echo times resulted in much larger areas of decr eased image intensity). Combined with validation by light microscopy on the same coverslip preparations, we are confident that we have detected single cells labeled with MION at

PAGE 146

146 this field. Here, we show for the first time that MION labeling, using this pr eparation, did not reduce viability of human stem cells; however, at higher MION concentrations, we observed an increase in early apoptosis. In vitro cell migration and behavior in short term culture was indistinguishable from that of unlabelled cells. Fur thermore, we demonstrate in vitro the detection of MION within single stem cells while, in vivo we found that very small numbers of stem cells could easily be localized by this method. Successful stem cell therapy is likely to depend on the localization o f small numbers of cells, therefore the results we present are of particular importance. These results support that MRI represents a technique that offers the advantage of non invasively monitoring a small number of stem cells which is essential for precis e definition of the distribution of injected cells, an endeavor which will allow optimization of stem cell therapy and answer key questions that may explain the failure and inconsistent results reported in many studies undertaken to date. Moreover, since s tem cells possess a tremendous ability to self replicate, it is important that they are transplanted in the correct location. Thus, strategies to localize stem cells have great therapeutic significance. The method we report will allow realization of that g oal. In our study, we employed a stem cell population that is known to differentiate into endothelial cells and directly revascularize areas of injury as well as provide a source of paracrine factors that promote vessel repair as well as maintain vascular health. EPC display a potential for repair of damaged vasculature, a major cause of morbidity and devastating pathology in eyes and other organs. The ability of these cells to migrate to the area where they are needed for repair is often altered in disease states that led to the initial damage, such as diabetic retinopathy, wounding or

PAGE 147

147 infection. Using a well defined model of vascular eye injury previously developed in our labs, we successfully employed MRI to assess the migration of stem cells within the e ye as they homed to foci of injury where they can facilitate vascular repair ( Grant et al., 2002 ) These results confirm our conclusion that MR imaging of stem cells is extremely effective in tracking the distribution and subsequent migration of stem cells in a delicate therapeutic setting, and will provide a means to allow further optimization of the localization of infused stem cells. Future applications of this technology may include in vivo monitoring of new cell based therapies, such as homing of mese nchymal stem cells in injured myocardium, homing of neurologic stem cells in impaired brain tissue, and evaluation of revascularized ischemic limbs. Specifically, MION labeled stem cells may be used in the future to monitor the therapeutic success of revas cularization in the myocardium and extremities. Furthermore, due to its high level of safety for human subjects, high field MRI is becoming optimum clinical standard in both the US and Europe. The recent decision by the United States Food and Drug Administ ration (FDA) to remove the In conclusion, iron oxide labeled human stem cells show excellent viability and function and can be detected non invasively in vivo with sub millimeter ana tomic resolution by using 11 T MR imaging equipment demonstrating that this cell tracking technique has clinical applicability. MION may provide biological insights relevant for the development of new cell based therapies and may be generally suited to mon itor stem cell homing and engraftment in patients.

PAGE 148

148 Figure 5 1. In vitro MR Imaging of labeled and unlabeled stem cells. (Li Calzi at al., 2009) A) Stem cells labeled with increasing concentrations of MION B) Close up of tube # 5 in panel A. MION lab eled cells are indicated (arrows). The larger dark spots represent air bubbles that formed during agar solidification. C D ) Stem cells cultured on fibronectin coated coverslips were incubated in medium without MION (C) and in the presence of MION (D)

PAGE 149

149 Figure 5 2. Injection of MION labeled stem cells in a mouse hind leg (Li Calzi at al., 2009). A ) I njected with varying levels of stem cell concentration. B, C, D ) Magnification of injection sites shows increased signal dropout with more stem cells (B) I 5: Injected 5 cells; (C) I 50: Injected 50 cells; (D) I 500: Injected 500 cells. MION concentration is same in all cases

PAGE 150

150 Figure 5 3 In vivo scan of a healthy control mouse eye Imaging parameters were TR=5000 ms, TE=9 ms, Matrix size = 51 2 x 256, NA=4, SW=101kHz, FOV=1.2 cm x 0.90 cm, slice thickness=0.25 mm

PAGE 151

151 Figure 5 4 In vivo tracking of stem cells injected in murine eyes with different conditions. Parameters used were TR=5000 ms, TE=10 ms, matrix size=512256, FOV=1.2 cm 0.9 cm and slice thickness=0.25 mm. Dotted circles enclose MION labeled cells. A) No Laser Injury Injection of MION labeled EPCs Imaged at time=24 hrs B) Laser Injury Injection of MION labeled EPCs Imaged at time=24 hrs. C) No Laser Injury In jection of MION labeled EPCs Imaged at time=24 hrs. D) Laser Injury Injection of MION labeled EPCs Imaged at time=24 hrs A B D C

PAGE 152

152 CHAPTER 6 CONCLUSIONS AND FUTU RE DIRECTIONS Magnetic Resonance Imaging has come a long way from its early days in the 1970s when t he first MR images were presented to the world. At that time it was considered a long shot from being taken seriously as an imaging modality of any clinical importance. Not only has that happened today, MRI today is considered as one of the premier methods of diagnosis, especially for detecting anatomical anomalies. Diffusion between nerves in the brain due to its capability to visualize water diffusion in restricted tissue Today it is widely used to understand anatomy and physiology which aid the medical field and drug discovery process. The topics studied in this dissertation focus on identifying and characterizing biomarkers for various pathologies which may contribute s ignificantly to early detection of diseases and treatment methods. Chapter 1 includes some important concepts and techniques that were employed in order to acquire data for this dissertation. Though the projects in this dissertation are essentially applic ation based studies, understanding and revisiting the basic physics of MRI is vital to have a clearer understanding of the overall scope of the studies. In chapter 2, t he ability of magnetic resonance microscopy to observe fine anatomical detail showing di stinct retinal cell layers without the need for tissue staining or destruction is explored Various methods to overcome artifacts such as truncation artifacts often observed in anatomical images were discussed. An important preliminary phase of the project was to test the feasibility of HARDM techniques in the control eye. This objective was achieved and anisotropic diffusion of water in the retina was measured. The FA in the retina was calculated to be significantly higher than that of the

PAGE 153

153 surrounding anat omy such as the vitreous humor. Our HARDM studies helped us establish the fact that in healthy control eyes, preferential diffusion occurs perpendicularly between the outer and inner layers of the retina rather than in a parallel fashion. Tensor ellipsoids fiber tracts and high resolution microscopic scans of the eye conform to established retinal anatomy. Moreover, a novel method was developed for analyzing the diffusion of water in the retina and calculating the focal point of primary eigenvectors in the center of the eye. In other words, lines extended from the eigenvectors in the photoreceptor cell layer of the healthy retina were found to converge near the center of the eye. This observation may be attributed to the uniform curvature of the retina and also its organized structure. The goal of c hapter 3 is essentially the crux of this dissertation. The overall aim of this chapter is to identify early biomarkers of diabetic retinopathy in mice. The studies aim to achieve this objective by investigating th e differences between the r etinas of healthy control mice and those of diabetic mice. Diabetic retinas are subjected to various changes in anatomy and physiology due to the progression of diabetic retinopathy. These changes cause a marked disruption in the natural organizational structure of the retinal layers. The question that is being addressed here is if we are able to identify this disruption using HARDM techniques. The hypothesis was that if we are able to identify an organized retinal structure, then we should also be able to observe lack of structure, or in other words, disruption caused by the progression of the disease. The HARDM techniques established are successful in detecting a lterations in the eye following damage due to d iabetic retinopathy FA values calculated in this group are significantly lower than in their control counterparts. The FA in the vitreous humor

PAGE 154

154 for both groups did not show any significant change. The new technique of calculating the focal points in the eyes were applied to b oth control and diabetic eyes. This technique was observed to be very effective in differentiating between the two groups In the case of control eyes, lines extended along the direction of the eigenvectors in the photoreceptor cell layer of the retina wer e found to converge near the center of the eye. In the case of diabetic eyes, the case was entirely different. Due to the disruption of the retina, the eigenvectors no longer pointed uniformly toward the center of the eye. This feature was mapped according to the technique developed earlier in this study. The plot of angles ( ) made by eigenvectors with the reference line provides a novel insight to distinguish between the two groups. The samples in the control group were found to follow a linear regression curve with much higher R 2 values than the diabetic group. This tendenc y of the control group to follow a linear regression suggests a correlation between the shape of the curve and the integrity of the retina However, w hen the animal has diabetes for a prolonged period of time, due to the progression of the retinopathy, the retina is continually disrupted. This causes loss of organization in the retina leading to consecutive eigenvectors in the retina being at odd angles to each other rather than being at a constant angular distance. This leads to curves with lower linear re gression. This hypothesis was tested and found to hold true as i t was observed value of R 2 This possibly signifies that the value of R 2 could be a direct indication of the extent of diabetic retinopathy. Further validation studies in this line could lead to this technique being used in practice to identify diabetic retinopathy at an early stage.

PAGE 155

155 Chapter 4 investigates the role of adult HSCs in the repair of the RPE layer The HARDM techniques developed in chapter 3 are applied in these studies as well. This is because the concept behind differentiating between healthy and damaged retinas is similar. The experiments are successful in detecting differences between the repai red versus unrepaired RPE layers. Further investigation using larger N numbers may reveal differences in the repair mechanism or rate of repair depending on the route of administration of stem cells. The mean FA in the group infected with LacZ gene is sign ificantly lower FA than that in groups 1, 2 and 3 which were expected to have higher FA values. As a high FA value is attributed to the presence of organized structure, and vice versa, this is a direct indication of loss of organized structure in the RPE l ayer. The technique involving plots of angular differences between consecutive eigenvectors is used successfully in this case. Groups 1 through 3 followed a linear regression with high values of R 2 Group 4 on the other hand, demonstrated curves which did not follow a linear regression, and consequently had much lower R 2 values as compared to the other groups. This comparison leads us to infer that the values of R 2 of these curves provide important clues to the condition of the retina. Variations of these t echniques may find use outside of ocular studies. Though these findings are at an initial stage, they hold promise to be applied to in vivo imaging, and ultimately in the clinical setting. Chapter 5 involves in vivo MRI of the progression of endothelial pr ogenitor cells as they migrate towards the site of injury in murine eyes. Iron oxide labeled human stem cells show excellent viability and function. The imaging techniques established show that stem cells can be detected non invasively in vivo with sub mil limeter anatomic resolution demonstrating that this cell tracking technique has clinical applicability. MION

PAGE 156

156 may provide biological insights relevant for the development of new cell based therapies and may be generally suited to monitor stem cell homing an d engraftment in patients. Future work in this line will include in vivo DTI of the mouse retina. Given the overall scope of this dissertation, it is essential to understand that though the ex vivo studies in chapters 2, 3 and 4 have been fairly successful in forming pathways into uncharted areas, there is a plethora of factors to be considered in order to transition into clinical studies, or for that matter, even in vivo studies. Some of the main advantages that ex vivo studies have over in vivo studies ar e increased scan time facilitating higher image quality, no motion artifacts, and coils suited to imaging a small volume of tissue. To suit in vivo time frames, scan times may be decreased by optimizing pulse sequences to suit the needs. Needless to say, t his will cause reduction of image quality, and this can be considered as one of the immediate issues to be addressed in order to move to the next logical step in this area of research. Also considering the technological advances being made in the field of MRI, it may just be a matter of time before optimal image quality is attained. Chapters 2, 3 and 4 only involve ex vivo experiments. Chapter 5 involves in vivo experiments but does not include DTI studies where the spatial resolutions of the scans were not top priority The aim for these experiments was to track the migration of the MION particles, and hence eye movement was not a hindrance to achieving the results. But if in vivo DTI of the retina is to be performed, then motion artifacts due to eye moveme nt may be an important factor to be considered. Muscle relaxants like ketamine xylazine have been used in the past to reduce eye movement in mice ( Erhardt et al., 1984) Recent studies have observed better results using a combination of 1% isoflurane and p ancuronium anesthesia (Nair

PAGE 157

157 et al., 2011). Various avenues may be explored to achieve minimal eye movement thereby reducing the chance of artifacts. Once the various hurdles to in vivo imaging are overcome, these studies have the potential to pave the way for the early diagnosis of retinal disorders in a clinical setting.

PAGE 158

158 LIST OF REFERENCES Aiello LP, Avery RL, Arrigg PG, Keyt BA, Jampel HD, Shah ST, Pasquale LR, Thieme H, Iwamoto MA, Park JE, and others. 1994. Vascular endothelial growth factor in ocul ar fluid of patients with diabetic retinopathy and other retinal disorders. N Engl J Med 331(22):1480 1487. Aiello LP, Pierce EA, Foley E, Takagi H, Chen H, Riddle L, Ferrara N, King GL, Smith LEH. 1995. Suppression of retinal neovascularization in vivo by inhibition of vascular endothelial growth factor (VEGF) using soluble VEGF receptor chimeric proteins. Proc. Natl. Acad. Sci. USA 92:10457 10461. Airey D, Robbins A, Enzinger K, Wu F, Collins C. 2005. Variation in the cortical area map of C57BL/6J and DBA /2J inbred mice predicts strain identity. BMC Neuroscience 6(1):18. Akbarzadeh A, Norouzian D, Mehrabi M, Jamshidi S, Farhangi A, Verdi A, Mofidian S, Rad B. 2007. Induction of diabetes by streptozotocin in rats. Indian Journal of Clinical Biochemistry 22( 2):60 64. Alberti KGMM, Zimmet PZ, Consultation WHO. 1998. Definition, diagnosis and classification of diabetes mellitus and its complications. part 1: Diagnosis and classification of diabetes mellitus. provisional report of a WHO consultation. Diabetic Me dicine 15(7):539 553. Allen MJ and Meade TJ. 2003. Synthesis and visualization of a membrane permeable MRI contrast agent. J Biol Inorg Chem 8(7):746 50. Amin RH, Frank RN, Kennedy A, Eliott D, Puklin JE, Abrams GW. 1997. Vascular endothelial growth factor is present in glial cells of the retina and optic nerve of human subjects with nonproliferative diabetic retinopathy. Invest. Ophthalmol. Vis. Sci. 38(1):36 47. Antonetti DA, Barber AJ, Hollinger LA, Wolpert EB, Gardner TW. 1999. Vascular endothelial grow th factor induces rapid phosphorylation of tight junction proteins occludin and zonula occluden 1. Journal of Biological Chemistry 274(33):23463 23467. Antonetti DA, Barber AJ, Khin S, Lieth E, Tarbell JM, Gardner TW. 1998. Vascular permeability in experim ental diabetes is associated with reduced endothelial occludin content: Vascular endothelial growth factor decreases occludin in retinal endothelial cells. penn state retina research group. Diabetes 47(12):1953 1959. Appukuttan B, McFarland TJ, Davies MH, Atchaneeyasakul L, Zhang Y, Babra B, Pan Y, Rosenbaum JT, Acott T, Powers MR, and others. 2007. Identification of novel alternatively spliced isoforms of RTEF 1 within human ocular vascular endothelial cells and murine retina. Investigative Ophthalmology & Visual Science 48(8).

PAGE 159

159 Archibald R and Gelb A ringing artifact in MRI scans while keeping tissue boundary integrity IEEE Transactions on Medical Imaging vol. 21, no. 4, pp. 305 319, 2002. Badea CT, Drangova M, Holdsworth D W, Johnson GA. 2008. In vivo small animal imaging using micro CT and digital subtraction angiography. Physics in Medicine and Biology 53(19):R319 50. Bae SK, Kim JY, Yang SH, Kim JW, Kim T, Lee MG. 2006. Pharmacokinetics of oltipraz in rat models of diabet es mellitus induced by alloxan or streptozotocin. Life Sciences 78(20):2287 2294. Baltes C, Radzwill N, Bosshard S, Marek D, Rudin M. 2009. Micro MRI of the mouse brain using a novel 400 MHz cryogenic quadrature RF probe. NMR in Biomedicine 22(8):834 842 Barbash IM, Chouraqui P, Baron J, Feinberg MS, Etzion S, Tessone A, Miller L, Guetta E, Zipori D, Kedes LH, and others. 2003. Systemic delivery of bone marrow derived mesenchymal stem cells to the infarcted myocardium: Feasibility, cell migration, and body distribution. Circulation 108(7):863 8. Barber AJ, Antonetti DA, Kern TS, Reiter CEN, Soans RS, Krady JK, Levison SW, Gardner TW, Bronson SK. 2005. The Ins2Akita mouse as a model of early retinal complications in diabetes. Invest. Ophthalmol. Vis. Sci. 46 (6):2210 2218. Barker PB, Breiter SN, Soher BJ, Chatham JC, Forder JR, Samphilipo MA, Magee CA, Anderson JH. 1994. Quantitative proton spectroscopy of canine brain: In vivo and in vitro correlations. Magnetic Resonance in Medicine 32(2):157 163. Barmpoutis A. 2009. fanDTasia, diffusion weighted MRI processing tool, open resource from http://www.cise.ufl.edu/~abarmpou/lab/fanDTasia/ Barmpoutis, A., Vemuri, B.C., Forder, J.R. 2006. Robust tensor splines for approximation of diffusion tensor MRI data. Basser PJ and Pierpaoli C. 1996. Microstructural and physiological features of tissues elucidated by quantitative diffusion tensor MRI. Journal of Magnetic Resonance. Series B 111(3):2 09 219. Basser PJ, Mattiello J, LeBihan D. 1994. MR diffusion tensor spectroscopy and imaging. Biophysical Journal 66(1):259 267. Basser PJ, Sinisa P, Carlo P, Jeffrey D, Akram A. 2000. In vivo fiber tractography using DT MRI data. Magnetic Resonance in Me dicine 44(4):625 632. Beck B, Plant DH, Grant SC, Thelwall PE, Silver X, Mareci TH, Benveniste H, Smith M, Collins C, Crozier S, and others. 2002. Progress in high field MRI at the University of Florida Magma 13(3):152 7.

PAGE 160

160 Berkowitz BA, Kowluru RA, Frank R N, Kern TS, Hohman TC, Prakash M. 1999. Subnormal retinal oxygenation response precedes diabetic like retinopathy. Invest. Ophthalmol. Vis. Sci. 40(9):2100 2105. Berkowitz BA, Sato Y, Wilson CA, de Juan E. 1991. Blood retinal barrier breakdown investigated by real time magnetic resonance imaging after gadolinium diethylenetriaminepentaacetic acid injection. Invest. Ophthalmol. Vis. Sci. 32(11):2854 2860. Berkowitz BA, Tofts PS, Sen HA. 1992. Accurate and precise measurement of blood retinal barrier breakdow n using dynamic gd DTPA MRI. Investigative Ophthalmology & Visual Science 33(13). Bjornerud A and Johansson L. 2004. The utility of superparamagnetic contrast agents in MRI: Theoretical consideration and applications in the cardiovascular system. NMR Biome d 17(7):465 77. Bloch F. 1946. Nuclear induction. Physical Review 70:460 474. Block KT, Uecker M, Frahm J 2008. Suppression of MRI Truncation Artifacts Using Total Variation Constrained Data Extrapolation, International Journal of Biomedical Imaging, Volum e 2008:184123 Bolzn AD and Bianchi MS. 2002. Genotoxicity of streptozotocin. Mutation Research/Reviews in Mutation Research 512(2 3):121 134. Brooks SE, Gu X, Samuel S, Marcus DM, Bartoli M, Huang PL, Caldwell RB. 2001. Reduced severity of oxygen induced retinopathy in eNOS deficient mice. Invest. Ophthalmol. Vis. Sci. 42(1):222 228. Brown MA and Semelka RC, 2003, MRI : B asic P rinciples and A pplications John Wiley & Sons, Inc Brown R. 1828. A brief account of microscopical observations made in the months of june, july and august, 1827, on the particles contained in the pollen of plants; and on the general existence of active molecules in organic and inorganic bodies. Phil. Mag. 4(161 173). Bulte JW and Kraitchman DL. 2004. Iron oxide MR contrast agents for molecular and cellular imaging. NMR Biomed 17(7):484 99. Butler JM, Guthrie SM, Koc M, Afzal A, Caballero S, Brooks HL, Mames RN, Segal MS, Grant MB, Scott EW. 2005. SDF 1 is both necessary and sufficient to promote proliferative retinopathy. Journal of C linical Investigation 115(1):86 93. Callaghan MF, Larkman DJ, Hajnal JV. Pad methods for reconstruction and feature extraction in magnetic resonance imaging. Magnetic Resonance in Medicine. 2005;54(6):1490 1502.

PAGE 161

161 Carr HY and Purcell EM. 1954. Effects of di ffusion on free precession in nuclear magnetic resonance experiments. Physical Review 94(3):630 638. Chalupa, L.M. and Finlay, B. 1998. Development and organization of the retina from molecules to function. Springer Verlag New York, LLC. Chalupa, L.M. and Williams, R.W. 2008. Eye, retina, and visual system of the mouse. MIT Press. Chan Ling T, Baxter L, Afzal A, Sengupta N, Caballero S, Rosinova E, Grant MB. 2006. Hematopoietic stem cells provide repair functions after laser induced bruch's membrane rupture model of choroidal neovascularization. Am J Pathol 168(3):1031 44. Chen J, Wang Q, Chen S, Wickline SA, Song S. 2010. In vivo diffusion tensor MRI of the mouse retina: A noninvasive visualization of tissue organization. NMR in Biomedicine 24(5):447 451. C hen J, Wang Q, Zhang H, Yang X, Wang J, Berkowitz BA, Wickline SA, Song S. 2008. In vivo quantification of T1, T2, and apparent diffusion coefficient in the mouse retina at 11.74T. Magnetic Resonance in Medicine 59(4):731 738. Cheng H, Nair G, Walker TA, K im MK, Pardue MT, Thul PM, Olson DE, Duong TQ. 2006. Structural and functional MRI reveals multiple retinal layers. Proceedings of the National Academy of Sciences of the United States of America 103(46):17525 17530. Crich SG, Biancone L, Cantaluppi V, Du o D, Esposito G, Russo S, Camussi G, Aime S. 2004. Improved route for the visualization of stem cells labeled with a gd /Eu chelate as dual (MRI and fluorescence) agent. Magn Reson Med 51(5):938 44. Cuthbertson R and Mandel T. 1986. Anatomy of the mouse re tina. endothelial cell pericyte ratio and capillary distribution. Invest Ophthalmol Vis Sci. 27(11):1659 64. Daldrup Link HE, Rudelius M, Piontek G, Metz S, Brauer R, Debus G, Corot C, Schlegel J, Link TM, Peschel C, and others. 2005. Migration of iron oxi de labeled human hematopoietic progenitor cells in a mouse model: In vivo monitoring with 1.5 T MR imaging equipment. Radiology 234(1):197 205. Damadian R. 1971. Tumor detection by nuclear magnetic resonance. Science 171(3976):1151 1153. Damon BM, Ding Z, Anderson AW, Freyer AS, Gore JC. 2002. Validation of diffusion tensor MRI based muscle fiber tracking. Magnetic Resonance in Medicine 48(1):97 104.

PAGE 162

162 Dawson DW, Volpert OV, Gillis P, Crawford SE, Xu HJ, Benedict W, Bouck NP. 1999. Pigment epithelium derived factor: A potent inhibitor of angiogenesis. Science 285(5425):245 248. de Gooyer TE, Stevenson KA, Humphries P, Simpson DAC, Gardiner TA, Stitt AW. 2006. Retinopathy is reduced during experimental diabetes in a mouse model of outer retinal degeneration. In vest. Ophthalmol. Vis. Sci. 47(12):5561 5568. Deans RJ and Moseley AB. 2000. Mesenchymal stem cells: Biology and potential clinical uses. Experimental Hematology 28(8):875 884. Denninghoff KR, Smith MH, Hillman L. 2000. Retinal imaging techniques in diabet es. Diabetes Technology & Therapeutics 2(1). Diabetic retinopathy, obtained from http://www.nei.nih.gov/health/diabetic/retinopathy.asp 2006 Dodd CH, Hsu H, Chu W, Yang P, Zhang H, Mountz JD, Zinn K, Forder J, Josephson L, Weissleder R, and others. 2001. Normal T cell response and in vivo magnetic resonance imaging of T cells loaded with HIV transactivator peptide derived superparamagnetic nanoparticles. Journal of Immunolog ical Methods 256(1 2):89 105. Dodd CH, Hsu HC, Chu WJ, Yang P, Zhang HG, D. MJ,Jr, Zinn K, Forder J, Josephson L, Weissleder R, and others. 2001. Normal T cell response and in vivo magnetic resonance imaging of T cells loaded with HIV transactivator peptid e derived superparamagnetic nanoparticles. J Immunol Methods 256(1 2):89 105. Donelly, R. and Horton, E.S. 2005. Vascular complications of diabetes: Current issues in pathogenesis and treatment, Second ed. Blackwell Publishing. Dong CJ and Werblin FS. 1998 Temporal contrast enhancement via GABAC feedback at bipolar terminals in the tiger salamander retina. The Journal of Neurophysiology 79(4):2171 2180. Duong TQ, Ngan S, Ugurbil K, Kim S. 2002. Functional magnetic resonance imaging of the retina. Invest. O phthalmol. Vis. Sci. 43(4):1176 1181. Duong TQ, Pardue MT, Thul PM, Olson DE, Cheng H, Nair G, Li Y, Kim M, Zhang X, Shen Q. 2008. Layer specific anatomical, physiological and functional MRI of the retina. NMR in Biomedicine 21(9):978 996. Erhardt W, Hebe stedt A, Aschenbrenner G, Pichotka B and Blmel G. 1984. A comparative study with various anesthetics in mice (pentobarbitone, ketamine xylazine, carfentanyl etomidate).Research in Experimental Medicine. Volume 184, Number 3

PAGE 163

163 Einstein A. 1905. Zur elektrody namik bewegter krper [on the electrodynamics of moving bodies]. Annalen Der Physik 14:891 921. El Remessy AB, Al Shabrawey M, Khalifa Y, Tsai N, Caldwell RB, Liou GI. 2006. Neuroprotective and blood retinal barrier preserving effects of cannabidiol in exp erimental diabetes. Am J Pathol 168(1):235 244. Elster, A.D. and Burdette, J.H. 2001. Questions and answers in magnetic resonance imaging, 2nd ed. Mosby. Espinosa Heidmann DG, Caicedo A, Hernandez EP, Csaky KG, Cousins SW. 2003. Bone marrow derived progeni tor cells contribute to experimental choroidal neovascularization. Investigative Ophthalmology & Visual Science 44(11):4914 4919. Espinosa Heidmann DG, Reinoso MA, Pina Y, Csaky KG, Caicedo A, Cousins SW. 2005. Quantitative enumeration of vascular smooth m uscle cells and endothelial cells derived from bone marrow precursors in experimental choroidal neovascularization. Experimental Eye Research 80(3):369 378. Feit Leichman RA, Kinouchi R, Takeda M, Fan Z, Mohr S, Kern TS, Chen DF. 2005. Vascular damage in a mouse model of diabetic retinopathy: Relation to neuronal and glial changes. Invest. Ophthalmol. Vis. Sci. 46(11):4281 4287. Ferrara N and Davis Smyth T. 1997. The biology of vascular endothelial growth factor. Endocr Rev 18(1):4 25. Flint JJ, Hansen B, F ey M, Schmidig D, King MA, Vestergaard Poulsen P, Blackband SJ. 2010. Cellular level diffusion tensor microscopy and fiber tracking in mammalian nervous tissue with direct histological correlation. NeuroImage 52(2):556 561. Flint JJ, Lee CH, Hansen B, Fey M, Schmidig D, Bui JD, King MA, Vestergaard Poulsen P, Blackband SJ. 2009. Magnetic resonance microscopy of mammalian neurons. NeuroImage 46(4):1037 1040. Forder JR, Bui JD, Buckley DL, Blackband SJ. 2001. MR imaging measurement of compartmental water diff usion in perfused heart slices. American Journal of Physiology Heart and Circulatory Physiology 281(3):H1280 H1285. Frank JA, Miller BR, Arbab AS, Zywicke HA, Jordan EK, Lewis BK, H. BL,Jr, Bulte JW. 2003. Clinically applicable labeling of mammalian and stem cells by combining superparamagnetic iron oxides and transfection agents. Radiology 228(2):480 7. Frank L. 2002. Characterization of anisotropy in high angular resolution diffusion weighted MRI. Magnetic Resonance in Medicine 47(6):1083 1099.

PAGE 164

164 Frank LR. 2001. Anisotropy in high angular resolution diffusion weighted MRI. Magn Reson Med 45(6):935 939. Frank RN. 1994. Vascular endothelial growth factor its role in retinal vascular proliferation. The New England Journal of Medicine 331(1519 1520). Frank RN. 2004. Diabetic retinopathy. N Engl J Med 350(1):48 58. Freeman WM, Bixler GV, Brucklacher RM, Lin CM, Patel KM, Vanguilder HD, Lanoue KF, Kimball SR, Barb er AJ, Antonetti DA, and others. 2009. A multistep validation process of biomarkers for preclinical drug development. The Pharmacogenomics Journal aop(current). Gardiner TA, Gibson DS, de Gooyer TE, de la Cruz VF, McDonald DM, Stitt AW. 2005. Inhibition of tumor necrosis factor (Gardiner et al.) improves physiological angiogenesis and reduces pathological neovascularization in ischemic retinopathy. Am J Pathol 166(2):637 644. Genetics Home Reference. http:// ghr.nlm.nih.gov/gene/RPE65 Glover P and Mansfield S. 2002. Limits to magnetic resonance microscopy. Reports on Progress in Physics 65(10):1489 1511. Gole G, Browning J, Elts S. 1990. The mouse model of oxygen induced retinopathy: A suitable animal model f or angiogenesis research. Documenta Ophthalmologica 74(3):163 169. Gollapalli D and Rando R. 2004. The specific binding of retinoic acid to RPE65 and approaches to the treatment of macular degeneration. Proc Natl Acad Sci USA 101:10030 10035. Grant MB, Dav is MI, Caballero S, Feoktistov I, Biaggioni I, Belardinelli L. 2001. Proliferation, migration, and ERK activation in human retinal endothelial cells through A2B adenosine receptor stimulation. Invest. Ophthalmol. Vis. Sci. 42(9):2068 2073. Grant MB, May WS Caballero S, Brown GAJ, Guthrie SM, Mames RN, Byrne BJ, Vaught T, Spoerri PE, Peck AB, and others. 2002. Adult hematopoietic stem cells provide functional hemangioblast activity during retinal neovascularization. Nat Med 8(6):607 612. Grant MB, Tarnuzzer RW, Caballero S, Ozeck MJ, Davis MI, Spoerri PE, Feoktistov I, Biaggioni I, Shryock JC, Belardinelli L. 1999. Adenosine receptor activation induces vascular endothelial growth factor in human retinal endothelial cells. Circ Res 85(8):699 706.

PAGE 165

165 Haacke, E.M. Brown, R.W., Thompson, M.R., Venkatesan, R. 1999. Magnetic resonance imaging: Physical principles and sequence design. John Wiley & Sons, New York. Hahn EL. 1950. Spin echoes. Physical Review 80(4):580 594. Hammes HP, Lin J, Bretzel RG, Brownlee M, Breie r G. 1998. Upregulation of the vascular endothelial growth factor/vascular endothelial growth factor receptor system in experimental background diabetic retinopathy of the rat. Diabetes 47(3):401 406. Harris JR, Brown GA, Jorgensen M, Kaushal S, Ellis EA, Grant MB, Scott EW. 2006. Bone marrow derived cells home to and regenerate retinal pigment epithelium after injury. Invest Ophthalmol Vis Sci 47(5):2108 13. Hawes NL, Chang B, Hurd RE, Nusinowitz S, Heckenlively JR, Davisson MT. 2002. A new mouse model of retinal degeneration (rd11). Invest. Ophthalmol. Vis. Sci. 43(12):3669. Hayes, C E, Edelstein, W A, Schenck, J F, Mueller, O M, Eash, M. 1985. An efficient, highly homogeneous radiofrequency coil for hole body NMR imaging at 1.5 T. J Magn Reson 63. Herschm an HR. 2003. Micro PET imaging and small animal models of disease. Current Opinion in Immunology 15(4):378 384. Heyn C, Ronald JA, Mackenzie LT, MacDonald IC, Chambers AF, Rutt BK, Foster PJ. 2006. In vivo magnetic resonance imaging of single cells in mous e brain with optical validation. Magn Reson Med 55(1):23 9. Holcman D and Korenbrot JI. 2004. Longitudinal diffusion in retinal rod and cone outer segment cytoplasm: The consequence of cell structure. Biophysical Journal 86:2566 2582. Horsfield M and Jones D. 2002. Applications of diffusion weighted and diffusion tensor MRI to white matter diseases a review. NMR in Biomedicine 15(7 8):570 577. Inglis BA, L. BE, L. BD, D. WE, Mareci TH. 2001. Visualization of neural tissue water compartments using biexpone ntial diffusion tensor MRI. Magnetic Resonance in Medicine 45(4):580 587. Jian B, Vemuri B, Ozarslan E, Carney P, Mareci T. 2007. A novel tensor distribution model for the diffusion weighted MR signal. NeuroImage 37(1):164 176. Kamble R, Naik A, Venkataram ana N, Rao S. 2011. Diffusion tensor imaging in spinal cord injury. 21:221 224.

PAGE 166

166 Kang YS, Risbud S, Rabolt JF, Stroeve P. 1996. ChemInform abstract: Synthesis and characterization of nanometer Fe2O3 particles. Kaplan JR and Wagner JD. 2006. Type 2 diabetes an introduction to the development and use of animal model s. Inst. for Lab Animal Research (ILAR) Journal 47(3). Kieffer TJ and Francis Habener J. 1999. The glucagon like peptides. Endocrine Reviews 20(6):876 913. Kiessling F. 2008. Noninvasive cell tracking. Handb Exp Pharmacol (185 Pt 2):305 21. Kim JH, Kim JH, Yu YS, Jun H, Kwon HJ, Park KH, Kim K. 2008. Inhibition of choroidal neovascularization by homoisoflavanone, a new angiogenesis inhibitor. Molecular Vision 14:556 561. Korte GE, Bellhorn RW, Burns MS. 1983. Infrastructure of blood retinal barrier permeabi lity in rat phototoxic retinopathy. Investigative Ophthalmology & Visual Science 24:962 971. Lally FJ, Ratcliff H, Bone AJ. 2001. Apoptosis and disease progression in the spontaneously diabetic BB/S rat. Diabetologia 44(3):320 324. Lambert V, Munaut C, Jos t M, Noel A, Werb Z, Foidart J, Rakic J. 2002. Matrix metalloproteinase 9 contributes to choroidal neovascularization. Am J Pathol 161(4):1247 1253. Lauterbur PC. 1973. Image formation by induced local interactions: Examples employing nuclear magnetic reso nance. Nature 242(5394):190 191. Lawrence RF. 2001. Anisotropy in high angular resolution diffusion weighted MRI. Magnetic Resonance in Medicine 45(6):935 939. Le Bihan D and Turner R. 1991. Intravoxel incoherent motion imaging using spin echoes. Magnetic Resonance in Medicine 19(2):221 227. Le Bihan D, Mangin J, Poupon C, Clark CA, Pappata S, Molko N, Chabriat H. 2001. Diffusion tensor imaging: Concepts and applications. Journal of Magnetic Resonance Imaging 13:534 546. Le Bihan D. 1991. Molecular diffusio n nuclear magnetic resonance imaging. Magnetic Resonance Quarterly 7(1):1 30. Le Bihan D. 2003. Looking into the functional architecture of the brain with diffusion MRI. Nature Reviews Neuroscience 4(6):469 480. Leiter EH. 1982. Multiple low dose streptozo tocin induced hyperglycemia and insulitis in C57BL mice: Influence of inbred background, sex, and thymus. Proc. Natl. Acad. Sci. USA 79:630 634.

PAGE 167

167 Lewin M, Carlesso N, Tung C, Tang X, Cory D, Scadden DT, Weissleder R. 2000. Tat peptide derivatized magnetic n anoparticles allow in vivo tracking and recovery of progenitor cells. Nat Biotech 18(4):410 414. Liang ZP Boada FB Constable RT Haacke EM Lauterbur PC and Smith M Reviews of Magnetic Resonance in Me dicine vol. 4, no. 2, pp. 67 185, 1992. Li Calzi S, Kent DL, Chang K, Padgett KR, Afzal A, Chandra SB, Caballero S, English D, Garlington W, Hiscott PS, and others. 2009. Labeling of stem cells with monocrystalline iron oxide for tracking and localizatio n by magnetic resonance imaging. Microvascular Research In Press, Corrected Proof. Lim KO and Helpern JA. 2002. Neuropsychiatric applications of DTI a review. NMR in Biomedicine 15(7 8):587 593. Liney G, 2005, MRI from A to Z: A Definitive Guide for Medi cal Professionals, Cambridge University Press Logan BH,Jr, Sergio C,Jr, Newell CK, Steinmetz RL, Watson D, Segal MS, Harrison JK, Scott EW, Grant MB. 2004. Vitreous levels of vascular endothelial growth factor and stromal derived factor 1 in patients with diabetic retinopathy and cystoid macular edema before and after intraocular injection of triamcinolone. Arch Ophthalmol 122(12):1801 1807. Mansfield P and Grannell PK. 1973. NMR 'diffraction' in solids? Journal of Physics C: Solid State Physics 6(22):L422 L426. Martin PM, Roon P, Van Ells TK, Ganapathy V, Smith SB. 2004. Death of retinal neurons in streptozotocin induced diabetic mice. Invest. Ophthalmol. Vis. Sci. 45(9):3330 3336. McEvoy RC, Andersson J, Sandler S, Hellerstrm C. 1984. Multiple low dose st reptozotocin induced diabetes in the mouse. evidence for stimulation of a cytotoxic cellular immune response against an insulin producing beta cell line. Journal of Clinical Investigation 74(3):715 722. Melhem ER, Mori S, Mukundan G, Kraut MA, Pomper MG, v an Zijl PCM. 2002. Diffusion tensor MR imaging of the brain and white matter tractography. Am. J. Roentgenol. 178(1):3 16. Mitchell, D.G. and Cohen, M.S. 1999. MRI principles, 2nd ed. Saunders, Philadelphia. Miyamoto K, Khosref S, Bursell S, Rohan R, Murat a T. 1999. Prevention of leukostasis and vascular leakage in streptozotocin induced diabetic retinopathy via intercellular adhesion molecule 1 inhibition. Proc. Natl. Acad. Sci. USA 96:10836 10841.

PAGE 168

168 Mizota, A and Adachi Usami, E ( 1997 ). Functional recovery of retina after sodium iodate injection in mice. Vision Res 37 : 1859 1865. Mori S and Zhang J. 2006. Principles of diffusion tensor imaging and its applications to basic neuroscience research. 51(5):527 539. Mori S and Zijl PCMv. 2002. Fiber tracking: Prin ciples and strategies a technical review. NMR in Biomedicine 15(7 8):468 480. Mori S, Itoh R, Zhang J, Kaufmann WE, Zijl PCMv, Solaiyappan M, Yarowsky P. 2001. Diffusion tensor imaging of the developing mouse brain. Magnetic Resonance in Medicine 46:18 2 3. Mori, S. 2007. Introduction to diffusion tensor imaging. Elsevier. Moseley ME, Cohen Y, Kucharczyk J, Mintorovitch J, Asgari HS, Wendland MF, Tsuruda J, Norman D. 1990. Diffusion weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 176(2):439 445. Moseley ME, Kucharczyk J, Asgari HS, Norman D. 1991. Anisotropy in diffusion weighted MRI. Magnetic Resonance in Medicine 19(2):321 326. Mller A, Schott Ohly P, Dohle C, Gleichmann H. 2002. Differential regulation of Th1 type and Th2 type cytokine profiles in pancreatic islets of C57BL/6 and BALB/c mice by multiple low doses of streptozotocin. Immunobiology 205(1):35 50. Mller Sieburg, C.E., Torok Storb, B., Visser, J.W.M., Storb, R. 1992. Hematopoietic stem cells: Animal models and human transplantation, 1st ed. Springer, New York. Nagai N, Oike Y, Izumi Nagai K, Koto T, Satofuka S, Shinoda H, Noda K, Ozawa Y, Inoue M, Tsubota K, and others. 2007. Suppression of choroidal neovascularization by inhibiting angiotensin conve rting enzyme: Minimal role of bradykinin. Invest. Ophthalmol. Vis. Sci. 48(5):2321 2326. Nair G, Kim M, Nagaoka T, Olson DE, Thul PM, Pardue MT, Duong TQ. 2011. Effects of common anesthetics on eye movement and electroretinogram. Documenta Ophthalmologica Vol 122(3):163 76. National Diabetes Statistics 2011. National diabetes statistics 2011, national diabetes information clearinghouse, http://diabetes.niddk. nih.gov/dm/pubs/statistics/index.aspx#fast Natio nal Institute of Health (NIH). 2011. Facts about diabetic retinopathy, obtained from http://diabetes.niddk.nih .gov/dm/pubs/statistics/index.aspx#fast Neil J, Miller J, Mukherjee P, Hppi PS. 2002. Diffusion tensor imaging of normal and injured developing human brain a technical review. NMR in Biomedicine 15(7 8):543 552.

PAGE 169

169 Nyengaard JR, Ido Y, Kilo C, Williamson JR. 2004. Interactions between hyperglycemia and hypoxia: Implications for diabetic retinopathy. Diabetes 53(11):2931 2938. Oca Cossio J, Mao H, Khokhlova N, Kennedy CM, Kennedy JW, Stabler CL, Hao E, Sambanis A, Simpson NE, Constantinidis I. 2004. Magnet ically labeled insulin secreting cells. Biochem Biophys Res Commun 319(2):569 75. Oldendorf, W. and Oldendorf, W.,Jr. 1991. MRI primer. Raven Press, New York. Ozaki H, Seo M, Ozaki K, Yamada H, Yamada E, Okamoto N, Hofmann F, Wood JM, Campochiaro PA. 2000. Blockade of vascular endothelial cell growth factor receptor signaling is sufficient to completely prevent retinal neovascularization. Am. J. Pathol. 156(2):697 707. zarslan E and Basser PJ. 2008. Microscopic anisotropy revealed by NMR double pulsed fiel d gradient experiments with arbitrary timing parameters. The Journal of Chemical Physics 128(15). zarslan E and Mareci T. 2003. Generalized diffusion tensor imaging and analytical relationships between diffusion tensor imaging and high angular resolution diffusion imaging. Magnetic Resonance in Medicine 50(5):955 965. zarslan E, Basser PJ, Shepherd TM, Thelwall PE, Vemuri BC, Blackband SJ. 2006. Observation of anomalous diffusion in excised tissue by characterizing the diffusion time dependence of the MR signal. Journal of Magnetic Resonance 183(2):315 323. zarslan E, Shepherd T, Vemuri B, Blackband S, Mareci T. 2006. Resolution of complex tissue microarchitecture using the diffusion orientation transform (DOT). NeuroImage 31(3):1086 1103. zarslan E, Vem uri B, Mareci T. 2005. Generalized scalar measures for diffusion MRI using trace, variance, and entropy. Magnetic Resonance in Medicine 53(4):866 876. Pierce EA, Avery RL, Foley E, Aiello LP, Smith LEH. 1995. Vascular endothelial growth factor/vascular per meability factor expression in a mouse model of retinal neovascularization. Proc. Natl. Acad. Sci. USA 92:905 909. Pierpaoli C and Basser PJ. 1996. Toward a quantitative assessment of diffusion anisotropy. Magnetic Resonance in Medicine 36(6):893 906. Pitt enger MF, Mackay AM, Beck SC, Jaiswal RK, Douglas R, Mosca JD, Moorman MA, Simonetti DW, Craig S, Marshak DR. 1999. Multilineage potential of adult human mesenchymal stem cells. Science 284(5411):143 147. Poretsky, L. (Ed.), 2002. Principles of DIabetes me llitus. Kluwer Academic Publishers, Boston.

PAGE 170

170 Purcell EM, Torrey HC, Pound RV. 1946. Resonance absorption by nuclear magnetic moments in a solid. Physical Review 69(1 2):37 38. Qaum T, Xu Q, Joussen AM, Clemens MW, Qin W, Miyamoto K, Hassessian H, Wiegand SJ Rudge J, Yancopoulos GD, and others. 2001. VEGF initiated Blood Retinal barrier breakdown in early diabetes. Investigative Ophthalmology & Visual Science 42(10). Quesenberry P and Levitt L. 1979. Hematopoietic stem cells. New England Journal of Medicine 301(14):755 760. Quillen, D.A. and Blodi, B.A. 2002. Clinical retina. AMA Press. Ramalho J, Anders R, Jaissle G, Seeliger M, Huxley C, Seabra M. 2002. Rapid degradation of dominant negative Rab27 proteins in vivo precludes their use in transgenic mouse mod els. BMC Cell Biology 3(1):26. Ramalho JS, Gregory Evans K, Huxley C, Seabra MC. 2004. Mouse genetic corneal disease resulting from transgenic insertional mutagenesis. British Journal of Ophthalmology 88:428 432. Ray D, Mishra M, Ralph S, Read I, Davies R, Brenchley P. 2004. Association of the VEGF gene with proliferative diabetic retinopathy but not proteinuria in diabetes. Diabetes 53(3):861 864. Reeder SB, Atalay MK, McVeigh ER, Zerhouni EA, Forder JR. 1996. Quantitative cardiac perfusion: A noninvasive spin labeling method that exploits coronary vessel geometry. Radiology 200(1):177 184. Rees DA and Alcolado JC. 2005. Animal models of diabetes mellitus. Diabetic Medicine 22(4):359 370. Ricci B. 1990. Oxygen induced retinopathy in the rat model. Documenta Ophthalmologica 74(3):171 177. Robinson GS, Pierce EA, Rook SL, Foley E, Webb R, Smith LEH. 1996. Oligodeoxynucleotides inhibit retinal neovascularization in a murine model of proliferative retinopathy. Proc. Natl. Acad. Sci. USA 93:4851 4856. Rogers WJ, Meyer CH, Kramer CM. 2006. Technology insight: In vivo cell tracking by use of MRI. Nat Clin Pract Cardiovasc Med 3(10):554 562. Rossini AA, Like AA, Chick WL, Appel MC, Cahill GF. 1977. Studies of streptozotocin induced insulitis and diabetes. Proceedings of the National Academy of Sciences of the United States of America 74(6):2485 2489. Ryan S. 1979. The development of an experimental model of subretinal neovascularization in disciform macular degeneration. Trans Am Ophthalmol Soc..

PAGE 171

171 Saudek, C.D., Rubin, R.R., Shump Cynthia S (Eds.), 1997. The johns hopkins guide to diabetes. The Johns Hopkins University Press, Baltimore. Schwesinger C, Yee C, Rohan RM, Joussen AM, Fernandez A, Meyer TN, Poulaki V, Ma JJK, Redmond TM, Liu S, and others. 2001. Intrachoroida l neovascularization in transgenic mice overexpressing vascular endothelial growth factor in the retinal pigment epithelium. Am J Pathol 158(3):1161 1172. Scollan DF, Holmes A, Winslow R, Forder JR. 1998. Histological validation of myocardial microstructur e obtained from diffusion tensor magnetic resonance imaging. Am J Physiol Heart Circ Physiol 275(6):H2308 2318. Sengupta N, Caballero S, Mames RN, Butler JM, Scott EW, Grant MB. 2003. The role of adult bone marrow derived stem cells in choroidal neovascula rization. Investigative Ophthalmology & Visual Science 44(11):4908 4913. Sengupta N, Caballero S, Sullivan SM, Chang L, Afzal A, Li Calzi S, Kielczewski JL, Prabarakan S, Ellis EA, Moldovan L, and others. 2009. Regulation of adult hematopoietic stem cells fate for enhanced tissue specific repair. Mol Ther 17(9):1594 1604. Shapiro EM, Sharer K, Skrtic S, Koretsky AP. 2006. In vivo detection of single cells by MRI. Magn Reson Med 55(2):242 9. Shepherd TM, zarslan E, King MA, Mareci TH, Blackband SJ. 2006. St ructural insights from high resolution diffusion tensor imaging and tractography of the isolated rat hippocampus. NeuroImage 32(4):1499 1509. Shepherd TM, Thelwall PE, Stanisz GJ, Blackband SJ. 2009. Aldehyde fixative solutions alter the water relaxation a nd diffusion properties of nervous tissue. Magnetic Resonance in Medicine 62(1):26 34. Silva MD and Chandra S. 2006. MRI in preclinical drug development. Methods in Molecular Medicine 124(1):299 322. Smith MR and Nichols ST .1990. A comparison of models use d as alternative magnetic resonance image reconstruction methods. Magnetic Resonance Imaging vol. 8, no. 2, pp. 173 183 Song S, Sun S, Ju W, Lin S, Cross AH, Neufeld AH. 2003. Diffusion tensor imaging detects and differentiates axon and myelin degenerati on in mouse optic nerve after retinal ischemia. NeuroImage 20(3):1714 1722. Sotak CH. 2002. The role of diffusion tensor imaging in the evaluation of ischemic brain injury a review. NMR in Biomedicine 15(7 8):561 569. Stadnik T, Luypaert R, Jager T, Oste aux M. 1999. Diffusion imaging: From basic physics to practical imaging. RSNA EJ Radiog 3.

PAGE 172

172 Stejskal EO and Tanner JE. 1965. Spin diffusion measurements: Spin echoes in the presence of a time dependent field gradient. The Journal of Chemical Physics 42(1):2 88 292. Stellmach V, Crawford SE, Zhou W, Bouck N. 2001. Prevention of ischemia induced retinopathy by the natural ocular antiangiogenic agent pigment epithelium derived factor. PNAS 98(5):2593 2597. Stroh A, Faber C, Neuberger T, Lorenz P, Sieland K, Jako b PM, Webb A, Pilgrimm H, Schober R, Pohl EE, and others. 2005. In vivo detection limits of magnetically labeled embryonic stem cells in the rat brain using high field (17.6 T) magnetic resonance imaging. Neuroimage 24(3):635 45. Sundstrom JB, Mao H, Santo ianni R, Villinger F, Little DM, Huynh TT, Mayne AE, Hao E, Ansari AA. 2004. Magnetic resonance imaging of activated proliferating rhesus macaque T cells labeled with superparamagnetic monocrystalline iron oxide nanoparticles. J Acquir Immune Defic Syndr 3 5(1):9 21. Tawfik A, Sanders T, Kahook K, Akeel S, Elmarakby A, Al Shabrawey M. 2009. Suppression of retinal peroxisome proliferator activated receptor in experimental diabetes and oxygen induced retinopathy: Role of NADPH oxidase. Invest. Ophthalmol. Vis. Sci. 50(2):878 884. Tee LBG, Penrose MA, O'Shea JE, Lai CM, Rakoczy EP, Dunlop SA. 2008. VEGF induced choroidal damage in a murine model of retinal neovascularisation. Br J Ophthalmol 92(6):832 838. Thompson M, Wall DM, Hicks RJ, Prince HM. 2005. In vivo tracking for cell therapies. Q J Nucl Med Mol Imaging 49(4):339 48. Titterington, D., Smith, A., Makov, U. 1985. Statistical analysis of finite mixture distributions. John Wiley & Sons. Tombran Tink, J. and Barnstable, C.J. 2007. Retinal degenerations : Bi ology, diagnostics, and therapeutics. Springer Verlag. Torrey HC. 1956. Bloch equations with diffusion terms. Physical Review Online Archive (Prola) 104(3):563 565. Tsuboi S, Fujimoto T, Uchihori Y, Emi K, Lizuko S, Kishido K, Manabe R. 1984. Measurement o f retinal permeability to sodium fluorescein in vitro Investigative Ophthalmology & Visual Science 25. Tsutsumi T, Hagino N, Liang X, Guo S, Kobayashi S. 2008. Effects of oral administration of stephania tetrandra S. moore on neovascularization of retinal and choroidal capillaries of diabetes in rats. Phytotherapy Research 22(5):591 596.

PAGE 173

173 Tuch DS, Reese TG, Wiegell MR, Makris N, Belliveau JW, Wedeen VJ. 2002. High angular resolution diffusion imaging reveals intravoxel white matter fiber heterogeneity. Magn etic Resonance in Medicine 48(4):577 582. Vemuri, B.C., Chen, Y., Rao, M., McGraw, T., Wang, Z., Mareci, T.H. 2001. Fiber tract mapping from diffusion tensor MRI. 81. Wang Z and Gleichmann H. 1998. GLUT2 in pancreatic islets: Crucial target molecule in d iabetes induced with multiple low doses of streptozotocin in mice. Diabetes 47(1):50 56. Wang, Z., Vemuri, B.C., Mareci, T.H. 2004. A constrained variational principle for direct estimation and smoothing of the diffusion tensor field from complex DWI. 23. Watanabe T, Michaelis T, Frahm J. 2001. Mapping of retinal projections in the living rat using high resolution 3D gradient echo MRI with Mn21 induced contrast. Magnetic Resonance in Medicine Weissleder R and Mahmood U. 2001. Molecular imaging. Radiology 219(2):316 33. Weissleder R. 1999. Molecular imaging: Exploring the next frontier. Radiology 212(3):609 14. Westbrook C, Roth CK, Talbot J, 2005, MRI in Practice, 3rd Ed, Blackwell Publishing Williams, G. and Pickup, J.C. (Eds.), 2004. Handbook of diabetes 3rd ed. Blackwell Publishing. Winnard P and Raman V. 2003. Real time non invasive imaging of receptor ligand interactions in vivo J Cell Biochem 90(3):454 63. Wu EX, Tang H, Jensen JH. 2004. Applications of ultrasmall superparamagnetic iron oxide contra st agents in the MR study of animal models. NMR Biomed 17(7):478 83. Zahner D and Malaisse W. 1990. Kinetic behaviour of liver glucokinase in diabetes. I. alteration in streptozotocin diabetic rats. Diabetes Research 14(3):101 8. Zaraiskaya T, Kumbhare D, Noseworthy MD. 2006. Diffusion tensor imaging in evaluation of human skeletal muscle injury. Journal of Magnetic Resonance Imaging 24(2):402 408. Zhang SX, Ma J, Sima J, Chen Y, Hu MS, Ottlecz A, Lambrou GN. 2005. Genetic difference in susceptibility to th e blood retina barrier breakdown in diabetes and oxygen induced retinopathy. Am J Pathol 166(1):313 321.

PAGE 174

174 Zhizhou W, Vemuri BC, Chen Y, Mareci TH. 2004. A constrained variational principle for direct estimation and smoothing of the diffusion tensor field fr om complex DWI. Medical Imaging, IEEE Transactions on 23(8):930 939.

PAGE 175

175 BIOGRAPHICAL SKETCH Saurav Chandra was born in Mumbai, India in 1979. He received his Bachelor of Engineering degree in 2001 from Mumbai University, India. He obtained his degree from the Department of Instrumentation Engineering with a special interest in biomedical instrumentation. This led him to acquire a Master of Science from the Department of Biomedical Engineering at the University of Florida. He specialized in medical imaging a nd image processing. He then worked as a research assistant at the Department of Psychiatry where his research involved functional MRI (fMRI). During this time he successfully designed paradigms for fMRI studies and performed image processing which led to his keen interest in MRI research. Eventually he decided to enroll for a PhD in b iomedical engineering in 2005. He worked under the tutelage of Dr. John Forder at the reputed McKnight Brain Institute (MBI) at UF, one of the premiere institutes in the field of MRI. The MBI houses some of the most powerful magnets in the country. special emphasis on Diffusion Tensor Imaging. His work also included extensive image processing, pulse sequence programming, animal handling techniques etc spanning various medical fields like n euroscience, o phthalmology, p harmacology, e ndocrinology etc. Since most of his work focused on the use of MRI to identify biomarkers for various pathophysiologies, h e developed a special interest in the applications of MRI in the drug discovery process in the pharmaceutical industry Saurav accepted a position as Scientist I within the Global Imaging Group at the pharmaceutical company, Novartis Institutes for Biomedi cal Research (NIBR) at Cambridge, MA. His work will involve preclinical MRI in the drug discovery process.