Citation
Optimizing High Field T1 and Diffusion Tensor Structural Magnetic Resonance Imaging

Material Information

Title:
Optimizing High Field T1 and Diffusion Tensor Structural Magnetic Resonance Imaging
Creator:
PADGETT, KYLE ROBERT ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Cocaine ( jstor )
Field strength ( jstor )
Gray matter ( jstor )
Image contrast ( jstor )
Imaging ( jstor )
Magnetic fields ( jstor )
Magnetism ( jstor )
Signals ( jstor )
Simulations ( jstor )
White matter ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Kyle Robert Padgett. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
5/31/2007
Resource Identifier:
436098739 ( OCLC )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

OPTIMIZING HIGH FIELD T 1 AND DIFFUSION TENSOR STRUCTURAL MAGNETIC RESONANCE IMAGING By KYLE ROBERT PADGETT A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Kyle Robert Padgett

PAGE 3

This document is dedicated to my family (Mom, Dad, Ryan, and Kelly).

PAGE 4

ACKNOWLEDGMENTS This dissertation would not have been possible without the help and input of many people. First, I would like to thank Professor Stephen J. Blackband for being my Ph.D. advisor and for providing the supportive environment where I could follow my research interests. I was very fortunate to work for such a good friend and kind mentor. I would especially like to thank my Ph.D. advisory committee for their patient guidance during my graduate studies: Wesley Bolch Ph.D., David Hintenlang Ph.D., Manuel Arreola Ph.D., and Ioannis Constantinidis Ph.D. Many thanks go out to the postdoctoral associates of the Blackband lab for the guidance and the discipline they provided, Sam Grant Ph.D., Peter Thelwall Ph.D., Tim Shepherd Ph.D., and Jon Bui Ph.D. Thanks to Tim Shepherd Ph.D. for providing guidance in relation to my graduate work and for assistance in formatting this dissertation. Peter Thelwall Ph.D. complemented my graduate studies by being a great resource of biological MR knowledge. Thanks to Jon Bui Ph.D. for nurturing my first steps into MRI research. I would especially like to thank Sam Grant Ph.D. for his continual guidance, friendship, and his intimate involvement in the high field T 1 contrast optimization research. The prenatal cocaine work could not have been completed without the resources and guidance of Project Care. Fonda Eyler Ph.D. and Dr. Marylou Behnke provided the means for the prenatal cocaine exposure study as well as guidance into the potential implications of such important work. Thanks to Ilona Schmalfuss M.D., Tiana Leonard iv

PAGE 5

Ph.D., Ken Crandall, and Tamara Warner Ph.D. for their contribution of their specialties to this difficult study. Cyndi Garvan Ph.D., Vijay Komaragiri, and Wei Hou provided excellent statistical analysis of the data. Special thanks to Ann Welch for eliciting excellent subject cooperation and creating an enjoyable work environment. I also would like to thank Thomas Mareci Ph.D., Evren Ozarslan Ph.D., and Ty Black for DTI support and many hours of computer time. Thanks to AMRIS for access to world class MRI facilities, and special thanks to Carol Tsekouras and Denise Mesa for their help. Above all, I would like to thank my mother and father who have supported me in endless ways throughout my life. They are my true mentors. I also would like to thank Ryan for tolerating his slightly eccentric younger brother. Besides her wealth of patience Kelly has been an invaluable resource of support and encouragement. My family provided me with everything I need to achieve my goals. Financial support for the prenatal cocaine exposure DTI research was provided by the National Institutes of Health to Dr. Behnke and Dr. Eyler (3 R01 DA05854-10S1). Financial support for the high field T 1 contrast optimization was provided by the National Institutes of Health to Dr. Blackband (P41-RR16105). I also have received travel stipends from the International Society of Magnetic Resonance in Medicine. v

PAGE 6

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES .............................................................................................................ix LIST OF FIGURES .............................................................................................................x ABSTRACT .......................................................................................................................xv CHAPTER 1 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: BACKGROUND AND SIGNIFICANCE.............................................1 Introduction...................................................................................................................1 Background and Significance.......................................................................................2 Diffusion MRI.......................................................................................................2 Diffusion in Biological Systems............................................................................4 Diffusion Tensor and Anisotropy..........................................................................7 Neurobiology and Development..........................................................................13 Mechanisms of Cocaine Damage........................................................................16 Subject Population Details..................................................................................19 2 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: METHODS...........................................................................................25 MRI and DTI Acquisition and Processing Methods...................................................25 MRI and DTI Final Protocol.......................................................................................27 White Matter Structure Segmentation........................................................................28 Global Segmentation Decisions..........................................................................29 Initial Segmentation Technique (Hand Drawn Method).....................................31 Initial Segmentation Technique (Core Method)..................................................32 Final Segmentation Technique (Region Shrink Method)....................................33 Thresholding Techniques....................................................................................36 Final Segmentation Technique (New Core Method)...........................................37 3 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: RESULTS.............................................................................................50 vi

PAGE 7

Introduction.................................................................................................................50 Preliminary Results.....................................................................................................50 Region Shrink Results................................................................................................52 Conclusion..................................................................................................................54 4 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: DISCUSSION......................................................................................62 5 HIGH FIELD T 1 CONTRAST OPTIMIZATION: BACKGROUND AND SIGNIFICANCE.........................................................................................................73 Introduction.................................................................................................................73 T 1 and T 2 and Field Strength Dependence..................................................................76 B 1 Homogeneity, Susceptibility, and In vivo MDEFT Imaging.................................78 6 METHODS.................................................................................................................83 Introduction.................................................................................................................83 In Vivo T 1 Measurements and the Construction of Biological Phantoms..................83 Simulation Methods Matched With Results...............................................................86 Solutions to RF Excitation In-Homogeneities............................................................89 7 SR, IR, AND MDEFT GRADIENT ECHO CNR PERFORMANCE.....................107 Introduction...............................................................................................................107 MRI Acquisition.......................................................................................................108 Data Analysis............................................................................................................108 Results.......................................................................................................................109 Discussion and Conclusion.......................................................................................121 8 INTER-MAGNET COMPARISONS.......................................................................139 Introduction...............................................................................................................139 Results.......................................................................................................................141 Discussion and Conclusion.......................................................................................148 9 SE BASED T 1 IMAGING........................................................................................160 Introduction...............................................................................................................160 Results.......................................................................................................................163 Conclusion................................................................................................................166 10 T 1 CONTRAST ENHANCED SE AND GE FAST IMAGING METHODS..........173 Introduction...............................................................................................................173 Results.......................................................................................................................176 Discussion.................................................................................................................180 vii

PAGE 8

11 HIGH FIELD T 1 CONTRAST OPTIMIZATION: CONCLUSION........................188 12 DISSERTATION SUMMARY................................................................................192 APPENDIX: PAPERS AND PRESENTATIONS.........................................................195 A Comparison of the Reliability of Two Techniques for White Matter Segmentation in a Clinical DTI Study.................................................................196 Diffusion Changes in Frontal White Matter in Prenatally Cocaine Exposed Children................................................................................................................197 A Comparison of the Reliability of Two Techniques for White Matter Segmentation in a Clinical DTI Study.................................................................198 Reduction of Fractional Anisotropy in Frontal White Matter in Prenatally Cocaine Exposed Children.................................................................................................198 Native T 1 Contrast Enhancement at 4.7, 11 and 17.6 T for Neuroimaging..............199 Enhancement of Native T 1 Contrast at 4.7, 11 and 17.6 T for Small Animal Imaging................................................................................................................200 Publications and Presentations Outside of Thesis Research.....................................201 Publications:......................................................................................................201 Presentations:.....................................................................................................202 LIST OF REFERENCES.................................................................................................203 BIOGRAPHICAL SKETCH...........................................................................................209 viii

PAGE 9

LIST OF TABLES Table page 2-1: This table lists the Intraclass correlations for both the hand drawn segmentation technique and the region shrink technique...............................................................41 2-2: This table shows the pixel threshold values used in the region shrink technique to segment out all structures, FA ranges from zero to one...........................................47 2-3: The ICC results from the old and new core segmentation method............................48 3-1: This is a table of the fractional anisotropy results in the white matter structures studied using the preliminary data...........................................................................56 3-2: This is a table of the Average diffusion results in the white matter structures studied using the preliminary data...........................................................................57 3-3: This table lists the results of the group comparisons using the region shrink data...........................................................................................................................59 3-4: This table lists the results of the FA group comparisons using the region shrink data...........................................................................................................................60 3-5: This table lists the results of the correlations between hemispheric volume and diffusion measures. Linear relationships between two variables were assessed using the Pearson correlation coefficient.................................................................61 6-1: Results of in vivo T 1 imaging at 4.7T, 11T, and 17.6T. Rows shaded in white are white matter areas, rows shaded in gray are gray matter areas, and rows shaded in blue are CSF areas................................................................................................95 8-1: This table lists the biologically representative T 1 s used for each system for white matter, gray matter and CSF tissue types...............................................................151 ix

PAGE 10

LIST OF FIGURES Figure page 1-1: A standard diffusion weighted spin echo pulse sequence. Motion of spins during the time period become dephased proportional to the amount of motion along the gradient direction applied...................................................................................21 1-2: Theoretical signal attenuation curves for increasing b value. The pink curve represents a diffusion curve reaching a limiting value due to restriction. The blue curve represents a free diffusion signal attenuation curve without any restriction. This figure adapted from Neil et al{Neil 1997 9 /id}...........................22 1-3: An axial fractional anisotropy image of the corpus callosum acquired from an adolescent of 10 years of age...................................................................................23 1-4: A diffusion weighted echo planar pulse sequence....................................................24 2-1: This is an example of the hand drawn technique in the right and left front callosal fibers.........................................................................................................................38 2-2: This figure demonstrates the core method. The top image displays the genu of the corpus callosum. The bottom image is zoomed in on the genu and has a core ROI drawn over the left genu...................................................................................39 2-3: This figure is an example of the encompassing region for the left genu and its resulting region shrink ROI. A pixel threshold of FA 0.45 was used.....................40 2-4: This figure shows the FA distribution of white matter cores. The data is obtained from white matter cores, which should provide little to no volume averaging, from several subjects................................................................................................42 2-5: This figure shows the FA distribution of CSF cores. The data is obtained from CSF cores segmented from maps from several subjects..................................43 2-6: This figure shows the FA distribution of gray matter cores. The data is obtained from gray matter cores from several subjects..........................................................44 2-7: This figure shows the FA distribution of the corpus callosum whole slice. The data is obtained drawing an ROI around the whole slice from several subjects......45 x

PAGE 11

2-8: This is an example of the region shrink method using a high threshold value. The resulting ROI is only of white matter pixels without volume averaging, but this ROI does not represent the shape of the left genu very well.............................46 2-9: This figure demonstrates the new core method. Image A displays the genu of the corpus callosum. Image B is zoomed in on the genu and has a core ROI drawn over the left genu, representing the old core technique. Image C represents the new core technique...................................................................................................49 3-1: The top image is an example of a core ROI inside the right genu of the corpus Callosum, this core measurement in the right genu had significantly reduced FA in the exposed population. The bottom image give examples of hand drawn ROIs in the right and left Callosal fibers in frontal white matter, these frontal white matter areas had significantly reduced FA in the exposed population...........55 3-2: This figure illustrates the four frontal white matter areas segmented using the Region Shrink Method. RFP and LFP are abbreviations for right and left frontal projection fibers. RFC and LFC are abbreviations for right and left frontal callosal fibers............................................................................................................58 5-1: MDEFT pulse sequence with the two variable time delays used to control T 1 contrast.....................................................................................................................81 5-2: IR pulse sequence with the inversion time which is used to control T 1 contrast......82 6-1: MSME_VTR pulse sequence with TR periods of 4000, 700, and 250ms................92 6-2: This is an example of segmented gray matter and CSF areas. This mouse image was collected on the 17.6T magnet..........................................................................93 6-3: This is an example of segmented white matter areas. This mouse image was collected on the 11T magnet....................................................................................94 6-4: This graph displays the trend of T 1 in the structures of interest as field strength increases...................................................................................................................96 6-5: T 1 phantom with wide ranging T 1 values..................................................................97 6-6: A MSME_VTR image of the 4.7T biologically representative T 1 phantom. The differences in T 1 between vials are apparent............................................................98 6-7: A MSME_VTR image of the 11T biologically representative T 1 phantom. The differences in T 1 between vials are apparent............................................................99 6-8: A MSME_VTR image of the 17.6T biologically representative T 1 phantom. The differences in T 1 between vials are apparent..........................................................100 xi

PAGE 12

6-9: This figure gives the comparison of 4.7T GE SR experimental results versus simulations.............................................................................................................101 6-10: This figure gives the comparison of 4.7T GE IR experimental results versus simulations.............................................................................................................102 6-11: This figure gives the comparison of 4.7T GE MDEFT experimental results versus simulations..................................................................................................103 6-12: This figure qualitatively illustrates the RF excitation profile in the biologically representative 17.6T phantom................................................................................104 6-13: This figure demonstrates the signal from each ROI of the 17.6T biologically representative phantom versus TR.........................................................................105 6-14: This figure shows the white versus gray matter CNR comparison. The blue curve is the uncorrected data, the yellow curve is the simulated data, and the purple curve is the RF intensity corrected data......................................................106 7-1: This figure illustrates the 4.7T biologically representative T 1 phantom..................125 7-2: CNR performance of white versus gray matter representative T 1 values at 4.7T. .126 7-3: CNR performance of white matter versus CSF representative T 1 values at 4.7T. .127 7-4: CNR performance of gray matter versus contrast enhanced tissue representative T 1 values at 4.7T.....................................................................................................128 7-5: This figure illustrates the contrast enhancement of the three pulse sequences at 4.7T........................................................................................................................129 7-6: This figure illustrates the 11T biologically representative T 1 phantom...................130 7-7: CNR performance of white versus gray matter representative T 1 values at 11T....131 7-8: CNR performance of white matter versus CSF representative T 1 values at 11T....132 7-9: CNR performance of white matter representative T 1 value versus a long T 1 value at 11T......................................................................................................................133 7-10: Illustrates contrast enhancement from both the SR_GE preparation and the MDEFT_GE preparation........................................................................................134 7-11: This figure illustrates the 17.6T biologically representative T 1 phantom..............135 7-12: CNR performance of white versus gray matter representative T 1 values at 17.6T......................................................................................................................136 xii

PAGE 13

7-13: CNR performance of white matter versus CSF representative T 1 values at 17.6T......................................................................................................................137 7-14: CNR performance of white matter versus contrast enhanced tissue representative T 1 values at 17.6T...........................................................................138 8-1: This figure demonstrates the RC performance of white versus gray matter for SR_GE at all field strengths. It is apparent that for both the 11T and 17.6T the RF excitation profile dominated RC performance.................................................152 8-2: This figure demonstrates the RC performance of white matter versus CSF for SR_GE at all field strengths. Again both the 11T and 17.6T demonstrate that the RF excitation profile dominated their RC performance...................................153 8-3: This figure demonstrates the RC performance of white matter versus gray matter for IR_GE at all field strengths. IR_GE was not as affected by the non-uniform RF profile as SR_GE only the 11T curve demonstrates a significant effect..........154 8-4: This figure demonstrates the RC performance of white matter versus CSF for IR_GE at all field strengths....................................................................................155 8-5: This figure demonstrates the qualitative CNR curve shape of IR_GE for all three magnets using the white matter versus CSF comparison. All curves are normalized to 1 because CNR cannot be directly compared across systems.........156 8-6: This figure demonstrates the RC performance of white matter versus gray matter for MDEFT_GE at all field strengths.....................................................................157 8-7: This figure demonstrates the RC performance of white matter versus CSF for MDEFT_GE at all field strengths..........................................................................158 8-8: This figure demonstrates the qualitative CNR curve shape of MDEFT_GE for all three magnets using the white matter versus CSF comparison. All curves are normalized to 1 because CNR cannot be directly compared across systems.........159 9-1: This is a MDEFT_SE scan on the 4.7T system using the biologically representative T 1 phantom. A time of 0.5s was used, TR 50ms, TE 6.5ms, 1 average, FOV 3x3cm, 2mm slice thickness, 128 2 , SW 50kHz..............................168 9-2: This is an example of the advantages of SE imaging over GE imaging. These images were collected on the 11T..........................................................................169 9-3: This figure demonstrates the contrast performance of both saturation recovery SE and GE pulse sequences at 4.7 T............................................................................170 9-4: This figure demonstrates the contrast performance of both inversion recovery SE and GE pulse sequences at 4.7 T............................................................................171 xiii

PAGE 14

9-5: This figure demonstrates the contrast performance of both MDEFT SE and GE pulse sequences at 4.7 T.........................................................................................172 10-1: This is a MDEFT_SE scan with the phantom used for the curves. A time of 1s was used, TR 50ms, TE 10ms, 1 average, FOV 1.5x1.5cm, 0.5mm slice thickness, matrix of 128x128, and SW 20kHz.......................................................183 10-2: This figure demonstrates the contrast performance of both segmented and non-segmented IR_GE imaging at 17.6T......................................................................184 10-3: This figure demonstrates the contrast performance of both segmented and non-segmented MDEFT_GE imaging at 17.6T............................................................185 10-4: This figure demonstrates the contrast performance of both RARE and non-segmented IR_SE imaging at 17.6T.......................................................................186 10-5: This figure demonstrates the contrast performance of both segmented and non-segmented MDEFT_GE imaging at 17.6T............................................................187 xiv

PAGE 15

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy OPTIMIZING HIGH FIELD T 1 AND DIFFUSION TENSOR STRUCTURAL MAGNETIC RESONANCE IMAGING By Kyle Robert Padgett May 2005 Chair: Wesley Bolch Major Department: Nuclear and Radiological Engineering Brain structure has been a major focus for MRI since its inception. Early imaging was conducted at much lower fields than are now available, and early imaging was heavily reliant on T 1 contrast for its structural images. These structural images have been used for the diagnosis of several disease states, and with higher field strengths now being applied the sensitivity of these diagnostic tools is increasing. With the added signal to noise ratio (SNR) from high field strengths, the superior performance of today’s gradient systems, and new techniques the structural imaging modality diffusion tensor imaging (DTI) has been developed. DTI provides 3D water diffusion maps of the brain, and with this information the structure of the brain may be probed with added detail. The combination of these two structural MRI techniques will increase the impact of subsequent findings. The research presented here includes two main topics, the first being an application of DTI to a group of in vivo children who were prenatal exposed to cocaine. The second is the development of T 1 contrast enhancement at high field xv

PAGE 16

strengths with the application of the pulse sequence modified driven equilibrium Fourier transform (MDEFT). xvi

PAGE 17

CHAPTER 1 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: BACKGROUND AND SIGNIFICANCE Introduction Diffusion tensor imaging (DTI) has become an important tool in analyzing brain structure. Although DTI has shown its usefulness in various animal models its human applications have been very limited. This work applied DTI to a group of 28 prenatally cocaine exposed children and a control group of 25 socioeconomic and age matched controls both groups containing children between 10 and 11 years of age. Neurobehavioral studies done by collaborators here at UF have begun to identify subtle clinical differences between the exposed and control groups. These clinical differences aroused interest in whether these differences can be linked to underlying structural brain changes as determined by magnetic resonance imaging (MRI) studies. This body of work contains chapters covering the background and significance of the problem and the technologies used, the MRI/DTI methods employed as well as the white matter segmentation techniques, the specific structural alterations elucidated in this study, and a chapter giving the potential implications of this work and future directions. DTI and the measures that it generates may give information of the microstructure on neural tissue. Fractional anisotropy (FA) and average diffusion coefficient () have shown changes in the microstructure in animal models(1). Animal studies have also shown microstructural alterations resulting from prenatal cocaine exposure(2-5). Some of the various alterations in neural tissue that may be detected by DTI are axonal 1

PAGE 18

2 thickness, axonal density, myelin thickness, axonal death, fiber connectivity, and other less investigated issues. Both FA and have the potential to give a scale to axonal thickness, axonal density, and myelin thickness, which are of interest in brain development and any condition that may affect brain development. Considering these issues DTI was employed in order to determine alterations to microstructure in white matter areas of the brain due to the chemical insult from prenatal cocaine insult. DTI changes detected in this study are indicative of microstructural alterations to frontal white matter areas suggesting specific areas of the brain at risk from prenatal cocaine exposure potentially linked to developmental outcomes. Background and Significance Presented in this chapter is a review on the topics essential to understanding potential diffusion changes from cocaine insult and details about the subject population. This chapter contains reviews on general diffusion, diffusion tensor and anisotropy that cover the basics of water diffusion in neural tissue and how diffusion is measured through MRI. Reviews on neurobiology and development as well as the mechanisms of cocaine damage are also covered giving the basics of neurobiology and normal neurodevelopment along with the perturbations expected from prenatal cocaine exposure and how these perturbations may influence the dynamics of diffusion. Lastly the details of the subject population and how it was formed are presented. Diffusion MRI The image contrast arising from diffusion-weighted imaging (DWI) is based on the measurement of Brownian motion of water molecules. This MRI technique gives a measure of diffusion in each pixel of an image. While qualitative contrast changes are the most widely used in vivo human application of DWI, quantitative methods are being

PAGE 19

3 developed and applied in many disease states and animal models. This rapidly evolving modality of MRI has vast potential in probing physiology with respect to diffusion effects. Diffusion weighting may be generated with an alteration to a standard spin-echo pulse sequence (Figure 1-1). A spin-echo pulse sequence begins with a 90 tip of magnetization into the transverse plane. After a user defined amount of time a 180 rotation of the magnetization occurs at (TE/2). This 180 pulse flips the spins to the opposite side of the transverse plane. After another TE/2 has elapsed the spins become coherent with one another and an echo occurs. This method greatly reduces effects from magnetic field in-homogeneities. Now if a gradient is applied between the 90 and the 180 pulses and an identical gradient is applied after the 180 pulse, the pulse sequence is now considered motion sensitive. In standard diffusion weighting only one gradient direction is applied, and therefore only motion along the direction of the applied gradient is weighted. If there is not motion along the direction of the diffusion gradient in the time between these two gradients then there is no net dephasing of the spins, but if motion has occurred then the amount of dephasing is proportional to the amount of motion(6). This dephasing of the spins due to motion is the image contrast mechanism essential to DWI. This signal attenuation caused by this dephasing may be described by equation 1-1 below where S is signal with the application of the diffusion gradients, and S o is the signal of the same sequence without the diffusion gradients while b defines the gradient characteristics. Equation 1-2 below gives the solution for b where is the

PAGE 20

4 gyromagnetic ratio, G is the gradient amplitude, is the duration of the gradient, and is the separation of the two gradient pulses. bDDGoeeSS)3/(222 [1-1] )3(222 Gb [1-2] Diffusion in Biological Systems Tissue water unlike free water resides in a confined environment demonstrated by the decreased diffusion coefficient compared with free water. Free diffusion no longer applies to the heterogeneous microstructure of tissue because of the presence of internal structure and barriers. Due to the heterogeneous microstructure of tissue the diffusion of water becomes an anisotropic process. Tissue also contains different sub-compartments, namely intra and extra cellular compartments, and exchange between these compartments influences diffusion properties. Due to the complexity of tissue, diffusion in biological systems is measured is no longer described as a diffusion coefficient. Instead apparent diffusion coefficient (ADC) is measured acknowledging that simple free diffusion no longer applies. Biological systems are inherently heterogeneous and contain several different sub-compartments. The main compartments of interest are the intra and extra cellular compartments. For an accurate model of biological diffusion one must account for the exchange of spins between these compartments(7). The observed signal attenuation in the diffusion experiment is dependant upon the rate of exchange and the diffusion measurement time ( D ) of the experiment. In a system with slow exchange, spins stay within their compartments during the measurement time. This situation is reflected by a

PAGE 21

5 multi-exponential signal attenuation curve in the diffusion MRI experiment. In a system with fast exchange, spins move back and forth between compartments several times during the measurement period. This causes a complete redistribution of spins resulting in mono-exponential signal attenuation. In between these two cases lies intermediate exchange, which is the most likely scenario for biological systems. Exchange times for certain tissues have been reported from several sources but they are inconsistent. For example, Pfeuffer et al. report the approximate exchange time for nervous tissue is (8). The exchange time for erythrocytes is approximately as reported by Thelwall et al(9). These numbers are relatively similar, but the more porous erythrocytes would intuitively allow for more exchange. ms4.55.26 ms03.01.22 Natural barriers in tissue such as membranes restrict the diffusion of water. The diffusion time determines the sensitivity of the experiment to the restriction of diffusion. When short diffusion times are used the majority of spins do not reach the barriers of the system, and diffusion will appear to be relatively free. As diffusion time increases the amount of spins that interact with the barriers will also increase, and diffusion will appear to be restricted. The morphology and permeability of membranes and other natural barriers have a significant effect upon the amount of restriction. See Figure 1-2 below for a theoretical signal attenuation curve for a restricted and non-restricted system(10). In the presence of impermeable obstacles, extracellular spins exhibit unrestricted diffusion if the spins are able to diffuse around the impermeable barriers. The mean path between two points is thereby lengthened and the measured ADC is lowered compared with an identical measurement in a medium without barriers. This aspect of hindered diffusion is described by the concept of tortuosity. Tortuosity describes how barriers

PAGE 22

6 within a biological system reduce the diffusion coefficient compared with the coefficient without such barriers(11). Tortuosity ( is defined by a dimensionless number that incorporates the diffusion coefficient in free water, D, and the apparent diffusion coefficient of the media of interest, D . has a potential range from one, where D matches D, to infinity, where the extracellular space restricts all movement; equation 1-3 below gives the solution for . Tortuosity is affected by many factors, and it is difficult to separate the contributions of these factors either qualitatively or quantitatively. Some of the more dominant factors affecting tortuosity are the size, shape, and density of the cellular obstacles(12). In biological systems a change in the extracellular volume fraction is the largest effect on changes in tortuosity. A decrease in the extracellular space will increase the effects of cellular obstacles. This decrease can originate from several different disease states including stroke and traumatic brain injury. )/(DD [1-3] Some have simplified tortuosity into two contributing categories of factors, the geometric factor and the viscous factor(13). The geometric aspect of tortuosity is the interaction with cellular obstacles and interaction with other macromolecules. The viscous factor arises from the inelastic interactions of the diffusing molecules with the cellular obstacles or macromolecules. This viscous factor is further increased with the increase in size of the diffusing particle. The viscosity component however is not a large contribution to the tortuosity factor of diffusing water.

PAGE 23

7 Diffusion Tensor and Anisotropy DTI can be used to probe diffusion orientations in tissue. In systems where water diffusion is isotropic a scalar diffusion coefficient contains the same information that a tensor does. A tensor is necessary in a situation where the diffusion properties of tissue are not isotropic. White matter in the brain and nervous system as well as many other tissues exhibit anisotropic diffusion properties. Diffusion in most cases is not an isotropic process; therefore information must be obtained in three-dimensional space in order to fully characterize the diffusion process. With isotropic diffusion MRI only one diffusion gradient direction is required for encoding. While this is sufficient in some clinical cases to achieve sufficient contrast, it does not represent three-dimensional diffusion. Anisotropy can easily be observed by applying these diffusion sensitive gradient pulses in many directions. The manners in which these different directions are applied vary greatly from experiment to experiment depending on the application and information of interest. White matter in the brain and nervous system is inherently anisotropic. The myelination of the long axons gives a very preferential diffusion direction along the length of the axon. When the axons of the neurons organize in one direction it forms a fiber bundle. These fiber bundles are where diffusion anisotropy arises from in white matter(6). This knowledge has been exploited in order to track fiber bundles in the white matter, under the assumption that the fastest diffusion direction is in the same direction as the orientation of the fibers. The developing brain exhibits larger ADC values than developed brain and as the brain matures this ADC begins to decrease(6). Not only does the ADC decrease, but diffusion becomes more anisotropic(14). It is generally accepted that the decrease in

PAGE 24

8 ADC is due to the decrease in the extracellular space from the increase of myelin and also the increase size of neurons and glial cells(14). Once the myelin sheath has been created the anisotropy of the brain increases significantly. It is generally accepted that the main cause of this is the preferential diffusion along the axon inside and outside of the myelin sheath. Perpendicular to the myelin membrane water diffusion is restricted, while parallel to the membrane it is not. While this is a major cause of diffusion anisotropy there are other contributing factors. One proposed factor is that the oligodendrocytes that surround the fascicles of axons possibly restrict water diffusion perpendicular to axons. Another possible factor contributing to anisotropic diffusion is that of macromolecules that are present between axons(14). With isotropic diffusion the signal attenuation resulting from the diffusion sensitive gradient is explained by the scalar value of A, shown in equation 1-4. D is the diffusion coefficient and b, shown in equation 1-5, is a factor that describes the gradient’s properties; for example amplitude, length of duration, and separation of the two gradient lobes. D being expressed as a scalar coefficient using one diffusion encoding direction is no longer sufficient when trying to describe anisotropic diffusion. Instead a matrix of diffusion values in several directions is needed in order to visualize the three dimensional diffusion properties, equation 1-6 below(15). A full tensor has seven unknowns; therefore echoes encoded with a minimum of six directions is required with one additional echo with no diffusion weighting are collected to solve for these unknowns. Most commonly the reference frame in which the diffusion gradients are applied coincides with the orientation of the magnet. )exp(bDA [1-4]

PAGE 25

9 )3/(222Gb [1-5] zzzyzxyzyyyxxzxyxxDDDDDDDDDD [1-6] Once the tensor has been acquired several different diffusion characteristics may be calculated. This diffusion matrix may be used to determine mean diffusivity or the degree of anisotropy in each voxel. The diffusion gradients used are only sensitive to the direction applied and cannot discriminate between positive and negative displacement. Therefore the D xy term is equal to the D yx ;because of this property the tensor is symmetric. Once the tensor has been diagonalized into a reference frame ),,(zyx that has the principal directions of diffusion as its axes the off-diagonal terms become zero and the signal attenuation becomes equation 1-7 below. )exp(zzzzyyyyxxxxDbDbDbA [1-7] Diffusion measurements are made in the lab coordinate system, i.e., being referenced to the orientation of the magnet, which does not coincide with the coordinate system associated with principle diffusion. Because of this property one must include coupling of the off-diagonal elements of the b matrix to the off-diagonal diffusion terms. For example when the lab coordinate system does not match the axes of principle diffusion attenuation along the x-axis of the lab frame also has contributions from diffusion along the y and z-axes. If the lab frame matches the principal diffusion orientations then there would be no coupling from one axis to another. The optimal sampling of the ADC involves proper b value selection and these optimal b values vary depending upon experimental conditions. In the case of scalar

PAGE 26

10 diffusion the collection of two b values and many averages has more accuracy than the collection of several b values. It is desirable to have accurate eigenvalues for anisotropy indices. Anisotropy is a directional invariant measure; therefore error in the eigenvector calculations has less relevance compared to minimizing the error in the eigenvalues. The eigenvalues may be more accurately obtained by following some guidelines for optimizing the measurement of ADC. The diffusion to noise ratio, DNR, is shown below in equation 1-8 and the maximization of DNR minimizes the error of D(16). The components and D of DNR are also defined in equations 1-9 and 1-10. In the equations below, D is the standard deviation of D, SNR 1 is the signal to noise ratio at the first b value, b 1 is the first b value, b 2 is the second b value, n 1 is the number of measurements at the first b value, and n 2 is the number of measurements at the second b value. DNR depends upon SNR 1 and k D , and in order to maximize DNR these parameters must be maximized. The optimization of k D will occur when is approximately 1.1. One can also optimize n 1 and n 2 for a given scan time. Collecting n 2 with a greater number of averages than n 1 may improve the accuracy of calculations because of lower SNR at higher b values(16). DDSNRDDNR1 [1-8] 212exp1nnD [1-9] bDbbD12 [1-10] One evaluation technique of the diffusion tensor is the mean diffusivity, , map. This method gives an averaged diffusion value per pixel. This technique gives similar

PAGE 27

11 results to that of just simple diffusion weighting, but it avoids artifacts of arbitrary directional selection by the collection of the full tensor. The full tensor is needed because of the dependence of the off-diagonal terms upon the diagonal ones when the reference frame does not match the principle diffusion directions, or if the system is not isotropic. In order to determine the the tensor must be diagonalized solving for the eigenvalues and eigenvectors. Summing the three eigenvalues and dividing by three gives , shown in equation 1-11 below(17). 3321 D [1-11] Another useful evaluation technique is that of the diffusion anisotropy maps. Essentially an algorithm is used to calculate the amount of anisotropy present per pixel and then this is used as the contrast mechanism. Some of the original methods of calculating diffusion anisotropy were plagued by the same problems as DWI. These methods did not use the full tensor for their calculations therefore they were directionally dependant. One of the results of this technique was that anisotropy was generally underestimated. In order to obtain the directionally invariant anisotropy indices the eigenvalues and the eigenvectors of the tensor must be used. Fractional anisotropy, FA, is a very useful index used to display the degree of anisotropy per pixel. This method utilizes the diagonalization of the full tensor in order to obtain the eigenvalues of the three principle diffusion axes. These eigenvalues are then used to calculate FA. The range of values for FA is from zero to one with zero being perfectly isotropic diffusion, and one being correlated to infinite anisotropy. FA has been used to display the effects of demyelination or increased isotropic diffusion in white matter following traumatic injury(18). Equation 1-12 below is the mathematical

PAGE 28

12 representation of FA where 1 , 2 , and 3 are the eigenvalues and <> is described by equation 1-13. Also below is an example of a FA map in the human brain, figure 1-3. )(2])()()[(3232221232221FA [1-12] 3)(321 [1-13] Relative anisotropy (RA) is a similar index used to show the degree of anisotropy. It is calculated by equation 1-14 in a very similar way to FA but the range of RA is from zero to 2 . Zero is again associated to perfect isotropic diffusion, and 2 being infinite anisotropy. RA represents the ratio of the anisotropic component of D to the isotropic component(19). Another index that is utilized on a less frequent basis is volume ratio (VR), defined in equation 1-15 below. VR is a ratio of the volume of an ellipsoid with the eigenvalues as the definition of the axes of the ellipsoid to the volume of a sphere with the radius . The range of VR is again from one to zero but with one representing isotropic diffusion and zero representing infinite anisotropy(19). There are several different methods of collecting a diffusion tensor. These different methods have their advantages and disadvantages, but are well suited to a range of applications. In an in vivo setting it is generally advantageous to make experiment time a minimum. In these cases something similar to a diffusion weighted echo planar sequence may be utilized. In an ex vivo setting a more traditional spin echo diffusion weighted experiment may be more advantageous. It requires a longer acquisition time, but there are less artifacts associated with this sequence. It also allows for higher matrix sizes which in turn provide superior resolution. EPI techniques may be employed in the in vivo

PAGE 29

13 setting in order to decrease scan times, but this method possess more difficulties than the traditional spin echo diffusion weighted sequence. 3)()()(232221RA [1-14] 3321 VR [1-15] The traditional spin echo diffusion weighted sequence, which was explained earlier, is just a variation upon the common spin echo sequence. This experiment is then repeated several times with different gradient directions until the entire tensor has been collected. This method of collecting the diffusion tensor is very time intensive but is not plagued with the artifacts and corrections that the other methods possess. Earlier in figure 1-1 an example of a diffusion weighted spin echo sequence was shown and below a diffusion weighted echo planar sequence is shown, figure 1-4. Neurobiology and Development Other than some contributions from the endocrine system, the nervous system is the control and integrating system of the body. The neuron is the principle building block of both the central and peripheral nervous systems. Alterations in the components or the function of neurons are the cause of many disease and injury states. The myelin sheath around the axon of a neuron plays a major role in the functionality of a neuron. Damage or elimination of the axon or myelin can severely alter the function of neurons and if this damage is extensive enough then major alterations in overall central nervous system (CNS) performance ensue. Although the CNS is a complex system it only contains two types of cells. The first cell type is the neurons themselves and the second being the neuroglia. There are

PAGE 30

14 several different types of neuroglia but they all serve the same purpose, which is to support, nurture, and protect the neurons. Neuroglia cells are typically smaller than neurons and they inhabit about half the CNS volume. There are six types of neuroglia only two of these are associated with myelination; oligodendrocytes in the CNS and neurolemmocytes in the peripheral nervous system, PNS. Neurons typically have three parts; the cell body, dendrites, and axon. The later two parts are considered processes of the neuron. The cell body contains the typical organelles and cytoplasm. A majority of neurons also contain cytoplasmic inclusions such as lipsfuscin pigment, which is most likely the end-product of lysosomal activity. Another unique feature is that the cytoplasm of neurons contains a chromatophilic substance called Nissl bodies, which is an orderly arrangement of rough endoplasmic reticulum. This is where protein synthesis occurs. These synthesized proteins are used the produce cytoskeletal structures for the growth of processes of neurons and the regeneration of damaged peripheral nerve axons(20). Dendrites are the input portion of the neuron. Dendrites are often a short and highly branched structure. Usually the dendrites form a tree shaped structure that emerges from the cell body. Most often dendrites are not myelinated. Dendrites also contain a chromatophilic substance called mitochondria, as well as other organelles. The third portion of a neuron is the axon, which is a long, thin, cylindrical projection that may be myelinated. Its function is to conduct nerve impulses toward another neuron, muscle fiber, or gland cell. The axon hillock is where it joins the cell body. The first portion of the axon is called the initial segment. Axons, like dendrites and the cell body, contain mitochondria, neurofibrils, and microtubules but there is no endoplasmic reticulum.

PAGE 31

15 This means that the axon cannot perform protein synthesis. An axon’s cytoplasm is called axoplasm, and is surrounded by a plasma membrane called the axolemma. The axon ends by dividing into many fine processes called axon terminals(20). At the ends of the axon terminals there is a ball shaped structure called synaptic end bulbs. This is where the axon will interact with its adjoining neuron, muscle, or gland. It interacts by releasing a neurotransmitter that is stored in synaptic vesicles found in the synaptic end bulbs. White matter axons are surrounded by a myelin sheath. A myelin sheath is a many layered lipid and protein covering which is produced by either neurolemmocyte or a oligodendrocyte. The myelin sheath electrically insulates the axon which increases the speed on conduction of nerve impulses(14). The oligodendrocyte is the microglial cell in the CNS which myelinates the axons. Unlike the neurolemmocytes in the PNS the oligodendrocytes in the CNS myelinate portions of several different axons. A single oligodendrovyte myelinates approximately portions of 15 different axons and can myelinate up to 50. The axons in the CNS have a similar myelin sheath to that of axons in the PNS, but the oligodendrocytes do not form a neurolemma. The absence of this neurolemma is thought to be a contributing factor to the lack of regeneration in the CNS axons. The myelin sheath is not continuous but is broken up into segments. Between these segments the bare axon is exposed to the interstitial space. These segments are called neurofibral nodes or nodes of Ranvier. These nodes are present in both PNS and CNS axons, but in the CNS axons they are fewer in number(20). Myelin has a large effect upon the conduction of nerve impulses between the neurofibral nodes. When the axon membrane is excited, the generated electrical impulse

PAGE 32

16 flows preferably down the length of the axon and not through the high-resistance of the myelin sheath. This nerve impulse flows down the axon and depolarizes the axonal membrane at the next node. Also the low capacitance of the myelin sheath allows for the depolarization of the remaining membrane between the nodes with little energy and with an increased speed(14). Normal myelination is a complex but orderly process. Myelination begins in a human during the fifth fetal month and it does not end till well into adulthood. While myelination does occur before birth the bulk of it occurs after birth and it continues to myelinate at a rapid rate for the first eighteen months after birth(21). Neurotransmitters play an important role in normal brain development by modulating growth of developing neurons. Perturbing the initial stages of cell proliferation and neurogenesis may produce major structural anomalies in the brain(2). Dopamine and serotonin have both been shown to greatly modify axonal and dendritic outgrowth. Several studies have shown that alterations in the neurotransmitter system during development alter structure. There is evidence that dopamine induces retraction of neurites in retinal neurons(3). It also has been shown that D1 receptor activation decreases axonal and dendritic outgrowth in vitro, while D2 activation stimulates axonal extension of cortical neurons(3). Therefore it is plausible that neurons in which D1 receptors exhibit reduced coupling would show uncontrolled growth of neurites(3). Mechanisms of Cocaine Damage Cocaine is a lipophilic substance, which allows it to easily cross the fetal blood-brain barrier of the fetus and subsequently accumulate in the CNS. This results in increases in extracellular levels of norepinephrine, dopamine, and serotonin interfering with fetal neurotransmitter uptake mechanisms(3,5). Cocaine binds to the presynaptic membrane transporters of the monoamine neurotransmitters and blocks their reuptake

PAGE 33

17 into the presynaptic terminal(2,22). Therefore cocaine initiates specific physiological changes, such as motor hyperactivity and reward behavior(2). There have been several studies that show specific structural anomalies resulting from prenatal cocaine exposure. The alterations to the neurotransmitter system in the developing brain bring about many permanent alterations to brain structure. Studies have shown structural anomalies in animals prenatally cocaine exposed which include a reduction in number of cortical cells, inappropriate positioning of cortical neurons, altered glial morphology, reduction in length of neurites, and apoptotic neural cell loss(22,23). Alterations to the function of the monoaminergic neurotransmitters, which affect synaptogenesis, neural growth, and cell proliferation, are thought to bring about these changes(22). Based on the regions from which these neurotransmitters originate; neurodevelopment functions such as reactivity, arousal modulation, attention regulation and memory, and fine motor control might be particularly vulnerable(24,25). One study shows that dendrites in regions of excess neurotransmitters course abnormally through the cortex, on and out of the plane of section in an irregular fashion rather than in the usual straight course(2). This unusual trajectory was reported to be due to less controlled growth forcing the neurites to extend in an undulating fashion to fit within the limits of the tissue. It was also shown that D1 receptor activation in developing cortical neurons decreases outgrowth of neurites. Therefore, neurons in which D1 receptors exhibit reduced coupling will show less controlled growth of neurites(2). A similar study shows that injections of cocaine caused dendrites to grow longer and wavier compared with the straight, single-plane path of a control saline injection(5). Clear differences between cocaine and saline groups existed because

PAGE 34

18 cocaine-exposed neurons exhibited longer and more complex fiber networks than did the saline neurons. Another study demonstrated a decreased bundling of pyramidal neuron dendrites(4). The exposed mice in this study failed to develop the normal barrel structures in the somatosensory cortex. This was attributed to an increase in serotonin levels during the sensitive period for barrel formation. These anatomical abnormalities are accompanied by a sustained reduction in coupling of the D1 dopamine receptor. It was also demonstrated that cocaine-induced changes in neuronal growth are initiated around mid-gestation(4). Prenatal exposure to cocaine causes specific and robust alterations to dopamine-rich areas of the brain, which put specific regions at risk. Early in fetal development monoamine pathways innervate forebrain regions of the brain. Early manipulation of monoamines due to cocaine exposure alters forebrain development, which causes long-term structural anomalies that persist into adulthood(3). Brain regions where all three monoamine neurotransmitters are expressed during development by incoming axon systems are clearly susceptible to the effects of cocaine, such as the cerebral cortex. The anterior cingulated and prefrontal cortices demonstrated pronounced structural alterations because both are limbic targets of dense dopamine input(2,4). Another potential mechanism of damage from prenatal cocaine insult is from hypoxia. Various animal studies have shown vasoconstriction and hypoxia of the fetus after cocaine administration to the mother(26-29). Some of the areas at risk from hypoxia include the hippocampus, white matter development involving interhemispheric communication, and the efficiency of tract connections between the subcortex and cortex. Information processing or problem solving are probable effects from hypoxia(30-34).

PAGE 35

19 With these studies in mind prenatal exposure to cocaine may have an effect upon myelination, axonal thickness, neuron density, or fiber connections during brain development. These neurobiological changes are hypothesized to affect the water diffusion properties in neural tissue, which in turn may be detected by DTI. Jones et al. report that dendrites in regions of excess neurotransmitters course abnormally through the cortex and these wavy dendritic profiles reflect a compensatory mechanism to counteract a more rapid rate of increase in the length of apical dendrites(5). The increase in waviness of dendrites and neurites may induce a reduction of FA in the developing brain. Stanwood et al. report that cocaine exposed mice fail to develop the normal barrel structures that are related to vibrissae representation in the somatosensory cortex due to increases in serotonin levels during the sensitive period for barrel formation(4). If the bundling of fibers is decreased or packed more loosely a decrease in the restriction of water diffusion would be probable resulting in an increase in . Subject Population Details This prospective longitudinal study was designed to evaluate women and their children from a mostly rural, understudied population. Most women received prenatal care, but had minimal access to drug treatment. However, they represented a wide range of amount of (mostly “crack”) cocaine use. To more accurately identify drug-use groups, urine specimens were requested at 2 unanticipated times: day of enrollment and delivery. Exclusion criteria included major illness diagnosed prior to pregnancy; chronic use/abuse of medications; illicit drug use except cocaine and marijuana, and heavy alcohol use. Cocaine users were more likely to smoke, drink and use marijuana, although many controls also used these “other drugs”. Statistical controls have been used in analyses to account for these and other confounding variables. To equate the user and control groups

PAGE 36

20 on other potential confounders, subjects were matched on: high vs low perinatal risk; level of Socioeconomic Status (SES); first vs subsequent births; African-American vs other reported race. To decrease variance, children who are left-handed were excluded, and to avoid pregnancy testing, girls who have reached puberty. Based on statistical power analyses, 154 users were enrolled during pregnancy, and 154 were matched as controls; the child subjects have been followed since birth and are now 10-11 years of age. From this group 28 cocaine exposed and 25 non-exposed children have been scanned, and their images processed and analyzed. During MR scanning and data processing researchers were blinded as to the assignment of each child to a particular group.

PAGE 37

21 90 180 G d i f f ADC AT TE RF Figure 1-1: A standard diffusion weighted spin echo pulse sequence. Motion of spins during the time period become dephased proportional to the amount of motion along the gradient direction applied.

PAGE 38

22 Free vs. Restricted Diffusion00.10.20.30.40.50.60.70.80.91020040060080010001200b valueSignal Free Diffusion Restricted Diffusion Figure 1-2: Theoretical signal attenuation curves for increasing b value. The pink curve represents a diffusion curve reaching a limiting value due to restriction. The blue curve represents a free diffusion signal attenuation curve without any restriction. This figure adapted from Neil et al(10).

PAGE 39

23 Figure 1-3: An axial fractional anisotropy image of the corpus callosum acquired from an adolescent of 10 years of age.

PAGE 40

24 Figure 1-4: A diffusion weighted echo planar pulse sequence.

PAGE 41

CHAPTER 2 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: METHODS MRI and DTI Acquisition and Processing Methods Before a final MRI and DTI acquisition protocol was decided on many tests were completed. Volunteer scans were completed first in order to test the initial protocol and to evaluate the performance of the DTI scans. The overall scanner performance was exceptional, the standard radiological scans operated well (including the EPI sequences). A water phantom was used to test that the diffusion EPI sequence and b-value calculations were operating correctly. A rapid 6-direction tensor with 2 b-values was run in order to do ADC fitting along several directions. When the ADC of all six directions was averaged, the results were 1.96 m 2 /ms and = 1.73%. This result matches the diffusion rate of free water at room temperature, and the stability of the ADC over the directions demonstrate that the gradients are performing consistently, which permits an accurate calculation of ADC values over all directions. There were several prerequisites that were considered before the creation of a protocol. The first of these prerequisites was that the scan time be kept less than forty minutes. This was decided upon because it was thought that this subject population would not be able to tolerate a long MRI protocol. The second prerequisite was to consider what MR methods would be appropriate for structural scans and what scans would be used to rule out pathology. T 1 MPRAGE pulse sequence was employed for structural scans (a 3D gradient echo sequence). Axial orientation was used for 25

PAGE 42

26 consistency with the orientation of the DTI scans, in case there is a need to co-register the DTI images with the structural scans. The scans chosen to rule out pathology were T 2 weighted fluid attenuated sequences. Appropriate settings for the DTI scans were then considered. It was decided that a significant portion of the subject population would not finish the entire protocol therefore a 6-direction tensor was placed early in the protocol. It was also decided that 3 b-values were to be obtained in order to ensure greater accuracy in the ADC calculations. The b-values chosen were 0, 250, and 1000 s/mm 2 and these were chosen to optimize the ADC measurement for typical ADC values experienced in neural tissue(17). It was also decided that for the cooperative subjects a 12-direction tensor would be acquired at the very end of the protocol. The 12-direction scan would allow a greater degree of accuracy in the diffusion measurements. With the DTI, structural, and radiological prerequisites considered several adult volunteers were scanned and an initial protocol was created. A subset of five child subjects was scanned using the initial protocol to see if further optimization was required for the switch from adult subjects to child subjects. The initial protocol required little alteration. In the initial protocol a field of view of 230mm was used for many of the scans. It was noticed that a smaller field of view could be achieved on the children. A measurement of several child brains showed that a FOV of 210mm could be employed on the children with a margin for error remaining. Therefore, a switch from 230mm to 210mm FOV was employed and this was the only major change to the initial protocol to make the final protocol. The final protocol is listed below. Due to the inherent logistical difficulties in imaging a pediatric population, approximately a year was required to acquire all datasets.

PAGE 43

27 MRI and DTI Final Protocol 1: The standard Siemens localizer: The collection method is gradient echo, the field of view (FOV) is (280x280)mm, the slice thickness (Sl_thk) is 10mm, the TR is 20ms, the TE is 5ms, 1 average (1 NA), 40 degree tip is used, the matrix size (MTX) is (256x128), and the acquisition time (TA) is 9s. 2: A higher resolution sagittal localizer scan. The purpose of this scan is to visualize both the Anterior and Posterior Commensure. 15 sagittal slices, Sl_thk is 3mm, the FOV is (220x220mm), the TR is 400ms, the TE is 9.1ms, the MTX is (256x128), and the TA is 1min59s. 3: The 3 rd scan is the 6-dir 2 b-value diffusion tensor. The 2 b-values are 250 and 1000, the FOV is (210x210mm), there are 28 slices, the Sl_thk is 3.5mm, there is 20% separation between the slices, the TR is 4200ms, the TE is 90ms, NA is 4, the MTX is (128x128), the TA is 3min and 26sec. 4: The 4 th scan is the b-0 scan for the diffusion tensor. This is just the zero point for the calculation. Similar to the 6-dir scan the FOV is (210x210mm), 28 slices, the Sl_Thk is 3.5mm, there is 20% separation, the TR is 4200ms, the TE is 90ms, NA is 4, the MTX is (128x128), the TA is 21sec. 5: The 5 th scan is the high resolution gradient echo 3D scan called T 1 MPRAGE. FOV of (250x172mm), MTX of (256x256x192), a TR of 2150ms, a TE of 4.38ms, an IT of 1100ms, the tip angle is 8 degrees, the voxel size is 1mm 3 , and the TA is 3min and 21sec. 6: The 6 th scan is an echo planar fluid attenuated scan. FOV of (220x220mm), 20 slices, a Sl_thk of 5mm, a TR of 8000ms, a TE of 66ms, an IT of 1750ms, NA of 3, the MTX is (128x128), and the TA is 32sec.

PAGE 44

28 7: This scan is a spin-echo version of scan 6. The FOV is (220x220mm), 20 slices, the Sl_thk is 5mm, the TR is 9000ms, the TE is 99ms, the IT is 2150ms, the MTX is (256x233), and the TA is 2min and 53sec. 8: The 8 th scan is a 12-direction tensor. This scan has all of the properties of the 6-dir scan except there are 12 directions collected, and the TA is 6min and 47sec. For DTI processing software developed by professor Thomas Mareci, Dr. Evren Ozarslan, and Ty Black here at the university of Florida was employed(35). This group has developed tensor evaluation software for several field strengths, but all coming from Bruker consoles. This software required alteration in order to read in the Siemens images and header information in order to correctly calculate the tensor. This software has many capabilities. It can calculate FA, RA, , ADCs, eigenvalues, eigenvectors, S o image, and fiber tracking. The S o image is simply taking the ADC of every pixel and back calculating its intensity as if there were no diffusion weighting. The main advantage of using this software package is the flexibility of having the programmers of the original code available for consultation and assistance with further development. White Matter Structure Segmentation “Classically, image segmentation is defined as the partitioning of an image into nonoverlapping, constituent regions that are homogeneous with respect to some characteristic such as intensity or texture(36).” The application of image segmentation for this project was to partition white matter areas of the brain from other structures such as gray matter or CSF. This study was designed to give a first look at potential microstructural changes instigated by prenatal cocaine exposure. Considering this was a preliminary study and lacks a certain outcome, segmentation techniques that were easy to implement and are commonly used in the field were chosen. The initial techniques

PAGE 45

29 chosen to segment out the white matter structures were a hand traced method and a core method, both manual techniques. These techniques were chosen in order to give an initial look at potential group differences without a large commitment of time and to provide knowledge on the specific difficulties of using FA maps to segment white matter structures in this unique population. These techniques were successful in obtaining measures of FA and for specific white matter areas, but the techniques were not as reliable as hoped. Even without optimal reliability the initial segmentation was quite useful in that it provided encouraging data, driving the study to continue and providing insight on the difficulties of segmenting this unique population. Considering this information, a semi-automated technique was adapted in order to improve reliability. This technique relies upon pixel thresholding and did improve reliability, while still obtaining measures of FA and as well as some significant group differences. Global Segmentation Decisions Some decisions were applied to all segmentation techniques and therefore were chosen before any segmentation of the data. Namely what structures were of interest and which imaging slices were most appropriate to segment these areas of interest. The areas of the brain that would be affected by prenatal cocaine exposure during embryonic growth are unclear. Areas of the brain that have shown diffusion changes in an adult cocaine exposure study(37) were investigated as well as areas exhibiting diffusion changes from a developmental delay DTI study(38). Areas of the corpus callosum were also investigated because these areas are very large, thus are easy to segment from other structures, and are linked to many of the psychometric testing administered to the subject population. The imaging slices selected for segmentation were the same for all techniques. The white matter areas chosen for this study lie approximately on the AC_PC

PAGE 46

30 plane and 8mm above the AC_PC plane. Structures studied on the AC_PC plane consisted of right and left callosal fibers as well as right and left projection fibers. Structures studied at 8mm above the AC_PC plane were the right and left genu, splenium, and internal capsule. A statistical approach was decided upon in order to compare the efficacy of different segmentation techniques. The procedure consisted of a neuroanatomist and the thesis author independently segmenting a subset of the subject population into ROIs for each technique in order to test reliability between users. Some structures investigated in this study are large enough to appear in more than one imaging slice. To ensure that the most appropriate slice was selected for segmenting and to reduce inconsistencies between researchers the slice information generated by the neuroanatomist was shared. The statistical test used for reliability testing of the different techniques was intraclass correlation, shown in equation 2-1 below(39,40). This measure gives a ratio of the variance between the measurements of the two researchers and the variance present in the data. While this statistical test was sufficient to show lack of agreement in some techniques and excellent agreement in others, it may not be the optimal test. Future work has been discussed with the statisticians focusing on the development of a statistical test where one directly compares the geometric overlap of the ROIs generated by the two researchers. This type of statistical test would be more appropriate than just comparing the variances. groupwithinianceusersbetweeniancegroupwithinianceICC_var_var_var [2-1]

PAGE 47

31 Initial Segmentation Technique (Hand Drawn Method) The hand drawn method was intended as a first attempt at obtaining reliable data from white matter structures. This technique has the strength of being easily implemented and closely mimicking the radiologist’s interpretation, which realistically is the only “valid truth” available for in vivo imaging(41). Manual segmentation also has the benefit of potentially improving accuracy by incorporating the prior knowledge of an operator(36). Manual tracing is also a very common segmentation technique used in clinical imaging studies. A weakness of methods that rely upon manual interaction is their greater susceptibility to reliability issues(36). These reliability issues can particularly be detected in intra observer variation(41). Furthermore, manual labeling is labor intensive and cannot be feasibly performed for a large number of image data sets(41). For the purposes of this preliminary study the number of subjects involved was at a level making the manual segmentation time commitment considerable but still manageable. The hand drawn method of segmentation used in this study relied heavily upon the user to properly outline the structure of interest. General guidelines were established on defining the boundaries of structures on the FA image, but specific rules on pixel intensity inclusion and exclusion were not considered. Below is an example set of anatomical rules used to trace frontal callosal fibers, and figure 2-1 is an example of the ROIs achieved from the hand drawn method. Frontal Callosal Fiber: Begin at the anterior midline and draw posterior to the postero-medial end of the callosal fiber. Draw anterior-laterally to the posterior junction of the callosal and projection fibers. Bisect the callosal fibers by drawing a 45 degree line toward the anterior midline. Connect with original line.

PAGE 48

32 The intraclass correlation, a measure of reliability, was not optimal for this technique, approximately 0.38 full results are listed on table 2-1. While the structure of interest was identified easily, the borders of the structure were difficult to define and were the main contributing factor to the poor reliability results. Insufficient detail in the anatomical guidelines for defining these boundaries and the lack of guidelines on pixel intensity thresholds were the likely causes for the variability in defining the structure boundaries. This technique has given the researches valuable initial information on the group differences in diffusion parameters and a first look at the segmentation issues involving this data, but its poor reliability is a definite shortcoming. Reliability may be improved by creating more rigorous rules for defining the borders of structures, but a semi-automated technique should require less time for segmentation, produce similar results, and achieve superior reliability. Initial Segmentation Technique (Core Method) Volume averaging of voxels due to the lack of resolution makes segmenting clinical MRI datasets more difficult. With voxels the size of 1mm 3 and larger for clinical MRI the volume averaging of two or more tissue types within one voxel is a common occurrence. The core ROI method attempted to provide a comparison of the least volume averaged voxels in the structures of interest between the two groups while also providing reliability. This was achieved by placing a small set of pixels in the center of white matter structures sufficiently removed from the edge of the structure. This method has simple and strict rules to follow making it more reliable than the hand drawn method. The user finds the structure of interest on the FA image, similar to the method used in the hand drawn method, and then finds four contiguous pixels with the highest intensity. Figure 2-2 is an example of a core ROI in the left genu of the corpus callosum. The core

PAGE 49

33 method provided superior reliability due to its simplistic and consistent rules. The intraclass correlation for this technique was approximately 0.71, see table 2-3 for full results. While the reliability of this technique was superior to the hand drawn method it was not optimal, therefore changes were made to the technique and subsequently used in the final analysis. The core technique also provided information about the maximum FA values between the groups, but its lack of coverage of the entire white matter structure makes interpretation of the results more difficult. Final Segmentation Technique (Region Shrink Method) Semi-automated techniques increased the segmentation reliability significantly. The concept is that the user and the computer work together to define the ROI. The user provides a rough estimate of the location of the white matter structure of interest and the computer used an algorithm to define the borders. Before the user needed to define the entire border of the structure, with a semi automated technique the users need only define the general area of the structure reliability should improve. The region shrink and region grow methods are slightly different in their execution. The region grow method needs two criteria in order to operate, a seed point within the structure of interest provided by the user and a pixel intensity threshold also provided by the user. The computer then will expand the ROI from the seed point one pixel at a time until the pixel intensity threshold is encountered. The region shrink method again requires two criteria to operate, an encompassing region in which the entire structure is contained and again a pixel intensity threshold provided by the user. The computer now shrinks in from the encompassing region one pixel at a time until the pixel threshold is encountered. Both semi-automated techniques reproduce ROIs more robustly than the Hand Drawn method in this study, but were not without some issues. Semi-automated

PAGE 50

34 segmentation methods require some interaction for specifying initial parameters whose values can significantly affect performance(36). Also, when using a semi-automated segmentation technique the level of operator supervision in the process will affect the stability of the segmentation methods(41). Region grow or region shrink techniques extract an image region that is connected based on some predefined criteria. These criteria can be based on intensity information and/or edges in the image(36). A disadvantage of region growing is that it requires manual interaction to obtain the seed point and this introduces a level of manual error(36). Particular to this study there were two main shortcomings for implementing the region grow technique. First the ROI usually grew outside the structure of interest. This mainly happened when another white matter structure borders the structure of interest. The second shortcoming was that pixels with intensity lower than the threshold but contained within the structure were excluded; basically the ROI grows around such pixels. The details on how pixel intensity thresholds were chosen will be discussed later in this chapter. The solution to both of these issues would have required an edit ROI feature and a user with the ability to realize and fix these issues. An example of this would be a paintbrush type tool where the user would erase the portion of the ROI that grew outside of the structure and conversely would add pixels contained within the structure that were below the pixel threshold. The region shrink method had one weakness when dealing with this data; it required the user to define the encompassing region. This only became an issue when the white matter structure of interest bordered another white matter structure. At these borders a user trained in differentiating between these structures was required. The

PAGE 51

35 strength of the method was that no further editing was required with a properly drawn encompassing region. Therefore both methods required the user to differentiate between two white matter structures bordering each other, but the region shrink method did not exclude pixels in the middle of a structure and the region shrink method has fewer steps from start to finish. With these issues in mind the region shrink method was chosen for segmentation over the region grow method. The potential for error using the region shrink method was greatest when defining the encompassing region. While the white matter structures of interest were mainly bordered by gray matter or ventricle most were bordered by some white matter. At these white matter on white matter borders are where errors in the region shrink method likely occurred. The creation of rigorous anatomical rules to define such borders minimized these errors. The anatomical rules for the Hand Drawn method contained instructions for defining these borders. These rules were adapted and expanded in order to minimize errors in the region shrink method. An example of anatomical rules to draw the left genu encompassing region is listed below. Left Genu: 1) Determine the midline using the anterior and posterior aspects of the longitudinal fissure. 2) Starting at the midline, bisect the genu of the corpus callosum and make a generous ROI around the white matter of the genu. 3) To determine the lateral border of the genu make a 45-degree line (toward the midline) from the intersection point of the anterior internal capsule and the genu. Figure 2-3 is an example of the encompassing region and subsequent ROI on the left genu using a pixel threshold of FA 0.45. Similar instructions for every white matter structure of interest were created. ROIs created using these methods have demonstrated

PAGE 52

36 sufficient reliability. See table 2-1 for intraclass coefficients for every white matter structure studied. Thresholding Techniques A thresholding procedure attempts to determine an intensity value, called the threshold, which separates the pixels into the desired classes(36). The segmentation is then achieved by grouping all pixels with intensities greater than the threshold into one class and all other pixels into another class. The determination of more than one threshold value to segment several structures is a process called multithresholding(36). The pixel threshold for the region shrink method determined the borders of the white matter structure of interest with surrounding gray matter and CSF. The edge of the structure of interest that borders other white matter was decided by using a set of anatomical rules. The distribution of FA pixel intensities of white matter compared with the distributions of pixel intensities of CSF and gray matter were quite separate from each other. See figure 2-4, 2-5, and 2-6 for the distributions of FA pixel intensities for white matter, CSF, and gray matter. While the FA values for white matter were greatly different from those of gray matter and CSF, the distribution of FA values from whole brain ROIs show no clear distinction between these types of tissues. See figure 2-7 for whole brain distribution of FA values. This arises from a spectrum of volume averaging white matter pixels with gray matter and CSF pixels. An obvious choice for a threshold would be to choose one that would provide pixels containing non-volume averaged white matter. While this may be an attainable goal it ruins the purpose of the region shrink method. An encompassing region with a threshold at this level would produce an ROI that does not resemble the structure of interest, see figure 2-8. A more logical approach is to set the threshold at the highest level that still maintains the shape of the structure of

PAGE 53

37 interest. A threshold at this level should not only provide an ROI without extensive volume averaging because the FA distributions from the three major tissue types are sufficiently separated, but also retain the shape of the structure of interest. This method of selecting the threshold was the one implemented. Table 2-2 lists the pixel threshold values used to segment out every structure. Final Segmentation Technique (New Core Method) The initial protocol for the Core ROI method produced a fairly reliable technique for sampling the least volume averaged white matter pixels. Improvements upon this technique were implemented in order to raise its intraclass correlation coefficient; Table 2-3, shows ICC measurements of the initial techniques. The basic rules of the original method remained; draw an ROI around the most intense four pixels that make a two by two square within the structure of interest. The alteration that was made to improve reliability was to draw five core ROIs within each structure of interest and use the ROI with the highest mean and delete the remaining four, see figure 2-9. This protocol was tested by two independent researchers on a subset of ten subjects. The intraclass correlation coefficient of approximately 0.89 resulted, see table 2-3 for complete results of ICC testing.

PAGE 54

38 Figure 2-1: This is an example of the hand drawn technique in the right and left front callosal fibers.

PAGE 55

39 Figure 2-2: This figure demonstrates the core method. The top image displays the genu of the corpus callosum. The bottom image is zoomed in on the genu and has a core ROI drawn over the left genu.

PAGE 56

40 Figure 2-3: This figure is an example of the encompassing region for the left genu and its resulting region shrink ROI. A pixel threshold of FA 0.45 was used.

PAGE 57

41 Table 2-1: This table lists the Intraclass correlations for both the hand drawn segmentation technique and the region shrink technique. Equation 2-1 defines the solution for ICC. An ICC of one indicates perfect agreement. Structure Hand Drawn ICC Region Shrink ICC Left Genu 0.16 0.91 Right Genu 0.41 0.96 Left Anterior Internal Capsule 0.46 0.94 Right Anterior Internal Capsule 0.31 0.96 Left Posterior Internal Capsule 0.55 0.92 Right Posterior Internal Capsule 0.40 0.98

PAGE 58

42 White Matter Cores Histogram0246810121416182000.10.20.30.40.50.60.70.80.91FAFrequency Figure 2-4: This figure shows the FA distribution of white matter cores. The data is obtained from white matter cores, which should provide little to no volume averaging, from several subjects.

PAGE 59

43 CSF cores Histogram0246810121416182000.10.20.30.40.50.60.70.80.91FAFrequency Figure 2-5: This figure shows the FA distribution of CSF cores. The data is obtained from CSF cores segmented from maps from several subjects. These maps make it easy to segment CSF from white and gray matter.

PAGE 60

44 Gray Matter Cores Histogram05101520253035404500.10.20.30.40.50.60.70.80.91FAFrequency Figure 2-6: This figure shows the FA distribution of gray matter cores. The data is obtained from gray matter cores from several subjects.

PAGE 61

45 Whole Slice Histogram020040060080010001200140016001800200000.10.20.30.40.50.60.70.80.91FAFrequency Figure 2-7: This figure shows the FA distribution of the corpus callosum whole slice. The data is obtained drawing an ROI around the whole slice from several subjects.

PAGE 62

46 Figure 2-8: This is an example of the region shrink method using a high threshold value. The resulting ROI is only of white matter pixels without volume averaging, but this ROI does not represent the shape of the left genu very well.

PAGE 63

47 Table 2-2: This table shows the pixel threshold values used in the region shrink technique to segment out all structures, FA ranges from zero to one. The same threshold was used of both right and left sides of all structures studied. Anatomical Structure FA Pixel Threshold Frontal Callosal Fibers 0.32 Frontal Projection Fibers 0.30 Genu of the corpus callosum 0.45 Anterior Internal Capsule 0.39 Posterior Internal Capsule 0.40 Splenium of the corpus callosum 0.49

PAGE 64

48 Table 2-3: The ICC results from the old and new core segmentation method. Structure Initial Core ICC New Core ICC Left Genu 0.91 0.97 Right Genu 0.92 0.99 Left Splenium 0.78 0.95 Right Splenium 0.22 0.65

PAGE 65

49 A B C Figure 2-9: This figure demonstrates the new core method. Image A displays the genu of the corpus callosum. Image B is zoomed in on the genu and has a core ROI drawn over the left genu, representing the old core technique. Image C represents the new core technique.

PAGE 66

CHAPTER 3 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: RESULTS Introduction The results from these studies implicate microstructural alterations in frontal white matter areas in children prenatally exposed to cocaine. While not everything investigated produced dramatic differences, both preliminary and final analyses resulted in alterations to diffusion parameters in this unique population compared with the control group. Preliminary analysis completed on approximately half of the subject population resulted in significant reduction of FA in frontal white matter areas in the exposed population, with trends of increased noticed. While the results were interesting, the reliability of the segmentation used to generate these results was poor. Employing semi-automated segmentation techniques on the entire population resulted in a significant increase in frontal white matter in the exposed population, with trends of decreased FA noticed. Diffusion measures generated from the Region Shrink Method were correlated with brain hemispheric volumes and significant correlations were found in the control group but not in the exposed group insinuating developmental differences between these two groups. With these results in mind an impression of microstructural differences in the exposed group revealed itself. Preliminary Results Preliminary data was needed in order to define the directions in which to pursue areas of data analysis and possible outcomes. The preliminary data consisted of twenty50

PAGE 67

51 two subjects, and began once half of the population had been collected. FA and values of these areas were compared with several test criteria, namely differences between exposed and control groups. Two methods of white matter segmentation were used, hand drawn structures and core measurements described in chapter 2, with varying results. Statistical significance between the two groups existed in two of the four frontal white matter areas studied. In those areas with significance the exposed group demonstrated lower FA values with no significant difference between values. However, in the regions without significance the trend of lower FA values was still present. White matter structures studied for preliminary analysis included right and left callosal fibers, right and left projection fibers, right and left genu, right and left splenium, right and left anterior internal capsule, and finally right and left posterior internal capsules also including core ROIs from all areas of the genu and splenium. From these ROIs both FA and means and standard deviations were collected. An unblinded statistician analyzed all data. The data was tested for normal distributions and passed, therefore parametric statistical tests were used. A two tailed student t-test employing a 95% confidence interval was used in order to test for significant differences in FA and between exposed and unexposed populations in all ROIs. Table 3-1 shows the results of FA differences between the two populations in the ROIs studied, and table 3-2 shows the changes. Two of the four frontal white matter areas showed a statistically significant lower FA in the exposed population with the other two showing a trend of lower FA values, the left genu core had significantly reduced FA in the exposed population. It is interesting that all four frontal white matter areas studied exhibited

PAGE 68

52 increased while none of these areas were significantly increased. The ROIs with significant differences in FA are displayed in figure 3-1. The areas that did not show significantly reduced FA leaned in the same direction in the exposed population. Ten of the thirteen remaining structures studied exhibited lower FA values seen in table 3-1. FA values for each ROI were also compared with IQ measurements from the subject population in order to determine if anisotropy values correlated with intelligence testing. No correlation was found within the groups or between groups. These preliminary results gave a better impression of what to expect once the whole population was studied, and gave valuable knowledge on different techniques of analysis with associated strengths and weaknesses. Region Shrink Results The Region Shrink data consisted of the entire population of fifty-six with three removed from the study because of MRI exclusion criteria. Similar to the preliminary results, FA and values of these areas were compared with several test criteria, including diffusion differences between exposed and control groups as well as other comparisons. The Region Shrink semi-automated segmentation technique was employed for white matter segmentation and described in chapter 2. Statistical significance between the two groups again existed in two of the four frontal white matter areas studied, but this time it was and not FA that was significant. In those areas with significance the exposed group demonstrated increased with no significant change in FA values. In the frontal white matter regions without significance the trend of increased was still present. With the intriguing results in frontal white matter from the preliminary studies other frontal areas were attempted to be examined. Due to the inherent poor resolution

PAGE 69

53 associated with clinical DTI these structures were too difficult to segment and therefore not included. Subsequently the structures studied were the same as in the preliminary analysis. Both means and standard deviations of FA and were again collected from the ROIs generated. Again two tailed student t-tests employing a 95% confidence interval were used in order to test for significant differences in FA and between exposed and unexposed populations in all ROIs. Table 3-4 shows the results of FA differences between the two populations in the ROIs studied, and table 3-3 shows the changes. Two of the four frontal white matter areas showed a statistically significant increase in in the exposed population with the other two showing a trend of increased . Of the four frontal white matter areas studied, three exhibited decreased FA values in the exposed population, but none were significantly different. All frontal white matter ROIs studied using the Region Shrink Method are shown in figure 3-2. Of the areas studied outside of the frontal white matter a trend of increased was also noticed; only two of these eight structures did not have increased in the exposed population. While no structures exhibited significantly different FA values compared with the control group, a trend of decreased FA values was noticed. Of all twelve structures studied nine displayed a decrease in FA in the exposed population. Using the high resolution structural scans collected on this population, right and left hemisphere volumes were measured. Correlations were measured between these hemispheric volumes and diffusion measures. These correlations were completed to potentially provide information on links between brain development and diffusion measures, full results are listed on table 3-5. Frontal white matter areas again provided interesting results. had a negative correlation in the control population with brain

PAGE 70

54 volume in four of the eight correlations measured in frontal areas. Three of the four areas without significance were negative leaning correlations. In the exposed population there were no significant correlations found between and hemispheric volume. Contrary to the control population six of the eight correlations measured were positive leaning, and the correlations that were significant in the control group were leaning in the opposite direction in the exposed group. Additionally hemispheric volumes were compared between the groups. There were no significant volume differences between the control and exposed groups. Correlating FA with hemispheric volume provided one significant positive correlation in the control population. Six of the remaining seven correlations measured were positive while not significant. There were no significant correlations of FA with hemispheric volume in the exposed population. Interestingly half of the correlations measured in the exposed population were positive leaning with the other half negative leaning, but the significant positive correlation in the control population was opposite leaning in the exposed. Conclusion Prenatal cocaine exposure has been proposed to potentially disrupt the monoaminergic system and cause hypoxia to the fetus. These mechanisms may alter white matter structure and development and may have a negative impact upon an individual during development. Frontal white matter areas were hypothesized to be most at risk from these mechanisms, and has shown to be damaged in an adult cocaine exposure study(37). The data presented here supports our hypotheses and the previous work done in adults. The broader implications of these results will be discussed in chapter 4.

PAGE 71

55 Figure 3-1: The top image is an example of a core ROI inside the right genu of the corpus Callosum, this core measurement in the right genu had significantly reduced FA in the exposed population. The bottom image give examples of hand drawn ROIs in the right and left Callosal fibers in frontal white matter, these frontal white matter areas had significantly reduced FA in the exposed population.

PAGE 72

56 Table 3-1: This is a table of the fractional anisotropy results in the white matter structures studied using the preliminary data. A two-tailed student t-test was used to test for differences between groups. Those rows highlighted in green have significantly different FA values, p-value < 0.05. White Matter ROI FA of control population FA of target population P value Left Frontal Callosal Fibers 0.713 0.678 0.041 Right Frontal Callosal Fibers 0.685 0.639 0.049 Left Frontal Projection Fibers 0.650 0.630 0.246 Right Frontal Projection Fibers 0.626 0.615 0.593 Left Corpus Callosum Genu 0.815 0.793 0.399 Left Corpus Callosum Genu core 0.887 0.888 0.979 Right Corpus Callosum Genu 0.812 0.780 0.203 Right Corpus Callosum Genu core 0.914 0.874 0.039 Left Anterior Internal Capsule 0.679 0.661 0.313 Right Anterior Internal Capsule 0.676 0.666 0.487 Left Posterior Internal Capsule 0.687 0.665 0.415 Right Posterior Internal Capsule 0.701 0.681 0.332 Left Corpus Callosum Splenium 0.810 0.819 0.548 Left Corpus Callosum Splenium core 0.897 0.888 0.581 Right Corpus Callosum Splenium 0.803 0.814 0.560 Right Corpus Callosum Splenium core 0.900 0.883 0.385

PAGE 73

57 Table 3-2: This is a table of the Average diffusion results in the white matter structures studied using the preliminary data. A two-tailed student t-test was used to test for differences between groups. No ROI studied had significantly different values p-value < 0.05. The right genu core is approaching significance and is highlighted in yellow. There is a trend to higher values in the exposed population. White Matter ROI of control population m 2 /ms of target population m 2 /ms P value Left Frontal Callosal Fibers 0.780 0.800 0.181 Right Frontal Callosal Fibers 0.773 0.776 0.222 Left Frontal Projection Fibers 0.776 0.782 0.699 Right Frontal Projection Fibers 0.790 0.814 0.067 Left Corpus Callosum Genu 0.776 0.800 0.359 Left Corpus Callosum Genu core 0.758 0.797 0.203 Right Corpus Callosum Genu 0.781 0.814 0.206 Right Corpus Callosum Genu core 0.746 0.806 0.054 Left Anterior Internal Capsule 0.746 0.754 0.423 Right Anterior Internal Capsule 0.752 0.752 0.952 Left Posterior Internal Capsule 0.755 0.752 0.735 Right Posterior Internal Capsule 0.760 0.754 0.579 Left Corpus Callosum Splenium 0.829 0.823 0.802 Left Corpus Callosum Splenium core 0.805 0.789 0.539 Right Corpus Callosum Splenium 0.850 0.838 0.626 Right Corpus Callosum Splenium core 0.829 0.802 0.361

PAGE 74

58 Figure 3-2: This figure illustrates the four frontal white matter areas segmented using the Region Shrink Method. RFP and LFP are abbreviations for right and left frontal projection fibers. RFC and LFC are abbreviations for right and left frontal callosal fibers. Of these areas LFC and RFP exhibited a significantly increased in the exposed population.

PAGE 75

59 Table 3-3: This table lists the results of the group comparisons using the region shrink data. A two-tailed student t-test was used to test for differences between groups. Those rows highlighted in green have significantly different FA values p-value < 0.05. White Matter ROI of Control Population m2/ms of Exposed Population m2/msp-valueLeft Front Callosal0.7680.8020.006Right Front Callosal0.7820.7970.193Left Front Projection0.7750.7830.396Right Front Projection0.7990.8200.013Left Genu0.7890.8020.317Right Genu0.8030.8200.256Left Anterior Internal Capsule0.7530.7530.945Right Anterior Internal Capsule0.7500.7540.714Left Posterior Internal Capsule0.7570.7470.115Right Posterior Internal Capsule0.7560.7540.677Left Splenium0.8230.8280.629Right Splenium0.8280.8390.401

PAGE 76

60 Table 3-4: This table lists the results of the FA group comparisons using the region shrink data. A two-tailed student t-test was used to test group differences. While no significant differences between the groups were found there is a trend of decreased FA in the exposed population. The left frontal callosal fiber approached significance, highlighted in yellow. White Matter ROIFA of Control PopulationFA of Exposed Populationp valueLeft Front Callosal0.6980.6790.079Right Front Callosal0.6830.6750.458Left Front Projection0.6370.6450.199Right Front Projection0.6330.6300.753Left Genu0.7960.7850.213Right Genu0.7880.7830.502Left Anterior Internal Capsule0.7070.7160.281Right Anterior Internal Capsule0.7030.7060.685Left Posterior Internal Capsule0.7070.7020.458Right Posterior Internal Capsule0.7110.7040.320Left Splenium0.8070.8020.414Right Splenium0.8070.7990.326

PAGE 77

61 Table 3-5: This table lists the results of the correlations between hemispheric volume and diffusion measures. Linear relationships between two variables were assessed using the Pearson correlation coefficient. Alpha was set at 0.05 for all statistical tests. The areas highlighted in green exhibited a significant correlation with hemispheric volume. StructureControl L _ Brain CorrelationControl R _ Brain CorrelationExposed L _ Brain CorrelationExposed R _ Brain CorrelationLFC FA0.058-0.0440.0060.109RFC FA0.3770.1890.0220.184LFP FA0.2680.161-0.0140.067RFP FA0.5000.507-0.188-0.187LFC 0.082-0.062-0.159-0.199RFC -0.486-0.4550.0230.062LFP -0.389-0.3420.0340.037RFP -0.594-0.5850.1650.120

PAGE 78

CHAPTER 4 DIFFUSION TENSOR IMAGING OF PRENATAL COCAINE EXPOSED CHILDREN: DISCUSSION The hypothesis that prenatal cocaine exposure induces microstructural changes in the developing brain is supported by the DTI results from this study. The reduction of FA and the increase in in the frontal white matter areas of the brain indicate subtle microstructural changes in these areas. This data is consistent with the work of Lim et al.(37) in which adult cocaine users showed a significant reduction of FA in frontal white matter. This data is also consistent with the proposed mechanisms of hypoxia and alterations of the monoaminergic system, where frontal white matter is at particular risk. Psychometric testing of the exposed population has shown a deficit in performance on tests that are controlled by frontal white matter, which provides evidence that the changes detected by DTI are not anomalous. A potential implication of this and similar work would be the creation of training/teaching methodology mitigating the specific deficits, which this high-risk subject population exhibits. Another potential implication of this and similar work would be its effect upon social policy, for example establishing new regulations on the removal of these children from their natural mothers. While much still needs to be completed to fully understand the effect of prenatal cocaine exposure on the developing brain, this work provides insight into specific microstructural changes due to this dangerous drug. A proposed mechanism of frontal white matter damage from prenatal cocaine exposure discussed in chapter 1 is the alteration of neurotransmitter levels. Cocaine 62

PAGE 79

63 exposure has a significant effect upon the neurotransmitter levels and cocaine easily crosses the fetal blood-brain barrier allowing it to accumulate in the CNS. Increased extra-cellular levels of norepinephrine, dopamine, and serotonin have been attributed to cocaine exposure. Neurotransmitters play an important role in normal brain development by modulating growth of developing neurons. Brain abnormalities reported in animal models of prenatal cocaine exposure include a reduction in the number of cortical cells, inappropriate positioning of cortical neurons, altered glial morphology, reduction in length of neurites, and apoptotic neural cell loss. These abnormalities stem from the interference with the fetal monoamine uptake mechanisms(22). One study shows that dendrites in regions of excess neurotransmitters course abnormally through the cortex, on and out of the plane of section in an irregular fashion rather than in the usual straight course(2). The wavy dendritic profiles reflect a compensatory mechanism to counteract a more rapid rate of increase in the length of apical dendrites(5). The increase in waviness of dendrites and neurites may induce a reduction of FA in the developing brain. If fibers within a bundle travel in a more undulating fashion a reduction of anisotropy would result. Another study demonstrated a decreased bundling of pyramidal neuron dendrites(4). The cocaine exposed mice fail to develop the normal barrel structures that are related to vibrissae representation in the somatosensory cortex due to increases in serotonin levels during the sensitive period for barrel formation(4). If the bundling of fibers is decreased or packed more loosely a decrease in the restriction of water diffusion would be probable resulting in an increase in .

PAGE 80

64 As shown in chapter 3 a reduction of FA in frontal white matter areas was detected. In the preliminary results two of the four frontal white matter areas studied showed a statistically significant lower FA in the exposed population with the other two showing a trend of lower FA values. The left genu core had significantly reduced FA in the exposed population. The areas that did not show significantly reduced FA leaned in the same direction in the exposed population. Ten of the thirteen remaining structures studied exhibited lower FA values. These preliminary results are listed in table 3-1 and described in full detail in chapter 3. In the region shrink results no structures exhibited significantly different FA values compared with the control group, but a trend of decreased FA values was noticed. Of all twelve structures studied nine displayed a decrease in FA in the exposed population. These region shrink results are listed in table 3-4 and described in full detail in chapter 3. The smaller study group used in the first analysis and/or the different segmentation techniques employed may explain this discrepancy in significance between the two sets of results. As previously discussed, alterations to the monoaminergic system in the developing brain may cause permanent structural alterations. Studies in animal models have shown that dendrite and neurite outgrowth occur in a more undulating fashion in the exposed animals. This results from a compensatory mechanism to counteract a more rapid rate of growth. This is a likely cause of the reduced FA seen in the exposed population in our study. FA is proportional to the organization of fiber bundles. The more wavy nature of the exposed animals’ neurites and dendrites indicate less organization than the unexposed counterparts(5). Potential damage originating from hypoxia could also reduce FA in areas affected. Cell injury or death resulting from hypoxia would likely result in less

PAGE 81

65 organization or cell density reduction, both potentially leading to FA reduction. This information suggests that these animal studies and our work are consistent with one another. Also, as previously described, these results match the adult cocaine study performed by Lim et al., and are consistent with our own hypothesis(37). In the hypothesis it was stated that would be increased in the exposed population and these data support that. In the Region Shrink results two of the four frontal white matter areas studied showed a statistically significant increase in in the exposed population with the other two areas showing a trend of increased . Of the areas studied outside of the frontal white matter a trend of increased was also noticed; only two of these eight structures did not have increased in the exposed population. These Region Shrink results are listed in table 3-3 and described in full detail in chapter 3. In the preliminary results it is interesting that all four frontal white matter areas studied exhibited increased while none of these areas were significantly increased. These preliminary results are listed in table 3-2 and described in full detail in chapter 3. Again, the smaller study group used in the first analysis and/or the different segmentation techniques employed may explain this discrepancy in significance between the two sets of results. It has been previously stated that there is a decreased bundling due to prenatal cocaine exposure(4). This animal study showed that the exposed mice fail to develop the normal barrel structures due to increased neurotransmitter levels. The increased seen in the exposed population in our study is potentially linked to this same phenomenon. The likely reduction in the restriction of water diffusion resulting from this abnormal barrel development would increase . Again, hypoxia damage due to

PAGE 82

66 prenatal cocaine exposure could also be a mechanism of alteration to . Hypoxia causing cell injury or death resulting in less organization or cell density reduction would cause a decrease in the restriction of water diffusion thus leading to an increase in . The results from our study are consistent with the animal work done by Stanwood et al. and support our hypothesis that prenatal cocaine exposure increases . As shown in chapter 3, hemispheric brain volumes correlated positively with FA values in the control population while this was not so in the exposed population. Although not as strongly, correlated negatively with hemispheric brain volume in the control population while again this was not so in the exposed population. As the brain develops brain volume increases. Therefore those children with decreased hemispheric brain volumes potentially are exhibiting delayed development. This potential delay could cause a reduction in anisotropy and increase in through less myelination or less organization of fiber bundles. This is only a potential explanation for this phenomenon since a literature review revealed no insight into these correlations. Note that no correlation between cocaine exposure and hemispheric brain volume was detected. This theory may explain the significant correlations in the control population, but it does not explain the lack of correlation in the exposed population. Potentially cocaine causes a developmental disruption at the micro-structural level explaining the lack of correlation in the exposed population. A potential confound for these results lie in an artificial increase in FA. Brains which are larger on average, should load the RF coil more. Excess loading of a coil should bring down the SNR of the data. A reduction in overall SNR artificially elevates FA(42). While this may explain the positive correlation between FA and brain volume it does not seem likely. The range of hemispheric brain

PAGE 83

67 volumes in this study was not excessive which would lead to a small change in loading from one subject to another. This small change in loading would have a small effect upon SNR, which in turn should have a minimal effect upon FA. These brain volume correlations were not anticipated and a literature search did not help in explaining this phenomenon. Future work may help understand the underlying mechanisms causing this phenomenon and what effect it might have on this at risk population. While these data suggest that prenatal cocaine exposure causes microstructural changes in frontal white matter there are some potential confounds to these findings. The home environment and labeling are potential causes of variation in these data. Potentially the two groups have different home environments when compared collectively. A poor home environment likely results in less motivation and stimulation of the child, home environment and other environmental factors are indicators of child performance(43). Similarly the labeling of a child as a crack baby potentially lowers the expectations of the child’s performance. Under these conditions a child is less likely to reach their full potential and live up to the expectations of their teachers or caregivers, leading to poorer performance on psychometric testing. IQ testing and FA values have been shown to be correlated demonstrating a link between reduced anisotropy in children with diminished performance on standardized testing(44). Whether a microstructural alteration causes a reduction in performance on testing or a more complex picture of home environment, child stimulation, and microstructure exists is unknown. This effect has been determined to be an unlikely confound in our study because the groups were matched on socio-economic status (SES), an indicator of home environment.

PAGE 84

68 Another potential confound to the findings presented here is that a frontal white matter structural difference may have been passed down from the mother. Frontal white matter influences executive functioning that controls the ability to perceive negative consequences(45,46). Inhibited executive functioning may lead to an increased probability of cocaine use by the mother because the perception of negative consequences from drug use is diminished. Therefore mothers may have frontal white matter differences before they become users of cocaine and subsequently pass this on to their children. The question becomes whether cocaine exposure results in reduced frontal white matter integrity or those with reduced frontal white matter integrity are more likely to use cocaine. Psychometric studies on this population have shown that the exposed population performs significantly worse on executive function tests than the control group. It is unlikely that the microstructural changes were present before cocaine exposure because it has been shown in several animal studies that cocaine exposure and prenatal cocaine exposure significantly alters brain microstructure(3-5). A likely confound to these findings would be from other drugs in the system such as alcohol, nicotine, and marijuana. While excessive alcohol, nicotine, and marijuana use were exclusion criteria, low levels of these drugs were present during the prenatal period in some mothers. This was accounted for by matching the groups on drug exposures other than cocaine use. With several drugs in the body at one time a potential interaction of these drugs with each other may have an unpredicted effect. While matching the groups on other drug use likely removed the individual effect of these drugs on the results, potential interactions of these drugs with each other and cocaine may have affected the results.

PAGE 85

69 The results of this study have several potential implications to this unique subject population. Psychometric testing of this population in the areas that frontal white matter controls has shown deficits in the exposed population. First this supports that the frontal white matter microstructural alterations seen are real and not random because the functions these areas control are showing reduced performance on standardized tests. Second this reduced performance may be a direct result from prenatal cocaine exposure. As discussed earlier the home environment has a large impact on child development. While this research does indicate that prenatal cocaine exposure does induce microstructural changes, environmental factors have also shown to be risk factors. A measure of home environment was compared with psychometric testing and was found to have a stronger correlation than prenatal cocaine exposure. This and similar research may elucidate that while prenatal cocaine exposure is a risk factor on child development all is not lost. Potentially placing children in more supportive and nurturing homes could mitigate many of the shortcomings these children exhibit. This work and similar studies have their greatest potential impact upon policy making. Results have indicated that the home environment is a very important risk factor for this population. Potential implications would be different guidelines dealing with the removal of children from their biological parents as well as different guidelines on which foster homes are best suited for the care. The government also has tough penalties on cocaine abuse creating a strong deterrent from mothers attending prenatal care or from revealing that a problem exists. Potentially confidential prenatal care may provide help for at risk mothers otherwise not willing to be helped(47).

PAGE 86

70 This work has presented subtle microstructural differences to frontal white matter in children prenatally exposed to cocaine. Microstructural alterations to white matter areas likely affect certain areas of brain function. Future work may potentially provide a clear picture of the specific deficits these children present. Once a clear picture of what specific areas demonstrate inhibited performance specific measures may be taken in order to compensate for these deficits. The labeling of a child as a crack baby also potentially has a deleterious effect upon self-esteem. This research and others may help mitigate the thought that children prenatally exposed to cocaine are hopeless, thus raising the performance expectations of the children in the eyes of caregivers and educators. This study has answered some questions presented by the researchers involved but it also has created many new ones. In order to pursue new directions the group must overcome the first obstacle which is obtaining more funding. With new funding potentially the entire subject population of 308 may be imaged. There would be some proposed changes to the MRI protocol if imaging of the entire subject population were to occur. Throughout the course of this study the amount of cooperation these children demonstrated was a pleasant surprise. This cooperation was unexpected and in future work it would be possible for longer scan times allowing for the collection of superior DTI data sets. With the potential for longer scan times and the ability of these children to lay still for at least an hour the 6 direction DTI acquisition would not be used and at a minimum a 12 direction tensor would be collected. Alternatively the proper licensing from Siemens would be obtained and the diffusion pulse sequence would be rewritten optimizing the gradient parameters and redefining the number of diffusion directions available. With longer scan times available thinner slices with no gaps may be collected.

PAGE 87

71 Thinner slices along with added diffusion directions will increase the power of fiber tracking. The amount of psychometric testing completed on this population along with the detailed case history of each subject will allow a very large number of hypotheses to be tested if DTI datasets were collected on the whole population. First, this large subject group will add to the statistical power of the findings presented here. Second, answers to many of the questions raised by this initial study can be attempted. One area that could be queried would be the effect of the home environment on diffusion parameters. Some comparisons that could be made using the entire subject population are home environment scores versus cocaine exposed males, cocaine exposed females, unexposed males, unexposed females, unexposed males with biological mothers, and many others. Also with such a large number of unexposed subjects many comparisons between diffusion characteristics and normal child development can be attempted. An example of this would be correlating certain brain regions’ diffusion characteristics with psychometric testing in unexposed males and females. The analysis of the data could also be improved upon. More sophisticated white matter segmentation techniques could be employed; an example would be a fully automated segmentation technique. This could potentially allow a much larger number of white matter structures to be included in the analysis. Fiber tracking also becomes a strong possibility for this subject population especially if the proposed MRI protocol changes are made making fiber tracking more powerful. Using fiber tracking a more in depth look at the white matter structural changes seen with this preliminary study can be attempted. Future work should also include a method to properly define the relationship

PAGE 88

72 between diffusion measures and brain hemispheric volume unexpectedly seen in the preliminary study. Prenatal cocaine exposure is a growing problem with the consequences not fully understood. This work has shown subtle alterations to frontal white matter microstructure resulting from prenatal cocaine exposure. Frontal white matter areas control many aspects of brain function and alterations to its microstructure may have significant effects to these areas of brain function. The microstructural changes most likely resulted from increased extra-cellular levels of norepinephrine, dopamine, and serotonin due to prenatal cocaine exposure. Neurotransmitters play an important role in normal brain development by modulating growth of developing neurons thus altering growth and development of areas with increased neurotransmitter levels. With a better understanding of these mechanisms and the effects that prenatal cocaine exposure has on them a clearer picture of the specific challenges exposed children are confronted with can be seen. With this knowledge a training/teaching methodology may be implemented mitigating specific deficits, which this high-risk subject population exhibits. Prenatal cocaine exposure is a sad social condition doubtful to go away in the near future. Until it does, research needs to be conducted to help understand the underlying mechanisms of damage and generate ways to help these children.

PAGE 89

CHAPTER 5 HIGH FIELD T 1 CONTRAST OPTIMIZATION: BACKGROUND AND SIGNIFICANCE Introduction The MR community has been pushing to higher and higher field strengths for many years primarily for the added SNR and increased spectral resolution. At higher field strengths several imaging issues become more apparent. As field strength increases the T 1 values of neural tissue become longer, as the difference in T 1 values between different neural structures appears to become smaller. Many clinical MRI protocols exploit T 1 contrast, and because high fields offer little T 1 contrast this has been a major obstacle in pursuing high field machines for clinical work. Also at higher fields B 1 effects become more complex with higher amounts of power being required as well as standing wave phenomenon becoming more significant. Susceptibility effects also become larger as field strength increases. These susceptibility effects make in vivo imaging more difficult as well as seriously degrading the utility of gradient recalled echo imaging in many cases. The work presented here characterizes the performance of three pulse sequences that provide T 1 contrast on three high field systems. The three pulse sequences of interest are saturation recovery (SR), inversion recovery (IR), and modified driven equilibrium Fourier transform (MDEFT). A methodology was created in order to assess the T 1 contrast provided by these pulse sequences at high field strengths. Neurological T 1 values at high field strengths needed to be determined for modeling and phantom construction, therefore neurological T 1 values from in vivo mice were measured on three 73

PAGE 90

74 high field systems, 4.7T, 11T, and 17.6T. Biologically representative T 1 phantoms were then created using the in vivo T 1 values by doping distilled de-ionized water (ddH 2 O) with copper sulfate. The copper sulfate shortened the T 1 of ddH 2 O to match the in vivo T 1 values obtained in mice. A simulations package was created for modeling contrast generated by the pulse sequence in order to optimize experimental settings and to confirm results. The simulations package provided the researchers with insight into optimal settings by modeling T 1 contrast experiments. The modeling of T 1 contrast relied on several variables: obviously the T 1 s from the two groups of spins being compared were considered, along with the T 2 s from the two groups of spins. Also used in the modeling was the TR, TE, the tip angles for all RF pulses, and the number of phase encodes employed. Using these user defined variables the T 1 contrast between two groups of spins were compared for a large range of preparation times, generating contrast curves. Steady state solutions for both SR and IR were employed. This was not required for MDEFT because of the driven equilibrium aspect of the pulse sequence. All contrast simulations were normalized to the number of averages and to acquisition time. This package was created using Microsoft Excel and all variables were dynamically linked to the T 1 contrast calculations. This allowed researchers to view the effect of one variable on the contrast of all the pulse sequences of interest. For example, to see the effect of the tip angle on the contrast a quick update of the tip angle variable shows the effect on all contrast curves. This simulations package not only gave insight into optimal settings, but also provided a powerful troubleshooting tool when results did not meet with expectations.

PAGE 91

75 Chapter 7 evaluates the contrast performance of the three pulse sequences focusing on each system individually describing the strengths and weakness of the pulse sequences. Chapter 8 evaluates the contrast performance of the pulse sequences as a function of field strength demonstrating the evolution of T 1 contrast with increasing field strength. Chapters 7 and 8 characterize the T 1 contrast performance of the pulse sequences of interest at high field, but do not deal with some of the obstacles of high field imaging. Chapters 9 and 10 evaluate spin echo (SE) imaging and fast imaging methods providing information on how to implement these T 1 contrast pulse sequences under practical experimental conditions. In vivo imaging becomes difficult at high fields due to increased susceptibility effects. Because of this spin-echo imaging is required more often at higher fields. Chapter 9 evaluates the performance of SE versions of the three pulse sequences compared with their gradient echo (GE) counterparts. T 1 contrast studies often require a significant time allotment and as T 1 s lengthen with increasing field strength acquisition times also lengthen. Due to this time requirement most in vivo T 1 contrast studies need to employ the fast imaging techniques of segmented gradient echo (segmented-GE) and rapid acquisition relaxation enhanced (RARE), a SE based fast imaging technique, in order to complete other imaging studies in a reasonable time frame. Chapter 10 evaluates the performance of these fast imaging techniques compared with their SE and GE counterparts. Chapters 7 and 8 demonstrate that sufficient neurological T 1 contrast may be achieved at high field strengths using three T 1 contrast pulse sequences and how the T 1 contrast performance evolved with increasing field strength. Chapters 9 and 10 demonstrate that sufficient biological T 1 contrast with good image quality at high field strengths may be achieved under practical conditions.

PAGE 92

76 T 1 and T 2 and Field Strength Dependence As we strive for higher field strengths to improve the SNR the T 1 values grow longer and in neural tissue the different structures have convergence of T 1 values(48). The lengthening of the T 1 values at these higher fields causes imaging time to increase due to the slower re-growth of longitudinal relaxation. The convergence of T 1 values in neural tissue causes a dramatic decrease in contrast. A T 1 preparation pulse used with appropriate selection in imaging techniques addresses both of these issues. Both issues of T 1 lengthening and T 1 neural values converging are related by a field dependant characteristic of the surrounding media. After a spin is excited it can only return to the ground state by releasing its excess energy to the surrounding media. This return to ground state requires a stimulating radio-frequency, RF, field. The surrounding nuclear environment generates these fields. The lattice field must fluctuate at a rate that matches the Larmor frequency in order for the spin to dissipate its energy to the lattice. The average rate at which molecules reorient is proportional to the size of the molecule. Larger molecules, such as lipids, have a frequency of rotation which is slower than H 2 O, but which more closely matches Larmor precession at typical clinical MR field strengths. This is why at typical MR field strengths fat has a T 1 on the order of a few hundred milliseconds while pure water is on the order of several seconds(49). T 1 values in tissue are determined by the relaxation rates of two different populations of spins, namely bound water and free water. These two populations have different physical motion properties due to one being bound to a very large protein or lipid and the other not bound at all. The water in biological tissues tends to relax to the ground state much quicker than pure water. A fraction of the water in tissues is bound to the surfaces of proteins. This binding slows the motion of the water. The slower

PAGE 93

77 molecular motion of this bound water more closely matches the Larmor frequency, which in turn gives more efficient relaxation to the ground state. As field strength increases the Larmor frequency increases. This causes the bound water’s molecular motion to be slower than the Larmor frequency. So as field strength increases, the bound water’s efficiency to return to ground state is lessened, and free water’s efficiency is relatively insensitive to change in frequency. Therefore, at higher field strengths free water, which has a longer T 1 than bound water, plays a more dominate role in longitudinal relaxation, which increases the overall T 1 value. Not only do the T 1 values increase as field strength increases, but the difference in T 1 values of biological tissues are diminished. The separation of T 1 values in biological tissues at lower field strengths is related to the different ratios of free water to bound water. As stated earlier as the Larmor frequency increases, bound water’s efficiency to return to ground state is lessened. As bound water’s effect upon T 1 is lessened, free water becomes a more dominating factor upon T 1 . Therefore, as field strength increases the ratio of bound water to free water become less significant and the T 1 values of the tissues becomes closer to the T 1 value of free water. Thus the T 1 values of different tissues converge at higher fields. The MDEFT T 1 contrast method uses RF preparation of longitudinal magnetization that improves T 1 contrast. The concept of the MDEFT method is that preceding the imaging sequence a 90 RF pulse is applied and immediately after this pulse de-phasing gradients are used. A user-defined time period elapses and an inversion 180 pulse is applied and again de-phasing gradients are applied. Another user-defined time period elapses before either spin-echo or gradient-echo imaging techniques are used. The T 1

PAGE 94

78 contrast mechanism for MDEFT is similar to that of inversion recovery (IR). The pulse diagrams for both MDEFT and IR sequences are shown in figures 5-1 and 5-2. Compared with IR MDEFT brings added benefits other than T 1 contrast. B 1 Homogeneity, Susceptibility, and In vivo MDEFT Imaging At the high frequencies used for the proposed research, the interaction between the B 1 field and the sample becomes important. Wave effects cause this interaction given that the effective wavelength of the B 1 field is comparable with the dimension of the sample. Assuming that the 1/4 wavelength being equal to or smaller than the dimension of the sample the magnetic and electric fields are tightly coupled, and these fields interact strongly with the structure of the sample through wave propagation processes. This interaction, distinct from, but related to, the spinB 1 interaction, gives rise to wave effects within the sample(50). This interaction degrades the B 1 field homogeneity and consequently image quality. At 8T, or 340 MHz for proton, the RF wavelength is approximately 10cm in distilled water. The RF wavelength will be even smaller in tissues giver that their permittivities are somewhat smaller than distilled water(50). The RF wavelength will continue to shorten from those measured at 8T to the field strengths on which this research was conducted. Very often high field imaging has both low T 1 contrast as well as B 1 in-homogeneity. For these cases T 1 preparation may be used to increase the T 1 contrast. Also, Norris et al. claim that the MDEFT pulse sequence not only provides T1 contrast but a reduction of B1 in-homogeneities(51,52). While high field imaging brings higher SNR and other benefits it is not without its drawbacks. Some experiments become difficult to complete at high fields due to increased susceptibility effects. As field strength increases the image artifacts due to B 0 variability increase accordingly. In vivo imaging at these higher fields becomes

PAGE 95

79 extremely difficult due to large B 0 variations caused by the sinus cavities and other susceptibility effects. Utilizing spin echo imaging techniques will minimize many of these artifacts. It is a goal of this proposed research to create a spin-echo based MDEFT preparation pulse sequence so high contrast in vivo experiments may be conducted. While several groups have addressed these issues at high fields there are extensive amounts of work to be done. While T 1 contrast methods have been developed at high fields they are not optimized and as of yet no rigorous comparison of these methods has been done. For example, while there have been several papers addressing MDEFT at high fields, inversion pulse sequences are hardly mentioned, and in those instances only superficial qualitative comparisons were made(51-54). Another area that needs to be addressed is the claims of B 1 homogeneity. While several papers have stated that MDEFT is less susceptible to B 1 variations little evidence is given to support this statement. While it is fairly obvious that a refocusing pre-pulse would have an effect on the B 1 field, some characterization of this effect would be extremely useful. In the literature reviewed no comparative images of MDEFT versus images with inferior B 1 homogeneity were shown. While one publication has done some numerical simulations, the results were not correlated with experimental results(52). To date no work regarding MDEFT has been on an in vivo animal model. The phantom work that has been done does not address the B 0 variability or B 1 variability stemming from in vivo samples. The in vivo human work(51,53,54) does not accurately address the susceptibility issues that are present in rodent models, mainly the proximity of sinus cavity and ear canal to tissues of interest. Optimum imaging parameters for a rodent model deviate from optimum human parameters, so these need to be explored.

PAGE 96

80 Also, no in vivo work has been done at the field strengths proposed. Therefore, characterization of high field issues of in vivo T 1 imaging will need to be accomplished.

PAGE 97

81 Figure 5-1: MDEFT pulse sequence with the two variable time delays used to control T 1 contrast.

PAGE 98

82 Figure 5-2: IR pulse sequence with the inversion time which is used to control T 1 contrast.

PAGE 99

CHAPTER 6 METHODS Introduction Before evaluation of the contrast performance of the pulse sequences of interest were performed several issues needed to be resolved. First, phantoms on which to perform the contrast evaluations needed to be created. This involved in vivo T 1 measurements on mice at all three field strengths and then, through an iterative process of doping ddH 2 0 (distilled de-ionized water) with copper sulfate, biologically representative phantoms were created. Also a simulation package was created to model SR, IR, and MDEFT pulse sequences in order to anticipate optimal settings prior to experimentation and to validate the performance of any results obtained. Finally preliminary studies uncovered discrepancies between simulations and experimental data. These discrepancies resulted from RF excitation in-homogeneities. This chapter explains the methodology resolving these three issues of biological representative T 1 phantoms construction, creation of a simulation package, and the characterization and correction of RF excitation in-homogeneities. In Vivo T 1 Measurements and the Construction of Biological Phantoms Since in vivo imaging will be one of the main applications of the T 1 enhancement techniques that are being developed, biologically matched T 1 phantoms were constructed. Therefore, T 1 measurements were made upon in vivo rodents at all field strengths of interest in order to determine the range of T1s encountered at these field strengths. Once the T 1 values of various brain structures were determined biologically representative T 1 83

PAGE 100

84 phantoms were created for each magnet. There are wide ranging T 1 values present in nervous tissue so that the biologically representative phantoms consist of an array of T 1 values spanning the in vivo results. These phantoms were then used to determine the T 1 contrast behavior of the pulse sequences of interest. Two techniques were employed to determine the T 1 values of neural tissue, and several animals were measured on each magnet to determine the range of T 1 values in the rodent brain. One method used a standard Bruker pulse sequence called multi-slice multi-echo variable TR (MSME_VTR) in order to make the T 1 measurements. It is a string of SE echo images with each subsequent image having a different TR. Figure 6-1 shows some sample in vivo images from the MSME_VTR sequence. The other method employed to determine in vivo T1 values was a localized spectroscopy technique employing STEAM acquisition. Similar to MSME_VTR the TR period was varied in order to sample the T1 growth. More accurate results were achieved using the MSME_VTR sequence and were subsequently used for T 1 curve fitting, because the MSME_VTR method was less sensitive to the motion of the subject. Using the MSME_VTR sequence white matter, gray matter, and CSF areas were segmented and T 1 fitting performed in these specific ROIs. Figure 6-2 is an example of segmented gray matter, and CSF areas, and figure 6-3 is an example of segmented white matter areas. The fitting of T 1 s was accomplished using the equation 6-1 below, (55), where S is signal, y 0 is a scaling constant, S 0 is initial signal, and TR is the repetition time. The T 1 fitting for the biologically representative T 1 phantoms was accomplished using the same method. Table 6-1 shows the results of in vivo T 1 imaging at 4.7T, 11T,

PAGE 101

85 and 17.6T. Figure 6-4 shows the trend of in vivo T 1 values of these structures versus field strength. 1100TTReSyS [6-1] The biologically representative phantoms were created using distilled de-ionized water, ddH2O, doped with different concentrations of copper sulfate. This variable amount of copper sulfate doping is how T 1 values were controlled. The first phantom created was made blindly not knowing what concentration of copper sulfate would produce what T 1 value. Due to this unknown, a wide range of concentrations was chosen in order to ascertain the relationship between copper sulfate concentration and T 1 at 11 Tesla. This phantom was also used for preliminary CNR curves using MDEFT, inversion, and non-contrast enhanced gradient echo pulse sequences at 11 Tesla. The phantom consisted of a 30mm vial filled with ddH2O with four small vials of copper sulfate doped ddH2O inserted into the larger vial. The concentrations of copper sulfate used were 0.5mM, 1mM, 2mM, and 4mM with T 1 values of 1547ms, 987ms, 577ms, and 322ms respectively. Figure 6-5 is an example MDEFT image of the T 1 phantom with wide ranging T 1 values. Using T 1 values from in-vivo rodent brain tissue and the T 1 values obtained from copper sulfate doped ddH 2 O the biologically representative T 1 phantoms were constructed. The procedure followed for the construction of these phantoms is as follows. First an existing T 1 phantom is scanned and the T 1 values determined. Using this information another T 1 phantom is constructed with copper sulfate concentrations altered in order to match the T 1 values obtained from the in vivo scans. This new phantom is then scanned and its T 1 values measured in order to ensure that it is

PAGE 102

86 biologically representative. Figure 6-6 illustrates the biologically representative 4.7 Tesla T 1 phantom including the T 1 values of each vial. Figure 6-7 and 6-8 illustrate the 11 Tesla T 1 and 17.6T phantoms with the T 1 values for each vial. Using this method of in vivo T 1 measurement, neural structure segmentation, T 1 quantification, and phantom construction a flexible T 1 modeling system has been created. These phantoms give the researcher the flexibility to model more complex systems and determine which experimental technique is best suited for T 1 contrast or quantification. Simulation Methods Matched With Results Mathematical simulations are used to confirm experimental results from contrast enhancement and to anticipate optimal settings prior to experimentation. The goal of these simulations is to test that the experimental results, and not to exactly model all aspects of the pulse sequence. For instance these experiments were designed to demonstrate T 1 contrast. Therefore, T 2 effects on the signal were not considered. Also noise was not simulated, and therefore simulations were scaled to match the experimental results. The simulations matched well with experimental results, which gave confidence that the experimental results were robust. Saturation recovery, Inversion Recovery, and MDEFT pulse sequences are modeled, and these simulations matching the 4.7T results are shown. Overall simulations matched well with experimental results. Slight discrepancies are mainly attributed to RF excitation in-homogeneities and errors in T 1 measurements. Figure 6-6 above shows the T 1 phantom used for these 4.7T experiments and simulations. The contrast comparisons evaluated here were ROIs 4vs5, which best represents white versus gray matter, ROIs 1vs5 which best represents white matter versus CSF, and ROIs 4vs7 which best represents gray matter versus contrast enhanced tissue.

PAGE 103

87 The modeling of the GE saturation recovery was straightforward because this is a very common sequence. Gradient echo saturation recovery required a steady state solution because low encoding tip angles were used for acquisition. With these low tip angles several phase encode steps pass before a steady state is achieved. Haccke et al. provided steady state solutions for saturation recovery(55). This steady state solution is seen in equation 6-2, where MzssA is the steady state solution and is the encoding tip angle. The solution for signal from saturation recovery is shown in 6-3, provided again by Haccke(55), where the steady state solution is multiplied to the signal calculation. These equations were used to model GE saturation recovery. Using these modeling equations the signal curves from two different T 1 values were simulated, and then these curves were subtracted to generate a contrast curve between areas with two different T 1 s. cosexp1exp111TTRTTRssAMz [6-2] cosexp1sinexp111TTRTTRssAMSz [6-3] The simulations matched well with experimental results at 4.7T. Figure 6-9 demonstrates the contrast performance of both experimental and simulated results. While the simulations and experimental results match well there are discrepancies, which can be explained by either small variations in the RF excitation profile or T 1 measurement error. Similar to the SR simulations the solution for IR_GE required proper steady state calculations in order to achieve a good simulation. Again the steady state calculations are required because a 22.5 encoding tip angle was used. Unlike the SR_GE, a steady state solution was not found in the literature. The steady state was solved for by simulating the

PAGE 104

88 magnetization following the inversion pulse and again after the encoding pulse, generalized equations found in Haacke et al(55). These simulations were iterative and were completed 128 times, once for each phase encode step. Once the steady state was reached the perpendicular magnetization was calculated following the encoding pulse. The steady state answer was used for the solution because the steady state was achieved quickly and dummy scans were used in all experiments. Equation 6-4 or [M z ] (n) represents the longitudinal magnetization present following the inversion pulse on the n th step. Equation 6-5 or [M z ] (n) represents the longitudinal magnetization present following the encoding pulse on the n th step. Equation 6-6 or steadyM represents the transverse magnetization present following the encoding pulse once steady state has been achieved. This series of solutions were used for all IR_GE contrast simulations. Again the simulations matched well with experimental results at 4.7T. Figure 6-10 demonstrates the contrast performance of both experimental and simulated results. Similar to the SR_GE the results and simulations match well, but slight discrepancies again most likely stem from variations in RF excitation and error in the T 1 measurements. As stated earlier this is a small effect, which is why the simulations and the experimental results match well. cos*exp11)(110TMMMnznz [6-4] cos*exp11)(10TTRMMMnznz [6-5] sin*exp11)(10TTRMMMsteadyzsteady [6-6]

PAGE 105

89 The simulation of MDEFT_GE was much easier than the other two simulations because it did not require a steady state solution. The MDEFT sequence is a driven equilibrium sequence, which ensures that the steady state is achieved for all phase encode steps. Lee et al. provided equation 6-7, which was used for all simulations(53). 2102exp21)(TMMz [6-7] Like both SR_GE and IR_GE the MDEFT_GE simulations matched well with experimental results. Figure 6-11 demonstrates the contrast performance of both experimental and simulated results for MDEFT_GE. Again the discrepancies between experimental and simulated results are small and are explained by the same mechanisms as SR_GE and IR_GE. The simulations accomplished that which was needed for the study; matching with experimental results demonstrating that the contrast provided by the magnetization preparation was following expectations. The simulations reproduced the experimental results well and will allow qualitative predictions of contrast performance. These simulations were used to help determine imaging settings for in vivo contrast experiments. Since time is not abundant for in vivo imaging optimal , TR and IT were determined and used as a starting place for in vivo contrast experiments at all field strengths used. Solutions to RF Excitation In-Homogeneities During the course of this work it was noticed that the RF excitation was not homogeneous for the higher field systems, 11T and 17.6T. As field strength increases it becomes more difficult to obtain homogeneous RF excitations due to coil effects and

PAGE 106

90 standing wave phenomenon(56,57). These RF excitation in-homogeneities manifest in alterations to the CNR and sometimes were quite dramatic. Figure 6-12 below demonstrates the RF in-homogeneity in the 17.6T biologically representative phantom. The top and bottom of figure 6-12 received a less intense RF excitation making ROI 1 followed by ROIs 3 and 4 less intense than ROIs 2 and 5, which were in the middle. ROI 1’s SNR curve was shifted lower because of this RF phenomenon generating contrast between other vials regardless of T 1 contrast. This problem manifested itself mainly in SR_GE acquisitions because the T 1 contrast decayed relatively quickly allowing contrast from non-uniform RF excitation to become obvious at the longer preparation times. This problem was fixed by scaling ROIs in areas of low RF excitation up to match ROIs in the areas of higher RF excitation. The 17.6T biologically representative phantom displayed the most dramatic effect from this phenomenon. Therefore the raw and RF intensity corrected SR_GE 17.6T results are presented here as an example of the magnitude of the problem and for the techniques employed to correct this problem. Figure 6-13 below demonstrates the magnitude of RF excitation differences in the 17.6T biologically representative T 1 phantom. Figure 6-13 are the SR_GE 17.6T results and this shows the signal versus TR for each ROI. At the longest TR period all the vials should essentially be fully relaxed and have the same value. It can easily be seen that this is not the case with each vial having a different plateau. As predicted from figure 6-12 above ROI 1’s SNR plateaus to the lowest value of all; ROIs 2, 5, and 6 have signals that plateau to the highest values; and finally ROIs 3 and 4 have signal in between these two extremes. In order to correct for the RF in-homogeneities each ROI was scaled up to the ROI with the most intense RF excitation forcing each ROI to plateau to the same value.

PAGE 107

91 Figure 6-14 below displays the CNR results from white versus gray matter comparison for SR_GE at 17.6T, the simulated results, and the RF intensity corrected results. The uncorrected CNR results do not follow the expected shape for SR_GE results where CNR is highest at short preparation times and then quickly decays to zero. Instead the CNR is high at short preparation times, quickly decays to zero, and finally re-grows to a constant CNR value. The white matter representative ROI has a shorter T 1 value but a less intense RF excitation. This gives the white matter ROI more SNR at early TR periods, but less SNR at the longer TR periods. The third TR employed is where the signal from the white and gray matter ROIs intersect after this third TR the contrast is dominated by the RF excitation profile. The simulations assume a constant RF excitation and have the expected shape for SR_GE results. The RF intensity corrected results almost exactly match the simulations demonstrating that the non-uniform RF excitation was the sole cause for the experimental results not matching simulations. The RF excitation profiles varied from system to system, but only demonstrated significant effects on the 11T and 17.6T systems. While this RF phenomenon can have a drastic effect upon SNR curves, but only the SR_GE data was significantly altered. The ability to remove the effect from the non-uniform RF profiles is possible by scaling areas according to their RF intensities. The data is presented without these RF intensity corrections in order to illustrate that the higher field systems are susceptible and that MDEFT_GE and IR_GE pulse sequences are less susceptible than SR_GE to these non-uniform RF profiles.

PAGE 108

92 Figure 6-1: MSME_VTR pulse sequence with TR periods of 4000, 700, and 250ms.

PAGE 109

93 Figure 6-2: This is an example of segmented gray matter and CSF areas. This mouse image was collected on the 17.6T magnet. Regions 1 and 2 are cortex gray matter areas. Regions 5 and 6 are Caudate-Putamen gray matter areas. Regions 3 and 4 are CSF areas.

PAGE 110

94 Figure 6-3: This is an example of segmented white matter areas. This mouse image was collected on the 11T magnet. Regions 1 and 4 are external capsule white matter areas. Regions 2 and 3 are corpus callosum white matter areas.

PAGE 111

95 Table 6-1: Results of in vivo T 1 imaging at 4.7T, 11T, and 17.6T. Rows shaded in white are white matter areas, rows shaded in gray are gray matter areas, and rows shaded in blue are CSF areas. 4.7T 4.7T 11T 11T 17.6T 17.6T T 1 (ms) (ms) T 1 (ms) (ms) T 1 (ms) (ms) External Capsule 1197 89 1596 95 2121 166 Corpus Callosum 1202 6 1680 23 1894 50 All White 1199 76 1617 90 2064 177 Cortex 1351 32 1815 47 2629 240 Caudate Putamen 1404 63 1855 75 2436 99 All Gray 1378 55 1835 61 2533 199 CSF 1948 14 2646 261 3550 476

PAGE 112

96 T1 versus Field Strength1000.001500.002000.002500.003000.003500.004681012141618Field Strength (Tesla)T1(ms) External Capsule Corpus Callosum All White Cortex Caudate-putamen All Gray CSF Figure 6-4: This graph displays the trend of T 1 in the structures of interest as field strength increases. While the error bars overlap several curves, the white matter, gray matter, and CSF areas do not overlap each other.

PAGE 113

97 Figure 6-5: T 1 phantom with wide ranging T 1 values. The top vial has the highest concentration of copper sulfate, 4mM, with a T 1 of 322ms with the right vial having the lowest concentration of copper sulfate, 0.5mM, with a T 1 of 1547ms. The settings used in the above image were a TR of 250ms, a TE of 5ms, a matrix size of 256x256, a slice thickness of 2mm, 1 average, and a time of 550ms.

PAGE 114

98 Figure 6-6: A MSME_VTR image of the 4.7T biologically representative T 1 phantom. The differences in T 1 between vials are apparent. Each ROI is labeled along with its corresponding T 1 value.

PAGE 115

99 Figure 6-7: A MSME_VTR image of the 11T biologically representative T 1 phantom. The differences in T 1 between vials are apparent. Each ROI is labeled along with its corresponding T 1 value.

PAGE 116

100 Figure 6-8: A MSME_VTR image of the 17.6T biologically representative T 1 phantom. The differences in T 1 between vials are apparent. Each ROI is labeled along with its corresponding T 1 value.

PAGE 117

101 4.7T GE Saturation Recovery vs simulation00.511.522.533.544.50100200300400500600700Tacq(s)CNR(a.u.) GE CNR(4-5) GE CNR(1-5) sim GE CNR(4-5) sim GE CNR(1-5) Figure 6-9: This figure gives the comparison of 4.7T GE SR experimental results versus simulations. The comparisons made were ROIs 4vs5, which best represents white versus gray matter and ROIs 1vs5 which best represents white matter versus CSF.

PAGE 118

102 4.7T GE Inversion Recovery vs simulations00.511.522.533.50100200300400500600700Tacq(s)CNR(a.u.) CNR(4-5) CNR(1-5) CNR(4-7) sim CNR(4-5) sim CNR(1-5) sim CNR(4-7) Figure 6-10: This figure gives the comparison of 4.7T GE IR experimental results versus simulations. The comparisons made were ROIs 4vs5, which best represents white versus gray matter, ROIs 1vs5 which best represents white matter versus CSF, and ROIs 4vs7 which best represents gray matter versus Gad enhanced tissue.

PAGE 119

103 4.7T MDEFT GE vs simulation00.511.522.533.5020040060080010001200Tacq(s)CNR(a.u.) GE CNR(4-5) GE CNR(1-5) GE CNR(4-7) sim CNR(4-5) sim CNR(1-5) sim CNR(4-7) Figure 6-11: This figure gives the comparison of 4.7T GE MDEFT experimental results versus simulations. The comparisons made were ROIs 4vs5, which best represents white versus gray matter, ROIs 1vs5 which best represents white matter versus CSF, and ROIs 4vs7 which best represents gray matter versus Gad enhanced tissue.

PAGE 120

104 Figure 6-12: This figure qualitatively illustrates the RF excitation profile in the biologically representative 17.6T phantom. The signal from the vials has been removed and only un-doped water is shown in this image. This was done to remove any effects that different T 1 s may have upon this image. The blue areas represent areas of the coil with less intense RF excitation, and the green areas represent areas with more intense RF excitation.

PAGE 121

105 17.6T SR_GE All ROIs0500001000001500002000002500003000003500004000004500000123456TRSignal Signal 1 Signal 2 Signal 3 (White) Signal 4 Signal 5 Signal 6 (Gray) Figure 6-13: This figure demonstrates the signal from each ROI of the 17.6T biologically representative phantom versus TR. It is apparent that the different ROIs plateau to different signal intensities. The yellow curve is the white matter representative ROI and the maroon curve is the gray matter representative ROI.

PAGE 122

106 Intensity Corrected Results W-G00.511.522.533.544.50100200300400500600700Tacq(s)CNR(a.u.) IC CNR(W-G) sim CNR(W-G) CNR(W-G) Figure 6-14: This figure shows the white versus gray matter CNR comparison. The blue curve is the uncorrected data, the yellow curve is the simulated data, and the purple curve is the RF intensity corrected data. It is apparent that once the RF intensity correction is applied that real data matches the simulated data.

PAGE 123

CHAPTER 7 SR, IR, AND MDEFT GRADIENT ECHO CNR PERFORMANCE Introduction The push to higher magnetic field strengths has necessitated the re-evaluation of standard MR protocols to account for changes in image contrast mechanisms. Of particular interest has been the degradation of T 1 relaxation contrast at high fields. As magnetic fields increase, tissue T 1 values tend to converge significantly reducing image contrast. Recently, several methods have been proposed to recover T 1 contrast at field strengths above 4T. Of these techniques, the modified driven equilibrium Fourier transform (MDEFT) preparation(52,53,58) has been proven to be especially useful in human neurological imaging because it provides T 1 contrast enhancement over a wide range of T 1 values. In this study, the MDEFT preparation is compared to standard saturation recovery (SR) techniques and inversion recovery (IR) preparation at three magnetic field strengths (4.7, 11 and 17.6 T) that are pertinent to high resolution, morphometric animal imaging. To assess the benefits and drawbacks of each T 1 preparation technique, physiologically relevant phantoms were constructed for the three field strengths. These phantoms cover a range of field dependant T 1 values matched to relaxation coefficients measured from the brains of normal C57BL6/J mice. The construction and evaluation of these phantoms was described in chapter 6. Particular attention was paid to gray and white matter contrast. Using these phantoms and the methods outlined in chapter 6, optimal preparation times were determined for the three T 1 contrast techniques at the three field strengths so as to provide optimal contrast 107

PAGE 124

108 enhancement over the given range of T 1 values. The contrast properties of SR_GE, IR_GE, and MDEFT_GE pulse sequences were evaluated over the field strengths of interest. 4.7, 11, and 17.6T magnets were employed to generate CNR curves of biologically representative T 1 phantoms. MRI Acquisition The MRI methods employed were fairly consistent from magnet to magnet. The MDEFT, IR, and GE acquisitions were imaged using a non-segmented gradient echo acquisition (NEX = 2; MTX = 128x128; TE = 5 ms; TR = 50 ms; = 22.5; FOV and slice thickness dependent on magnet system). Some parameters were optimized depending on the system in use. Detailed MRI acquisition parameters are listed with the results from each system. The MDEFT and IR sequences apply adiabatic hyperbolic secant pulses during the preparation period. To assess contrast enhancement, the time was incremented (0.05-5 sec) for MDEFT acquisitions, the inversion time (TI; 0.05-5 sec) was incremented for IR acquisitions and the TR was incremented for saturation recovery GE acquisitions. Data Analysis Using the images collected from each magnet ROIs were placed in each of the sample containers for each experiment. The mean signal from each ROI was recorded as a function of the total acquisition time. The total acquisition time is calculated differently for each pulse sequence. The total acquisition time, Tacq, for MDEFT_GE uses the following expression: Tacq = NEX *PE*(TR + 2*) where NEX is the number of averages and PE is the number of phase encodes. The Tacq for IR_GE uses this expression: Tacq = NEX *PE*(TR + IT). Finally the Tacq for SR_GE acquisitions uses

PAGE 125

109 this expression: Tacq = NEX *PE*(TR+). The signal to noise ratio (SNR) of each ROI was determined by the ratio of the mean signal of each ROI to the standard deviation of a noise ROI and then divided by the square root of Tacq. Dividing by the square root of Tacq was done to weight the shorter preparation times heavier than the longer preparation times because of the time gain they provide. For example, a preparation time that requires 1/4th the time as a longer preparation time may average four times as long and be equal in acquisition time. The signal increases with the root of averages therefore the SNR from each ROI was divided by the square root of Tacq in order to weight these different preparation times. After the SNR has been calculated and divided by the square root of Tacq the contrast to noise ratio (CNR) between two vials was determined by taking the absolute difference of the SNR of different ROIs. CNR curves are presented for each magnet and each pulse sequence of interest. These curves are used to determine the contrast performance of each pulse sequence over the T 1 s of interest. Results The results matched well with expectations giving a clear picture of the performance of T 1 preparation pulse sequences at high magnetic field strength, however there were some issues in collecting these data. For instance the higher field systems, 11T and 17.6T, showed significant RF profile non-uniformity that altered some of the SR_GE results significantly. Another issue was a console malfunction from the 11T, which caused the SR_GE results not to be collected, therefore the SR_GE results shown here are from a different imaging session. Also there are no 17.6T in vivo images showing T 1 enhancement due to difficulties imaging in vivo rodents on this system. Despite of these small issues the different pulse sequences showed a diverse T 1

PAGE 126

110 enhancement allowing for optimization of T 1 contrast under varied experimental conditions. The 4.7T magnet performed well and provided very stable and consistent results. This was shown with the simulations matching the results well in chapter 6. Presented here are the contrast performances of the three pulse sequences evaluated. Matching predictions the SR_GE provided excellent contrast at short preparation times, but the contrast quickly dropped off. Also matching predictions, the IR_GE provided good contrast peaking at an earlier preparation time than MDEFT and having slightly more CNR. Figure 7-1 below lists the T 1 s of each ROI in the T 1 representative phantom for 4.7T. The CNR performances of the three pulse sequences were evaluated using three biologically representative comparisons ROIs (4vs5), (1vs5), and (4vs7). The comparison of (4vs5) is the best representation of white versus gray matter, (1vs5) is the best representation of white matter versus CSF, and (4vs7) is the best representation of gray matter versus contrast-enhanced tissue. Figure 7-1 also demonstrates an example of the contrast performance of MDEFT_GE and IR_GE with a and IT time of 500ms. The acquisition parameters used were 2 averages, a matrix of 128 by 128, a TE of 5ms, a TR of 50ms, an encoding tip angle of 22.5, a FOV of 3 by 3cm, and a slice thickness of 2mm. Figure 7-2 below demonstrates the contrast performance of the three pulse sequences when comparing white versus gray matter representative T 1 values. The SR_GE CNR curve exhibits the most CNR of the three pulse sequences, but the CNR quickly drops and maintains near zero contrast for the longer magnetization preparation times. This is due to the saturation of the magnetization only being present at very short

PAGE 127

111 preparation times. The IR_GE CNR curve displays good contrast at relatively short preparation times and the contrast then decays to near zero at the longer preparation times. As expected the MDEFT_GE CNR curve is similar in shape to that of the IR_GE. The MDEFT_GE demonstrates approximately 80% the maximum contrast of the IR_GE CNR curve. While the contrast is slightly less than the maximum of IR the MDEFT curve maintains contrast over a wider range of preparation times, this being one of the advantages of the MDEFT sequence. The three pulse sequences displayed good white versus gray matter contrast but at different preparation times. This variety in performance allows for contrast between white and gray matter over a range in experimental conditions. Figure 7-3 below demonstrates the contrast performance of the three pulse sequences when comparing white matter versus CSF representative T 1 values. The three pulse sequences evaluated all exhibit increased CNR compared with the white versus gray matter CNR. This was due to the larger difference in the T 1 values between white matter and CSF than the white and gray matter T 1 s. The SR_GE CNR curve is similar to its white versus gray matter counterpart in figure 7-2, the main difference being more CNR between white matter and CSF. The CNR was also maintained for longer than the SR_GE CNR in white versus gray matter. The IR_GE CNR curve is also similar to the

PAGE 128

112 white versus gray matter curve in figure 7-2 above. The white matter versus CSF curve peaks at a slightly later preparation time and also has more CNR as to be expected. Similar to the other two pulse sequences the MDEFT_GE CNR curve is similar in shape to the white versus gray matter comparison. Again the CNR peaks at a slightly later preparation time and has significantly more CNR than the white versus gray matter comparison. The MDEFT_GE white matter versus CSF comparison demonstrates approximately 80% the maximum CNR of the IR curve, similar to the white versus gray matter comparison. The three pulse sequences behaved in a predictable manner displaying more CNR when the T 1 s compared have a larger difference. Figure 7-4 below demonstrates the contrast performance of the three pulse sequences when comparing gray matter versus contrast enhanced tissue representative T 1 values. Overall CNR was higher than seen with either the white versus gray matter comparison or the white matter versus CSF comparison. This gray matter versus contrast-enhanced comparison has the largest percentage difference in T 1 values, which gave it the most overall CNR. The SR_GE CNR curve is similar in shape to the other two SR_GE comparisons, but it achieves zero CNR quickly and then proceeds to settle in to a constant low CNR after that. This can be explained by a non-uniform RF profile described in chapter 6. The non-uniform RF profile imparts more signal to one of the two ROIs compared giving CNR at the later preparation times where the T 1 contrast has dissipated allowing the effects from this non-uniform RF profile to be seen. The gray matter versus contrast-enhanced tissue IR_GE CNR curve has a similar shape to the other two comparisons made. The differences are that this comparison has the largest CNR of the three IR_GE comparisons made and it achieved maximum CNR the earliest due to the

PAGE 129

113 contrast-enhanced representative T 1 being the shortest. The gray matter versus contrast-enhanced tissue MDEFT_GE CNR curve is also similar in shape to the other two comparisons made. This comparison gave the most CNR of the three comparisons made and achieved its maximum CNR earliest. The MDEFT_GE gray matter versus contrast-enhanced tissue comparison also demonstrated approximately 80% the maximum CNR of the IR_GE curve, similar to the other two comparisons. In vivo images were collected at 4.7T to demonstrate the contrast enhancement of SR_GE, IR_GE, and MDEFT_GE. Figure 7-5 below has in vivo rat images of the three pulse sequences at three different magnetization preparation times, 250ms, 375ms, and 500ms. MDEFT_GE exhibits the best image contrast of the three pulse sequences, while IR_GE exhibits the least image contrast. The MDEFT_GE most likely has the most CNR because of its T 1 contrast enhancement over a wide range of magnetization preparation times. The fact that the MDEFT_GE images display good contrast in all three images in figure 7-5 is evidence of this. The lack of good contrast for the IR_GE images was most likely caused by the relatively narrow range of preparation times, which provide good T 1 contrast. The SR_GE images have abundant signal but are lacking sufficient image contrast. The SR_GE images obtain the most image contrast at the shortest preparation times and the times used in these images were not short. These 4.7T images demonstrate the contrast performance was consistent between both in-vivo and phantom T 1 contrast experiments. The 11T magnet also performed well but some unique challenges presented themselves, which were not present at 4.7T. The 11T is over twice as strong as the 4.7T magnet at this higher field strength making it more difficult to obtain uniform RF

PAGE 130

114 excitations. This less uniform RF profile manifested itself primarily in the SR_GE contrast evaluations. As discussed earlier SR_GE achieves significant T 1 contrast at short TR times and minimal T 1 contrast at longer TR times. At these longer TR times the T 1 contrast has dissipated allowing the effect of this non-uniform RF profile to be seen. Another confounding factor the SR_GE data presented here was not collected during the same imaging session as the MDEFT_GE and IR_GE. The consistent imaging parameters and reproducibility of these experiments should generate similar results. Slightly different experimental settings such as different tuning and matching of the coil and different sample positioning will slightly alter the CNR between imaging sessions, but the curve shape of the SR_GE should not alter. Figure 7-6 below lists the T 1 s of each ROI in the T 1 representative phantom for 11T. For MDEFT_GE and IR_GE the contrast performance was characterized using comparisons made between ROIs (1vs2), (1vs4), (2vs8). The comparison of (1vs2) is the best representation of white versus gray matter, (1vs4) is the best representation of white matter versus CSF, and (2vs8) represents gray matter versus the longest T 1 of the phantom. For SR_GE similar comparisons were made during a different imaging session, these comparisons are ROIs (4vs7), (1vs7), (4vs9). The comparison of (4vs7) is the best representation of white versus gray matter, (1vs7) is the best representation of white matter versus CSF, and (4vs9) represents gray matter versus the longest T 1 of the phantom. ROI 7 representing white matter had a T 1 of 1.6s, ROI 4 representing gray matter had a T 1 of 1.8s, ROI 1 representing CSF had a T 1 of 2.6s, and ROI 9 had the longest T 1 of 2.8s. Figure 7-6 also demonstrates an example of the contrast performance of MDEFT_GE and IR_GE with a and IT time of 750ms. The acquisition parameters used were 1 average, a matrix of 128 by 128, a TE of 5ms, a TR

PAGE 131

115 of 50ms, an encoding tip angle of 22.5, a FOV of 3 by 3cm, and a slice thickness of 2mm. Figure 7-7 below demonstrates the contrast performance of the three pulse sequences when comparing white versus gray matter representative T 1 values. The SR_GE CNR curve exhibits slightly more CNR than the other two pulse sequences, and the CNR is less at the longer preparation times but does not get to zero. This comparison does not achieve zero CNR at an early preparation time and stay there at longer preparation times because of the non-uniform RF profile. The IR_GE CNR curve displays good contrast at relatively short preparation times and the contrast then decays at a slow rate. The CNR at the longest preparation times used in IR_GE has a contribution from the non-uniform RF profile. The evolution of T 1 contrast is significantly longer than SR_GE, which is why only at the longest preparation times, when T1 contrast has dissipated, does CNR from the non-uniform RF profile become a contributing factor. As expected the MDEFT_GE CNR curve is similar in shape to that of the IR_GE. The MDEFT_GE demonstrates approximately 85% the maximum contrast of the IR_GE CNR curve. The MDEFT_GE CNR curve has a similar non-uniform RF profile contribution at the longest preparation times to that of IR_GE. Again at the longest preparation times the T1 contrast ceases to dominate letting the RF profile become a significant contributing factor. The three pulse sequences displayed good white versus gray matter contrast but over different preparation times. The added complexity of the non-uniform RF profile only significantly affected the SR_GE curve. The contrast performance of the pulse sequences is shown in figure 7-8 below for white matter versus CSF representative T 1 values at 11T. IR_GE and MDEFT_GE TR = 2000 ms

PAGE 132

116 exhibited increased CNR compared with the white versus gray matter CNR. This was due to the larger difference in the T 1 values between white matter and CSF than the white and gray matter T 1 s. The SR_GE CNR has attenuated CNR compared with the other two pulse sequences. This is most likely from large differences in RF performance in the two vials compared. The IR_GE CNR curve is similar to the white versus gray matter curve in figure 7-7 above, but with significantly more CNR. This is to be expected due to the increase in the difference between the T 1 s compared here. The MDEFT_GE CNR curve is similar in shape to the MDEFT_GE white versus gray matter comparison. Again the CNR is significantly more than the white versus gray matter comparison. The MDEFT_GE white matter versus CSF comparison also demonstrates approximately 85% the maximum CNR of the IR curve, similar to the white versus gray matter comparison. The MDEFT_GE and IR_GE pulse sequences exhibited increased CNR as expected compared with the white versus gray matter CNR curve and also demonstrated that T 1 differences dominate the contrast. The SR_GE showed that it was more susceptible to the non-uniform RF profile than the other pulse sequences. Figure 7-9 below demonstrates the contrast performance of the three pulse sequences when comparing the biologically representative gray matter versus a long T 1 value. Overall CNR was similar to what was just shown between white matter and CSF. This was expected because the ratio between these T 1 s is similar to the ratio between white matter and CSF T 1 s. The SR_GE CNR curve is similar to the SR_GE curve of white matter versus CSF in that the CNR was quite low again most likely from the nonuniform RF profile. The IR_GE gray matter versus a long T 1 value CNR curve has a similar shape to the other two IR_GE comparisons made. The CNR from this IR_GE

PAGE 133

117 comparison fell off at a faster rate than the other two comparisons. This was most likely caused by the T 1 contrast dominating the CNR curves not allowing the RF profile to contribute to the CNR. The MDEFT_GE CNR curve is also similar in shape to the other two comparisons made. Similar to the IR_GE CNR curve the CNR fell off from maximum at a faster rate than the other two comparisons made. This MDEFT_GE comparison also demonstrated approximately 85% the maximum CNR of the IR_GE curve, similar to the other two comparisons. In vivo images were collected at 11T to demonstrate the contrast enhancement of MDEFT_GE compared with SR_GE. Figure 7-10 below displays in vivo mouse images of SR_GE and MDEFT_GE with the magnetization preparation time of 750ms for the SR_GE image and 1250ms for the MDEFT_GE image. As expected the MDEFT_GE exhibits better image contrast than the SR_GE although the SR_GE does have some contrast. The MDEFT_GE has demonstrated T 1 contrast enhancement over a wide range of magnetization preparation times. The good image contrast most likely resulted from this. The in vivo SR_GE images do not have the same issues as the biologically representative phantoms because while there may be RF non-uniformity the tissue boundaries lie adjacent to each other, i.e. white matter gray matter boundary. Over these short differences the RF profile is consistent therefore not reducing CNR between the two tissues. Decent image contrast can be seen for the SR_GE in figure 7-10. It is also apparent that the SR_GE image has significantly more SNR than the MDEFT_GE. These in vivo 11T images demonstrate the good CNR obtained with the phantom experiments can be obtained in an in vivo setting.

PAGE 134

118 The 17.6T magnet performed well and provided very stable and consistent results. Matching predictions the SR_GE provided excellent contrast at short preparation times, but the CNR quickly dropped off followed by a re-growth of CNR to a plateau. This unanticipated re-growth stems from a non-uniform RF profile, discussed in chapter 6. This unwanted RF non-uniformity at 17.6T is caused because the coil is linear not a quadrature coil as in the other two magnets and because higher operational frequencies reduces coil homogeneity(56). As expected this RF non-uniformity mainly affects the SR_GE images as it did in the other magnets because the T 1 contrast does not exist at longer preparation times. The CNR seen at these later times is solely from signal variations due to RF non-uniformity. The SR_GE CNR curves did give the predicted contrast at early preparation times. The IR_GE CNR curves performed as expected providing good contrast and peaking at an earlier preparation time than MDEFT. Figure 7-11 below lists the T 1 s of each ROI in the T 1 representative phantom for 17.6T. The CNR performance of the three pulse sequences was evaluated using three biologically representative comparisons of ROIs (3vs6), (3vs5), and (1vs6). The comparison of (3vs6) is the best representation of white versus gray matter, (3vs5) is the best representation of white matter versus CSF, and (1vs6) is the best representation of gray matter versus contrast-enhanced tissue. Figure 7-11 also demonstrates an example of the contrast performance of MDEFT_GE and IR_GE with a and IT time of 1000ms. The acquisition parameters used were 1 average, a matrix of 128 by 128, a TE of 5ms, a TR of 50ms, an encoding tip angle of 22.5, a FOV of 1.5 by 1.5cm, and a slice thickness of 0.5mm.

PAGE 135

119 Figure 7-12 below demonstrates the contrast performance of the three pulse sequences when comparing white versus gray matter representative T 1 values at 17.6T. The SR_GE CNR curve exhibits the most CNR of the three pulse sequences, but the CNR quickly drops and then returns to a significant level due to the non-uniform RF profile. If the CNR curves were intensity corrected this phenomenon would cease to be a problem and the SR_GE curves would return to their common shape. The IR_GE CNR curve displays good contrast at relatively short preparation times and the contrast then decays to near zero at the longer preparation times. As expected the MDEFT_GE CNR curve is similar in shape to that of the IR_GE but maintaining CNR at longer preparation times as expected. The MDEFT_GE demonstrates slightly more CNR than the maximum contrast of the IR_GE CNR curve. The three pulse sequences displayed good white versus gray matter contrast but at different preparation times. The flexibility in experimental design that these pulse sequences give allows researchers to select the appropriate pulse sequence for their needs. Figure 7-13 below demonstrates the contrast performance of the three pulse sequences when comparing white matter versus CSF representative T 1 values. The SR_GE CNR curve is similar to its white versus gray matter counterpart in figure 7-12 above, normally this comparison would display more CNR than the white versus gray matter comparison, but the T 1 that represents in vivo CSF is not much larger than the T 1 which represents gray matter. The biologically representative CSF T 1 should be longer but at this high field strength un-doped water at room temperature cannot match the T 1 of CSF at in vivo temperatures. The IR_GE CNR curve is also similar to the white versus gray matter curve in figure 7-12 above, but slightly more CNR is maintained at the longer

PAGE 136

120 preparation times. Similar to the other two pulse sequences the MDEFT_GE CNR curve is similar in shape to the white versus gray matter comparison. Again the CNR is slightly more maintained at the longer preparation times compared with the white versus gray matter curve. Here the MDEFT_GE white matter versus CSF comparison demonstrates approximately the same maximum CNR of the IR_GE curve. This most likely results from less variation in RF profile between the white matter and CSF representative vials. The three pulse sequences behaved in a predictable manner displaying similar CNR performance to the white versus gray matter comparisons because the T 1 s were not dramatically different in this comparison. Figure 7-14 below demonstrates the contrast performance of the three pulse sequences when comparing gray matter versus contrast enhanced tissue representative T 1 values. Overall CNR was higher than seen with either the white versus gray matter comparison or the white matter versus CSF comparison. This gray matter versus contrast-enhanced comparison has the largest difference in T 1 values, which gave it the most overall CNR. The SR_GE CNR curve is similar in shape to the other two SR_GE comparisons, but the CNR at longer preparation times from the non-uniform RF profile appears less dominant because of the high level of CNR at early preparation times. The gray matter versus contrast-enhanced tissue IR_GE CNR curve has a similar shape to the other two comparisons made, but this comparison showed dramatically more CNR than the other two comparisons and the CNR peaked at an earlier preparation time. This is due to the contrast-enhanced representative T 1 being the shortest. The gray matter versus contrast-enhanced tissue MDEFT_GE CNR curve is also similar in shape to the other two comparisons made. This comparison gave the most CNR of the three comparisons made

PAGE 137

121 and achieved its maximum CNR earliest. The MDEFT_GE gray matter versus contrast-enhanced tissue comparison also demonstrated approximately 80% the maximum CNR of the IR_GE curve, this was because the CNR in this comparison was so high in both the IR_GE and MDEFT_GE curves that differences in the RF profile are less dominant giving the IR_GE more CNR as expected. Discussion and Conclusion Overall the pulse sequences performed well giving good image contrast on all three systems. The contrast performance of the three pulse sequences varied from each other, which allows for optimal pulse sequence selection for different experimental conditions. The RF excitation was not uniform on all three systems and is expected for high field magnets. The SR_GE pulse sequence dealt with this RF non-uniformity poorly while the other two sequences sufficiently attenuated the effects. Here the performance of each pulse sequence is discussed and the experimental conditions of when each is appropriate are reviewed. The SR_GE performance across the different systems was quite varied. It was obvious that the SR_GE pulse sequence is the most sensitive to the RF profile characteristics. The SR_GE also preformed best at the lowest field strength employed, 4.7T. This was so because at lower field strengths RF excitations are more uniform than at the higher field strengths. The 17.6T RF profile was non-uniform for an additional reason other than wave phenomenon in that the coil was a linear coil not a quadrature coil. A quadrature coil can be described as two coils orthogonal to each other where the coils are designed to interact with each other as little as possible. Most times this means the RF profiles of the coils are averaged together giving a more uniform RF excitation than a linear coil. For these reasons the SR_GE results were inconsistent at both 11T and

PAGE 138

122 17.6T. As shown in chapter 6 if one removes the unwanted effects from the non-uniform RF profile the SR_GE results then become smooth and match with simulations. There are a few situations where SR_GE would be advantageous over the other T 1 preparation techniques. These situations require a homogenous RF excitation, extremely short TR times, and abundant signal. Short TR times can be difficult to achieve if multi-slice experiments are desired and at short TR times image artifacts are more likely from non-perfect refocusing pulses and improper crushing. Nevertheless, stable systems with well built coils and well written pulse sequences can provide quality SR_GE images with good image contrast. Unlike the SR_GE results the IR_GE performed consistently from system to system providing good image contrast in all situations. While the IR_GE was less affected by non-uniform RF profiles, this phenomenon had some effect at the longer preparation times on the high field magnets. As stated earlier the non-uniform RF excitation significantly affects image contrast when the T 1 contrast ceases to dominate. T 1 contrast was much longer lived in IR_GE than SR_GE because the T 1 preparation draws out the T 1 re-growth. Situations where the anticipated T 1 s are relatively well known and the RF excitation is not dramatically non-uniform are where IR_GE can be used and good image quality can be obtained. The good image contrast of IR_GE and the stability of the pulse sequence make IR_GE a good choice for experiments that require T 1 contrast. Similar to the IR_GE results the MDEFT_GE performed consistent from system to system and provided good image contrast in all experiments. MDEFT_GE was the least sensitive of the three pulse sequences to RF excitation non-uniformity. The T 1 contrast is

PAGE 139

123 longest lived using the MDEFT_GE pulse sequence, which minimizes the effect from the RF non-uniformity until the longest T 1 preparation time. The contrast is easier to predict using MDEFT_GE than the other two pulse sequences because of the resistance to non-uniform RF effects and the simple steady state solution. One application of MDEFT is that for experiments that are looking at quantification of T 1 contrast MDEFT_GE would be superior to IR_GE or SR_GE because it is less sensitive to experimental conditions and mainly dependant upon T 1 compared with the other techniques. Situations where MDEFT_GE is advantageous to the other two sequences are when the RF excitation is significantly non-uniform and when the T 1 s are not well known before the experiment. Experiments where T 1 contrast agents, gadolinium, are used the T 1 s are very difficult to predict because the amount of T 1 contrast agent that passes through the blood-brain barrier is not known. MDEFT_GE is optimal in these situations because of its broad contrast enhancement MDEFT provides good image contrast over a large range in T 1 values. A similar situation where MDEFT is advantageous is when contrast between more than two tissue types is desired. Again MDEFT_GE provides a broad contrast enhancement profile, which provides good image contrast to several T 1 groups simultaneously. The broad image contrast enhancement, pulse sequence stability, near the same maximum CNR to IR_GE, and simple steady state calculation make MDEFT_GE a good choice for many T 1 imaging applications and optimal in many high field situations. The push to higher magnetic field strengths provides many advantages but is not without certain drawbacks. The degradation of T 1 relaxation contrast at high fields has necessitated the evaluation of T 1 contrast techniques on these high field systems in order

PAGE 140

124 to appropriately deal with this reduction in T 1 contrast. The variety of contrast performance and their ability to deal with other high field imaging issues is apparent. The MDEFT_GE and IR_GE pulse sequences allow for the inconsistent RF profile to not be a dominant factor in image contrast. The SR_GE contrast performance at extremely short preparation times allows for fast imaging with good image contrast. The broad T 1 contrast profile of MDEFT_GE allows for uncertainty in T 1 values prior to imaging and gives contrast to multiple tissue types simultaneously. The diversity in the contrast performance of the three pulse sequences along with their varied susceptibility to other high field imaging issues allows for a large range in experimental conditions to be imaged with sufficient quality.

PAGE 141

125 Figure 7-1: This figure illustrates the 4.7T biologically representative T 1 phantom. The top figure labels each ROI with its corresponding T 1 values. The bottom left image is an IR_GE image and the bottom right image is an MDEFT_GE image both have a preparation time of 500ms.

PAGE 142

126 4.7T CNR(4-5) All Sequences00.20.40.60.811.20200400600800100012001400Tacq(s)CNR(a.u.) SR_CNR(4-5) IR_CNR(4-5) MDEFT_CNR(4-5) Figure 7-2: CNR performance of white versus gray matter representative T 1 values at 4.7T. The y-axis is CNR as defined on earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 143

127 4.7T CNR(1-5) All Sequences00.511.522.533.544.50200400600800100012001400Tacq(s)CNR(a.u.) SR_CNR(1-5) IR_CNR(1-5) MDEFT_CNR(1-5) Figure 7-3: CNR performance of white matter versus CSF representative T 1 values at 4.7T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 144

128 4.7T CNR(4-7) All Sequences00.511.522.533.50200400600800100012001400Tacq(s)CNR(a.u.) SR_CNR(4-7) IR_CNR(4-7) MDEFT_CNR(4-7) Figure 7-4: CNR performance of gray matter versus contrast enhanced tissue representative T 1 values at 4.7T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 145

129 Figure 7-5: This figure illustrates the contrast enhancement of the three pulse sequences at 4.7T. Top row of images : MDEFT_GE images with t times = 250, 375 and 500 ms from left to right. Middle row of images : SR_GE images with TR times = 250, 375 and 500 ms from left to right. Bottom row of images : IR_GE images with TI times = 250, 375 and 500 ms from left to right.

PAGE 146

130 Figure 7-6: This figure illustrates the 11T biologically representative T 1 phantom. The top figure labels each ROI with its corresponding T 1 values. The bottom left image is an IR_GE image and the bottom right image is an MDEFT_GE image both have a preparation time of 750ms.

PAGE 147

131 11T CNR(1-2) MDEFT and IR00.10.20.30.40.50.60.70.80200400600800100012001400Tacq(s)CNR(a.u.) MDEFT CNR(1-2) IR CNR(1-2) SR CNR(4-7) Figure 7-7: CNR performance of white versus gray matter representative T 1 values at 11T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 148

132 11T CNR(1-4) MDEFT and IR00.511.522.530200400600800100012001400Tacq(s)CNR(a.u. ) MDEFT CNR(1-4) IR CNR(1-4) SR CNR(1-7) Figure 7-8: CNR performance of white matter versus CSF representative T 1 values at 11T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 149

133 11T CNR(2-8) MDEFT and IR00.511.522.530200400600800100012001400Tacq(s)CNR(a.u.) MDEFT CNR(2-8) IR CNR(2-8) SR CNR(4-9) Figure 7-9: CNR performance of white matter representative T 1 value versus a long T 1 value at 11T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 150

134 Figure 7-10: Illustrates contrast enhancement from both the SR_GE preparation and the MDEFT_GE preparation. Left image: SR_GE image at TR = 750 ms. Right image: MDEFT_GE image at = 1250 ms

PAGE 151

135 Figure 7-11: This figure illustrates the 17.6T biologically representative T 1 phantom. The top figure labels each ROI with its corresponding T 1 values. The bottom left image is an IR_GE image and the bottom right image is an MDEFT_GE image both have a preparation time of 1000ms.

PAGE 152

136 17.6T CNR (3-6) All sequences00.20.40.60.811.21.41.61.80200400600800100012001400Tacq(s)CNR(a.u.) SR CNR(3-6) MDEFT CNR(3-6) IR CNR(3-6) Figure 7-12: CNR performance of white versus gray matter representative T 1 values at 17.6T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 153

137 17.6T CNR (5-3) All sequences00.20.40.60.811.20200400600800100012001400Tacq(s)CNR(a.u.) SR CNR(5-3) MDEFT CNR(5-3) IR CNR(5-3) Figure 7-13: CNR performance of white matter versus CSF representative T 1 values at 17.6T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 154

138 17.6T CNR (1-6) All sequences012345670200400600800100012001400Tacq(s)CNR(a.u.) SR CNR(1-6) MDEFT CNR(1-6) IR CNR(1-6) Figure 7-14: CNR performance of white matter versus contrast enhanced tissue representative T 1 values at 17.6T. The y-axis is CNR as defined earlier in this chapter, and the x-axis is Tacq also defined earlier in this chapter.

PAGE 155

CHAPTER 8 INTER-MAGNET COMPARISONS Introduction The contrast performance of a pulse sequence does not stay constant with increasing field strength. As shown earlier the T 1 s of in vivo neural tissue change with field strength. This work attempts to give a picture of what to expect of T 1 contrast trends and magnitude when taking T 1 contrast experiments to higher field strengths. Using the three pulse sequences of interest this chapter demonstrates the field dependence of T 1 contrast over three high field systems. While qualitative changes are easy to compare from system to system, it is difficult to compare quantitative pulse sequence contrast performance between systems because many aspects including hardware and hardware performance are not constant. Therefore relative contrast, RC, is used for comparisons instead of CNR because the noise cannot be directly compared between magnets. A ratio of the difference of signal between two regions and the sum of these regions is used to determine RC. This removes SNR dependence so T 1 contrast between systems may be compared regardless of aspects, which ca not be held constant from system to system. The construction and evaluation of these phantoms was described in chapter 6, and the MRI methods employed and the raw data are the same as was used in chapter 7. Here it is just presented as a function of field strength instead of a function of pulse sequence. The general MR methods employed were 2 averages, a matrix size of 128 by 128, a TE of 5ms, a 22.5 encoding tip angle, and the FOV was dependent on the system used. All 139

PAGE 156

140 scans employed a non-segmented GE acquisition. To assess contrast enhancement, the time was incremented (0.05-5 sec) for MDEFT acquisitions, the inversion time (TI; 0.05-5 sec) was incremented for IR acquisitions, and the TR was incremented for saturation recovery GE acquisitions. Detailed MRI acquisition parameters for each system can be found in chapter 7. Using the images collected from each magnet, ROIs were placed in each of the sample containers for each experiment. Unlike the within magnet comparisons the mean signal from each ROI was not divided by the square root of the total acquisition time. SNR is dependent on acquisition time and because noise cannot be held constant from system to system the acquisition time was not considered when calculating RC. The RC was calculated for each comparison using equation 8-1 listed below. Where S a is the mean signal from ROI a and S b is the mean signal from ROI b . Using this equation RC has a range of zero to one. RC was plotted against the Tacqs for white versus gray matter and white matter versus CSF for the three pulse sequences. The biologically representative T 1 s used for each tissue type across the field strengths is listed in table 8-1 below, a description on how these values were obtained is available in chapter 6. Gray matter versus contrast-enhanced tissue was not compared between systems because T 1 values for contrast-enhanced tissue could not be standardized therefore comparisons across magnets were not possible. These curves helped determine the contrast performance of each pulse sequence across field strengths using biologically representative T 1 phantoms. babaSSSSRC [8-1]

PAGE 157

141 Results It was originally anticipated that the biological T 1 values would lengthen and converge with increasing field strength. Chapter 6, which covered the determination of the in vivo T 1 values, shows this to be partly true. The T 1 values did become longer with increasing field strength but the T 1 s did not appreciably converge. The absolute difference in T 1 s between biological tissues remained similar but the percent difference decreased with field strength. Given information obtained form these studies, RC should slightly decrease as field strength increases. The RC did slightly decrease on the higher field systems as predicted because of the difference between biological T 1 values remained relatively consistent and the non-uniform RF excitation on high field systems. SR_GE contrast performance was dominated by the RF excitation profile at the higher field strengths. At 4.7T the SR_GE RC curve had the expected shape of maximum contrast at short preparation times and losing contrast at the longer preparation times. The other two magnets did not display this expected shape because the non-uniform RF profile dominated the contrast, especially at the longer preparation times where more signal means a greater contrast between different regions of the coil. This was also magnified by not weighting RC by Tacq, therefore not amplifying the shorter preparation times for their ability to average longer than the longer preparation times. As expected from the results in chapter 7, SR_GE behaved in a predictable manner at low field strengths but at higher field strengths its performance was dominated by the RF profile of the coil. Figure 8-1 below demonstrates the RC performance of SR_GE comparing white versus gray matter representative T 1 values on all three magnets. The SR_GE RC curve for the 4.7T system exhibits its maximum RC at short preparation times. The 11T curve

PAGE 158

142 is not similar to the 4.7T curve displaying low RC at the shortest preparation times and subsequently increasing RC at the longer preparation times. This is solely due to the non-uniform RF excitation between these regions of the coil. The 17.6T SR_GE RC curve exhibits high RC at the shorter preparation times then a reduction of contrast as the contrast from the saturation diminishes and a subsequent increase in RC when the signal becomes great enough to show the RF excitation differences between these ROIs. The dependence of SR_GE on the RF homogeneity limits the functionality of this pulse sequence at the higher field strengths. Figure 8-2 below demonstrates the RC performance of SR_GE comparing white matter versus CSF representative T 1 values on all three magnets. Similar to the white versus gray matter comparison, the SR_GE RC curve for the 4.7T system exhibits its maximum RC at short preparation times. The 4.7T white matter versus CSF provides more RC than the white versus gray matter comparison. The added RC originates in increased separation of T 1 s between white matter and CSF when compared with white versus gray matter. The 11T and 17.6T curve displays similar characteristics. At early preparation times both were dominated by T 1 contrast then the RC decayed only to return at the longer preparation times. This again happened because the signal became large enough to show the RF excitation differences between these ROIs. Both the 11T and 17.6T curves display more RC at the earlier preparation times than the white versus gray matter comparison. This is also due to the increased separation in T 1 values, but this increased separation does not compensate for the non-uniform RF profile at the longer preparation times. The white matter versus CSF comparison provided increased separation in T 1 values subsequently providing increased RC at the earlier preparation

PAGE 159

143 times, but the SR_GE sensitivity to the RF characteristics dominated the longer preparation times on the higher field systems. The RC for IR_GE was not dominated by the RF excitation profile as it was for SR_GE. Figure 8-3 below demonstrates the contrast performance of white versus gray matter representative T 1 s using IR_GE for the three magnets. The 4.7T RC curve demonstrates the smoothest shape with the maximum amount of RC. This smoothness was anticipated because of the RF uniformity at low fields but surprisingly the white versus gray matter RC should not be the maximum at 4.7T and will be discussed later. The 11T IR_GE RC curve demonstrates the least maximum RC of the three systems as expected because the 11T exhibited the smallest ratio of white to gray matter T 1 s. The 11T curve is also not very smooth, an artifact of using RC. The signal from both the white matter and gray matter representative vials was smooth but the RC between the vials did not vary much and was quite low across the preparation times used. RC may be relatively high at short preparation times where signal was low because RC is a ratio and independent of the amount of signal. If longer preparation times were sampled the RC curve would appear to be smoother because these longer times would have reduced RC. The 17.6T IR_GE RC curve, while not as smooth as the 4.7T IR_GE white versus gray matter comparison, is sufficiently smoother than the 11T curve. The white matter representative ROI exhibited a 14 percent increase in RF excitation compared with the gray matter representative ROI. At 17.6T the T 1 s between white and gray matter were sufficiently separated that this dominated the RC not the RF excitation profile. The 17.6T curve also demonstrates a larger maximum RC than the 11T, but smaller maximum RC than the 4.7T. RC simulations showed that the 17.6T maximum RC should be larger

PAGE 160

144 than both the 4.7T and 11T maximums. However, CNR simulations with the noise held constant between magnets showed that the 4.7T and 17.6T have near identical maximum CNR with the 11T being the lowest. The 14 percent reduction of the white matter representative ROI compared with the gray matter ROI at 17.6T was the cause of the lower maximum RC seen in the 17.6T RC curve compared with the 4.7T curve. While it is counter intuitive that the 17.6T should exhibit increased RC compared with 4.7T and 11T it should be noted that the accuracy of our in vivo T 1 measurements was inferior at 17.6T compared with 4.7T and 11T, and CNR simulations showed that 17.6T and 4.7T had similar maximum CNR. Figure 8-4 below demonstrates the contrast performance of white matter versus CSF representative T 1 s using IR_GE for the three magnets. The three systems performance using the white matter versus CSF comparison shows the expected trend of less RC at the higher field strengths and at the longer preparation times the RC drops at a slower rate on the 11T and 17.6T systems because the T 1 s are significantly longer at high fields. Similar to the white versus gray matter comparison the white matter versus CSF comparison at 4.7T displays the maximum amount of RC of the three systems and exhibits a predictable and smooth shape. This performance was expected because the 4.7T T 1 s should give the most RC compared to the other systems and the RF excitation at 4.7T is smooth. While the 11T IR_GE RC curve is significantly smoother than the white versus gray matter comparison and demonstrates more RC, the RF profile did have a small effect. The white matter representative vial was located in a more intense area of the coil receiving more signal than the CSF representative vial. The result of this was a small amount of RC added over the whole RC curve. The RC curve shape is smooth and

PAGE 161

145 similar to the expected shape because the white matter representative ROI should have more signal because of the shorter T 1 value, therefore the RF profile added RC to the curve without changing the shape. If the RF excitations of these two vials had been switched the curve would not be smooth. Nonetheless the 11T IR_GE curve demonstrates the expected performance of reduced maximum RC and a slower reduction of RC at the longer preparation times compared with the 4.7T curve. The 17.6T IR_GE RC curve follows expectations with its maximum RC the least of the three systems and a slow rate of decay of RC at the longer preparation times. The maximum RC of the white matter versus CSF curve should be slightly higher than reported here because the 17.6T CSF representative T 1 value was lower than the recorded in vivo 17.6T CSF value. Un-doped ddH 2 O exhibited the longest T 1 using the 17.6T biologically representative phantom but was shorter than the CSF T 1 measured in vivo. Only heating of the phantom would allow un-doped ddH2O to achieve the in vivo CSF T 1 measurement, but heating would drastically change the T 1 s of the other vials. Heating of a sample increases T 1 periods by decreasing the correlation times. This makes it more difficult to undergo a transition, which relaxes the spin to equilibrium. However even with exact biologically representative T 1 s, the 17.6T would still display the least maximum RC of the systems. Although CNR was not used to demonstrate contrast performance between systems CNR gives the curve shape one would see experimentally. Therefore the white matter versus CSF comparison of IR_GE across field strengths using CNR instead of RC is shown below in figure 8-5. As stated CNR cannot be compared from system to system, therefore the curves from the three systems are normalized to one. This figure demonstrates the qualitative performance of IR_GE curve shape over three field

PAGE 162

146 strengths. Also the curves from the three systems were RF intensity corrected using the procedure explained in chapter 6. This was done to demonstrate the CNR curve shape of the three systems ignoring the effects from non-uniform RF excitation profiles. As expected the 4.7T curve’s CNR maximizes the earliest with the 11T maximizing next and the 17.6T last, due to increasing T 1 s at higher field. The 4.7T CNR curve’s CNR decays at a more rapid rate than the higher field systems, also due to the increased T 1 s at higher field. Figure 8-5 provides a qualitative comparison of CNR curve shape of the three systems employed in this study. Similar to IR_GE the RF excitation profiles did not dominate the performance of MDEFT_GE. The RC performance of MDEFT_GE comparing white versus gray matter representative T 1 s is shown in figure 8-6 below for all systems. The 4.7T RC curve demonstrates the smoothest shape with its maximum amount of RC between the other systems. This smoothness was anticipated because of the RF uniformity at low fields and the maximum RC at 4.7T was larger than 11T and smaller than the 17.6T as simulations predicted. Similar to the 11T IR_GE, the 11T MDEFT_GE RC curve demonstrates the least maximum RC of the three systems as expected because the 11T has the smallest ratio of white to gray matter T 1 s. The 11T MDEFT_GE curve was smoother than the 11T IR_GE curve because over the preparation times sampled the RC was larger for the MDEFT_GE. The 17.6T MDEFT_GE RC curve is relatively smooth and its maximum RC is the largest of the three systems as simulations predicted. Again, the 17.6T white matter representative ROI exhibited a 14 percent increase in RF excitation compared with the gray matter representative ROI, however this difference did not reduce the maximum RC of the 17.6T MDEFT_GE curve below the 4.7T MDEFT_GE curve as happened with

PAGE 163

147 the 17.6T IR_GE curve compared with the 4.7T IR_GE curve. Similar to the IR_GE simulations of CNR, the 4.7T and 17.6T MDEFT_GE had similar maximum CNRs while 11T had the lowest maximum CNR. As stated earlier RC was not weighted by acquisition time, the 17.6T maximum RC resides at a longer acquisition time than the 4.7T. If acquisition time were a weighting factor this would cause the 4.7T to have a larger maximum RC than the 17.6T. Figure 8-7 below demonstrates the contrast performance of white matter versus CSF biologically representative T 1 s using MDEFT_GE for the three systems. Similar to IR_GE the contrast performance of white matter versus CSF using MDEFT_GE showed the expected trend of less RC at the higher field strengths. The RC also decayed at a slower rate for the 11T and 17.6T at the longer preparation times because of the longer T 1 s at higher fields. The 4.7T MDEFT_GE exhibits the maximum amount of RC compared with the other systems as expected. The smoothness of the RC curve shape was also expected because the RF excitation at 4.7T is uniform. The 11T MDEFT_GE RC curve also follows expectations with its maximum RC less than the 4.7T MDEFT_GE curve and more than the 17.6T MDEFT_GE curve. The 11T white matter representative vial was located in a more intense area of the coil receiving more signal than the CSF representative vial. Similar to the 11T IR_GE RC curve this resulted in a small amount of RC added over the whole curve. The 17.6T MDEFT_GE RC curve follows expectations with its maximum RC the least of the three systems and a slow rate of decay of RC at the longer preparation times. As stated earlier the CSF representative T 1 value was lower than the measured in vivo value resulting in reduced RC across the whole curve. Simulations using the exact 17.6T CSF biological T 1 value still give the

PAGE 164

148 17.6T MDEFT_GE curve the least maximum RC of the systems. The 17.6T MDEFT_GE RC curve also peaks at a longer preparation time than the other two systems because of the longer T 1 s at 17.6T. Figure 8-8 below demonstrates the qualitative performance of MDEFT_GE CNR curve shape over three field strengths. As was done for IR_GE the CNR curves are normalized to one and all curves are intensity corrected to remove effects from the RF intensity profile. As expected the 4.7T curve’s CNR maximizes the earliest with the 11T maximizing next and the 17.6T last, this due to increasing T 1 s at higher field. Also expected the 4.7T CNR curve’s CNR decays at a more rapid rate than the higher field systems this was also due to the increased T 1 s at higher field. Discussion and Conclusion While it is difficult to compare T 1 contrast performance across systems the experimental data presented here provides a picture of this performance at different field strengths. RC was chosen to remove noise from the comparison because the differences in the hardware and many other considerations make it very difficult to scale noise from system to system. Also the calculation of RC is based on a ratio and this drastically changes the shape of the curve compared with the CNR curves seen in chapter 7. Thus, normalized CNR curves were shown to demonstrate the practical curve shape changes with field strength. Also confounding the evaluation of T 1 contrast versus field strength was the decreasing RF homogeneity with increasing field strength. MDEFT_GE and IR_GE were less susceptible to the RF excitation in-homogeneities but SR_GE was very sensitive making T 1 contrast comparisons across systems difficult. T 1 contrast imaging at high field becomes increasingly difficult but T 1 contrast is still obtainable, thus not taking away from the other benefits that high field imaging provides.

PAGE 165

149 The performance of the three pulse sequences of interest were quite varied but each demonstrated dramatic changes with increasing field strength. As shown in figures 8-1 and 8-2 SR_GE’s RC performance was dominated by the RF excitation profile. SR_GE at 11T and 17.6T did not follow the expected trend of high contrast at short preparation times and low contrast at long preparation times. Instead the non-uniform RF profile dominated the contrast especially at the longer preparation times where more signal means a greater contrast between different regions of the coil. IR_GE performed more consistently with the non-uniform RF profile only dominating the 11T white versus gray matter comparison. IR_GE generally started at a low RC then quickly rising to achieve maximum relatively early. Once maximum RC was reached it decayed at a slower rate for the higher field systems. The 4.7T IR_GE exhibited the highest maximum RC for all three systems for both the white versus gray matter comparison and the white matter versus CSF comparison. The CNR curve shape shown in figure 8-5 above shows the trend of CNR peaking at longer preparation times as field strength increases and the slower decay of CNR for the higher field magnets. MDEFT_GE performed the most consistent of the three pulse sequences evaluated with smooth curves for all comparisons. Similar to IR_GE RC started low at short preparation times quickly maximizing then decaying away at the longer preparation times. The RC peaked at slightly longer preparation times for MDEFT_GE than IR_GE. Also the MDEFT_GE at 17.6T maximized later than MDEFT_GE for the other two systems because of the longer T 1 s at 17.6T. Also like IR_GE once maximum RC was achieved the RC decayed at a slower rate for the higher field systems. The MDEFT_GE CNR curve shape shown in figure 8-8 above demonstrates the CNR peaking at longer preparation times as field strength

PAGE 166

150 increases and a slower decay from maximum CNR was also seen for the 11T and 17.6T systems. Both the amount and dynamics of T 1 contrast are affected by field strength. At the lower field strengths the T 1 s of biological tissue are relatively short and the ratio between tissue types is relatively large giving high T 1 contrast with the short T 1 values providing short scan times. At higher field strengths these T 1 s lengthen and the ratio between them decreases causing maximum T 1 contrast to diminish and making scan times longer. High field systems partially compensate for reduced contrast by the added SNR from these systems thus increasing CNR, but the lower field systems may average more to bring up SNR because of the shorter acquisition times. A uniform RF excitation profile also becomes more difficult to obtain at high field strength limiting the choices for obtaining T 1 contrast at high field. There are obstacles to high quality T 1 contrast imaging on high field systems, but with knowledge about these obstacles and specific techniques that deal with these issues quality T 1 contrast imaging may still be obtained.

PAGE 167

151 Table 8-1: This table lists the biologically representative T 1 s used for each system for white matter, gray matter and CSF tissue types. Tissue Type 4.7T 11T 11T SR_GE 17.6T White Matter T 1 1.08s 1.61s 1.56s 1.99s Gray Matter T 1 1.29s 1.88s 1.81s 2.53s CSF T 1 1.88s 2.61s 2.62s 2.57s

PAGE 168

152 SR_GE (W-G) All Magnets00.010.020.030.040.050.060.070.080.090200400600800100012001400Tacq(s)RC 4.7T (W-G) 11T (W-G) 17.6T (W-G) Figure 8-1: This figure demonstrates the RC performance of white versus gray matter for SR_GE at all field strengths. It is apparent that for both the 11T and 17.6T the RF excitation profile dominated RC performance.

PAGE 169

153 SR_GE (W-V) All Magnets00.020.040.060.080.10.120.140.160200400600800100012001400Tacq(s)RC 4.7T (W-CSF) 11T (W-CSF) 17.6T (W-CSF) Figure 8-2: This figure demonstrates the RC performance of white matter versus CSF for SR_GE at all field strengths. Again both the 11T and 17.6T demonstrate that the RF excitation profile dominated their RC performance.

PAGE 170

154 IR_GE (W-G) All Magnets00.020.040.060.080.10.120100200300400500600700Tacq(s)RC 4.7T (W-G) 11T (W-G) 17.6T (W-G) Figure 8-3: This figure demonstrates the RC performance of white matter versus gray matter for IR_GE at all field strengths. IR_GE was not as affected by the non-uniform RF profile as SR_GE only the 11T curve demonstrates a significant effect.

PAGE 171

155 IR_GE (W-CSF) All Magnets00.050.10.150.20.250.30.350100200300400500600700Tacq(s)RC 4.7T (W-CSF) 11T (W-CSF) 17.6T (W-CSF) Figure 8-4: This figure demonstrates the RC performance of white matter versus CSF for IR_GE at all field strengths.

PAGE 172

156 IR_GE Field Dependence W vs CSF00.20.40.60.811.20100200300400500600700Tacq(s)CNR(normalized) 4.7 (W-CSF) 11T (W-CSF) 17.6 (W-CSF) Figure 8-5: This figure demonstrates the qualitative CNR curve shape of IR_GE for all three magnets using the white matter versus CSF comparison. All curves are normalized to 1 because CNR cannot be directly compared across systems.

PAGE 173

157 MDEFT (W-G) All Magnets00.020.040.060.080.10.120.140.160.180.20200400600800100012001400Tacq(s)RC 4.7T (W-G) 11T (W-G) 17.6T (W-G) Figure 8-6: This figure demonstrates the RC performance of white matter versus gray matter for MDEFT_GE at all field strengths.

PAGE 174

158 MDEFT (W-CSF) All Magnets00.050.10.150.20.250.30.350.40.450.50200400600800100012001400Tacq(s)RC 4.7T (W-CSF) 11T (W-CSF) 17.6T (W-CSF) Figure 8-7: This figure demonstrates the RC performance of white matter versus CSF for MDEFT_GE at all field strengths.

PAGE 175

159 MDEFT_GE IC field Dependence W vs CSF00.20.40.60.811.20200400600800100012001400Tacq(s)CNR(normalized) 4.7 (W-CSF) 11T (W-CSF) 17.6 (W-CSF) Figure 8-8: This figure demonstrates the qualitative CNR curve shape of MDEFT_GE for all three magnets using the white matter versus CSF comparison. All curves are normalized to 1 because CNR cannot be directly compared across systems.

PAGE 176

CHAPTER 9 SE BASED T 1 IMAGING Introduction The contrast benefits from T 1 preparation have been shown in chapter 7, but only associated with GE experiments. There are many applications where GE experiments are not optimal and a SE version of the pulse sequence should be used. Many applications of SE involve in vivo imaging due to its insensitivity to susceptibility artifacts compared with GE imaging. SE imaging also provides inherently more signal than GE imaging which is an advantage. Another benefit of SE imaging is that the different contrast mechanisms associated with SE are easier to separate. The contrast is also easier to predict because it is less dependant upon experimental conditions, such as sample positioning and shimming. Some of the reasons not to use SE are the minimum TR is longer than that of GE and diffusion effects can become more pronounced in SE imaging. Because of the advantages of SE imaging a MDEFT_SE pulse sequence was evaluated. The contrast benefits shown with the GE studies were duplicated with SE experiments. With this information one can apply either SE or GE imaging to a particular experiment based upon which technique is best suited for the procedure and not based upon T 1 enhancement performance. As the need for SNR increases and technology advances, the strength of imaging magnets has risen significantly. While high field imaging brings greater SNR and other benefits it also brings increased susceptibility artifacts. In vivo imaging at these higher fields becomes difficult due to large B 0 variations caused by the sinus cavities and other 160

PAGE 177

161 susceptibility effects. Any sample that causes B 0 variability will also be difficult to image at high field strengths. Utilizing spin echo imaging techniques will minimize many of these artifacts. The research presented here shows the contrast performance of MDEFT_SE and IR_SE pulse sequences. The reduction of susceptibility artifacts by using a SE pulse sequence is also demonstrated. SE pulse sequences inherently provide more SNR than their GE counterparts. This arises from the refocusing aspect of the 180 pulse compared with the gradient refocusing used in GE. When a SE sequence is used the signal decreases from excitation to the time of echo dependant upon T 2 . While in a GE sequence the signal decreases from excitation to the time of echo dependant upon T 2 *, which always has a faster decay. This faster decay stems from local B 0 in-homogeneities and non-optimal shimming. When using SE these in-homogeneities are refocused while they are not for GE. In cases where signal is not abundant, SE imaging may be a more prudent choice. SE experiments also have the advantage of contrast simplicity compared with GE experiments. For basic SE experiments T 1 and T 2 play the dominant role in determining image contrast. For GE experiments, T 1 and T 2 * are the dominant contrast mechanisms, with T 2 * being more complex than its counterpart T 2 . T 2 * is a combination of T 2 which is a molecular process and with variations of B 0 throughout the sample. Variations of B 0 may be caused by local macroscopic or microscopic susceptibility changes in the sample, and non-optimal shimming. For GE sequences it may become difficult to differentiate these contrast mechanisms apart, especially when some may change during the experiment, for example if the sample/subject moves. For these reasons image contrast is

PAGE 178

162 easier to predict with a SE experiment because T 2 is more static from experiment to experiment than is T 2 *. Spin echo experiments have a larger minimum repetition time and echo time than its gradient echo counterpart. This occurs for two reasons, first the time requirement for the 180 pulse is longer than that of a refocusing gradient and secondly the time gap between the 90 and 180 pulses is repeated before the echo occurs. Typical minimum repetition time and echo time achieved using a 2000s sinc3 pulse spin echo experiment would be approximately 26ms and 6.5ms respectively with the equipment used for these experiments. Gradient echo experiments typically have minimum repetition time and echo time of approximately 9ms and 3ms respectively using the same equipment. These minimum TR and TE times are dependent upon maximum RF amplifier power, maximum gradient power, and bandwidth, as well as rise and fall times of the gradients. These minimum times become major factors with extremely short T 1 , T 2 , or T 2 * times, and for 3D imaging. Spin echo experiments have the disadvantage of pronounced effects from water diffusion. For SE imaging the rephrasing lobe of the readout gradient is typically right after the 90 pulse, which is quite separated from the readout gradient, allowing for large diffusion effects. The second main contributing factor to unwanted diffusion effects stems from the increased minimum TE time in SE imaging. This extra time allows spins to migrate farther from their origin, thus increasing diffusion effects. A main application of ultra high field imaging is MR microscopy. MR microscopy utilizes much stronger gradients than standard clinical imaging, diffusion’s effect upon the MR signal is proportional to the square of gradient strength, causing the diffusion effects to increase in

PAGE 179

163 these situations. Presented here are SE versions of the pulse sequences evaluated in chapters 7 and 8. The construction and evaluation of these phantoms was described in chapter 6, as well as analysis techniques. Results As expected the SE experiments conducted in this study reveal similar contrast behavior to that of the GE results. The contrast behavior matched well with all T 1 preparation pulse sequences employed; saturation recovery, MDEFT, and IR. While the contrast performance was similar between GE and SE the overall CNR of the SE results were significantly higher; this can be explained by the added SNR in SE experiments. All experiments were conducted on the 4.7T magnet. Figure 9-1 is an MDEFT SE image of the phantom used to generate the contrast curves displayed in this chapter. The comparisons looked at here are vials 4vs5 and 4vs7, vial 4 has a T 1 of 1.29s, vial 5 T 1 of 1.08s, and vial 7 T 1 of 0.64s. The 4vs5 comparison is most similar to a white matter versus gray matter comparison and 4vs7 is most similar to a gray matter versus contrast-enhanced tissue. With the good matching of the SE results with the GE results the contrast benefits observed with the GE sequences are retained and can subsequently be applied to situations where SE imaging provides benefits over GE imaging. Figure 9-2 below demonstrates why a SE version of the MDEFT pulse sequence is advantageous. One can easily see the poor image quality in the GE image and the improvement in the SE image. This does not mean that GE images cannot produce sufficient image quality at these field strengths, but images showing good in vivo MDEFT_GE contrast in chapter 7 were chosen specifically because little distortion was apparent on those images.

PAGE 180

164 SR_SE contrast curves were not similar to their GE counter parts as predicted. The modified MDEFT_SE pulse sequence was used in favor of traditional SR_SE imaging techniques. This was done in order to maintain the highest level of consistency between the images. The MDEFT_SE sequence had the 90 and 180 adiabatic pulses turned off and the times were changed in order to alter the TR time. This was repeated with the MDEFT_GE sequence. The SR_SE experiments used a matrix of 128x128, a FOV of 3x3cm, a 2mm slice thickness, a TR of 50ms, a TE of 6.5ms, and the times were varied in order to produce the contrast curves. The SR_GE experiments used similar settings except the TE was 5ms and the tip angle was 22.5 instead of 90. The same TR times were used to generate the contrast curves. Figure 9-3 shows the GE and SE results for the saturation recovery imaging. One can see that the curves do not match well with the SE imaging having superior CNR due to the extra signal in these experiments. It is apparent that the SE contrast curves maintain contrast at longer preparation times than do their GE counterparts. Both the added CNR of the SR_SE and different curve shapes can be attributed to the 90 encoding tip angle used in the SE experiments compared with the 22.5 tip angle used in the GE experiments. With a large degree tip angle saturation will be maintained at longer TR times than for smaller tip angles. A 22.5 tip angle was used for all GE experiments throughout the entire study to keep all GE experiments consistent. The 22.5 tip angle was chosen because most GE experiments employ very short TR periods therefore a smaller tip angle is used for SNR reasons; this is particularly so for segmentation applications. Saturation recovery imaging may be completed with either GE or SE encoding without sacrificing the inherent T 1 contrast, but different preparation times for each pulse sequence need to be employed to maximize the CNR.

PAGE 181

165 Unlike the saturation recovery results the IR_SE results match well with the GE results. Again the MDEFT_SE sequence was modified to produce the IR_SE sequence. The 90 adiabatic preparation pulse was turned off and the first time was eliminated for both SE and GE imaging. The IR_SE experiments used the same imaging parameters as the SR_SE experiments; similarly, the IR_GE experiments had the same imaging parameters as the SR_GE experiments. Identical IR times were used for both GE and SE experiments to generate the contrast curves. Figure 9-4 shows the GE and SE results for the inversion imaging. One can see the curves match generally well with the SE imaging having superior CNR due to the extra signal in these experiments. The IR_SE CNR curves peaked slightly earlier than their GE counterparts, which is due to the amount of magnetization present at the time of the inversion. The SE experiments utilized a 180 inversion pulse followed by a 90 pulse where the GE experiments used a 22.5 pulse instead of the 90 pulse used in SE. Once the steady state was achieved this resulted in less magnetization at time of inversion for the IR_SE experiments. With less magnetization at time of inversion, the CNR curves of the IR_SE experiments should peak at a shorter inversion time than their GE counterparts. As stated earlier, the 22.5 tip angle was chosen because typical GE experiments utilize shorter pulse repetitions therefore smaller tip angles are employed for SNR considerations. The 22.5 tip angle was selected for this study in order to keep all GE experiments consistent. Inversion recovery imaging may be completed with either GE or SE encoding without sacrificing the inherent T 1 contrast provided by the T 1 preparation. While the SE and GE experiments matched well using saturation recovery and inversion recovery the MDEFT_SE results matched almost identically with the GE

PAGE 182

166 results. Figure 9-5 shows the GE and SE results for the MDEFT imaging. The MDEFT_SE experiments had the same imaging parameters as the MDEFT_GE experiments except for the encoding tip angle, which was 22.5 for GE and 90 for SE. The same times were used for both GE and SE experiments to generate the contrast curves. The magnitude of the CNR curves is again larger in the SE experiments but the shape of the MDEFT_SE and the MDEFT_GE curves are quite well matched. The SE and GE MDEFT experiments match better than the saturation recovery and inversion recovery experiments due to the driven equilibrium aspect of the MDEFT T 1 preparation. The driven equilibrium forces the steady state immediately and is not dependant upon experimental parameters such as the encoding tip angle, TR, or TE time. Therefore, the different tip angles used in the SE and GE sequences would have a minimal effect upon the steady state so the shape of the curves will match well. Conclusion This work showed that SE encoding does not diminish the contrast provided by the T 1 preparation. The dynamic range of the contrast benefits of T 1 preparation over the several field strengths using GE imaging was also shown. The 4.7T work demonstrated that SE encoding does not significantly alter this contrast compared with the GE encoding. While T 1 values change with field strength, SE encoding should have similar contrast performance compared with GE regardless of field strength. Therefore these experiments were not completed on other systems. The SE experiments were successful because they demonstrated that the SE pulse sequences maintained the desired contrast achieved with the GE pulse sequences. The SE and GE versions of saturation recovery and inversion recovery pulse sequences matched quite well and the only discrepancy

PAGE 183

167 between the shapes of these curves was due to the different encoding tip angles used. The MDEFT_SE pulse sequence almost perfectly matched with the MDEFT_GE sequence. As explained earlier this was due to the driven equilibrium aspect of the MDEFT T 1 preparation. Given these results SE encoding may be used instead of GE encoding without loss of T 1 contrast.

PAGE 184

168 Figure 9-1: This is a MDEFT_SE scan on the 4.7T system using the biologically representative T 1 phantom. A time of 0.5s was used, TR 50ms, TE 6.5ms, 1 average, FOV 3x3cm, 2mm slice thickness, 128 2 , SW 50kHz.

PAGE 185

169 Figure 9-2: This is an example of the advantages of SE imaging over GE imaging. These images were collected on the 11T. The same slice was used and it is apparent that the SE image on the left has superior image quality. The SE image was collected with a matrix of (256x128), a FOV of (4x2cm), a slice thickness of 0.7mm, a TR of 500ms, and a TE of 12ms. The GE image was collected with a matrix of (128x128), a FOV of (3x2cm), a slice thickness of 0.7mm, a TR of 25ms, and a TE of 7.5ms.

PAGE 186

170 4.7T GE vs SE saturation recovery00.511.522.533.544.550100200300400500600Tacq(s)CNR(a.u.) GE CNR(4-5) GE CNR(1-5) GE CNR(4-7) SE CNR(4-5) SE CNR(1-5) SE CNR(4-7) Figure 9-3: This figure demonstrates the contrast performance of both saturation recovery SE and GE pulse sequences at 4.7 T. Both experiments varied TR to generate contrast and used the following parameters: matrix of (128x128), a FOV of (3x3cm), and a slice thickness of 2mm. Only TE was different between the two with the SE having a TE of 6.5ms and the GE having a TE of 5ms.

PAGE 187

171 4.7T GE vs SE Inversion Recovery01234560100200300400500600700Tacq(s)CNR(a.u.) GE CNR(4-5) GE CNR(1-5) GE CNR(4-7) SE CNR(4-5) SE CNR(1-5) SE CNR(4-7) Figure 9-4: This figure demonstrates the contrast performance of both inversion recovery SE and GE pulse sequences at 4.7 T. Both experiments varied inversion time to generate contrast and used the following parameters: matrix of (128x128), a FOV of (3x3cm), and a slice thickness of 2mm. Only TE was different between the two with the SE having a TE of 6.5ms and the GE having a TE of 5ms.

PAGE 188

172 4.7T GE vs SE MDEFT00.511.522.533.50200400600800100012001400Tacq(s)CNR(a.u.) GE CNR(4-5) GE CNR(1-5) GE CNR(4-7) SE CNR(4-5) SE CNR(1-5) SE CNR(4-7) Figure 9-5: This figure demonstrates the contrast performance of both MDEFT SE and GE pulse sequences at 4.7 T. Both experiments varied times to generate contrast and used the following parameters: matrix of (128x128), a FOV of (3x3cm), and a slice thickness of 2mm. Only TE was different between the two with the SE having a TE of 6.5ms and the GE having a TE of 5ms.

PAGE 189

CHAPTER 10 T 1 CONTRAST ENHANCED SE AND GE FAST IMAGING METHODS Introduction The time allotment of T 1 preparation comprises the bulk of the TR period. Many T 1 contrast studies could not be completed without segmentation. Presented here are the contrast performances of segmented MDEFT_GE and IR_GE as well as the contrast performances of MDEFT_RARE and IR_RARE compared with their non-segmented counterparts. The reason for segmenting T 1 contrast studies is for the increased speed factor gained by the segmentation process. In vivo T 1 contrast studies often cannot be completed without segmentation, this is particularly so if other studies are planned for the same imaging session. Segmentation is achieved in both RARE and segmented-GE by acquiring multiple phase encodes after the T 1 preparation period. Using this technique the time dominating T 1 preparation period is not repeated for the additional phase encodes substantially reducing acquisition time. This research shows that with the scan time reduced by one-fourth, one can still achieve good contrast for T 1 s in the biological range. Drawbacks to segmenting are less T 1 contrast than with no segmentation and that T 2 or T 2 * contrasts are also included in the image. It is shown that by utilizing both RARE and segmented-GE overall image quality and image contrast are maintained while acquisition times are greatly reduced. As previously discussed, the main application of segmentation is the time bonus achieved. RARE, rapid acquisition relaxation enhanced, imaging is a string of spin echoes following a 90 excitation(59). Due to the time requirement of a spin echo RARE 173

PAGE 190

174 encoding seldom has an echo train longer than 10-16. If a large RARE factor is used then SNR is sacrificed, especially when short T 2 s are present. Therefore a large RARE factor is not optimal in poor SNR conditions. For animal imaging RARE factors of four thru eight are commonly seen giving a scan time of approximately one-fourth to one-eighth as long as SE. Segmented-GE can achieve shorter acquisition times than RARE. GE segmentation is simply GE imaging following T 1 preparation, one can place any number of GE phase encodes following T 1 preparation sacrificing mainly contrast and depending on the TR and T 1 s involved SNR. For animal imaging at high field, sufficient image quality can be seen with all levels of GE segmentation, but in order to maintain T 1 contrast significant GE segmentation must be avoided. Similar to RARE four thru sixteen GE phase encodes following T 1 preparation are common giving a scan time of approximately one-fourth to one-sixteenth that of non-segmented GE. For most in vivo T 1 applications segmentation needs to be employed if any other imaging studies are desired. There are several tradeoffs associated with segmentation but if dealt with acceptable image quality with sufficient contrast can be achieved. The most compelling tradeoff with segmentation is the loss of T 1 contrast. The mechanism of contrast loss in RARE experiments is that the magnetization difference present due to the T 1 preparation is attenuated by T 2 . Therefore RARE studies where there is a long echo train and the T 2 s are short T 1 contrast will quickly be diminished. The mechanism of T 1 contrast loss due to GE segmentation arises from the reduction of longitudinal magnetization in the echoes following the T 1 preparation. If large encoding tip angles are used the first echo in the GE chain has all the T 1 contrast, if small tip angles are used the contrast is less at the first

PAGE 191

175 phase encode but the contrast persists further down the chain. Using less GE phase encodes per T 1 preparation can mitigate this T 1 contrast loss. Another tradeoff when segmenting is the combining of difference contrast mechanisms. When segmentation is used one introduces either T 2 or T 2 * contrast depending on whether RARE or segmented-GE is used. In RARE the phase encodes that are further removed from the T 1 preparation have a longer TE, which introduces T 2 contrast. In segmented-GE the phase encodes which are further removed from the T 1 preparation have the same T 2 * weighting but much reduced T 1 weighting. These effects give different phase encodes different contrast mechanisms making the whole image a combination of these contrasts. In situations where T 1 contrast is preferred the combination is undesired, but in certain situations the combination of the contrasts may provide superior image contrast. Either way this effect from segmentation makes image contrast more difficult to predict. Another major tradeoff due to segmentation is SNR loss. The first phase encode following a T 1 preparation has the most signal with each subsequent encode losing signal. This loss may be compensated for with signal averaging, but this diminishes the time bonus that segmentation provides. The SNR loss due to segmentation is more pronounced in segmented-GE imaging where there is increased saturation of the signal further along the echo train, therefore the more segmentation used the more the SNR is reduced. The SNR loss in RARE imaging is significantly less than in segmented GE if a short TE time is employed. Another tradeoff when segmenting is the phase encode artifacts associated with inconsistent phase amplitudes. The reduced signal in each subsequent echo following a T 1 preparation is the cause of inconsistent phase amplitudes. This inconsistency manifests itself in a fashion similar to a motion artifact. Certain phase

PAGE 192

176 sampling schemes minimize this effect by keeping the phase encodes near the center of k-space similar in amplitude. Also due to the inconsistent phase encode amplitudes the steady state solution is more complex. This complexity makes the contrast harder to predict making simulations more difficult to create and T 1 curve fitting from segmented experiments difficult. When employing RARE imaging there are issues when imperfect 90 and 180 pulses are employed. When using RARE encoding the effect from imperfect pulses is amplified compared with SE because the pulses are cumulative. This is more common when an experiment is multi-slice and the RF profile is not consistent from slice to slice. This effect also makes contrast harder to predict, and makes the steady state solution more complex. Presented here are comparisons of the fast imaging techniques RARE and segmented GE with SE and GE non-segmented acquisitions. The biologically representative phantoms described in chapter 6 are again employed to evaluate the performance of these fast imaging techniques. Results The contrast results for segmented MDEFT_GE, MDEFT_RARE, segmented IR_GE, and IR_RARE at 17.6T are presented here. As anticipated the segmentation results were similar to the non-segmented results. The 17.6T in vivo T 1 representative phantom was utilized in evaluating the contrast performance of segmented T 1 preparation. Figure 10-1 is an image of this phantom. The comparisons looked at here are vials 3vs6, 3vs5, and 6vs1. Vial 3 has a T 1 of 1.99s, vial 6 2.53s, vial 5 2.57s, and vial 1 0.97s. The 3vs6 comparison is most similar to a white matter versus gray matter comparison, the 3vs5 is most similar to a white matter versus CSF, and the 6vs1 is most similar to gray matter versus contrast enhanced tissue. The Bruker developed MDEFT and segmented IR_GE pulse sequences were used for the segmented-GE contrast

PAGE 193

177 evaluations. The MDEFT_RARE pulse sequence that was created for SE experiments was used for the MDEFT_RARE contrast evaluations and modified for the IR_RARE contrast evaluations. Segmented IR_GE contrast curves matched well with their non-segmented GE counter parts. The standard Bruker IR sequence was utilized with the inversion time being altered to generate the curves. These experiments were completed again with the segmentation turned on to generate the segmentation results. The settings used were a matrix of 128x128, a bandwidth of 19kHz, a TR of 50ms, a TE of 5ms, 1 average, FOV 1.5cm by 1.5cm, and a slice thickness of 0.5mm. The inversion times used were 50, 125, 250, 500, 750, 1000, 2000, 3000, 4000, and 5000ms. For the segmented experiments there was an echo train of four after each T 1 preparation. Figure 10-2 shows the segmented-GE and non-segmented results for IR imaging. One can see the curves match generally well with the non-segmented data having superior CNR due to the magnetization preparation for each phase encode instead of every four encodes. This data shows that IR_GE imaging may be used with or without segmentation. The gain in time which segmentation provides is mitigated by the loss in CNR. Similar to the segmented IR_GE results the segmented MDEFT_GE contrast curves matched well with the non-segmented MDEFT_GE. Both segmented and non-segmented experiments altered time to generate the contrast curves. The settings used were a matrix of 128x128, a bandwidth of 19kHz, a TR of 50ms, a TE of 5ms, 1 average, FOV 1.5cm by 1.5cm, and a slice thickness of 0.5mm. The times used were 50, 125, 250, 500, 750, 1000, 2000, 3000, 4000, and 5000ms. All segmentation experiments utilized a segmentation factor of four. The GE segmented and non-segmented results for

PAGE 194

178 MDEFT imaging are shown in figure 10-3. One can see the curves match well, with the non-segmented data having superior CNR due to the magnetization preparation for each phase encode instead of every four encodes. This data shows that MDEFT imaging may be used with or without segmentation. Similar to the segmented IR_GE experiments the gain in imaging speed which segmentation provides is counteracted by a loss in CNR. IR_RARE contrast curves matched well with their non-segmented SE counter parts. The homemade MDEFT_RARE pulse sequence was customized to generate the IR_SE data with the inversion time being altered to generate the curves. These experiments were completed again with the RARE factor set at four for the segmentation results. The settings used were a matrix of 128x128, a bandwidth of 19kHz, a TR of 50ms, a TE of 10ms, 1 average, FOV 1.5cm by 1.5cm, and a slice thickness of 0.5mm. The inversion times used were 50, 125, 250, 500, 750, 1000, 2000, 3000, 4000, and 5000ms. Figure 10-4 shows the RARE and non-segmented SE results for IR imaging. One can see the RARE curves match well with the non-segmented data, but the RARE results have slightly more CNR than the SE counterparts. This was caused by lower noise measurements during the IR_RARE acquisition this caused an increase in CNR. The T 2 values between all vials were quite similar, approximately 450ms, and the TE was short, 10ms, therefore the CNR between IR_SE and IR_RARE should be near equal. This is so because the attenuation of the T 1 contrast during a 40ms IR_RARE echo train is negligible when the T 2 s of both ROIs are over 400ms. This data shows that IR imaging may be used with or without RARE segmentation. For a RARE factor of four the contrast was not reduced significantly as it was in the segmented-GE. The time gain provided by RARE in these circumstances is advantageous considering the similar

PAGE 195

179 contrast performance to non-segmented data. However, in situations where the T 2 s are significantly shorter a large RARE factor may cause dramatic contrast reduction. MDEFT_RARE contrast curves matched well with their non-segmented SE counter parts. A modified RARE sequence was used to create the MDEFT_RARE sequence and then was utilized to generate the contrast curves. The times were altered in order to generate the different contrasts and the experiments were completed with a RARE factor of one for the SE experiments and a RARE factor of four for the segmented experiments. The settings used were a matrix of 128x128, a bandwidth of 19kHz, a TR of 50ms, a TE of 10ms, 1 average, FOV 1.5cm by 1.5cm, and a slice thickness of 0.5mm. The times used were 50, 125, 250, 500, 750, 1000, 2000, 3000, 4000, and 5000ms. Figure 10-5 shows the RARE and non-segmented SE results for MDEFT imaging. One can see the RARE results match well with the non-segmented SE data. Similar to the IR_RARE the MDEFT_RARE results have slightly more CNR than the non-segmented MDEFT_SE results, for similar reasons as the IR_RARE results. This data shows that MDEFT imaging may be used with or without segmentation. Similar to the IR_RARE the time gain provided by MDEFT_RARE is advantageous considering the similar contrast performance to the non-segmented data. Consistent with predictions both the segmented-GE and RARE matched well with their respective non-segmented counterparts. The only result, which was unexpected, was the slight increase in CNR for the RARE results. It was anticipated that the SE would have a slight edge in CNR compared with RARE imaging. This was most likely due to random variation between experiments. The segmented and non-segmented results matched so well partially because the segmentation was set at a reasonable level

PAGE 196

180 of four phase encodes for every T 1 preparation. This is a common segmentation level resulting in a time reduction of one-fourth, this segmentation factor was kept consistent between both segmented-GE and RARE. As larger amounts of segmentation are used the results from segmented data should deviate further from non-segmented data. As discussed previously, segmentation causes inconsistent phase amplitudes, which may lead to image artifacts in the phase direction. Using centered phase sampling, which smoothes the amplitude from one line of k-space to another, compensated for this potential image artifact. Centered phase sampling has maximum signal for the lines passing through the center of k-space and minimum signal for the lines passing through the outer edge of k-space with gradual steps in between. Segmentation experiments were not completed on the 4.7T and 11T magnets because the performance of segmentation is not significantly affected by field strength. Therefore this work needs only to be completed on one magnet to show its effectiveness. Discussion Some of the tradeoffs given up by segmentation for the increase in speed cannot be modeled by these phantom experiments. These phantoms were created to model in vivo T 1 values for the magnets used, but were not matched on T 2 values. The pulse sequences being evaluated are dominated by T 1 contrast; which is why simulations fit so well with experimental results. Nevertheless T 2 still plays a role and with segmentation, especially RARE, its contribution while not dominant increases. The tradeoff of CNR due to segmentation was shown with the segmented-GE results, but not so for the RARE results. The loss of CNR in the RARE will present itself for in vivo experiments where the T 2 s are shorter than the approximate T 2 of 450ms of the phantoms. Biological T 2 s at 17.6T

PAGE 197

181 are approximately 30ms with these reduced T 2 s the loss of contrast will become more apparent because of the loss of signal following the first phase encode in the echo train. Earlier it was discussed that when using segmentation techniques the image contrast becomes more complex. In vivo experiments will exhibit more complexity than these experiments because the T 2 and the T 2 * are shorter and are not as similar from vial to vial thus having a more pronounced effect. GE-segmentation will not compound the different contrast mechanisms as much as RARE because the TE period during the echo train is kept short therefore T 2 * contrast will remain low. RARE imaging will have significant added T 2 contrast when the T 2 s become shorter and not similar to each other. With a segmentation factor of four these effects are not likely to be too significant, but if larger factors are used then it may become more apparent. The SNR loss due to GE-segmentation has been shown compared with non-segmented data, but the magnitude of loss will not change much when T 2 *s are shortened. Regardless of segmentation use, the loss of signal from T 2 * decay will be similar because the same TE period is used. The SNR will be more dramatically reduced in RARE because it is more dependent upon T 2 values of the sample than segmented-GE. In RARE as the echo train following the T 1 preparation pulse gets longer, more segmentation, the effective T 2 of the experiment gets longer. Therefore, the biological T 2 values of approximately 30ms at 17.6T will have a more dramatic effect of SNR loss than the approximate 450ms T 2 seen with the phantom. As discussed the bulk of the acquisition time of T 1 contrast experiments arises from the T 1 preparation. In vivo T 1 contrast studies often cannot be completed without segmentation, this is particularly so if other studies are planned for the same imaging session. Segmentation provides the means to maintain T 1 contrast and to also reduce the

PAGE 198

182 acquisition time. The drawbacks to segmenting are that there is less T 1 contrast than without, especially so in segmented-GE, that T 2 or T 2 * contrasts are also included in the image contrast, and that SNR is reduced. The segmented results presented here match very well with the non-segmented results. The biologically representative T 1 phantom matched well with in vivo T 1 values, but was not matched on biological T 2 or T 2 * values. This lack of biological T 2 and T 2 * representation does not affect the matching of the segmented versus the non-segmented results. The effect it does have is that some of the marginal T 1 contrast changes one would see in vivo due to segmentation are not modeled with these experiments, specifically the loss SNR due to RARE imaging because biological T 2 s are shorter than the experimental conditions. This is not a major shortcoming of this work because the marginal SNR loss due to RARE imaging employing biological T 2 values does not have a dominant effect upon overall contrast. This research shows that with the scan time reduced by one-fourth one can still achieve good image contrast and image quality for T 1 s in the biological range, using both segmented-GE and RARE imaging.

PAGE 199

183 Figure 10-1: This is a MDEFT_SE scan with the phantom used for the curves. A time of 1s was used, TR 50ms, TE 10ms, 1 average, FOV 1.5x1.5cm, 0.5mm slice thickness, matrix of 128x128, and SW 20kHz.

PAGE 200

184 IR_GE seg vs noseg00.511.522.533.50100200300400500600700Tacq (s)CNR (a.u.) noseg_CNR(3-6) noseg_CNR(3-5) noseg_CNR(6-1) seg_CNR(3-6) seg_CNR(5-3) seg_CNR(1-6) Figure 10-2: This figure demonstrates the contrast performance of both segmented and non-segmented IR_GE imaging at 17.6T. Both experiments varied IT to generate contrast and used the following parameters: matrix of (128x128), a FOV of (1.5x1.5cm), and a slice thickness of 0.5mm.

PAGE 201

185 MDEFT_GE seg vs noseg00.511.522.533.50200400600800100012001400Tacq (s)CNR (a.u.) noseg_CNR(3-6) noseg_CNR(5-3) noseg_CNR(1-6) seg_CNR(3-6) seg_CNR(5-3) seg_CNR(1-6) Figure 10-3: This figure demonstrates the contrast performance of both segmented and non-segmented MDEFT_GE imaging at 17.6T. Both experiments varied times to generate contrast and used the following parameters: matrix of (128x128), a FOV of (1.5x1.5cm), and a slice thickness of 0.5mm.

PAGE 202

186 IR_SE vs IR_rare00.511.522.533.544.50100200300400500600700Tacq (s)CNR (a.u.) SE_CNR(3-6) SE_CNR(5-3) SE_CNR(1-6) rare_CNR(3-6) rare_CNR(5-3) rare_CNR(1-6) Figure 10-4: This figure demonstrates the contrast performance of both RARE and non-segmented IR_SE imaging at 17.6T. Both experiments varied IT to generate contrast and used the following parameters: matrix of (128x128), a FOV of (1.5x1.5cm), and a slice thickness of 0.5mm.

PAGE 203

187 MDEFT_SE vs MDEFT_Rare00.511.522.533.50200400600800100012001400Tacq (s)CNR (a.u.) SE_CNR(3-6) SE_CNR(5-3) SE_CNR(1-6) rare_CNR(3-6) rare_CNR(5-3) rare_CNR(1-6) Figure 10-5: This figure demonstrates the contrast performance of both segmented and non-segmented MDEFT_GE imaging at 17.6T. Both experiments varied times to generate contrast and used the following parameters: matrix of (128x128), a FOV of (1.5x1.5cm), and a slice thickness of 0.5mm.

PAGE 204

CHAPTER 11 HIGH FIELD T 1 CONTRAST OPTIMIZATION: CONCLUSION As previously discussed the diminishing T 1 contrast as field strength increases becomes a limiting factor for many applications. Considering this along with the many other high field issues there is much work to be done in order to make the transition to higher field strengths. This work has dealt with some of the issues of diminishing T 1 contrast at high field namely a characterization of T 1 contrast at 4.7T, 11T, and 17.6T using three T 1 contrast pulse sequences. These pulse sequences, SR, IR, and MDEFT, have shown varied performance over the field strengths studied giving a wide range of options to optimize image quality and contrast. Also developed was a methodology for the creation of biologically representative T 1 contrast phantoms. These phantoms allow for contrast evaluation, protocol creation, and troubleshooting without the use of animals. A simulation package was also created which allows for prediction of T 1 contrast prior to imaging experiments reducing scan time and assisting in detection of technical issues. These developments help in the creation and testing of high field imaging protocols specifically dealing with reduction of T 1 contrast. The within magnet comparisons chapter demonstrated the contrast performance of the pulse sequences at each field strength employed. Overall the pulse sequences performed well giving a range of image contrast on all three systems. The SR_GE performed well at short preparation times allowing for fast imaging with good image contrast, but its sensitivity to RF excitation uniformity can significantly affect contrast performance. IR_GE gave excellent contrast between tissue types and was not as 188

PAGE 205

189 sensitive to RF non-uniformity as SR_GE. The broad T 1 contrast profile of MDEFT_GE allows for uncertainty in T 1 values prior to imaging and gives contrast to multiple tissue types simultaneously. The diversity in the contrast performance of the three pulse sequences along with their varied susceptibility to other high field imaging issues allows for a large range in experimental conditions to be imaged with sufficient quality. The T 1 contrast enhanced in vivo images from the within magnet comparisons chapter demonstrated the ability to obtain quality T 1 contrast images at high field. The between magnets comparisons chapter demonstrated the evolution of each pulse sequences’ contrast performance as field strength increases. Increasing field strength alters several experimental parameters, which determine the performance of the pulse sequences. At the lower field strengths the T 1 s of biological tissue are relatively short and the ratio between tissue types is relatively large giving high T 1 contrast with the short T 1 values providing short scan times. At higher field strengths these T 1 s lengthen and the ratio between them decreases causing maximum T 1 contrast to diminish and making scan times longer. A uniform RF excitation profile also becomes more difficult to obtain at high field strength limiting the choices for obtaining T 1 contrast at high field. The performance of the three pulse sequences was quite varied but some similarities presented themselves. For example, all the pulse sequences demonstrated a shift of maximum contrast to longer preparation times, and a trend of decreased T 1 contrast on the higher field systems. Also the three pulse sequences exhibited a slower decay from maximum contrast on the higher field systems. The SR_GE pulse sequence differed from the others by demonstrating sensitivity to the homogeneity of the RF excitation. IR_GE was less sensitive to non-uniform RF profiles than SR_GE, but

PAGE 206

190 MDEFT_GE demonstrated the best performance in the presence of RF in-homogeneities. T 1 contrast imaging at high field becomes increasingly difficult but T 1 contrast is still obtainable, thus not taking away from the other benefits that high field imaging provides. Because of increased susceptibility artifacts on high field systems spin echo versions of these T 1 contrast pulse sequences were characterized and compared with the GE versions. Chapter 9 showed that SE encoding does not diminish the contrast provided by the T 1 preparation. Some differences between SE and GE encoding were apparent. The SR_GE and SR_SE exhibited differences in the shape of their CNR curves. The SR_SE retained CNR at longer preparation times than the GE version of the pulse sequence because of the larger tip angle used in SE experiments. This also presented itself in the IR_SE compared with the IR_GE but to a much smaller extent. The MDEFT_SE and the MDEFT_GE experiments matched very well because of the driven equilibrium aspect of the magnetization preparation. This work demonstrated that SE encoding does not significantly alter the shape of the contrast curves compared with the GE encoding and that SE encoding may be used instead of GE without loss of T 1 contrast. In vivo T 1 contrast studies often cannot be completed without fast imaging techniques, this is particularly so if other studies are planned for the same imaging session. Chapter 10 presented the T 1 contrast performance of both SE and GE fast imaging techniques. Fast imaging techniques can greatly reduce acquisition time, but the potential tradeoffs can be a reduction in T 1 contrast, addition of T 2 or T 2 * contrasts, or a reduction in SNR. Using a biologically representative T 1 phantom these fast imaging techniques were tested and showed a similar performance to the non-segmented results.

PAGE 207

191 This research shows that with the scan time reduced by one-fourth one can still achieve good image contrast and image quality for T 1 s in the biological range, using both SE and GE fast imaging techniques. There are obstacles to high quality T 1 contrast imaging on high field systems, but this body of work has demonstrated that with knowledge of these difficulties, modeling, and specific techniques dealing with these issues quality T 1 contrast imaging may still be obtained at high field strengths.

PAGE 208

CHAPTER 12 DISSERTATION SUMMARY While the push to higher field strengths has accomplished its main objective of obtaining higher SNR, it has also brought along with it several issues. As discussed earlier the amount of T 1 contrast decreases with increasing field strength. Traditionally structural imaging has utilized T 1 contrast, and the decrease in T 1 contrast becomes an issue at field strengths now being used in some clinical magnets and especially for field strengths that current research is being conducted on. Other high field imaging issues include increased B 1 in-homogeneities and increased B 0 variability due to susceptibility artifacts. DTI has benefited from the increase in field strengths now being able to probe water diffusion with more flexibility, but there are certain drawbacks. New methods must be developed to deal with high field clinical DTI studies since so very few have attempted to conduct experiments above 1.5T. The research presented here applied high field DTI to a unique subject population of prenatal cocaine exposed children in order to better understand an interesting human developmental outcome. Another focus of the research was the development of pulse programs and imaging techniques in order to maintain T 1 contrast at high field strengths. The hypothesis that prenatal cocaine exposure induces microstructural changes in the developing brain is supported by the DTI results from this study. The reduction of FA and the increase in in the frontal white matter areas of the brain indicate subtle microstructural changes in these areas. While much still needs to be completed to fully understand the effect of prenatal cocaine exposure on the developing brain, this work 192

PAGE 209

193 provides insight into specific microstructural changes due to this dangerous drug. The future of clinical structural imaging may include much higher field systems, along with the development of new techniques capitalizing on the benefits that high field imaging provides. In particular advancements such as high angular resolution diffusion imaging (HARDI) or q-space imaging may be implemented on high field clinical systems in the future. These techniques are superior in sampling diffusion in a biological system compared with current tensor techniques employed. Techniques such as these may provide a deeper understanding of the effects prenatal cocaine exposure has upon the developing brain. With a better understanding of these mechanisms and the effects that prenatal cocaine exposure has on them a clearer picture of the specific challenges exposed children are confronted with can be seen. The reduction of T 1 contrast with increasing field strength has become a limiting factor for many applications. This work has dealt with some of the issues of diminishing T 1 contrast at high field namely a characterization of T 1 contrast at 4.7T, 11T, and 17.6T using three T 1 contrast pulse sequences. These pulse sequences, SR, IR, and MDEFT, have shown varied performance over the field strengths studied giving a wide range of alternatives for the optimization of image quality and contrast. It has been demonstrated that sufficient T 1 contrast may be obtained at field strengths up to 17.6T. With this knowledge biological T 1 contrast may be obtained using even higher field systems than the ones employed in this study. T 1 contrast imaging on high field systems is not without its difficulties, but it has been demonstrated that with knowledge of these difficulties, modeling, and specific techniques dealing with these issues quality T 1 contrast imaging may still be obtained at high field strengths.

PAGE 210

194 There are benefits to high field structural imaging compared with imaging at typical clinical field strengths, but there are difficulties as well. The T 1 contrast research presented here has demonstrated that some of these difficulties can be overcome and that quality T 1 contrast images may be produced. The prenatal cocaine exposure research has demonstrated that high field structural imaging may be employed to elucidate microstructural changes in an in vivo setting.

PAGE 211

APPENDIX PAPERS AND PRESENTATIONS This appendix lists the publications, proposed publications, and presentations originating from the work presented in this dissertation. First, publications from the prenatal cocaine exposure DTI study are described. Listed below is a paper that will be submitted to Radiology titled “A Comparison of the Reliability of Two Techniques for White Matter Segmentation in a Clinical DTI Study.” A paper to the journal “Magnetic Resonance in Medicine” describing the diffusion changes in frontal white matter tissue in prenatally cocaine exposed children is planned. Four presentations at meetings covering the DTI study have also been made and the abstracts are reproduced. With respect to the high field T 1 contrast optimization research, a paper to the journal “Magnetic Resonance in Medicine” describing the native T 1 contrast enhancement at 4.7T, 11T and 17.6T is planned. Also planned is a paper to the journal “Magnetic Resonance in Medicine” describing the trend of rodent T 1 values at ultra high magnetic field strengths. Two presentations regarding this work have been made at international meetings and the abstracts reproduced below. In summary, one publication is close to submission, with three more planned shortly, and 6 presentations have been made at international meetings. Additionally I have been involved in several other research projects during my Ph.D. that utilized my MR expertise. Although these did not form part of my dissertation, they are listed 195

PAGE 212

196 following the papers and presentations derived from the dissertation work to provide a more complete description of the total work performed during my studies. A Comparison of the Reliability of Two Techniques for White Matter Segmentation in a Clinical DTI Study Author List: (Projected) K. R. Padgett, C. S. Garvan, K. M. Crandall, T. A. Black, F. D. Eyler, M. Behnke, C. M. Leonard, I. M. Schmalfuss, S. J. Blackband. Abstract: (Submitting to Radiology May 2005) Purpose: Various segmentation techniques for analysis of diffusion tensor images (DTI) have been in use ranging form manual tracing to fully automated segmentation techniques. Some of the advantages and disadvantages of these methods are known, however, their reliabilities have not been reported. Therefore, the purpose of this study was to determine the inter-rater reliability of the different segmentation techniques based on the hypothesis that the semi-automated method provides higher reliability than the hand drawn technique. Material and Methods: Fractional anisotropy (FA) maps were generated from 22 DTI sets collected on a Siemens 3T Allegra scanner. A spin-echo echo planar imaging DTI pulse sequence in 6 directions was used with b values of 0, 250, and 1000 s/mm2. Two raters applied different segmentation procedures (hand drawn versus semi-automated “Region Shrink” intensity based segmentation methods) to various white matter structures. The inter-rater reliability was determined using interclass correlations (ICC). Results: The ICC for the hand drawn method was significantly lower than for the semi-automated “Region Shrink” method with an average ICC of 0.38 and 0.94,

PAGE 213

197 respectively. While the anatomical structures of interest were easily identified with both segmentation techniques, their borders were difficult to reproduce using the hand drawn method. Conclusion: While manual segmentation is a common and easy to implement technique it has significantly lower inter-observer reliability than the semi-automated method. The inherent low resolution of clinical DTI datasets creates difficulty for manual segmentation. Therefore, caution should be applied in interpretation of DTI results acquired with a hand drawn method. Diffusion Changes in Frontal White Matter in Prenatally Cocaine Exposed Children Author List: K. R. Padgett, M. Behnke, F. D. Eyler, C. M. Leonard, K. M. Crandall, I. M. Schmalfuss, C. S. Garvan, T. D. Warner, S. J. Blackband. Abstract: Oral presentation at ISMRM 2005 Cocaine easily crosses the fetal blood-brain barrier and accumulates in the CNS(24,25). Various mechanisms such as hypoxia or disruption of the monoaminergic system as potential causes of brain injury in children exposed to cocaine prenatally have been therefore suggested. Hypoxia places the hippocampus and frontal white matter at risk, and may also affect white matter development. The neurotransmitters of the monoaminergic system greatly influence the developmental trajectory of their targets. These possible mechanisms of injury may alter tissue microstructure. Therefore, the purpose of our study was to examine the brains of the cocaine exposed children versus their control group for microstructural abnormalities that may relate to emerging developmental issues. DTI was employed for its ability to detect such changes.

PAGE 214

198 A Comparison of the Reliability of Two Techniques for White Matter Segmentation in a Clinical DTI Study Author List: K. R. Padgett, C. S. Garvan, K. M. Crandall, T. A. Black, F. D. Eyler, M. Behnke, C. M. Leonard, I. M. Schmalfuss, S. J. Blackband. Abstract: Poster presentation at ISMRM 2005 Developing a reliable, valid segmentation technique for DTI datasets poses many technical issues. While manual tracing techniques of segmentation closely mimic the radiologist’s interpretation it is susceptible to poor intra and inter operator reliability(41). Conversely, fully automated segmentation techniques are robust in reproducing regions of interest (ROIs), but are more open to segmentation artifacts. Semi-automated segmentation techniques are hypothesized to be more reliable than hand drawn methods but less susceptible to segmentation artifacts. The purpose of our study was to demonstrate the superiority of a semi-automated segmentation over the hand drawn method in regard to intra-rater reliability. Reduction of Fractional Anisotropy in Frontal White Matter in Prenatally Cocaine Exposed Children Author List: K. R. Padgett, F. D. Eyler, M. Behnke, C. M. Leonard, K. M. Crandall, T. H. Mareci, T. A. Black, I. M. Schmalfuss, C. S. Garvan, S. J. Blackband Abstract: Oral presentation at ISMRM 2004 We have tested the hypothesis that Diffusion Tensor Imaging (DTI) shows a decrease in fractional anisotropy (FA) in frontal white matter areas of children who were prenatally exposed to cocaine. The subject population consists of a group of prenatally cocaine exposed children and a control group; children in both groups are now 10 to 11

PAGE 215

199 years of age. There have been no drug-group differences in perinatal or subsequent deaths(60) nor were there signs at birth of withdrawal(61) or congenital abnormalities(62,63). Only a few drug-group differences have been found in overall growth and development since birth, from 6 months to 7 years(64,65). Because of these findings, we wanted to examine the brains of the population for abnormalities that may relate to emerging developmental issues. Cocaine easily crosses the fetal blood-brain barrier and accumulates in the CNS(24,25). Possible methods of injury are hypoxia and disruption of the monoaminergic system. Hypoxia puts the hippocampus and frontal white matter at risk, and may also affect white matter development. The neurotransmitters of the monoaminergic system greatly influence the developmental trajectory of their targets. These possible mechanisms of injury may alter tissue microstructure. DTI was employed for its ability to detect such changes. This study suggests prenatal cocaine exposure induces microstructural changes in frontal white matter because of significant reduction of FA in these regions in the exposed population. Native T 1 Contrast Enhancement at 4.7, 11 and 17.6 T for Neuroimaging Author List: K. R. Padgett, S. J. Blackband, S. C. Grant Abstract: Poster presentation at ISMRM 2005 The push to higher magnetic field strengths has necessitated the re-evaluation of standard MR protocols to account for changes in image contrast mechanisms. Of particular interest has been the degradation of T 1 relaxation contrast at high fields due to the increase in tissue T 1 s. Recently, several methods have been proposed to recover T 1 contrast above 3 T. Among these techniques, the modified driven equilibrium Fourier transform (MDEFT) preparation (52,53,58) has proven to be especially useful in human

PAGE 216

200 neurological imaging because it provides T 1 contrast enhancement over a wide range of T 1 values and is less sensitive to B 1 inhomogeneities. In this study, the MDEFT preparation is compared to standard saturation recovery (SR) and inversion recovery (IR) techniques at three ultra-high magnetic field strengths (4.7, 11 & 17.6 T) that are pertinent to high resolution, morphometric imaging. To assess the benefits and drawbacks of each T 1 preparation technique, physiologically relevant phantoms were constructed for the three field strengths. The phantoms cover a range of field dependant T 1 values matched to measurements from the brains of normal C57BL6/J mice. Using these phantoms, optimal preparation times were determined for the three T 1 contrast techniques at the three field strengths to provide optimal contrast enhancement over a range of T 1 values. These optimal values were utilized to generate enhanced MDEFT images of in vivo mouse and rat brains for comparison to standard imaging protocols. Enhancement of Native T 1 Contrast at 4.7, 11 and 17.6 T for Small Animal Imaging Author List: Kyle R. Padgett, Stephen J. Blackband, and Samuel C. Grant Abstract: Poster presentation at ISMAR 2004 The push to higher magnetic field strengths has necessitated the re-evaluation of standard MR protocols to account for changes in image contrast mechanisms. Of particular interest has been the degradation of T 1 relaxation contrast at high fields. As magnetic fields increase, tissue T 1 values tend to coalesce, significantly reducing image contrast quality. Recently, several methods have been proposed to recover T 1 contrast at field strengths above 4 T. Of these techniques, the modified driven equilibrium Fourier transform (MDEFT) preparation(52,53,58) has been proven to be especially useful in human neurological imaging because it provides T 1 contrast enhancement over a wide

PAGE 217

201 range of T 1 values and is less sensitive to B 1 inhomogeneities. In this study, MDEFT preparation is compared to standard saturation recovery (SR) techniques and inversion recovery (IR) preparation at three magnetic field strengths (4.7, 11 and 17.6 T) that are pertinent to high resolution, morphometric animal imaging. To assess the benefits and drawbacks of each T 1 preparation technique, physiologically relevant phantoms were constructed for the three field strengths. These phantoms cover a range of field dependant T 1 values matched to relaxation coefficients measured from the brains of normal C57BL6/J mice. Particular attention was paid to gray and white matter contrast. Using these phantoms, optimal preparation times were determined for the three T 1 contrast techniques at the three field strengths to provide optimal contrast over the given range of T 1 values. Publications and Presentations Outside of Thesis Research Publications: Cerebellar Nuclei Afferents and Efferents in Joubert Syndrome. B. L. Maria , K. R. Padgett , L. L. Olson , S. J. Blackband , J. G. Gleeson, E. Stashinko, A. H. Hoon, L. M. Nagae-Poetscher, and S. Mori. Annals of Neurology [In Press]. Increased susceptibility of S100B transgenic mice to perinatal hypoxia-ischemia. Wainwright MS, Craft J, Griffin S, Marks A, Pineda J, Padgett KR, Van Eldik. Annals of Neurology. 2004 Jul; 56(1):61-7. Observation of significant signal voids in images of large biological samples at 11.1 T. Beck BL, Jenkins K, Caserta J, Padgett K, Fitzsimmons J, Blackband SJ. Magn Reson Med. 2004 Jun;51(6):1103-7. Ex vivo High Resolution MRI of the Brain in Joubert Syndrome. Kyle. R. Padgett, Bernard L. Maria, Anthony T. Yachnis, Stephen J. Blackband. J Child Neurol. 2002 Dec;17(12):911-3.

PAGE 218

202 Presentations: Human Head Imaging at 11 Tesla. Michael B. Smith, Qing X. Yang, Christopher M. Collins, Barbara L. Beck, Kyle R. Padgett, Samuel C. Grant, G. Randy Duensing, Ron Smith, and Stephen J. Blackband. International Society of Magnetic Resonance in Medicine. Accepted for poster presentation at 2005 meeting in Miami, Fl. Cerebellar Nuclei Afferents and Efferents in Joubert Syndrome. B. L. Maria , K. R. Padgett , L. L. Olson , S. J. Blackband , J. G. Gleeson, E. Stashinko, A. H. Hoon, L. M. Nagae-Poetscher, and S. Mori. Child Neurology Society. Poster presentation at 2004 annual meeting in Ottawa. S100B Transgenic Mice Show Increased Vulnerability to Cerebral Ischmic Injury. C. Anderson, M. Mannix, J. Pineda, K. Padgett, D. Watterson, L. Van Eldik, M. Wainwright. Poster presentation at Society for Neuroscience 2003. Observation of B1 Inhomogeneities on Large Biological Samples at 11.1 Tesla. Barbara L. Beck, Kelly Jenkins, Kyle R. Padgett, Jeffrey Fitzsimmons, S.J. Blackband. International Society of Magnetic Resonance in Medicine. Oral presentation at 2003 meeting in Toronto. Very High-Field (17.6 Tesla) Magnetic Resonance Imaging and Spectroscopy at Tissue Resolution. T. H. Mareci*, S. J. Blackband, S. Germain, S. Grant, Kyle R. Padgett, H. D. Plant, E. Ozarslan, P. Patel, T. M. Shepherd, X. S. Silver, P. Thelwall, R. Walton, and E. D. Wirth. SEMRC 2001 Talk @ UFBI, Gainesville, FL

PAGE 219

LIST OF REFERENCES 1. Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209-219. 2. Levitt P, Reinoso B, Jones L. The critical impact of early cellular environment on neuronal development. Prev Med 1998;27:180-183. 3. Levitt P, Harvey JA, Friedman E, Simansky K, Murphy EH. New evidence for neurotransmitter influences on brain development. Trends Neurosci 1997;20:269-274. 4. Stanwood GD, Washington RA, Shumsky JS, Levitt P. Prenatal cocaine exposure produces consistent developmental alterations in dopamine-rich regions of the cerebral cortex. Neuroscience 2001;106:5-14. 5. Jones LB, Stanwood GD, Reinoso BS, Washington RA, Wang HY, Friedman E , Levitt P. In utero cocaine-induced dysfunction of dopamine D1 receptor signaling and abnormal differentiation of cerebral cortical neurons. J Neurosci 2000;20:4606-4614. 6. Mori S, Barker PB. Diffusion magnetic resonance imaging: its principle and applications. Anat Rec 1999;257:102-109. 7. Cercignani M, Horsfield MA. The physical basis of diffusion-weighted MRI. Journal of the Neurological Sciences 2001;186:11-14. 8. Pfeuffer J, Broer S, Broer A, Lechte M, Flogel U, Leibfritz D. Expression of aquaporins in Xenopus laevis oocytes and glial cells as detected by diffusion-weighted 1H NMR spectroscopy and photometric swelling assay. Biochim Biophys Acta 1998;1448:27-36. 9. Thelwall PE, Grant SC, Stanisz GJ, Blackband SJ. Human erythrocyte ghosts: Exploring the origins of multiexponential water diffusion in a model biological tissue with magnetic resonance. Magn Reson Med 2002 Oct ;48 (4):649 -57 2002;48:649-657. 10. Neil JJ. Measurement of Water Motion (Apparent Diffusion) in Biological Systems. Concepts Magn Reson 1997;9:385-401. 203

PAGE 220

204 11. Norris DG. The effects of microscopic tissue parameters on the diffusion weighted magnetic resonance imaging experiment. NMR Biomed 2001;14:77-93. 12. Nicholson C, Sykova E. Extracellular space structure revealed by diffusion analysis. Trends in Neuroscience 1998;21:207-215. 13. Rusakov DA, Kullmann DM. Geometric and viscous components of the tortuosity of the extracellular space in the brain. Neurobiology 1998;95:8975-8980. 14. Barkovich AJ. Concepts of myelin and myelination in neuroradiology. AJNR Am J Neuroradiol 2000 Jun -Jul ;21(6 ):1099 -109 2000;21:1099-1109. 15. Basser PJ, Mattiello J, LeBihan D. Estimation of the effective self-diffusion tensor from the NMR spin echo. J Magn Reson B 1994;103:247-254. 16. Armitage PA, Bastin ME. Utilizing the diffusion-to-noise ratio to optimize magnetic resonance diffusion tensor acquisition strategies for improving measurements of diffusion anisotropy. Magn Reson Med 2001 Jun ;45 (6 ):1056 -65 2001;45:1056-1065. 17. Le Bihan D, Mangin JF, Poupon C, Clark C, Pappata S, Molko N, Chabriat H. Diffusion Tensor Imaging: Concepts and Applications. Journal of Magnetic Resonance Imaging 2001;13:534-546. 18. Gulani V, Webb AG, Duncan ID, Lauterbur PC. Apparent diffusion tensor measurements in myelin-deficient rat spinal cords. Magn Reson Med 2001 Feb ;45 (2):191 -5 2001;45:191-195. 19. Pierpaoli C, Basser PJ. Toward a Quantitative Assessment of Diffusion. Magnetic Resonance in Medicine 1996;36:893-906. 20. Tortora G, Principles of Human Anatomy, Harper Collins College Publishers, New York, 1995. 21. Osborn AG, Tong K.A., Handbook of Neuroradiology: Brain and Skull (Corra E., Kleinberg A. S., Eds.), p. 497, Mosby, St. Louis, 1996. 22. Warner TD, The effects of prenatal cocaine exposure on attention and reading. University of Florida, Ph.D. dissertation; (2003). 23. Lidow MS. Prenatal cocaine exposure adversely affects development of the primate cerebral cortex. Synapse 1995;21:332-341. 24. Dow-Edwards DL, Mothers, babies, and cocaine : the role of toxins in development (Michael Lewis, Margaret Bendersky, Eds.), p. 5, Lawrence Erlbaum Associates, Hillsdale, N.J., 1995.

PAGE 221

205 25. Mayes LC, Bornstein MH, Mothers, babies, and cocaine : the role of toxins in development (Michael Lewis, Margaret Bendersky, Eds.), p. 251, Lawrence Erlbaum Associates, Hillsdale, N.J., 1995. 26. Covert RF, Schreiber MD, Tebbett IR, Torgerson LJ. Hemodynamic and cerebral blood flow effects of cocaine, cocaethylene and benzoylecgonine in conscious and anesthetized fetal lambs. J Pharmacol Exp Ther 1994;270:118-126. 27. Iida H, Gleason CA, O'Brien TP, Traystman RJ. Fetal responses to acute fetal cocaine injection in sheep. Am J Physiol 1994;267:H1968-H1975. 28. Schreiber MD, Madden JA, Covert RF, Torgerson LJ. Effects of cocaine, benzoylecgonine, and cocaine metabolites in cannulated pressurized fetal sheep cerebral arteries. J Appl Physiol 1994;77:834-839. 29. Woods JR, Jr., Plessinger MA, Clark KE. Effect of cocaine on uterine blood flow and fetal oxygenation. JAMA 1987;257:957-961. 30. Longo LD, Hermans RH. Behavioral and neurochemical sequelae in young rats of antenal hypoxia. Early Hum Dev 1992;29:83-90. 31. Bondy SC, Nakla M, Ali SF, Ahmad G. Cerebral ornithine decarboxylase levels following gestational exposure to cocaine. Int J Dev Neurosci 1990;8:337-341. 32. Seidler FJ, Slotkin TA. Prenatal cocaine and cell development in rat brain regions: effects on ornithine decarboxylase and macromolecules. Brain Res Bull 1993;30:91-99. 33. Spraggins YR, Seidler FJ, Slotkin TA. Cocaine exacerbates hypoxia-induced cell damage in the developing brain: effects on ornithine decarboxylase activity and protein synthesis. Biol Neonate 1994;66:254-266. 34. Webster WS, Brown-Woodman PD, Lipson AH, Ritchie HE. Fetal brain damage in the rat following prenatal exposure to cocaine. Neurotoxicol Teratol 1991;13:621-626. 35. Ozarslan E, Mareci TH. Generalized diffusion tensor imaging and analytical relationships between diffusion tensor imaging and high angular resolution diffusion imaging. Magn Reson Med 2003;50:955-965. 36. Pham DL, Xu C, Prince JL. Current methods in medical image segmentation. Annu Rev Biomed Eng 2000;2:315-337. 37. Lim KO, Choi SJ, Pomara N, Wolkin A, Rotrosen JP. Reduced frontal white matter integrity in cocaine dependence: a controlled diffusion tensor imaging study. Biol Psychiatry 2002 Jun 1;51 (11):890 -5 2002;51:890-895.

PAGE 222

206 38. Filippi CG, Lin DD, Tsiouris AJ, Watts R, Packard AM, Heier LA, Ulug AM. Diffusion-tensor MR imaging in children with developmental delay: preliminary findings. Radiology 2003;229:44-50. 39. Howell DC, Intraclass Coefficient: For Unordered Pairs. (2002), Web Page. http://www.uvm.edu/~dhowell/StatPages/More_Stuff/icc/icc.html, accessed August 2003. 40. Ott RL, An Introduction to Statistical Methods and Data Analysis, Wadsworth, Belmont, 1992. 41. Clarke LP, Velthuizen RP, Camacho MA, Heine JJ, Vaidyanathan M, Hall LO, Thatcher RW, Silbiger ML. MRI segmentation: methods and applications. Magn Reson Imaging 1995;13:343-368. 42. Anderson AW. Theoretical analysis of the effects of noise on diffusion tensor imaging. Magn Reson Med 2001;46:1174-1188. 43. Arendt RE, Short EJ, Singer LT, Minnes S, Hewitt J, Flynn S, Carlson L, Min MO, Klein N, Flannery D. Children prenatally exposed to cocaine: developmental outcomes and environmental risks at seven years of age. J Dev Behav Pediatr 2004;25:83-90. 44. Schmithorst VJ, Wilke M, Dardzinski BJ, Holland SK, IQ Correlates with Diffusion Tensor Imaging Parameters in Normal Children. Presentation at International Society of Magnetic Resonance in Medicine: Annual meeting; Honolulu, Hawaii, (2002). 45. Capilla A, Romero C, Maestu F, Campo P, Fernandez S, Gonzalez-Marques J, Fernandez A, Ortiz T. [Emergence and brain development of executive functions]. Actas Esp Psiquiatr 2004;32:377-386. 46. Demakis GJ. Frontal lobe damage and tests of executive processing: a meta-analysis of the category test, stroop test, and trail-making test. J Clin Exp Neuropsychol 2004;26:441-450. 47. Jackson J, The Myth Of The 'Crack Baby'. Extra! The Magazine of FAIR. Sept/Oct 1998;8:9. 48. Fischer HW, Rinck PA, Van Haverbeke Y, Muller RN. Nuclear relaxation of human brain gray and white matter: analysis of field dependence and implications for MRI. Magn Reson Med 1990;16:317-334. 49. Stark DD, Bradley W, Magnetic Resonance Imaging, Mosby Company, St. Louis, 1988.

PAGE 223

207 50. Kangarlu A, Baertlein BA, Lee R, Ibrahim T, Yang L, Abduljalil AM , Robitaille PM. Dielectric resonance phenomena in ultra high field MRI. J Comput Assist Tomogr 1999;23:821-831. 51. Norris DG. Reduced power multislice MDEFT imaging. J Magn Reson Imaging 2000 Apr;11(4):445 -51 2000;11:445-451. 52. Ugurbil K, Garwood M, Ellermann J, Hendrich K, Hinke R, Hu X, Kim SG, Menon R, Merkle H, Ogawa S. Imaging at high magnetic fields: initial experiences at 4 T. Magn Reson Q 1993;9:259-277. 53. Lee JH, Garwood M, Menon R, Adriany G, Andersen P, Truwit CL, Ugurbil K. High contrast and fast three-dimensional magnetic resonance imaging at high fields. Magn Reson Med 1995;34:308-312. 54. Kangarlu A, Abduljalil AM, Robitaille PM. T1and T2-weighted imaging at 8 Tesla. J Comput Assist Tomogr 1999;23:875-878. 55. Haacke EM, Brown RW, Thompson MR, Venkatesan R, Magnetic Resonance Imaging, John Wiley & Sons, New York, 1999. 56. Hoult DI, Phil D. Sensitivity and power deposition in a high-field imaging experiment. J Magn Reson Imaging 2000 Jul ;12 (1):46 -67 2000;12:46-67. 57. Beck BL, Jenkins K, Caserta J, Padgett K, Fitzsimmons J, Blackband SJ. Observation of significant signal voids in images of large biological samples at 11.1 T. Magn Reson Med 2004;51:1103-1107. 58. Deichmann R, Schwarzbauer C, Turner R. Optimisation of the 3D MDEFT sequence for anatomical brain imaging: technical implications at 1.5 and 3 T. Neuroimage 2004 Feb ;21(2):757 -67 2004;21:757-767. 59. Hennig J, Nauerth A, Friedburg H. RARE imaging: a fast imaging method for clinical MR. Magn Reson Med 1986;3:823-833. 60. Eyler FD, Behnke M, Conlon M, Woods NS, Wobie K. Birth outcome from a prospective, matched study of prenatal crack/cocaine use: I. Interactive and dose effects on health and growth. Pediatrics 1998;101:229-237. 61. Eyler FD, Behnke M, Garvan CW, Woods NS, Wobie K, Conlon M. Newborn evaluations of toxicity and withdrawal related to prenatal cocaine exposure. Neurotoxicol Teratol 2001 Sep -Oct ;23(5):399 -411 2001;23:399-411. 62. Behnke M, Eyler FD, Garvan CW, Wobie K. The search for congenital malformations in newborns with fetal cocaine exposure. Pediatrics 2001 May ;107 (5):E74 2001;107:E74.

PAGE 224

208 63. Eyler FD, Behnke M, Conlon M, Woods NS, Wobie K. Birth outcome from a prospective, matched study of prenatal crack/cocaine use: II. Interactive and dose effects on neurobehavioral assessment. Pediatrics 1998;101:237-241. 64. Eyler FD, Behnke M, Garvan CW, Wobie K, Warner TD, Hou W, Effects of prenatal cocaine exposure and other risk factors on outcome in a prospectively enrolled matched cohort now entering school. Presentation at Society for Research in Child Development: Biennial meeting; Tampa, Florida, (2003). 65. Behnke M, Eyler FD, Garvan CW, Wobie K, Hou W. Cocaine exposure and developmental outcome from birth to 6 months. Neurotoxicol Teratol 2002 May -Jun ;24 (3):283 -95 2002;24:283-295.

PAGE 225

BIOGRAPHICAL SKETCH Kyle R. Padgett was born in Plantation, Florida, in April 1977. He has lived and attended school in Florida his entire life. After completion of his high school degree at Martin County High School, Florida (1995), he attended the University of Florida and graduated in 1999 with a B.S. in physics. He then enrolled in the Ph.D. medical physics program at the Department of Nuclear and Radiological Engineering at the University of Florida. After completing the medical physics program Mr. Padgett plans to pursue a career in academic or industrial applications of magnetic resonance imaging. 209