The  Vanderbilt  University  Institute  of  Imaging  Science    

2010  Research  Retreat  June  21  –  23,  2010  

 Huntsville,  AL  

   

   

Cover  Image  by  R.  Adam  Smith  

Validation of molecular imaging through innovative registration methods

       

Vanderbilt  University  Institute  of  Imaging  Science  Research  Retreat  

2010      

     June    10,  2010    On  behalf  of  the  Program  Committee,  I  am  delighted  to  welcome  you  to  Huntsville,  Alabama   for   the   2010   Vanderbilt   University   Institute   of   Imaging   Science   (VUIIS)  Research  Retreat.    This  year  marks  the  seventh  consecutive  annual  retreat  and  it  is  our  sincere  hope  that  this  year’s  events  will  be  as  successful  as  the  previous  six.    The  VUIIS  continues  to  be  one  of  the  largest,  integrated  imaging  research  institutes  in  the  country.    This  statement  reflects  what  may  very  well  be  the  most   important  accomplishment  realized  during  the  first  six  years  of  the  Institute:  the  assembly  of  a  diverse,   multidisciplinary   team   of   collegial   investigators,   all   under   a   single   roof,  whose  cutting  edge  research  spans  nearly  all  aspects  of   imaging  science.  Beyond  a  simple   collection   of   labs   and   individuals   housed   in   a   common   building,   we   are   a  team  purposefully  assembled   to  ask  and  address   the  groundbreaking  questions   in  biomedical   imaging   science   and   medicine.     To   capitalize   upon   this   inherent  potential,   we   must   intentionally   and   frequently   seek   innovative   ways   to   work  together  towards  common  goals  and  the  Research  Retreat  is  an  important  vehicle  to  facilitate   this   exercise.   For   this   reason,   the   goals   of   the   2010   retreat   have   not  deviated  from  previous  Retreats  and  are:    

1) To  learn  everyone’s  name  and  know  what  aspect  of  Imaging  Science  they  study.  

2) To   foster   new   collaborative   interactions   with   Institute   colleagues,  particularly  those  outside  your  major  area  of  study.  

3) To   learn  about  new  and  existing   resources,  upcoming   initiatives,   and   to  identify  potential  institutional  needs.  

4) To  relax  and  have  a  bit  of  fun  away  from  campus.    The   Scientific   Program   for   2010   features   a   mixture   of   focused   talks   and   poster  sessions   that   fall   into   four   categories,   Functional   and   Structural   Neuroimaging,  Imaging  Physics  and  Analysis  Methods,  Molecular  and  Cellular  Imaging,  and  Cancer  and  Metabolic  Imaging.    Though  we  are  a  large  group  and  continue  to  grow,  it  is  our  hope   everyone  will   remember   that   the   retreat   atmosphere   is   intended   to  be  non-­‐intimidating  for  both  presenters  and  those  seeking  to  learn  something  new.    As  such,  presenters  are  highly  encouraged  to  utilize  every  opportunity  to  garner  constructive  feedback   on   their   current   projects   and   future   directions.   Similarly,   questions   and  comments   from   the   audience   participants   are   also   highly   encouraged.   If   do   not  understand  something-­‐ASK!    Outside  the  Scientific  Program,  we  have  planned  two  social  events  that  I  know  you  are  going  to  enjoy.  These  include  a  reception  at  the  Jackson  Center  Pub  and  dinner  at  the  Davidson  Space  Center.    

As  former  Program  Committee  members  know,  it  takes  a  lot  of  work  and  time  out  of  busy   schedules   to  plan   the   retreat   each  year.  Please   remember   to   thank   the  2010  Program   and   Scientific   Committee  members  when   you   see   them   around   over   the  next  couple  of  days  and  especially  Nancy  Hagans.    I  hope  you  enjoy  the  meeting  as  well  as  the  social  events.    Best  Regards,    

   C.  Chad  Quarles,  Ph.D.  2010  Program  Chairman    2010  Retreat  Committee  Nancy  Hagans    Seth  Smith  Jack  Virostko  Adrienne  Dula  Lori  Arlinghaus  Robert  Barry  Nellie  Byun  

Special Thanks to Our Supporters

The Department of Radiology & Radiological Sciences, Vanderbilt University Medical Center

Presentation and poster prizes generously sponsored by the

The Department of Radiology & Radiological Sciences

The Department of Radiology & Radiological Sciences

Monday, June 21, 2010Monday, June 21, 2010 Speaker

1:30 Check-in at Jackson Conference Center / Poster setup

Session 1

2:00 Welcome Chad Quarles

2:15 Director’s Opening Remarks John Gore

2:30 Functional and Structural Neuroimaging Adam Anderson

2:45 Evaluation of 2D EPI and 3D PRESTO for BOLD fMRI at 7 Tesla Robert Barry

3:00 The Relationship Between White Matter Connectivity and Children’s Reading Abilities Qiuyun Fan

3:15 Unsupervised Methods for Analyzing fMRI Data Santosh Katwal

3:30 Development of CEST Imaging at 7.0 Tesla for Examination of Amide Proton Transfer (APT) in Multiple Sclerosis Adrienne Dula

3:45 Break / Poster setup

4:00 Poster Session 1 (4:00 - 5:00)

Session 2

5:15 From Tensors and Cortical Surfaces to Platforms and Databases : Large-Scale NeuroImaging Informatics Bennett Landman

5:30 Modulation of Dopaminergic Neurotransmission by Muscarinic Acetylcholine Receptor Activation Nellie Byun

5:45 Study of Temporal Changes in Metabolites as a Function of Visual Stimulation Subechhya Pradhan

6:00 Quantitative MRI of the PNS: Development and Reproducibility Richard Dortch

6:30 Reception at Jackson Center

Tuesday, June 22, 2010Tuesday, June 22, 2010

8:15 Poster setup

Session 3

8:30 Imaging Physics and Analysis Methods Daniel Gochberg

8:45 High Resolution Multi-shot SENSE DWI at 7T Ha-Kyu Jeong

9:00 Physical Basis of Acousto-Optical Imaging Chris Jarrett

9:15 Correlation of 1H NMR Characteristics and Mechanical Properties in Human Cortical Bone Adam Horch

9:30 Compartmental Measurements of Extracellular Volume Fraction in a Graded Muscle Edema Model Jack Skinner

9:45 Break / Poster setup

10:00 Poster Session 2 (10:00 - 11:00)

Session 4

11:15 Let's Get Small - Steve Martin 1977 Seth Smith

11:30 Designing Pulsed-CEST Imaging Sequence Zhongliang Zu

11:45 Advancing Scintimammography Through the Development of Breast-Specific High Purity Germanium Detectors Desmond Campbell

12:00 Ultra-high Field MRI with Series of Spatially-selective Sub-pulses Jay Moore

12:15 Lunch (12:15 - 1:45) / Poster setup

Tuesday, June 22, 2010 (continued)Tuesday, June 22, 2010 (continued)

Session 5

2:00 In-Vivo and Numerical Studies of Multi-Exponential T2 Decay in Rat Spinal Cord Kevin Harkins

2:15 Effect of Registrations on DBM Analysis in Corpus Callosum Zhaoying Han

2:30 T1-rho Dispersion Measurements for Estimating Chemical Exchange Rates in Tissue Models Jared Cobb

2:45 Molecular and Cellular Imaging Charles Manning

3:00 Design and Synthesis of Hemicholinium-3 PET Probes Mike Nickels

3:15 Pyrazolo-pyrimidinyl TSPO Ligands for Cancer Imaging Dewei Tang

3:30 Break / Poster setup

3:45 Poster Session 3 (3:45 - 4:45)

Session 6

5:00 Parahydrogen Induced Polarization (PHIP) and Beyond Ed Chekmenev

5:15 Transcriptional and Translational Regulation of TK1 are Determinants that Affect [18F]FLT-PET in Oncology Eliot McKinley

5:30 Nanodendrons for Detection & Treatment of Breast Cancer Lynn Samuelson

5:45 Evaluation of Imaging Agents Targeting the Pancreatic Beta Cell Using Co-registered Multimodal Imaging Jack Virostko

7:00 Dinner at Space Center

Wednesday, June 23, 2010Wednesday, June 23, 2010

Session 7

9:30 Cancer and Metabolic Imaging Tom Yankeelov

9:45 In Vivo Skeletal Muscle Glycogen Measured by Chemical Exchange Saturation Transfer (glycoCEST) and 13C MRS at 7T Ted Towse

10:00 Modeling Magnetic Field Perturbations for Arbitrary Geometries Natenael Semmineh

10:15 Efficacy of a JAK-2 Inhibitor Assessed Using DW- and DCE-MRI Mary Loveless

10:30 Microenvironmental Influences in Early Breast Cancer Metastasis to Liver Michelle Martin

10:45 Break

Session 8

11:00 Predicting Breast Cancer Treatment Response with Functional and Molecular Imaging Techniques Jennifer Whisenant

11:15 A Comparison of DCE-MRI Models in Human Breast Cancer Xia Li

11:30 Improving Diffusion-weighted Imaging of the Breast at 3T Lori Arlinghaus

11:45 Closing Remarks

Poster Session 1 (Monday, 4:00 - 5:00)Poster Session 1 (Monday, 4:00 - 5:00) Presenter

Effects of Membrane Integrity and Edema in Magnetic Resonance Imaging Studies of Inflammatory Myopathies Nathan Bryant

Validation of DTI-Tractography-based Measures of Primary Motor Area Cortical Connectivity Yurui Gao

Parametric Mapping of Biological Tissues Using Temporal Diffusion Spectroscopy Susan Kost

Comparison of qMT Techniques with Optimal Schemes Ke Li

Synthesis and Evaluation of COX-2 PET Imaging Probes Don Nolting

MR Measurement of Cerebral Blood Volume in the Hippocampus Swati Rane

Synthesis of Metabolic Tracers for Real Time Hyperpolarized MRI of Breast Cancer Roman Shchepin

Development of a Clinically Relevant 3 T Breast Protocol David Smith

Imaging of Hyperpolarized 13C and 15N Tracers Diana Smith

Subject-specific CFD Modeling of the Vertebro-basilar System Amanda Wake

Detecting Tumor Early Response to Chemotherapy Using Temporal Diffusion Spectroscopy: How Early Can We Get? Junzhong Xu

Poster Session 2 (Tuesday, 10:00 - 11:00)Poster Session 2 (Tuesday, 10:00 - 11:00)

Microcalcification Detection using Susceptibility Weighted Phase Imaging: Cross-correlation and Relative Magnetic Susceptibility Difference Methods

Richard Baheza

Preclinical Evaluation of TSPO Ligand [18F]PBR06 for PET Imaging of Glioma Jason Buck

Structural Complexity of Cortex and its Implications on Cortical Connectivity Measurements using MR Tractography Ann Choe

Development of NMR Probe Heads for Hyperpolarization of 13C via Para Hydrogen Induced Polarization (PHIP) at 12.0 mT Aaron Coffey

Molecular Basis of TSPO as a Cancer Imaging Biomarker Saffet Guleryuz

Probing Demyelination using High Resolution 3D Quantitative Magnetization Transfer (qMT) and Diffusion Tensor Imaging (DTI) in a Lipopolysaccharide (LPS) Model of Multiple Sclerosis (MS)

Vaibhav Janve

Poster Session 2 (Tuesday, 10:00 - 11:00) (continued)Poster Session 2 (Tuesday, 10:00 - 11:00) (continued)

An Information Theory Approach to Parameter Estimates from Multi-Compartment Models of MRI Contrast Chris Lankford

Magnetic Resonance Spectroscopic Imaging of Brain at 7 T Indrajit Saha

Parallel Image Reconstruction for 7T MRI Sepideh Shokouhi

Correlating 18FLT Uptake with Drug Delivery using MALDI-IMS Adam Smith

Increased Differentiation of the Lateral Pain Network by High Resolution fMRI at 7T Elizabeth Ann Stringer

High Resolution FMRI Mapping of Cortical Plasticity Following Spinal Cord Injury in Non-Human Primates Xiang Ye

Poster Session 3 (Tuesday, 3:45 - 4:45)Poster Session 3 (Tuesday, 3:45 - 4:45)

Extracting Proliferation Rates from ADC Data Nkiruka Atuegwu

Quality Assurance of B1 RF Pulses for Parahydrogen Induced Polarization Raul Colon Moreno

High Resolution MRI at 7 Tesla to Evaluate the Anatomy of the Human Midbrain Dopamine System Mariam Eapen

Correlating MALDI and MRI Biomarkers of Breast Cancer Amelie Gillman

Development of Materials for TSPO-Directed HTS Matthew Hight

An Optimized Composite Refocusing Pulse for 7T MRI Marcin Jankiewicz

Radiation Dose-Based Comparison of PET and SPECT for Bone Imaging Lindsay Johnson

Resting State Functional Connectivity Analysis: a Potential Model for Human fMRI Studies Arabinda Mishra

Identifying EEG Correlates of Activity in the Working Memory Network Measured with fMRI Allen Newton

Partially Loaded Travelling Wave MRI Sasidhar Tadanki

In Vivo Mouse Kidney Imaging Feng Wang & Noor Tantawy

Comparison of Reduced-FOV Techniques at 7T Chris Wargo

Evaluation of 2D EPI and 3D PRESTO for BOLD fMRI at 7 Tesla Robert L. Barry, Stephen C. Strother, J. Christopher Gatenby, John C. Gore

Introduction

Blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) is commonly performed using 2D single-shot echo planar imaging (EPI). Although EPI provides excellent BOLD contrast, severe geometric distortions often arise. A recent study at 3 Tesla (T) showed that the multi-shot pulse sequence 3D PRESTO provided superior BOLD contrast compared to 2D EPI (1), but it is not yet known if this finding translates to higher fields. The goal of this work is to investigate the use of 3D PRESTO in lieu of 2D EPI for BOLD fMRI at 7T.

Methods

Twelve volunteers participated in a functional study using a robust visual stimulus (flashing checkerboard wedge in left visual field). MRI acquisition parameters were matched across pulse sequences. Functional data were optimally processed via independent and un-biased data-driven metrics of prediction and reproducibility using NPAIRS (Non-parametric Pre-diction, Activation, Influence and Reproducibility re-Sampling) (2).

Results

Group activation maps are presented with a common threshold of |z| > 5 and dynamic range from z = -6.67 to z = 19.0 (Fig. 1). Very good agreement in sensitivity and specificity is observed for positive BOLD activation, as well as a robust ‘negative’ response. A voxel-wise comparison shows that PRESTO, in general, provides higher z-scores than EPI (Fig. 2).

Discussion

In addition to exhibiting increased sensitivity to BOLD signal changes (z-scores), PRESTO also had a phase encode bandwidth that was 3.3 times higher than EPI (142.5 vs. 42.5 Hz). This extra bandwidth resulted in pixels shifts (distortions) that were less than one-third as severe compared to EPI. In conclusion, PRESTO produced images with high sensitivity and geometric stability, suggesting that it has potentially significant advantages for BOLD fMRI at 7T.

References 1. S.F.W. Neggers et al., NMR Biomed. 21, 663 (2008). 2. S.C. Strother et al., NeuroImage 15, 747 (2002).

Figure 1. Activation maps for data acquired using (A) EPI and (B) PRESTO (|z| > 5).

Figure 2. Voxel-wise comparison of unthresholded data from Fig. 1. A point above the line of unity represents a voxel with a higher z-score in PRESTO than EPI.

The relationship between white matter connectivity and children’s reading abilities

Qiuyun Fan, Nicole Davis, Donald Compton, Laurie E. Cutting, John C. Gore, Adam W. Anderson

Introduction Reading is a complex cognitive process mediated by a specific network of brain areas. Information transfer between cortical and sub-cortical areas in the network influences reading skill. Individual differences in reading skill relate to differences in white matter integrity1. The goal of this project is to study this relationship using diffusion tensor imaging (DTI).

Methods

We obtained information on each participant’s reading skills and acquired structural MRI and DTI scans. As shown in figure 1 a, our analysis focused on cortical and sub-cortical regions of interest (ROI) that were identified in previous studies as related to reading skill. Probabilistic tractography was performed between pairs of ROIs to calculate the connectivity strength between brain regions. The outcome from this analysis was correlated with participant’s reading scores.

Results Several pathways demonstrated significant correlations with behavioral test scores. The analysis suggested that the strength of each connection correlates differently with participants’ behavioral scores. This reveals the specificity of different brain areas in mediating reading performance. As shown in figure 1.b, we found a strong, significant correlation between performance on the Word Attack subtest (WAT) and connectivity strength between the left superior temporal and fusiform gyri. A scatterplot of this correlation (Fig.1.c) shows a linear relation and a group clustering effect.

Discussion In this study, we identified the contribution of individual cortical regions in mediating reading performance using DTI and probabilistic tractography. To enhance the stringency of the conclusions gained, we need to adopt a sample of larger size.

References 1. C. Beaulieu et al., NeuroImage. 25, 1266–1271 (2005).

Figure 1: Data processing methods (a) and results (b, c). Significance of correlation (b) is color-coded in log scale, and the arrow indicates the pathway analyzed in (c).

Unsupervised Methods for Analyzing fMRI Data Santosh B. Katwal, John C. Gore, and Baxter P. Rogers

Introduction Conventional model-based or statistical approaches such as Statistical Parametric Mapping (SPM) require a-priori knowledge of the task paradigm, precise timing of stimulus onsets and an assumed hemodynamic response function. These introduce biases in the model thereby increasing the possibility of suboptimal or fallacious inferences. Model-free data driven techniques such as cluster analysis take advantage of the internal relationships in actual fMRI data and are devoid of such biases. We use unsupervised clustering techniques as a fully data driven approach to analyze and cluster activations in fMRI. In particular, a cascade of Self-Organizing Mapping (SOM) and Hierarchical Clustering (HC) is considered [1].

Methods Following [2], five subjects were scanned each for five latency runs in a visual checkerboard task with known latencies between left and right hemifield stimulation onsets. The data were analyzed with an unsupervised clustering technique involving SOM and HC [1]. SOM is composed of several nodes arranged in a hexagonal or rectangular grid lattice. We used 100 nodes arranged in a 10x10 2-D map. SOM constitutes two stages: training and mapping. At the end, it produces a 2-D map of input data keeping the topology intact. To optimize clustering, similar SOM nodes were merged using HC with a spatio-temporal metric [1]. Average time series were obtained for each cluster and the one which gave the highest trial repeatability was chosen as activation.

Results Figure 1(a) shows the SOM output for a subject. The nodes shown in red correspond to activated cluster. The activations are overlaid on the functional image as shown in figure 1(c). Figure 1(b) shows the average trial of the BOLD responses from right and left V1 with 112 ms latency in one subject. With this technique, latency as short as 28ms was fully resolved in fMRI at 7T (Figure 1(d)), much better than results in [2].

DiscussionThe unsupervised clustering techniques in fMRI do not consider spatial connectivity. However, activated regions in brain tend to occur in clusters of neighboring voxels. In future, methods to include spatial connectivity information in SOM for fMRI will be studied. It is also important to include time dependencies between signals for clustering. Along that line, methods such as Hidden Markov Model will be explored.

References 1. W. Liao et al., IEEE Trans. Med. Imag., 27(10), 1472-83 (2008).2. S.B. Katwal et al., Proc. Intl. Soc. Mag. Reson. Med., 17, 3677 (2009).

(a) (b)

(c) (d)

Figure 1: (a) SOM output nodes (b) V1 BOLD response (c) Right and left V1 activation map (d) Influence of right V1 on left V1.

Figure 1: CEST imaging at 7T. A) Anatomical image of MS patient, B) Amide proton transfer asymmetry map, C) ROI analysis with CEST spectra (right y-axis) and CESTasym (left y-axis).

Development of CEST Imaging at 7.0 Tesla for Examination of Amide Proton Transfer (APT) in Multiple Sclerosis

Adrienne N. Dula, Elizabeth M. Asche, Bennett A. Landman, E. Brian Welch, John C. Gore, Seth A. Smith

Introduction

Chemical exchange saturation transfer (CEST) can generate MRI contrast based on exchange rates and concentration of mobile metabolites using their exchange properties with water. Examination of amide proton transfer (APT) using CEST benefits from higher Bo, coinciding with increased field inhomogeneities. These can be mitigated using the water saturation shift referencing (WASSR) method to center the z-spectra [1]. The application of CEST imaging at 7T could provide an advanced and quantitative metric to non-invasively study neuronal tissue.

Methods

A 7T CEST imaging sequence to quantify APT was developed and implemented on 10 healthy and 4 multiple sclerosis (MS) subjects. The WASSR method was used for z-spectra centering [1]. CEST and WASSR acquisitions were acquired with fast field echo. APT asymmetry analysis (3.5 ppm) was performed on registered and normalized images resulting in APTasym maps. APTasym was calculated for various regions of interest (ROIs) in white matter (WM), gray matter (GM) and lesions.

Results

Contrary to application at lower field strengths, APT imaging at 7T produced contrast between healthy WM (7.1+0.7%) and GM (3.8+0.5%). Necrotic lesions had decreased APTasym (3.3%) while that of NAWM (6.2%) was increased. APT imaging on MS pathology produced varying results for the APT asymmetry based on the lesion type.

Discussion

The discrepancy in APTasym found among various MS lesion types may be a signature for different disease processes simultaneously occurring. Asymmetry analysis of 7T CEST spectra at the amide proton resonance is reflective of the neural tissue properties, providing defining information of healthy tissue, as well as NAWM, inflammatory lesions, and T1 holes in MS patients. These studies may provide important new insights regarding the pathophysiology of MRI contrast.

References

1. M. Kim, et al., Magnetic Resonance in Medicine 61. 1441 (2009).

Modulation of dopaminergic neurotransmission by muscarinic acetylcholine receptor activation

Nellie E. Byun, Robert L. Barry, Carrie K. Jones, P. Jeffrey Conn, John C. Gore

Introduction Pharmacological MRI (phMRI) has been used to assess the effects of specific agents on regional changes in brain activity and their associated hemodynamic effects. Dopamine is a neurotransmitter that plays a critical role in reward, cognition, and motor output, and disorders such as schizophrenia, addiction, and Parkinson’s disease are characterized by dopaminergic imbalance. Dopaminergic modulation via muscarinic acetylcholine receptors (mAchRs) appears to be a viable therapeutic mechanism for treating such disorders. However, we need to elucidate how each of the different mAchRs modulate dopamine signaling. We are currently investigating the role of M1 and M4 mAchRs on dopaminergic output in the brain by administering amphetamine to exert a hyperdopaminergic state and evaluating how pretreatment with M1/M4 agonists modulate this activation. Specifically, we are using phMRI to identify the brain regions on which a M1/M4 agonist acts, and to quantify drug-induced cerebral blood volume (CBV) changes in specific regions of interest (ROIs).

Methods Studies are being performed on a 9.4T Varian scanner with a Doty Litz 38 coil. Adult male Sprague-Dawley rats are intubated, mechanically ventilated (33%:67% O2:N2O; isoflurane 0.88%) and paralyzed (pancuronium bromide 1 mg/kg) during the scan. Functional images are acquired with a T2-weighted fast spin echo sequence. After acquiring 12 initial images to determine the pre-contrast agent baseline, monocrystalline iron oxide nanoparticles (MION, 20 mg/kg, i.v.) are injected. The post-MION functional scan consists of a 15 min baseline, drug (M1/M4 agonist or saline) administration, followed by amphetamine (1 mg/kg, i.p.) or saline challenge 30 min later. After this, images are acquired for another 45 min. Fractional CBV change, ΔCBV/CBV o, is calculated as [ln (S(t)/S)]/[ln (So/Spre)], where So is the average baseline signal and Spre is the average pre-MION signal1. Time courses are extracted from ROIs and mean post-injection ΔCBV/CBVo are statistically compared.

Results Compared to saline treatment, amphetamine decreases signal in striatum and increases regional CBV. We are currently evaluating a M1/M4 agonist to test the hypothesis that activation of M1/M4 suppresses dopaminergic output in striatum. Other ROIs, including prefrontal cortex and thalamus, will be analyzed.

Discussion Functional MRI can be used to study the modulation of one neurotransmitter system on another at a systems level. This is especially intriguing for studying the effects of the cholinergic system on dopamine transmission as this may point to therapeutic efficacy, especially for schizophrenia and addiction. We will also be comparing analysis techniques (principal component analysis, linear detrending) and will evaluate how M1/M4 activation changes functional connectivity between regional pairs. These results will be correlated with on-going neurobehavioral studies.

References 1. J.B. Mandeville et al., Magn Reson Med. 39, 615 (1998).

Study of temporal changes in metabolites as a function of visual stimulation

Subechhya Pradhan, James M. Joers, Kevin W. Waddell, John C. Gore

Introduction Functional MRS studies have been used to provide insight into local metabolite changes in activated regions using block paradigms. However, the existing literature reports conflicting results on the changes in metabolite profiles during activation [1, 2]. Earlier studies have looked for changes using 6-8 minutes block acquisitions. We aim to gain better understanding of potential metabolite fluctuations using a binning method which yields higher temporal resolution, and combined with the higher SNR at 7T will allow fMRS with event-related paradigms.

Methods Healthy volunteers were scanned with a 7 Tesla Philips Achieva MR system using protocols approved by the Vanderbilt University Institutional Review Board. Activation maps of visual cortex acquired using conventional fMRI were used to guide voxel placements for fMRS scans. The stimulus for all functional scans used a checkerboard flashing at 8 hz frequency. The parameters for water unsuppressed spectra were 5 cycles of 36 s rest followed by 36 s stimulus at TE/TR = 60/6000 ms, and for water suppressed spectra at TE/TR = 73/4000 ms. All spectra were acquired using a PRESS sequence. The metabolite spectra are then re-sorted and binned based on the time acquired after the onset of each stimulus, thereby giving a single average time course throughout each cycle. The resulting water spectra are analyzed using programs written in MATLAB and the metabolites analyzed using LCModel.

Results Figure 1 shows how the water spectra and their line-width and peak heights vary in time with the presented (block design) visual stimulus. We observed a 4.9% change in line-width and 5.6% in peak height along with a 1.3% change in area under the water peak.

Discussion We are currently in the process of acquiring metabolite spectra from the activated region in the visual cortex using the same task. The acquisition of high quality spectra from the visual cortex is hindered by the high chemical shift displacement coupled with low maximum B1 (15µT), which results in significant lipid contamination of the spectra due to the proximity of the region to the scalp. We are working on overcoming this issue with the use of higher B1 (18µT) and a STEAM sequence, both of which will help reduce the chemical shift displacement errors.

References 1. S. Mangia et al., J Cereb. Blood Flow Metab, 27, 1055 (2007). 2. M.H. Baslow et al., J Mol Neurosci, 32, 235 (2007).

Figure 1: Top: Water spectra from the visual cortex during stimulation (red) and rest (blue), Bottom: Corresponding water peak height (blue) and linewidths (red) during the paradigm.

0 10 20 30 40 50 60 70 8015

16

17.5

19

21

Quantitative MRI of the PNS: Development and Reproducibility Richard D. Dortch, Peter D. Donofrio, Jun Li, John C. Gore, Seth A. Smith

Introduction Quantitative MRI methods have been employed in the central nervous system for

many years; however, similar studies in the peripheral nervous system (PNS) have

been limited (1,2). With the advent of higher field strengths, it is now possible to

perform high-resolution MRI of relatively small nerves. Additionally, novel contrast mechanisms allow for the possibility to quantitatively probe morphological changes

(e.g., demyelination) accompanying pathology in nerve. Thus, the goal of this study

is to develop and test the reproducibility of quantitative MRI methods in the PNS.

Methods Quantitative methods — diffusion (diffusion tensor, q-space), magnetization transfer (MT), and multiexponenital T2 (MET2) — will be developed at 3T for the sciatic nerve.

To test the reproducibility of each developed method, data will be obtained in ten

healthy subjects at two times points spaced approximately one week apart.

Results Pilot data were obtained in the sciatic nerve of two healthy volunteers at 3T (Fig. 1). Diffusion-weighted (DW) data were acquired over a range of b-values (600-6000

s/mm2) with a DW single-shot EPI sequence and SPAIR fat saturation. MT data were

acquired using an MT-prepared gradient-echo sequence with binomial (1331) water excitation. MET2 data were acquired with a multiple spin-echo pulse sequence and

SPAIR fat saturation. The resulting data are in agreement with previously published

values (1,2) and demonstrate the feasibility of performing the proposed studies.

Discussion

The long-term goal of these studies is to optimize these quantitative measures to

maximize the ability to study peripheral neuropathies in the clinic. We hypothesize

that combination of myelin- (diffusion, MT, MET2) and axon-sensitive measures (diffusion) will provide information about the morphological changes associated with

disease progression that is not available from individual measures.

References 1. M.F. Meek et al. Exp. Neurol. 198 , 479 (2006).

2. G. Gambarota et al. J. Magn. Reson. Imaging. 29, 982 (2009).

1 10 100 10000

200

400

600

800

1000

1200

1400

T2 (ms)

Co

mp

. Mag

. (a.

u.)

0 0.1 0.2 0.3 0.4G

diff

a b cDiffusion MTR T2 Spectrum

Δ = 2kHzθMT = 1000°b = 1000 s/mm2

SciaticNerve

Figure 1: Representative axial (a) diffusion data, (b) magnetization transfer ratio (MTR) map, and (c) T2 spectrum (from inverse Laplace transform of MET2 data).

Figure 1. Reconstructed images for b = 0 (a), b = 700 s/mm2

before (b) and after (c) phase correction and FA map (d).

High resolution multi-shot SENSE DWI at 7 T Ha-Kyu Jeong, John C. Gore, Adam W. Anderson

Introduction

Diffusion weighted images (DWI) are commonly acquired with single-shot methods [1] combined with parallel imaging [2] to reduce acquisition times and susceptibility effects. High spatial resolution DWI, however, requires multi-shot acquisition and correction of phase variations using an additional navigator [3] or self-navigated methods [4], which measure erroneous phase terms due to subject motion during the diffusion gradients. We proposed and have developed a simplified acquisition and reconstruction method for multi-shot SENSE DWI using a conventional interleaved EPI sequence with an additional navigator acquisition. A preliminary evaluation of DWI at 7 Tesla is presented.

Methods

Diffusion weighted multi-shot SENSE spin-echo echo-planar brain images were acquired for a healthy volunteer on a 7 Tesla Philips Achieva whole body scanner using a 16 channel receiver head coil. Image- and navigator-echoes were acquired with a 180 degree refocusing RF pulse between the two echoes. Imaging parameters used are FOV = 240×240 mm, diffusion b-value = 0 and 700 s/mm2, 15 diffusion gradient orientations, 7 slices with no slice gap, TR/(TEimg,TEnav) = 4029/(74,112) ms, NSA = 3, SENSE factor = 3 and # shots = 4. Fat signal was saturated using a SPAIR adiabatic pulse. Acquisition voxel size in the read/phase/slice direction = 1.00/1.21/6.00 mm and reconstructed with 1 mm in-plane voxel dimensions. Custom pulse sequence and reconstruction algorithms were developed for the whole analysis of the data.

Results

Diffusion weighted images and corresponding diffusion properties were reconstructed without apparent ghost artifacts after phase correction using the navigator data (Fig 1).

Discussion

High resolution DW images were successfully reconstructed using the pulse sequence and reconstruction methods we developed. However, B1 inhomogeneity, which causes relatively low SNR in some regions of the brain, hampers accurate measurement of the DW signal and reconstruction as shown in Fig 1d.

References

1. R. Bammer et al., Magn. Reson. Med. 48, 136 (2002) 2. K.P. Pruessmann et al., Magn. Reson. Med. 42, 952 (1999) 3. A.W. Anderson et al., Magn. Reson. Med. 32, 379 (1994) 4. C. Liu et al., Magn. Reson. Med. 54, 1412 (2005)

Physical Basis of Acousto-Optical Imaging Christopher W. Jarrett, Bibo Feng, John C. Gore

Introduction

Small animal optical Imaging, is of increasing importance in preclinical research, but in vivo optical techniques are limited by the attenuation and scattering of light in tissues, which reduce the sensitivity and spatial resolution. To overcome these limitations, several hybrid imaging methods have emerged in recent years. One such approach is Acousto-Optical Imaging which uses ultrasonic waves to modulate the light signal of a fluorescent imaging system in an effort to improve the spatial resolution [1]. However, the mechanisms for this effect are unclear. Our research aims to investigate ultrasonic modulation of incoherent light produced by fluorescence within tissues and establish what factors affect the magnitude of such effects. We are currently addressing the first specific aim of this research which is to design and develop instrumentation to quantify Acousto-Optical interactions.

Methods

Our experimental system (Fig 1) includes a 635 nm diode laser as the fluorescence excitation source. The laser light is focused on the sample. Fluorescent emissions are modulated at 1 Mhz using a continuously driven or pulsed ultrasound transducer with a 1.5 inch focal length. The emitted light passes to the PMT and the excitation light is rejected utilizing a dichroic mirror and a bandpass filter. The PMT signal is amplified via a wideband pre-amp. The amplified signal is received by the lock-in amplifier and is compared to the reference signal created by the pulse generator. The system is controlled by a PC and Labview.

Results The system has been built and parts individually tested, but the appropriate PMT and PMT preamplifier, as well as the excitation/emission filters, have not arrived. Thus, no data have yet been collected.

Discussion This study will provide insight into the mechanism of ultrasonic modulation of

incoherent light in tissues while providing a technique better suited for detection of optical molecular imaging agents.

References 1. M. Kobayashi et al., App. Phys. Lttrs. 89, 181102 (2006).

Figure 1: Acousto-Optical System set-up as designed.

0 0.05 0.1 0.15 0.2 0.250

40

80

120

160

200

T2 Pool Volume Fraction

Stre

ss (M

Pa)

Figure 2. Yield Stress vs. T2 Pool size

VFAy = 1900x + 72

r2 = 0.48

VFBy = 940x − 30

r2 = 0.77

VFCy = −530x + 190

r2 = 0.37

10 −410 −5 10 −3 10 −2 10 −1 10 0

T2 (s)

Spec

tral

Inte

nsity

(a.u

.)

Pool ACollagen

Pool BCollagen-bound Water

Pool CLipids and Pore Water

Figure 1: Human Cortical Bone T2 Spectrum and Physiological Origins

Correlation of 1H NMR Characteristics and Mechanical Properties in Human Cortical Bone

R. Adam Horch, Jeffry S. Nyman, Daniel F. Gochberg, and Mark D. Does Introduction

Modern MRI methods such as ultra-short echo time (uTE) imaging are capable of imaging proton signals from human cortical bone [1], providing new information on water and macromolecular distribution in bone that cannot be assessed with conventional X-ray based methods. In our previous NMR characterization of human cortical bone [2], T2 components ranging from 50µs to 1s were attributed to collagen, collagen-bound water, lipids, and mobile water in porous spaces. Herein, we extend this characterization to include mechanical testing for studying NMR/mechanical property relationships. Sensitivity of T2 features to mechanical properties in bone would provide a useful contrast mechanism for diagnostic bone MRI.

Methods

Human cortical bone specimens were harvested from the mid-shafts of 17 male and female donor femurs (5 young donors, 26.2 ± 5.4 Y.O.; 8 middle-age donors, 52.8 ± 4.2 Y.O.; and 4 old donors, 88.8 ± 7.1 Y.O.). Specimens were machined into separate pieces of uniform cortical bone for mechanical testing and NMR analysis. Mechanical testing via 3-point bending (35mm span) measured flexural modulus, yield stress, ultimate stress, and toughness. Quantitative NMR measurements were performed at 4.7T (CPMG with 100 μs echo spacing, 10000 total echoes, and 90°/180° hard pulses of 7.5/15 μs) to generate T2 spectra [3]. All measurements were compared pair wise with a Pearson’s linear correlation.

Results

All human cortical bone specimens exhibited three T2 pools: two sub-millisecond components (Pools A&B) and a broad range of decays spanning 1ms-1s (Pool C). Figure 1 shows the three pools in a typical T2 spectrum with previously-established signal origins [2]. T2 pool sizes were linearly correlated to many mechanical properties; an example of correlation to yield stress is shown in Figure 2.

Discussion

The size of Pool B was the NMR parameter most strongly correlated to mechanical properties, indicating that the amount of collagen-bound water is linked to mechanical integrity in human cortical bone. Thus, Pool B may be a diagnostically-useful target for NMR studies, which we will soon be studying in a diabetes model.

References 1. A. Techawiboonwong et al., Radiology 248, 824-833 (2008). 2. R. Horch et al., ISMRM 17th Annual Meeting, #1942 (2009). 3. K. Whittall and A. Mackay, J. Magn. Res. 84, 134-152 (1989).

Figure 1: SPECT-MR image showing muscle edema.

101

102

0

0.5

1

T2 (ms)

Com

p.

Ma

g.

(a.u

.)

Edematous Muscle

T2

CE-T2

Figure 2: T2 spectrum pre- and post-contrast.

Compartmental Measurements of Extracellular Volume Fraction in a Graded Muscle Edema Model

Jack T. Skinner, Todd E. Peterson, Mark D. Does

Introduction Previous studies of injured skeletal muscle have shown multi-exponential relaxation with edema [1, 2]. The observed signals have been thought to correspond to intra- and extracellular tissue water. Water exchange between tissue compartments, however, affects the observed relaxation times and corresponding volume fractions. Inverting the two compartment model, to resolve exchange rates, is an ill-posed problem and requires the knowledge of at least one of the intrinsic model parameters. To this end, contrast enhanced T2 (CE-T2) measurements can be made to calculate the intrinsic value of the extracellular volume fraction, fb. To help validate the value of fb, SPECT imaging can be implemented, as this method is insensitive to inter-compartmental water exchange.

Methods Rats received a hindlimb injection of λ-carrageenan at varying concentrations to induce inflammation. Six hours later, a single slice multiple spin-echo measurement (TR=15s, NE=36) was made prior to administration of Gd-DTPA, and repeated 20 minutes post-injection. The T2 decay data was fitted to a two compartment model to extract values of R2' and f'. From these observed measures, the change in the mean relaxation rate (∆R2m') and sum of relaxation rates (∆R2s') were calculated and the value of fb was found from the ratio ∆R2m'/ ∆R2s'. Immediately following MR imaging, the renal vessels of the animal were ligated to allow the radio-tracer to reach a state of equilibrium between the vascular and interstitial space. 111In-DTPA (~1mCi) was injected and ~20-30 minutes later, SPECT images of the hindlimb and a blood sample were collected. The concentration of the tracer in the vascular and extracellular-extravascular space was found using the activities recorded from the co-registered SPECT images (Fig.1).The aforementioned concentrations were then used to calculate fb.

Results

Introduction of contrast resulted in a decrease of T2b' (Fig. 2). Both methods showed an increase in fb (fb ≈ 0.2-0.5) with increasing injection concentration and inflammation. In general, fb was found to be smaller than fb'. Model inversion revealed exchange rates, Rx, ranging from 2-6s-1.

Discussion

In the context of a graded edema model, knowledge of compartmental volume fractions and exchange rates might be combined with diffusion and relaxation data to help form a more comprehensive model of muscle injury in vivo.

References 1. Z. Ababneh et al., Magn. Reson. Med. 54, 524 (2005). 2. R.H. Fan and M.D. Does, NMR Biomed. 21, 566 (2008).

Designing Pulsed-CEST Imaging Sequence Zhongliang Zu, ke Li, Vaibhav Janve, Mark D. Does, and Daniel F. Gochberg

Introduction

Opt

imal

Bav

g pow

er (

μT)

a

1.5

1

0.5

Figure 1. (a) The optimal Bavg

power vs irradiation FA and duty cycle. (b) The optimal iFA vs dc. (c) The CEST contrast vs dc.

10 20 30 40 50 60 70 80 900

5

10

Duty cycle (%)

CE

ST

con

tras

t (%

)

c

50 60 70 80 90120

140

160

180

200

220

240

Duty Cycle (%)

Opt

imal

iFA

(de

gree

)

b

10 30 50 70 90140160180200220

0

Chemical exchange saturation transfer (CEST) has shown promise in greatly increasing the imaging sensitivity to metabolite concentration. However, amplifier limits often preclude clinical application of the traditional continuous wave (CW)-CEST imaging method [1], and instead require a pulsed-CEST approach. However, designing optimum pulsed-CEST imaging sequences entails complicated and time consuming numerical integrations of differential equations. In this work, we provide a simplified and computationally efficient technique to optimize the pulsed-CEST imaging sequence. The contrast is optimized when: 1) The average irradiation power (Bavg power) equals (the easily calculated) optimal irradiation power (Bcw) in CW-CEST imaging, 2) irradiation flip angle (iFA) equals 180°, and 3) duty cycle (dc) is as high as possible with consideration of the limits of irradiation pulse and spoil gradient durations.

Methods iFA (degree) Duty cycle (%)

CW- and pulsed-CEST imaging experiments were performed on a creatine/agar phantom at 9.4T. The sample parameters were determined via a 3-pool model fitting of the CW-CEST z-spectrum, and the parameters were then utilized for simulations of the CEST signal by integrating the coupled Bloch equations.

Results

Figure 1a plots the optimal Bavg power as a function of the iFA and the dc in pulsed-CEST experiment. Note that the optimal Bavg power is independent of iFA and dc. The optimal Bavg

power is also approximately equal to the optimal Bcw (not shown). Figure 1b plots the simulated pulsed-CEST optimal iFA as a function of dc. Note that the optimal iFA is roughly independent of dc. Figure 1c plots the simulated CEST contrast as a function of dc. Note that the contrast increases with the dc.

Discussion

The metric Bavg power and the optimal iFA of 180° provide a computationally efficient optimization technique for designing pulsed-CEST imaging sequences. In addition, simulations indicate that the proposed optimization method is also likely to work in vivo, where solid pool asymmetry effects complicate CEST data interpretation.

References

1. P. Z. Sun et al., Magn Reson Med. 60, 834 (2008).

Advancing Scintimammography through the Development of Breast-Specific High Purity Germanium Detectors

Desmond Campbell, Todd Peterson

Introduction Scintimammography, or Molecular Breast Imaging (MBI), is a nuclear imaging technique that utilizes specifically designed high-sensitivity semiconductor gamma cameras for breast cancer detection. One benefit of semiconductor elements in imaging systems is their superior energy resolution, which allows for better separation of scattered photons from primary counts to improve image quality.1 If scatter rejection can be improved using these detectors, then alternative collimator configurations that increase the field of view become viable options for MBI. In this work we conducted simulations to compare the imaging performance of High-Purity Germanium (HPGe) and Cadmium Zinc Telluride (CZT) systems with 1% and 3.8% energy resolutions at 140keV, respectively. The objective was to investigate whether the better energy resolution offered by HPGe would translate into improved imaging performance.

Methods Using the Monte Carlo N-Particle (MCNP5) simulation software, the LumaGem 3200s system was modeled using both 5mm-thick CZT and 10mm-thick HPGe detectors for the imaging of a breast/torso phantom with three spherical tumors. Simulated energy spectra for scintimammography scans were generated, and images were created for various energy windows around the 140keV photopeak. Scatter and torso fractions were calculated along with tumor contrast.2

Results Simulations showed that utilizing a 5% energy window with an HPGe system suppressed background originating from small-angle scattered photons better than a comparable CZT system using a 15% energy window. As illustrated in Figure 1, sensitivity and scatter rejection were improved in the HPGe system. Tumor contrast was enhanced using narrower energy windows in each imaging system.

Discussion The results suggest that scatter and torso fractions have definite dependencies on selected energy windows. Because of the superior energy resolution of HPGe detectors, narrower energy windows can be utilized to eliminate non-primary photons and reduce background. Even though scatter and torso fractions are smaller in our HPGe system, tumor contrast was slightly better with CZT acquired images. Thus, energy resolution may not have a strong influence on tumor contrast. The energy resolution of HPGe may be even more beneficial when using alternative collimation schemes, so exploring them in the future will be a priority.

References 1. C. B. Hruska, M. K. O’Connor, Physica Medica. 21, 72-75 (2006) 2. C. B. Hruska, M. K. O’Conner, IEEE Trans. Nuc. Sci. 55, 491-500 (2008)

Figure 1: Resulting breast images from simulations using a) a CZT system with a -5%/+10% energy window and b) a HPGe system with a ±2.5% energy window.

Figure 1: Top: slice profile as a function of RF field strength (B1

+) for 60° sinc (left) and optimized (right) pulses. Bottom: corresponding flip-angle maps in a 17 cm dielectric phantom at 7T.

Ultra-high Field MRI with Series of Spatially-selective Sub-pulses Jay Moore, Marcin Jankiewicz, Adam W. Anderson, John C. Gore

Introduction Inhomogeneous radiofrequency (RF) fields arising from interference and attenuation pose a serious challenge for ultra-high field human MRI. Many RF pulse designs and costly hardware modifications have been proposed to mitigate resulting spatial variations in signal intensity, but all come with drawbacks such as the necessity for time-consuming, subject-specific field mapping, the deposition of large amounts of RF energy in the body, or the lack of spatial selectivity. Outlined here is an RF design that addresses such shortcomings by producing a uniform and spatially selective excitation via short, numerically optimized composite pulses transmittable on a single channel.

Methods Loosely based on the method of “sparse spokes” [1], this pulse design uses an oscillating slice-selection gradient in conjunction with Gaussian sub-pulses. Numerical optimization of sub-pulse amplitudes and phases is performed on a 2-dimensional grid with indices corresponding to 1) relative RF field magnitudes and 2) spatial position in the slice-selection direction (see top row of Figure 1). With the pulse optimization taking place in this abstract parameter space, pulses can be designed to perform well given a range of RF field magnitudes, regardless of subject-specific field geometries [2].

Results A 60° optimized composite pulse consisting of 8 sub-pulses is highlighted in Fig. 1. This example pulse is 4.5 ms in duration and is characterized by a smoothly varying phase with each sub-pulse having amplitude near the maximum allowed value of 15 µT. Flip-angle maps (Fig. 1) in a 17 cm dielectric phantom at 7 T indicate superior flip-angle uniformity for the optimized pulse as compared to a Gaussian-sinc pulse.

Discussion Such optimized composite pulses can be designed for arbitrary flip-angles and within the practical limits of many imaging sequences calling for slice-selective excitation. The pulse design accomplishes much of what single-channel “sparse spokes” sequences are capable, but without the need for subject-specific pulse optimization. A potential limitation is the susceptibility of slice profiles to static field variations.

References 1. M. Jankiewicz et al., J. Magn. Reson. 203, 294 (2010). 2. J. Moore et al., Proc. Intl. Soc. Mag. Reson. Med. 18, 2856 (2010).

In-Vivo and Numerical Studies of Multi-Exponential T2 Decay in Rat Spinal Cord

Kevin D. Harkins, Adrienne N. Dula, Mark D. Does Introduction

Multi-exponential T2 MRI has revealed the presence of two T2 components in white matter – a fast-relaxing component believed to originate from the water within myelin and a slower component believed to originate from water in both the extra- and intra-axonal spaces. A recent ex-vivo study showed that the myelin water fraction (MWF), which is defined as the relative area of the myelin water peak, varied within regions of the rat spinal cord with similar myelin content [1]. This study measures MET2 in-vivo within the rat spinal cord, and uses numerical models to investigate the biophysical characteristics necessary to explain the variation in MWF.

Methods

Histology images of four rat spinal cord tracts – vestibulospinal (VST), rubrospinal (RST), funiculus gracilis (FG), and dorsal corticospinal (dCST) tracts were segmented into regions of myelin, intra-axonal and extra-axonal space. MET2 images were collected of the rat spinal cord within 5 rats. ROIs of the VST, RST, FG, and dCST spinal cord tracts were drawn on the first echo images. Signal each oft the ROIs were fitted to a T2 spectrum, and an image-based finite difference model of transverse relaxation was fitted to the experimental T2 spectrum to estimate intrinsic T2 and exchange characteristics of tissue.

Results Fig. 1 shows an example T2 weighted image, where the four white matter tracts are drawn onto the image (VST = teal, RST = red, FG = green, dCST = blue). Example T2 spectra are given in Fig. 2, as well as 4 model fits: a finite difference model without exchange (FD), using a permeable boundary between compartments (FD + P), and myelin diffustion (FD + Dm) to limit compartmental exchange. A Bloch-McConnell compartmental model (Comp) was also used to fit the data.

Discussion

This work demonstrates a difference in the MWF estimated in-vivo from MET2 data in white matter tracts with similar myelin content. Further, models of transverse relaxation can account for the change in MWF by allowing exchange between tissue compartments. Future work will focus on models of white matter disease in rat spinal cord.

References

1. A.N. Dula et al., Mag. Res. Med. 73, 902 (2010).

Figure 1: T2-weighted image of the rat spinal cord, with 4 ROIs drawn on different tracts of white matter

Figure 2: Example T2 spectra from the 4 white matter tracks in Fig. 1, along with 4 model fits

Effect of Registrations on DBM Analysis in Corpus Callosum Zhaoying Han, John C. Gore, Benoit M. Dawant

Introduction

Deformation Based Morphometry (DBM) involves detecting differences between populations via the analysis of the deformation fields that register 3D MR image volumes to reference image volumes. Few studies have compared the effect of the registration algorithms on DBM. In this study, we compare DBM results obtained with five well established non-rigid registration algorithms on the Corpus Callosum (CC) in subjects with Williams Syndrome (WS) and Normal Control (NC) subjects.

Methods

Thirteen WS subjects and thirteen NC subjects were firstly aligned together with 9-parameter affine registrations after intensity correction. The group nonrigid atlas was created iteratively using the method proposed by Guimond et al1. The deformation fields from all subjects to the atlas were obtained with five different nonrigid registrations methods: ABA2, IRTK3, FSL, ART4, and SPM normalization. At each voxel in the mid-slice of the CC, the Jacobian determinant (JAC) was calculated.

Results

Figure 1 shows the color-coded mean JAC maps over the entire CC area for the NC (top) and WS (bottom) groups obtained with the five NR registrations. The mean JAC value was calculated over the Genu regions for each subject and these values were averaged over all the volumes in each population. Table 1 lists the overall JAC for each method and the ANOVA test results in both groups.

Table 1: Mean JAC value over Genu for all subjects in the NC and WS groups. Discussion

Jacobian value is used to estimate local tissue compression (JAC<1) and expansion (JAC>1). Our study over the Genu area indicates that some NR methods show increased volumes while others show decreased volumes. The ANOVA test shows that JAC differences between NR methods are statistical significant (p<0.05). These results suggest that it is important to consider the effect of registration when using DBM to compute morphological differences in populations.

References

1. A. Guimond et al, Comput. Vis. Image. Und. 77, 1470 (2000) 2. G. K. Rohde et al. IEEE TMI 22, 1470 (2003) 3. D. Rueckert, et al. IEEE Trans. Med. Imaging 18, 721 (1999) 4. B. A. Ardekani et al. J. Comput. Assist. Tomogr 23, 800 (2005)

ABA IRTK FSL ART SPM AVONA: pNC 0.99 ±0.08 1.03±0.04 1.09±0.04 0.87±0.03 0.94±0.00 0.048WS 0.95±0.07 1.06±0.03 1.15±0.04 0.79±0.06 0.95±0.00 0.0005

Figure 1: The mean JAC maps for NC and WS with ABA, IRTK, FSL, ART and SPM

SPMIRTK FSL ARTABA

T1ρ Dispersion Measurements for Estimating Chemical Exchange Rates in Tissue Models

Jared G. Cobb, Jingping Xie, John C. Gore

Introduction A more complete understanding of proton relaxation in heterogeneous tissues may improve the interpretation of magnetic resonance images (MRI) and lead to imaging methods that are more specific for pathological changes. The measurement of relaxation rates in the rotating frame relaxation, R1ρ (=1/T1ρ), at different locking field strengths, can in principle be used to quantify the effects of chemical exchange and of macromolecular organization on T1ρ measurements. To explore such effects we have studied relaxation in tissue phantoms of cross-linked polyacrylamide gels that can be manipulated to vary the contributions of different relaxation processes.

Methods Acrylamide and N,N'-methylene-bis-acrylamide (BIS, normally used for cross-linking gels)) co-monomers were dissolved in water at 5% total weight and then polymerized to form cross-linked gels. The % BIS ranged from 0 to 100 in 20% increments. The gels were then soaked in buffers from pH of 2-11 in 5 increments. T1ρ dispersion was measured at 9.4T with a 10 mm loop-gap coil. A spin-locking pulse (2) was placed in front of a fast spin-echo imaging sequence. The locking field time was arrayed from 20 ms to 1 s and varied from 250 Hz to 8 kHz. Temperature was maintained at 21 °C. The resulting dispersion data were then fit with a least-squares fitting algorithm to the two-site fast exchange model proposed by Chopra et al. (3) to estimate exchange rates and other parameters for each gel.

Results R1ρ values increased with pH and with %BIS at all locking field strengths. Above 40% BIS there appears to be a more rapid increase in R1ρ. Calculated data from the fit to the Chopra model shows rapid increases in exchange rates for both low and high %BIS (Fig. 1) above pH 7.

Discussion The increase in measured R1ρ values and calculated exchange rates with pH is consistent with the hypothesis that chemical exchange mediates rotating frame relaxation. The most likely sites for chemical exchange in this model system are the amide protons on the polymer. Therefore, it is likely that the increase in R1ρ is related to an increase in exchange between water and amide protons at high pH. An increase in R1ρ values with increased %BIS may be attributed to an increase in gel rigidity and in the size of the macromolecular pool. The measured increase in exchange rates with %BIS may reflect a subtle influence on magnetization transfer of different amide and cross-relaxation mechanisms, as postulated in Kennan et al (4).

References 1. B.P. Hills, Mol. Phys. 76, 509 (1992). 2. R.E. Sepponen et al., JCAT. 9, 1007 (1985). 3. S. Chopra et al., JMR. 59, 361 (1984). 4. R.P Kennan et al., JMR, 110, 267 (1996).

0 2 4 6 8 10 120

1000

2000

3000

4000

5000

6000

Exchange Rate vs. pH at 97.5% BIS

pH

Exch

. R

ate

[H

z]

Fig 1: Exchange rates increase with increasing pH.

DESIGN AND SYNTHESIS OF HEMICHOLINIUM-3 PET PROBES Synthetic Progress towards Fluorine-18-Labeled Hemicholinium-3

Mike Nickels; Ning Guo; John C. Gore; Ronald Price; Wellington Pham Introduction

Acetylcholine (Ach) is an important neurotransmitter that plays a critical role in both the central and peripheral nervous systems. It serves as the primary neurotransmitter at the vertebrate neuromuscular junction, and as a modulator of cognitive function.(1) A shortage of acetylcholine in the brain leads to a number of neurodegenerative diseases including myasthenia gravis and Alzheimer’s disease.(2) Several lines of research suggest that Ach synthesis depends on efficient reuptake of choline in the pre-synaptic membrane via the sodium dependent, hemicholinium-3 (HC3)-sensitive (Figure 1), high-affinity choline transporter (CHT). The objective of this work is to develop hemicholinium-3 PET probes for mapping the expression and neuronal activity-dependent trafficking of CHT.

Methods Incorporation of a radiolabel onto HC-3 can theoretically be accomplished on any of the available carbons found throughout the molecule. In determining where to place the fluorine-18 label, we took into account the fact that the target molecule possesses two different classes of potential fluorine attachment, the aliphatic and the aromatic. With this in mind, we envisioned the development of two distinct synthetic routes, which simultaneously place the fluorine in the desired position and facilitate future elaboration of the target molecule. Synthesis of both products begins with commerically available biphenyl, which then undergoes multiple synthetic transformations including Friedel-Crafts acylation and various electrophilic and nucleophilic substitutions. The fluorination precursors are specifically designed to enable fast and efficient fluorination through a nucleophilic displacement reaction, which uses either radioactive (hot) fluorine (18F) or non-radioactive (cold) fluorine (19F). In the case of aliphatic fluorine formation, a highly reactive fluorinated precursor must be prepared and then attached to the HC-3 precursor.

Discussion Currently, modest quantities of these precursors have been synthesized with the desired regiochemistry. Several of the non-radioactive target molecules have also been synthesized as reference standards. Efforts are underway to develop the in-house technologies needed to synthesize the radiolabeled precursor, as well as the analytical techniques necessary to characterize and purify the target molecules properly. When these syntheses have been completed successfully and can be reproduced at a low activity level (< 50 mCi), the synthesis process will be adapted into an automated reaction module that will enable the formation of high activity products (> 100 mCi) without the undesirable exposure to high levels of radioactivity.

References 1. M. Sarter et al., Brain Research Reviews 23, 28 (1997). 2. D. M. Fambrough et al., Science (New York, N.Y.) 182, 293 (1973)

N

OO

N

OH HO

CH3H3C CH3

CH3

Figure 1: Chemical structure of Hemicholinium-3

Pyrazolo-pyrimidinyl TSPO Ligands for Cancer Imaging Dewei Tang, Jason R. Buck, Eliot T. McKinley, Saffet Guleryuz, Matthew R.

Height, R. Adam Smith, Allie Fu, Ping Zhao and H. Charles Manning

Introduction Translocator protein (TSPO), an 18kDa outer-mitochondrial membrane protein, is highly expressed in numerous types of cancer and is regarded as an effective target for drug and molecular imaging probe development [1,2,3]. We are synthesizing and evaluating novel TSPO ligands that are structural analogues of the pyrazolo-pyrimidine DPA-714. When labeled with 18F, a number of our TSPO ligands appear suitable for assaying TSPO expression in tumors. Importantly, our diversity oriented synthetic approach has afforded us an improved understanding of structure activity relationships (SAR) inherent to pyrazolo-pyrimidine-TSPO binding thereby facilitating further TSPO-specific probe development.

Methods We have coupled microwave assisted organic synthesis (MAOS) with traditional and microfluidic radiochemical methods to facilitate high-throughput, library-based, positron emission tomography (PET) imaging probe synthesis [4]. Pre-clinical PET imaging of C6 glioma-bearing rats, a TSPO-expressing tumor model, is currently being employed for initial characterization of the imaging probes in vivo. Imaging studies being carried out are complemented by in vitro radioligand binding assays, in vivo radiometabolite assays, and in vivo displacement studies.

Results To date, approximately 25 pyrimidinyl-based TSPO ligands have been synthesized. A portion of these compounds appear to have superior in vitro binding and in vivo imaging characteristics when compared to [18F]DPA-714 (Figure 1). Overall, we find pyrazolo-pyrimidines to be particularly stable in vivo, exhibiting significantly less defluorination in rats when compared with aryloxyanilide-based tracers such as [18F]PBR06.

Discussion

We anticipate that numerous pyrazolo-pyrimidines being developed in this project could be successfully developed as PET imaging ligands suitable for assaying TSPO expression in tumors. Such imaging probes could be particularly useful towards elucidating the role of TSPO in cancer biology.

References 1. Chen, M. K.et al., Pharmacol Therapeut 2008, 118 (1), 1-17. 2. Grabowski, P. et al., H. Gastroenterology 2005, 128 (4), A162-A162. 3. Maaser, K. et al., H. Clinical Cancer Research 2002, 8 (10), 3205-3209. 4. Tang, D. et al. Submitted to Tetrahedron Letters April 2010.

Figure 1: PET imaging of [18F]DPA-714 uptake in C6 glioma-bearing rat prior to- (left) and following (right) infusion of cold DPA-714.

Transcriptional and Translational regulation of TK1 are determinants that affect [18F]FLT-PET in oncology

Eliot T. McKinley, Saffet Guleryuz, Ping Zhao, Allie Fu, Nathan J. Mutic, M. Noor Tantawy, John C. Gore, Robert J. Coffey, and H. Charles Manning

Introduction

The PET tracer 3’-deoxy-3’[18F]-fluorothymidine ([18F]-FLT) potentially serves as a biomarker of proliferation by reporting utilization of thymidine salvage via thymidine kinase 1 (TK1) activity [1]. [18F]-FLT PET has shown promise in some clinical trials, yet it remains unclear what molecular processes affect TK1 regulation in cancer and thus [18F]-FLT uptake in tumors. Therefore, we designed preclinical studies to explore the relationship between [18F]-FLT-PET imaging, cellular TK1 levels, and established metrics of proliferation in mouse models of human cancer.

Methods Tumor xenografts were generated in nude mice from human colorectal cancer cell lines (HCT-116, HCT-116p21-/-, Lim2405, DiFi). [18F]-FLT PET images were acquired in prognostic and therapeutic settings focusing primarily on EGFR inhibition. Tumors were harvested for molecular analysis, including Western blot, qRT-PCR, and immunohistochemistry (IHC). Correlations were noted between tumor [18F]-FLT uptake and TK1 levels at cellular, proteomic and genomic scales and compared to IHC. Molecular regulation of TK1 during EGFR targeted therapies was explored in DiFi cells and xenograft tumors.

Results Tumor uptake of [18F]-FLT agreed well with TK1 protein levels (figure 1). [18F]-FLT-PET and TK1 levels exhibited variable agreement with Ki67. Treatment of DiFi cells and xenografts with mAb-C225 showed that transcriptional and translational regulation of TK1 were required to reduce [18F]-FLT uptake. Despite clinical response to mAb-C225 in vivo at all dosages evaluated, DiFi cells could escape translational regulation of TK1 through an mTOR-driven, cap-dependant mechanism resulting in failure of [18F]-FLT-PET to signal decreased tumor proliferation at lower dosages.

Discussion Molecular determinants affecting TK1 regulation and [18F]-FLT-PET are poorly understood across tumor types and therapeutic settings. Further basic validation studies may inform the correct clinical utilization and interpretation of [18F]-FLT -PET in oncology.

References 1. A. Shields et al., Nature Medicine 4, 1334 - 1336 (1998).

Figure 1:   [18F]-FLT uptake in a high TK1 expressing cell line (HCT-116) and low TK1 expressing cell line (DiFi)

Nanodendrons for Detection & Treatment of Breast Cancer Lynn E. Samuelson, Randy Scherer, Kathy Carter, Darryl J. Bornhop, J. Oliver

McIntyre and Lynn M. Matrisian Introduction

Proteinases including matrix metalloproteinases (MMPs) are important in cancer progression and other pathologies. A variety of specific proteinases, including a number of the MMPs, found in the microenvironment of tumors and tumor metastases, present opportunities for imaging and prodrug therapy (1,2). MMP expression can be used to distinguish benign from malignant tumors and identify aggressive tumors associated with poor outcome. MMP9, one of two basement membrane-degrading type-IV collagenase/gelatinases, is associated with tumor invasion and metastasis (3). Here we describe a new class of nano-dendron (ND) based imaging(ND-PB)/therapeutics with molecular recognition and targeting capabilities that when coupled self-report drug delivery. These NDs are designed specifically for enhanced treatment of breast cancer and, in particular, breast cancer metastases.

Methods

The beacon was synthesized from a polyester dendron scaffold using methods analogous to our PAMAM beacons. Imaging was completed using a xenograft mouse model with an MMP inhibitor or a d-amino acid peptide sequence (uncleavable) and tumors in a mammary mouse model. Data was acquired with both the IVIS (Caliper Life Sciences) and the Li-COR PEARL® imaging systems. A pilot study using a therapeutic ND was accomplished.

Results

In vivo imaging studies demonstrated that the ND-PB is activated selectively with MMP9. Using a D-amino acid peptide sequence and an MMP inhibitor (MMPI) did not produce significant signal from the beacon compared to cleavable beacon in the absence of MMPI. Initial experiments with the therapeutic dendron indicate that therapy can be delivered and monitored with these agents.

Discussion

A new class of nanoparticle-based imaging and therapeutic compounds has been developed with molecular recognition and targeting capabilities. These NDs have been designed with the ability to link various combinations for multifunctional purposes in early detection and targeted treatment of metastatic breast cancer. The development of this novel class of NDs expands upon the current capabilities of modern proteinase-based optical beacons and prodrugs, and is a step towards developing personalized medicine, particularly for treatment of cancer metastases.

References

1. Coussens, L. M. et al., Science, 295, 2387-2392 (2002). 2. McCawley, L. J. et al., Current Opin in Cell Bio, 13, 534-540 (2001). 3. Liotta, L. A., et al., Nature, 284, 67-68 (1980). 4. Martin, M.D., et al., Clinical Lab. Invest., 28(6), 16-18 (2005).

Evaluation of Imaging Agents Targeting the Pancreatic Beta Cell Using Co-registered Multimodal Imaging

Jack Virostko, Kevin Wilson, Ronald Baldwin, M. Sib Ansari, Todd Peterson, John Gore, and Alvin Powers

Introduction

The ability to non-invasively assess beta cell mass in vivo would increase our understanding of the pathogenesis of diabetes and aid the development of new regenerative therapies. Several targets have been proposed for imaging the beta cell including the vesicular monoamine transporter 2 (VMAT2). A positron emission tomography (PET) ligand to VMAT2, dihydrotetrabenazine ([18F]DTBZ), has shown promise for imaging the pancreatic beta cell [1]. However, clearance of [18F]DTBZ in organs near the pancreas hampers accurate PET analysis. Here we show how the use of optical data from a bioluminescent model can be used to improve the quantification of such nuclear tracers.

Methods In order to image beta cell mass, we generated mice expressing the luciferase optical reporter gene under control of the insulin promoter. An algorithm reconstructing bioluminescence source locations from their surface projections was employed to localize the pancreas in three-dimensional space [2]. The tomographic bioluminescence source reconstruction was then co-registered with the [18F]DTBZ PET image to identify the pancreas on the PET image and guide placement of a region of interest (ROI) encompassing the pancreas.

Results

The ROI generated from the tomographic bioluminescence image allowed accurate and unambiguous delineation of the pancreas in the [18F]DTBZ image, enabling accurate quantification of the pancreatic PET signal without any confounding influence of [18F]DTBZ in adjacent organs (Fig 1). Additionally, the bioluminescence intensity provided a quantitative measure of beta cell mass for correlation with [18F]DTBZ PET.

Discussion

Three-dimensional reconstruction of the optical signal from luciferase-expressing mice can delineate cells or organs of interest and be co-registered to PET images to distinguish organ-specific radiotracer uptake from other organs.

References 1. M.P. Kung et al., J. Nucl. Med. 49, 1171 (2008). 2. H. Dehghani et al., Med. Phys. 35, 4863 (2008).

Figure 1: Bioluminescence tomography (A) and co-registered [18F]DTBZ PET (B) axial images. A region of interest encompassing the pancreas is shown in red.

A.

B.

In vivo Skeletal Muscle Glycogen Measured by Chemical Exchange Saturation Transfer (glycoCEST) and 13C MRS at 7T

Theodore Towse, Sam Bearden, Adrienne Dula, Brian Welch, Jim Joers, Seth Smith, Calum Avison, and Bruce Damon.

Introduction

GlycoCEST is a molecular imaging technique that allows indirect detection of -OH protons associated with glycogen (1). With glycoCEST, the magnetization of -OH protons of glycogen is saturated, then transferred to bulk water by way of chemical exchange. This reduces the bulk water signal in proportion to the glycogen content. GlycoCEST may provide several advantages to conventional approaches for measuring glycogen including improved temporal and spatial resolution, and the ability to measure glycogen in multiple muscles simultaneously. Due to the small chemical shift difference between the -OH protons and the bulk water resonance the CEST effect may not be easily determined from the z-spectrum due to direct saturation of water (2). However, at ultra-high fields such as 7T, the spectral resolution between the glycogen -OH protons and the bulk water protons is larger, which facilitates an easier detection of the glycoCEST effect. Therefore the purpose of this study was to determine the feasibility of glycoCEST imaging in human skeletal muscle in vivo at 7T. Further, we compared the asymmetry in the CEST spectrum due to glycogen, glycoCESTMTRasym, to 13C MRS measures of muscle glycogen.

Methods

Imaging data were acquired using a dual-tuned TR 13C/1H partial volume coil. Single slice glycoCEST images were acquired from the largest cross section of the calf. ROIs were manually drawn from the calf muscles; medial - and lateral-gastrocnemius (MG, LG), and soleus (SOL) using the anatomical images as a guide. GlycoCEST was characterized as the integral of the z-spectrum within the limits 0.0-1.7 ppm. 13C MRS data were acquired from the calf muscles using a pulse acquire sequence. The integrated area of the C-1 resonance of glycogen was expressed relative to the integrated area of a 13C-Urea phantom, and corrected for muscle size.

Results

The mean (SD) CEST effect from the calf muscles was 1.12 (0.46) %. The glycoCEST effect was 1.00(0.45), 1.23(0.43), and 1.13(0.46) % for the MG, LG, and SOL respectively. There was a weak, non-significant correlation (R2 = 0.04) between the muscle-averaged glycoCESTMTRasym and the 13C MRS measures of glycogen.

Discussion With high spatial and temporal resolution, glycoCEST imaging may provide means for probing chronic diseases which have a metabolic component e.g., diabetes and McArdles’ disease. Future studies, utilizing phantoms of varying glycogen concentrations and in vivo manipulations of glycogen content, will be conducted to determine the relationship between glycoCEST and 13C MRS measures of glycogen over a broader dynamic range of values.

References

1. P.C.M. van Zijl, et al., PNAS. 11, 4359 (2007). 2. S.A. Smith, et al., MRM. 62, 384 (2009).

Figure 1: FPFDM estimated vessel size dependence of ∆R2

*

and ∆R2 on perturber size .

Modeling Magnetic Field Perturbations for Arbitrary Geometries Natenael Semmineh, Junzhong Xu, C. Chad Quarles

Introduction

Dynamic Susceptibility Contrast (DSC) MRI is a method to assess physiological characteristics of tumor vasculature such as the blood volume. It is generally assumed in DSC-MRI that the relationship between the measured transverse relaxation rate and contrast agent concentration, characterized by the susceptibility calibration factor (kp), is linear and the same for all tissue. Given the heterogeneous nature of blood vessels within normal and tumor tissue and the dependence of susceptibility field gradients on vessel geometry this assumption could significantly impact the reliability of DSC-MRI measurements. The goal of this study is to develop a computational approach to model susceptibility-based contrast mechanisms and characterize kp for computer simulated three-dimensional vascular networks.

MethodsA computer simulation technique called the Finite Perturber Finite Difference Method (FPFDM) has been developed and used to estimate the intravascular and extravascular magnetic field perturbations induced by magnetic susceptibility variations between arbitrarily shaped mesoscopic scale compartments, and the resulting gradient echo (∆R2*) and spin echo (∆R2) transverse relaxation rate changes. A computer simulated three-dimensional vascular network constructed to reflect varying vascular features (i.e. branching patterns, diameter, volume-fraction, vascular composition) will be used as an input in the FPFDM and the transverse relaxation rates will be computed for a range of contrast agent concentrations. The relationship between the computed relaxation rates and contrast agent concentration will be used to determine the relationship between kp and the underlying vascular features and to estimate the extent of kp heterogeneity across simulated normal and tumor vascular networks.

Results The computational efficiency of the FPFDM was validated by computing the perturber size dependence of ∆R2 and ∆R2* for randomly distributed cylinders. The result in (Fig. 1) is an excellent agreement with previous work [1]. Ongoing studies include the inclusion of realistic three-dimensional vascular structures and the determination of the relationship between kp and vascular structural properties.

Discussion In future studies, we aim to characterize the susceptibility calibration factor experimentally using the FPFDM for physical phantoms of known geometry and for vascular trees extracted from µCT based angiograms of normal and tumor tissue.

References 1. Boxerman, J.L., et al., Mag.  Reson.  Med.  34,  555–566 (1995).

Efficacy of a JAK-2 Inhibitor Assessed Using DW- and DCE-MRI Mary E. Loveless, Jane Halliday, Marie Pinzon-Ortiz, Corinne Reimer, Dennis

Huszar, John C. Waterton, John C. Gore, Thomas E. Yankeelov

Introduction The novel JAK-2 inhibitor AZD1480 has been shown to inhibit growth of numerous tumor xenografts. Suppression of Jak/Stat signaling by AZD1480 has been observed in tumor cells/stroma, and reducing Stat activation may inhibit tumor growth. AZD1480 has also demonstrated anti-angiogenic effects (1). We propose to use diffusion weighted MRI (DW-MRI) and dynamic contrast enhanced MRI (DCE-MRI) to compare changes in cellularity and vascularity, respectively (2,3), in tumors treated with AZD1480 to that of a VEGFR inhibitor, AZD2171.

Methods

Thirty nude mice will be injected with 106 Calu-6 cancer cells in the flank. Mice will be distributed into three treatment groups receiving 50 mg/kg of AZD1480, 6 mg/kg of AZD2171, and vehicle doses (po qd). Animals will be imaged at days 0, 3, and 5 with the following imaging protocol. DW-MRI: A PGSE sequence (gated and navigated) with a 1282 x 15 imaging matrix over a (35 mm)2 x 1 mm FOV, with b-values of 150, 500, and 800 s/mm2. T1-mapping: A saturation recovery FSE protocol with varying TRs with the same FOV as above. DCE: A GRE sequence with a temporal resolution of 25.6 s/data set. Forty image sets will be acquired during and after a bolus injection of Gd-DTPA. Both DW-MRI and DCE-MRI data will be fit to extract the apparent diffusion coefficient (ADC) (2), and vascular parameters Ktrans and ve (3). Tumor tissues will be removed and stained for H&E, Caspase3, CD31 and pSTAT-3.

Results Protocols for DW-MRI and DCE-MRI have been developed as evident from Figure 1, and preliminary data collection is currently underway.

Discussion We hypothesize that both Ktrans and ADC will be sensitive metrics to AZD1480 treatment response. While changes in Ktrans may be more significant with AZD2171 since it is an anti-angiogenic agent, significant changes in ADC may be noted earlier with AZD1480 due to the Stat3 inhibition. Additionally, results found here may suggest useful imaging metrics for subsequent AZD1480 clinical trials.

References

1. M. Hedvat et al., Cancer Cell. 16, 487 (2009). 2. T. L. Chenevert et al., Clin Cancer Res. 3, 1457 (1997). 3. Yankeelov T.E., et al. Curr Med Imaging Rev. 3, 91 (2009).

Figure 1: Ktrans, ve, and ADC maps of a mouse before treatment with AZD 1480.

Microenvironmental Influences in Early Breast Cancer Metastasis to Liver

Michelle D. Martin, Gert-jan Kremers, Kurt W. Short, John V. Rocheleau, David W. Piston, Thomas E. Yankeelov, Chad C. Quarles, D. Lee Gorden, and Lynn

M. Matrisian Introduction

Two-thirds of women with metastatic breast cancer commonly have spread to the liver, and those patients that develop liver metastases are known to make up a poor prognosis group with median survival rates of less than 6 months and a reduced response rate to systemic therapy [1,2]. The goal of this project was to assess the impact of the liver microenvironment on the initial arrest and growth of the breast tumor cells using innovative imaging methods including two-photon microscopy and cellular MRI.

Methods

For all studies, polyoma virus middle T antigen (PyVT)-derived mammary tumor cells expressing GFP were injected into the spleen of female FVB/n mice. For microscopy, mice were injected via the spleen with a tomato lectin-quantum dot complex that binds to and labels the vasculature of the liver at specific time points, and the liver was imaged immediately. For studies involving MRI, mice were injected via the spleen with PyVT cells loaded with a .9µm MPIO agent containing a rhodamine tag. Mice were sacrificed at 72 hrs post-injection and livers were removed and prepared for MRI imaging.

Results

Two-photon analysis of the livers revealed that the majority of tumor cells moved from intravascular to extravascular between 12 and 24 hours. MRI analysis determined that livers of mice containing tumor cells loaded with MPIO particles displayed dark voids that are characteristic of cells containing iron oxide, in contrast to control livers (Fig 1).

Figure 1. Ex-vivo MRI of control liver (A) and liver loaded with MPIOs (B). Inserts show magnified images, with arrows in (B) denoting signal voids.

Discussion

The use of both microscopy and cellular MRI enabled the study of early metastasis both at the single cell level and throughout the entire organ to study cellular location and seeding of tumor cells in the liver. The data gained from these studies will add to the knowledge of breast cancer metastasis in a clinically relevant site.

References 1. G. Pentheroudakis et al., Breast Can. Res. Treat. 97, 237 (2006). 2. P. Pouillart et al., Ann. Gastro. Hepatol. 21, 87 (1985).

Predicting Breast Cancer Treatment Response with Functional and Molecular Imaging Techniques

Jennifer G. Whisenant, Michelle D. Martin, Mary E. Loveless, Todd E. Peterson, H. Charles Manning, John C. Gore, Thomas E. Yankeelov

Introduction

While functional and molecular imaging techniques have the ability to quantify treatment-induced changes in cance, effective translation of these techniques into clinical practice has yet to be realized as we do not adequately understand the sensitivity and reliability of imaging protocols can predict therapeutic efficacy.1 This project will systematically evaluate clinically relevant imaging methods to define a set of techniques that can predict tumor response to treatment in a murine model of breast cancer. Multimodal imaging studies combining MRI, PET, and SPECT will be performed to observe changes in vascular, cellular, and molecular characteristics of primary tumors in response to treatment. We will correlate imaging data with histology to investigate the underlying cellular changes that accompany response.

Methods Female athymic nude mice implanted with either trastuzumab responsive or resistant breast cancer cells will be imaged with MRI (DWI, DCE), and PET (FDG, FLT, FMISO) or SPECT (99mTc-Annexin-V). Mice will receive two doses of trastuzumab: an initial dose and a second dose 72 hours later. Mice will be imaged at three time points: baseline, 24 hours after first dose, and 24 hours after the second. At each time point ADC will be calculated from DWI data, DCE-MRI data will be analyzed to obtain estimates of vascular characteristics, and nuclear probe accumulation will be quantified via the standardized uptake value (SUV). Imaging data will be co-registered, and ROIs within the tumor as well as parametric maps will be compared between the responsive and nonresponsive cohorts. After the third imaging time point, mice will be sacrificed and excised tumors will be stained to quantify cell density, vascularity, proliferation, hypoxia, and apoptosis. The sensitivity and accuracy of the imaging data in predicting treatment outcome will be assessed.

Results Accurate image co-registration between modalities is required to compare identical regions within the tumor. Figure 1 presents an SUV map from FLT-PET data (a) co-registered with the corresponding ADC map (b). The contour line along the tumor boundary and crosshairs indicate the accuracy of the registration results.

Discussion

The main objective of this study is to define a set of imaging techniques that can separate non-responders from responders at the earliest possible time point. Furthermore, this work will provide a paradigm by which other imaging protocols as well as current and emerging breast cancer treatments might be evaluated.

References 1. V.N. Harry et al., Lancet Oncol. 11, 92 (2010).

Figure 1: SUV map of FLT uptake (a) co-registered with the corresponding ADC map (b).

a b

A Comparison of DCE-MRI Models in Human Breast Cancer Xia Li, E. Brian Welch, Lei Xu, Lori Arlinghaus,

John C. Gore, Thomas E. Yankeelov Introduction

Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) can be used to provide information on tumor physiological status. By fitting DCE-MRI data to a proper pharmacokinetic model, physiological parameters related to vessel perfusion and permeability (Ktrans) or extravascular extracellular volume fraction (ve) can be extracted. Four models have been proposed with different assumptions: standard and extended Tofts models [1] with the fast exchange limit assumption, and the fast exchange regime models (FXR) [2] with and without the shorter T1 component (T1s). In this study we perform a rigorous comparison between these four models by applying four common statistical measures used to assess model accuracy.

Methods

A Philips 3.0 T Achieva MR scanner was used to obtain DCE-MRI data on ten breast cancer patients prior to and after one cycle of neoadjuvant chemotherapy, yielding 16 usable data sets. All available tumor voxels for each data set are input to the four models described above to extract pharmacokinetic parameters. The standard chi-square test (χ2) was used to assess the goodness of fit. The Durbin-Watson statistic (DW) was computed to detect serial correlation of residuals. Both the Akaike Information Criteria (AIC) and the Bayesian Information Criterion (BIC) are used to detect the balance between the goodness of fit and the model complexity, with the BIC applying a heavier penalty on the complexity.

Results

The extended Tofts model resulted in 63.20% voxels with smallest χ2, while the other three models led to 12.20%, 23.91%, and 0.70%, respectively. The extended Tofts model also resulted in a reduction in percentage of voxels showing serial correlation of residuals: 0.94%, compared with other 3 models (7.81%, 2.30%, and 3.57%). The AIC and BIC also suggested the extended Tofts model obtained the best balance between the complexity and goodness of fit. Details are shown in the table.

Tofts (Standard) Tofts (Extended) FXR without T1s FXR with T1s

Durbin-Watson 7.81% 0.94% 2.30% 3.57%χ2 12.20% 63.20% 23.91% 0.70%AIC 19.28% 58.17% 22.03% 0.53%BIC 28.94% 51.37% 19.26% 0.43%

Discussion In order to accurately assess changes in Ktrans and ve during therapy, it is important to choose a proper model. The statistical metrics show that the extended Tofts model and the FXR model without the shorter T1 component are statistically superior to other DCE-MRI models in most voxels across all patients. This has important implications for analysis of the breast DCE-MRI data.

References

1. P.S. Tofts et al., J. Magn Reson Imaging. 10, 223-232 (1999). 2. Woessner DE, J Chem Phys. 35, 41-48 (1961).

Improving Diffusion-weighted Imaging of the Breast at 3T Lori R. Arlinghaus, E. Brian Welch, John C. Gore, Thomas E. Yankeelov

Introduction

The apparent diffusion coefficient (ADC) is a potential biomarker for response to chemotherapy in breast cancer tumors (1). However, ADC maps are calculated from diffusion-weighted images (DWIs), and high quality data using conventional approaches are difficult to acquire in the breast. The overall goal of this project is to improve image acquisition and post processing methods in a longitudinal study of 3T DWI of the breast.

Methods DWIs of the breast were acquired at 3T with a single shot spin echo (SE) echo planar imaging (EPI) sequence. Spectrally selective attenuated inversion recovery (SPAIR) fat suppression was used to reduce image artifacts, and affine registration was investigated for motion correction. Nonlinear registration (2) and B0 field maps (3) were investigated as potential tools for correcting distortion, and sensitivity encoding (SENSE) was used to decrease the single-shot EPI factor, which reduces distortion.

Results Improved fat suppression was achieved by using SPAIR instead of SPIR (spectral presaturation with inversion recovery) fat suppression. Motion correction reduces the variance by 90% on average between individual ADC maps used to calculate the mean ADC; however, it does not significantly change the mean ADC value averaged over an ROI. B0 field map correction and nonlinear registration failed to provide consistent distortion correction. SENSE reduced image distortion due to susceptibility differences by increasing the bandwidth in the phase encoding direction from 13 Hz to 26 Hz; however, it introduced subtle ghosting artifacts.

Discussion Significant improvements in image quality have been made by implementing SENSE and SPAIR fat suppression. Although subject motion does not appear to significantly affect the mean ADC value averaged over the tumor ROI, motion correction may be necessary for voxel-wise analyses. Other goals of this project include work to improve image acquisition through the implementation of improved shimming techniques, such as higher-order, image-based shimming or dynamic shimming, and the implementation of multi-shot EPI or spiral imaging.

References

1. J. Galons, et al., Neoplasia 1, 113-117 (1999). 2. G. K. Rohde, et al., IEEE Trans. Med. Imag. 22, 1470 (2003). 3. P. Jezzard, R.S. Balaban, Magn. Reson. Med. 34, 65 (1995). 4. T. Netsch, A. van Muiswinkel, IEEE Trans. Med. Imag. 23, 789 (2004).

2mm /s

10e-8

5

1

mm /s

4

2

2.5e-3

1.5

0.5

A B C

D E

Motion correction of DWIs acquired with SPAIR and SENSE. A) Tumor (white outline) shown in the b = 0 mm2/s image. Variance in the unregistered (B) and registered (C) ADC maps acquired in the x, y, and z directions. Mean ADC values in the unregistered (D) and registered (E) images.

Poster Session 1 (Monday, 4:00 - 5:00)Poster Session 1 (Monday, 4:00 - 5:00) Presenter

Effects of Membrane Integrity and Edema in Magnetic Resonance Imaging Studies of Inflammatory Myopathies Nathan Bryant

Validation of DTI-Tractography-based Measures of Primary Motor Area Cortical Connectivity Yurui Gao

Parametric Mapping of Biological Tissues Using Temporal Diffusion Spectroscopy Susan Kost

Comparison of qMT Techniques with Optimal Schemes Ke Li

Synthesis and Evaluation of COX-2 PET Imaging Probes Don Nolting

MR Measurement of Cerebral Blood Volume in the Hippocampus Swati Rane

Synthesis of Metabolic Tracers for Real Time Hyperpolarized MRI of Breast Cancer Roman Shchepin

Development of a Clinically Relevant 3 T Breast Protocol David Smith

Imaging of Hyperpolarized 13C and 15N Tracers Diana Smith

Subject-specific CFD Modeling of the Vertebro-basilar System Amanda Wake

Detecting Tumor Early Response to Chemotherapy Using Temporal Diffusion Spectroscopy: How Early Can We Get? Junzhong Xu

Poster Session 2 (Tuesday, 10:00 - 11:00)Poster Session 2 (Tuesday, 10:00 - 11:00)

Microcalcification Detection using Susceptibility Weighted Phase Imaging: Cross-correlation and Relative Magnetic Susceptibility Difference Methods

Richard Baheza

Preclinical Evaluation of TSPO Ligand [18F]PBR06 for PET Imaging of Glioma Jason Buck

Structural Complexity of Cortex and its Implications on Cortical Connectivity Measurements using MR Tractography Ann Choe

Development of NMR Probe Heads for Hyperpolarization of 13C via Para Hydrogen Induced Polarization (PHIP) at 12.0 mT Aaron Coffey

Molecular Basis of TSPO as a Cancer Imaging Biomarker Saffet Guleryuz

Probing Demyelination using High Resolution 3D Quantitative Magnetization Transfer (qMT) and Diffusion Tensor Imaging (DTI) in a Lipopolysaccharide (LPS) Model of Multiple Sclerosis (MS)

Vaibhav Janve

Poster Session 2 (Tuesday, 10:00 - 11:00) (continued)Poster Session 2 (Tuesday, 10:00 - 11:00) (continued)

An Information Theory Approach to Parameter Estimates from Multi-Compartment Models of MRI Contrast Chris Lankford

Magnetic Resonance Spectroscopic Imaging of Brain at 7 T Indrajit Saha

Parallel Image Reconstruction for 7T MRI Sepideh Shokouhi

Correlating 18FLT Uptake with Drug Delivery using MALDI-IMS Adam Smith

Increased Differentiation of the Lateral Pain Network by High Resolution fMRI at 7T Elizabeth Ann Stringer

High Resolution FMRI Mapping of Cortical Plasticity Following Spinal Cord Injury in Non-Human Primates Xiang Ye

Poster Session 3 (Tuesday, 3:45 - 4:45)Poster Session 3 (Tuesday, 3:45 - 4:45)

Extracting Proliferation Rates from ADC Data Nkiruka Atuegwu

Quality Assurance of B1 RF Pulses for Parahydrogen Induced Polarization Raul Colon Moreno

High Resolution MRI at 7 Tesla to Evaluate the Anatomy of the Human Midbrain Dopamine System Mariam Eapen

Correlating MALDI and MRI Biomarkers of Breast Cancer Amelie Gillman

Development of Materials for TSPO-Directed HTS Matthew Hight

An Optimized Composite Refocusing Pulse for 7T MRI Marcin Jankiewicz

Radiation Dose-Based Comparison of PET and SPECT for Bone Imaging Lindsay Johnson

Resting State Functional Connectivity Analysis: a Potential Model for Human fMRI Studies Arabinda Mishra

Identifying EEG Correlates of Activity in the Working Memory Network Measured with fMRI Allen Newton

Partially Loaded Travelling Wave MRI Sasidhar Tadanki

In Vivo Mouse Kidney Imaging Feng Wang & Noor Tantawy

Comparison of Reduced-FOV Techniques at 7T Chris Wargo

Extracting Proliferation Rates from ADC Data Nkiruka C. Atuegwu, Daniel C. Colvin, Mary E. Loveless, John C. Gore,

Thomas E. Yankeelov Introduction

Mathematical models of tumor growth are often parameterized by quantities that can be obtained only invasively. Our goal is to use non-invasive imaging data to model and predict tumor growth [1]. Here we show how to use data from diffusion-weighted magnetic resonance imaging (DW-MRI) to estimate proliferation rates in a rat model of brain cancer.

Methods Twelve rats underwent DW-MRI at 12, 13, and 15 days post inoculation with 9L tumor cells. Eight animals received treatment with 1, 3-bis (2-chloroethyl)-1-nitrosourea(BCNU) on day 12, immediately after the first imaging session. Apparent diffusion coefficient (ADC) values were calculated and used to estimate the cell number in each voxel as given by Eq. (1):

ADC(r,t) = ADCw – N(r,t), (1) where ADC(r, t) is the ADC of the voxel located at position r and time t, ADCw is the ADC of

free water, N(r,t) is the cell density at position r and time t, and is a proportionality constant. This model was then used in conjunction with the logistic model of tumor growth [2] to estimate each voxel’s proliferation rate; the logistic model is given by Eq. (2):

0

*

0 0

,0,

,0 ,0k r t

N rN r t

N r N r e ,

where N0 is the initial number of cells, k is the proliferation rate, and is the voxel’s cell carrying capacity. For each animal the pre-treatment and 72 hour imaging data were co-registered and used to calculate k for each voxel within the tumor. The ratio, R, of negative

to positive proliferation values for each rat was calculated.

Results

The distribution of proliferation values is shown in Figure 1. A student’s t-test on the R values showed a significant difference (p = 0.03) between the control and treatment rats.

Discussion We have shown that sequential ADC data can be used to estimate proliferation values of a tumor and these values are significantly different between treated and untreated animals. Future work will involve serial DW-MRI during therapy over a two week period. The proliferation values from the early DW-MRI data will be used to predict tumor response to treatment as measured at later time points. Since all data to drive the model is obtained noninvasively, we have the ability to experimentally validate (or refute) the mathematical model; something not possible with most models of tumor growth.

References 1. N. C. Atuegwu, et al., Physics in Medicine and Biology 55, 2429 (2010). 2. H. M. Bryne, Cancer Modelling and Simulation, 75 (2003).

Figure 1:Histogram of proliferation values for the control and the treated rats.

(2)

Figure 1. The CMMV (panel A) and Dc r (panel B) for the

simulated and experimental data are plotted as function of SNR.

|Dcr|

SNR

0 10 20 30 40 50 60 70 80

|Dc

r| (p

pm

)

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

Sim

Exp

CCMV

SNR

0 10 20 30 40 50 60 70 80

CC

MV

0.80

0.85

0.90

0.95

1.00

1.05

1.10

Sim

ExpA B

Microcalcification Detection using Susceptibility Weighted Phase Imaging: Cross-correlation and Relative Magnetic

Susceptibility Difference Methods Richard Baheza, E. Brian Welch, John C. Gore, Thomas E. Yankeelov

Introduction Calcium deposits in the breast can be early indicators of cancer. Since calcium has a different magnetic susceptibility than water and tissue, we evaluated, in simulations and in experiments, the conditions under which microcalcifications can be detected in practice by susceptibility weighted imaging (SWI).

Methods Simulations. First, a “template” for a 1 mm3 calcification is constructed as described in (2). A series of 3D phase image “data” sets are then generated using different SNR levels. We then compute the cross correlation matrix between the template and simulated data sets. The cross correlation matrix maximum value (CCMV) was used to identify the calcification signature. We also computed the relative magnetic

susceptibility difference map, Dcr, using the Salomir method (3).

Phantom studies. Phantoms were constructed using a 1 mm glass bead with c = -11

ppm, immersed in agar gel with c = -9 ppm; thus, |Dcr|= 2 ppm, just as for calcium and water. 4.7T, 7T and 9.4T Varian MRI scanners obtained 3D gradient echo

images similar to those employed in the simulations. The CCMV and the Dcr values were then computed for the experimental images

Results Fig. 1 depicts the

CCMV and |Dcr| mean and standard error as a function of SNR for simulated and experimental data. For both the

CCMV and Dcr, the experimental results appear to follow the response predicted by the simulations. These results suggest

little change in Dcr and CCMV are to be expected for SNR values above 20.

Discussion We have compared two different techniques to locate the susceptibility induced

signature of a 1 mm object by computing the Dcr values and the cross-correlation between phase data and a template. The results suggest a SNR ≥ 20 are required for both methods to work. Ongoing studies are exploring the SNR and resolution needed to locate calcium like objects as small as 0.5 mm.

References

1. E. M. Haacke, X. Yingbiao, Magn. Reson. Med. 52, 612-618 (2004). 2. R. Baheza, E. B. Welch, Proc. ISMRM 18th, 4467 (2009). 3. R. Salomir, B Denis, Conc. in Magn. Reson. B 19B, 26-34 (2003).

Effects of Membrane Integrity and Edema in Magnetic Resonance Imaging Studies of Inflammatory Myopathies

Nathan Bryant, Jane Park, and Bruce Damon Introduction

Idiopathic inflammatory myopathies (IIM) exhibit pathology and structural aberrations that are complex and include loss of membrane integrity, edema, adipose infiltration and fibrosis in muscle tissue. Magnetic resonance imaging (MRI) and spectroscopy (MRS) are sensitive to these biophysical changes and are valuable in both research and clinical application [1]. Yet correlations between the MR-based observations and clinical presentation of IIM are inconsistent. The goal of this study is to elucidate the pathological basis for these MR data using simplified animal models followed by correlative validation using quantitative histology. To do this, we will first assess the effects of membrane permeability and edema in healthy mice and in a subsequent experiment we will compare those results to MR data from a prednisone treated and untreated mouse model of the IIM polymyositis (PM).

Methods

Mice overexpressing aquaporin-4, to increase membrane permeability [2], and/or injected with 1% λ-carageenan, to elicit edema, will be compared to control mice by MRI/MRS and quantitative histological analysis. The MR data will assess edema, disruption of myofibrils, membrane integrity, fibrosis, and perfusion/volume fraction parameters. These data will be registered to and complemented by quantitative histology evaluating the same features. The same MR and histological measures will be collected on a mouse model of PM (Syt VII-/) and will again be compared to C57/BL6 mice. In addition to the pathology of the Syt VII -/- mice, we will also evaluate the effects of treatment with prednisone and the effects of disease progression by repeating the measures 8 weeks later in a second set of animals.

Results The edema condition should result in bi-exponential T2 relaxation and an increase in extracellular water volume; as calculated from histology and MR. In the animals with aquaporin overexpression, the increased membrane permeability should be reflected in the diffusion data (λ3, long ∆). No other differences from control mice are expected in the remaining outcome measures. In the following experiment, we expect all histological and MR markers of myopathy to increase in the Syt VII-/- at both time points and should be reduced by treatment with prednisone.

Discussion

Many aspects of the pathology observed in IIM are common to neuromuscular disorders in general [3]. An enriched understanding of the underlying basis of these MR parameters and their subsequent translation into clinical application would have great impact for development of treatments and management of care for individual patients.

References 1. J. Qi et al., JMRI. 27, 1 (2008) 2. Y. Wakayama et al., Micron. 38, 3 (2007) 3. J. Tidball and M. Wehling-Henricks, Curr Opin Rheumatol. 17, 6 (2005)

Preclinical Evaluation of TSPO Ligand [18F]PBR06 for PET Imaging of Glioma

Jason R. Buck, Eliot T. McKinley, Matthew R. Hight, Saffet Guleryuz, Ping Zhao, Allie Fu, Todd E. Peterson, M. Noor Tantawy, Dewei Tang, and H. Charles Manning

Introduction

Translocator protein (TSPO), formerly known as the peripheral-type benzodiazepine receptor (PBR), is a crucial trans-mitochondrial membrane protein involved in numerous cellular functions including cholesterol metabolism, steroidogenesis, and apoptosis. Elevated expression of TSPO is found in numerous diseases that range from neuroinflammation to cancer. As an imaging target, TSPO is well-documented in neuroscience [1], but its role in oncology is less established. Here we report quantitative PET imaging of glioma with [18F]PBR06, a high-affinity TSPO imaging ligand, in preclinical rodent models.

Methods

[18F]PBR06 was synthesized with high specific activity using published methods [1]. Two weeks prior to imaging, Wistar rats were stereotactically inoculated in the right hemisphere with 10K tumor cells (C6/9L rat glioma) or vehicle. Healthy and tumor-bearing rats were imaged in a microPET Focus 220 system and a 90 min dynamic acquisition was started simultaneously with the injection of ~ 7.4 MBq/0.2 mL [18F]PBR06. Arterial blood was collected to derive the arterial input function (AIF). HPLC radiometabolite analysis was performed on selected blood samples. Compartmental modeling of PET data was performed using a corrected AIF and PMOD program.  

Results

[18F]PBR06 was found to preferentially accumulate in C6 and 9L tumors with little uptake in healthy brain, facilitating excellent contrast between tumor and normal tissue. In contrast to human studies, [18F]PBR06 radiometabolites were not observed over a 90 min acquisition. Time-activity curves (TACs) deduced from tumor and healthy brain regions could be fit to a 3-compartment, 4-parameter model, providing estimates of K1, k2, k3, and k4 in tumor and healthy brain.

Discussion [18F]PBR06 appears to be a promising PET tracer to visualize TSPO expressing tumors such as glioma. Additionally, we anticipate that the [18F]PBR06 scaffold may represent an effective lead for future probe development for cancer imaging.

References

1. E. Briard et al., J. Med. Chem. 52, 688–699 (2009).

Figure 1. [18F]PBR06 PET image of C6 glioma (A); [19F]PBR06 displacement (B).

(a) (b)

Structural Complexity of Cortex and Its Implications on Cortical Connectivity Measurements Using MR Tractography

Ann S. Choe, Yurui Gao, Iwona Stepniewska, Xia Li, Zhaohua Ding, Adam W. Anderson

Introduction

The study of anatomical connections often involves tracing fiber bundles to or from the cortex using Magnetic Resonance (MR) tractography. Two major challenges to this approach exist. One is the low diffusion anisotropy in gray matter, and the high directional uncertainty this causes. This problem is often circumvented by including the border of white matter in seed regions for fiber tracking. However, effectiveness of this solution is limited by the second challenge, which is the structural complexity of the cortex itself. In this abstract the challenges of cortical connectivity measurements using MR tractography are investigated by analyzing histological sections at the border between white matter and gray matter.

Methods

A fixed squirrel monkey brain was scanned using a multi-slice, pulse gradient spin echo sequence (32 weighting directions, b = 0 and 1022 s/mm2, 0.3 x 0.3 x 0.3 mm3 voxels resolution). After scanning, the brain was sectioned and stained for myelin for comparison with MR data. A few subcortical regions were observed with light microscopy to reveal the underlying microstructure.

Results

Fiber structure in subcortical regions varies greatly from one region of cortex to another. Figure 1 shows an example of such structures. Figure 1(a) and (b) shows a region where fibers from the gray matter cross a large bundle after traveling only a short distance from the border. Figure 1(b) shows overlaid diffusion fibers that fail to track the bundles stemming from gray matter. Figure 1 also reveals the possibility of following fibers that do not originate in the seed volume.

Discussion

It was shown that the underlying microstructure of the gray/white matter interface is complex and placing seed regions in the subcortical white matter for the purpose of fiber tracking from cortical regions may result in incorrect connectivity information.

References

1. A.S. Choe et al., in 15th ISMRM. Berlin, Germany (2007) 2. A.S. Choe et al., in 16th ISMRM. Toronto, Canada (2008) 3. H. Jiang et al., Comput Methods Programs Biomed. 81, 2 (2006)

Figure 1. High resolution micrographs of the gray and white matter interface. (a) Interface where fibers stemming from gray matter cross a large tangential bundle. (b) Diffusion fibers overlaid on the same region. Incorrect tracing of the original fiber bundles stemming from gray matter can be observed.

Development of NMR Probe Heads for Hyperpolarization of 13C via Para Hydrogen Induced Polarization (PHIP) at 12.0 mT

Aaron M. Coffey, Raul D. Colon, Kevin W. Waddell, and Eduard Y. Chekmenev Introduction

Hyperpolarization techniques can increase the nuclear spin alignment from several parts per million to the order of unity, leading to significant signal enhancement on MRI scanners. 13C-enriched tracer compounds are hyperpolarized via spin order transfer from parahydrogen to 13C through a suitable pulse sequence. An efficient radio frequency (RF) circuit is required that is double tuned for irradiation of 1H and 13C, generating intense RF fields without sacrificing sensitivity for the 13C detection.

Methods

A double resonance probe head for 1H and 13C was constructed according to previous work by Zhang et al (1998). At a B0 field strength of 12.0 mT, the 1H and 13C frequencies corresponded to 510 kHz and 128 kHz. The addition of a parallel LC bandstop filter centered at the 1H frequency achieved greater channel isolation.

Results

As seen in Figure 1, the NMR probe head at the requisite frequencies was tuned and matched on the two channels.

Discussion

RF power calibration yields 1H (510 kHz) B1 = 7 kHz at 20 W and 13C (128 kHz) B1 = 1.3 kHz at 30 W at 12.0 mT. Proper calibration of RF on a two channel spectrometer to maintain phase relationships between successive pulses in the hyperpolarization transfer sequence is expected to increase spin transfer beyond the 1% currently observed. This RF coil is sufficient to yield hard pulses compensating for B0 inhomogeneities and NMR sensitivity, permitting hyperpolarization detection at 12.0 mT within the current system. Future work includes automation of the production of hyperpolarized metabolic contrast agent for reproducibility and quality assurance for in vivo work.

References

1. Zhang et al., J. Magn. Reson. 132, 167 (1998). 2. Hoevener et al., Magn. Reson. Phys. Biol. Med. 22, 123 (2009).

Figure 1: a) circuit diagram and b) frequency sweep response of dual resonant circuit operating at 12.0 mT.

13C1H filter

CT13C

CM13C CT1H 1HLS

LG LM1H

0.1 0.2 0.3 0.4 0.5 0.6 0.7

15

20

25

Frequency sweep response

Frequency (MHz)

Prob

e ou

tput

am

plitu

de (a

.u.)

ba

Quality Assurance of B1 RF pulses for Parahydrogen Induced Polarization

Raúl D. Colón Moreno, Eduard Y. Chekmenev, Kevin W. Waddell Introduction

The parahydrogen induced polarization (PHIP) method has proven to overcome the sensitivity limitation of MR spectroscopy by achieving signal enhancement factors of up to 100,000 [1]. However, PHIP polarizing equipment is not commercially available and, therefore, our lab is focused on development of new PHIP polarizer. The new instrumentation requires a series of quality assurance experiments in order to achieve reproducibility in hyperpolarized MR experiments. The goal of this study was to perform RF pulse (B1) power calibrations for 1H and 13C channels of dual tuned RF circuit of the polarization transfer pulse sequence of PHIP.

Methods Power calibrations for transmit RF pulses were performed using an 11.7 T high resolution NMR spectrometer to polarize standard samples and detect 13C or 1H. The NMR samples used were (i) a deuterated chloroform (residual 1H T1 = 73 ± 2 s at 11.7 T) for 1H channel and (ii) a saturated solution of sodium 1-13C-acetate-d3 in D2O (13C T1 = 59 ± 2 s at 4.7 T) for 13C. After polarizing the sample for at least 3*T1 in the high-field NMR spectrometer, the sample was moved to the low-field spectrometer (12.0 mT), where an excitation pulse (1H or 13C) was applied using a dual tuned RF circuit operating at 510 kHz and 128 kHz respectively. The sample was taken back to the high-field spectrometer for MR detection using “pulse and acquire” pulse sequence. The total time that the polarized sample spends in low field conditions is significantly shorter than T1. The procedure was repeated by incrementing RF power of excitation pulse with constant width. The experiment is fitted using a cosine function, Fig. 1.

Results and Discussion For 1H calibrations, the optimum voltages obtained for a 141.2 µs long pulse were 0.85 Vpp (18W) for a 180° RF pulse. For the 13C B1 calibrations using a pulse width of 375 µs, the optimum voltages obtained were 1.05 Vpp (28W) for a 180°. Using these optimum parameters of the PHIP method ensures that the RF pulses in the polarization transfer sequence are indeed applied correctly, therefore, yielding the desired signal enhancements.

References 1. P. Bhattacharya et al., J. Magn. Reson. 186, 150-155 (2007). 2. J.B. Hoevener et al., Magn. Phys. Biol. Med. 22, 123-124 (2009).

Figure 1: Power calibration of RF pulses in 12.0 mT polarizer for 13C channel at 128 kHz.

Figure1: 3D FFE scans in a single subject with midbrain structures segmented to show volume of the VTA, SN and RN. The scan has a voxel resolution of 0.4 x 0.4 x 2mm.

High Resolution MRI at 7 Tesla to Evaluate the Anatomy of the Human Midbrain Dopamine System

Mariam Eapen, David H. Zald, J. Chris Gatenby and John C. Gore Introduction

The ventral tegmental area (VTA) and substantia nigra pars compacta (SNc) are subcortical structures in midbrain containing neurons that synthesize the neurotransmitter dopamine (DA). These dopaminergic neurons are known to be associated with behaviors (in rodents and non-human primates) such as reward related learning and novelty processing1. Accurately characterizing the anatomical contours of these DA structures in humans can be useful for localizing blood oxygen level dependent responses in fMRI studies and for evaluating structural changes associated with neuropsychiatric diseases. The current study aims at identifying optimal methods for evaluating the anatomical architecture of the midbrain DA regions comparing 2D Gradient and Spin Echo (GRASE) and 3D Fast Field Echo (FFE) MR pulse sequences at 7 Tesla (7T).

Methods

Ten normal healthy participants (18-40years of age) were imaged on the Philips Achieva 7T MRI scanner with a 16 channel receive head coil. Transverse images were acquired with slices spanning the dorso-ventral axis of the midbrain from the basal ganglia to the pons. GRASE and FFE MR images were compared to i) observe contrast to noise ratios (CNRs) and ii) delineate boundaries of midbrain structures for segmentation.

Results

Both GRASE and FFE scans showed MR contrast between the SNc, VTA and adjacent regions. The FFE scan was also able to detect subtle vascular details. The FFE CNR calculations were slightly higher between the VTA and SNc compared to the GRASE scans. Segmentation of these midbrain structures using an in-house region growing algorithm3 enabled the creation of a volume template map (See Figure 1).

Discussion

Recent anatomical MR studies at 4T have shown distinctions between tissues in the midbrain for T2 weighted pulse sequences4. In the present study at 7T, both GRASE and FFE scans provide finer anatomical detail of structures and vasculature in the midbrain regions. The contrast and resolution at 7T promises to provide new insights into the functional architecture of this important brain region.

References

1. Kakade, S et al., Neural Netw. 15, 549 (2002). 2. Li, et al., IEEE Trans Image Process. 17, 1940(2008) 3. Manova, E.S et al., Am J Neuroradiol. 30(3):569-74 (2009).

Validation of DTI-Tractography-based Measures of Primary Motor Area Cortical Connectivity

Yurui Gao, Ann Choe, Iwona Stepniewska, Lisa Li, Adam Anderson

Introduction DTI based tractography is used to investigate cortical connectivity of brain noninvasively. However, the accuracy of this measure is not well validated. The goal of this project was to evaluate the agreement between connectivity derived from DTI tractography and from corresponding histological information from the squirrel monkey.

Methods

A bidirectional neural tracer (biotinylated dextran amine, BDA) was injected into the forelimb movement representation territory within the M1 area of a squirrel monkey. After sacrifice, ex-vivo DTI imaging of the brain on a 9.4T scanner was performed. Then the brain was sectioned in the coronal plane and every sixth section was reacted for BDA. For analysis, the injection region was segmented from histological space and aligned to the DTI space as a seed region for tractography. Then the FACT method [1] was performed to obtain the DTI fibers terminating in projection regions. BDA stained fibers were obtained by segmenting histological images and stained neuron cells were counted.

Results Comparing the DTI fibers to the BDA stained fibers revealed that DTI tractography may provide false positive and false negative connectivity to M1 (Fig 1). Comparison in the true positive projection regions shows that DTI fibers have roughly the same distribution ratio with stained neuron cells which approximately represent M1 afferent fibers.

Discussion

Quantitative comparison between anatomical and DTI-based connectivity is difficult due to limitation of the histological and image processing procedures. To make more reliable comparisons, more brain samples should be analyzed and more DTI tractography methods for measuring connectivity should be validated by this proposed strategy. Causes for the failure of DTI tractography should be identified and made the focus of future improvement of tractography methods.

References 1. S. Mori et al., Ann. Neurol. 45, 265 (1999).

Figure 1. Visualization of locations of DTI fiber terminals in 3D brain displaying injection/ projection regions from (A) front view and (B) lateral view. (I-Injection sites, P-Projection region, T-DTI fiber terminals)

PSMA

PSMA

PM1

PM1

IM1

IM1

PM1

TM1

PPMc

TSMA

TPMc

PCl PBg-Tha

A

．

B

．

Correlating MALDI and MRI Biomarkers of Breast Cancer Amelie Gillman, Kevin Wilson, Erin Seeley, Julie Sterling, Rachelle Johnson,

Julie Sterling, Thomas Yankeelov, Lynn Matrisian, John Gore

Introduction Specialized magnetic resonance imaging (MRI) methods can non-invasively assess tissue properties that change with pathology or in response to treatment. Matrix-assisted laser desorption ionization (MALDI) mass spectrometry methods can identify protein signatures in cancerous tissue. The up-regulation of numerous proteins has been associated with tumorigenesis and may be an independent indicator of patient prognosis in several kinds of cancer [1]. The correlation of MRI and MALDI measurements may assist the interpretation of MALDI data and clarify what changes in protein expression influence the contrast in MR imaging.

Methods In the first phase of this study, MALDI and diffusion-weighted (DW) MR data were acquired using C6 brain tumors in rats. Data were coregistered via methods described in [2] and correlations between diffusion rates and protein profiles in anatomical structures were calculated. In the second phase, protocols will be developed to image and correlate data from breast cancer metastases to bone. MRI data acquisition will be expanded to include gadolinium contrast-enhanced, DW, and relaxometric data in an intra-tibial mouse model of metastatic breast cancer. Hybrid MALDI/MRI hind limb datasets will be generated as in Figure 1. Data analysis will focus on identification of specific (groups of) proteins that most strongly correlate with variations in multi-parametric MRI data.

Results From Phase I of this project, protein and diffusion metrics correlate significantly (p ≤ 0.05) in 44.0% of 2280 comparisons (uncorrected for multiple comparisons). Further analyses revealed that significant correlations between protein and MR data extend to metrics other than diffusion.

Discussion Coregistration of these data sets offers opportunities for novel insights into how spatial variations in MR parameters correlate with protein changes in tissues. Some of the protein changes may directly influence changes in MR properties such as relaxation rates, while others may indicate changes in protein expression that lead to biophysical changes to which MR is sensitive. Computational methods using the tibia as a fiducial marker within deformable tissue will provide a foundation upon which to base further multi-modal and longitudinal coregistration algorithm development.

References 1. N Nishida et al., Vascular Health and Risk Management, 2, 213 (2006). 2. T. Sinha et. al., Nature Methods 5, 57 (2008).

Figure 1: Coregistered blockface (RBG), MRI (grayscale), and MALDI (false color) data. Left colorbar: relative protein concentration. Right color-bar: relative MR signal intensity. (Axes in cm).

Molecular Basis of TSPO as a Cancer Imaging Biomarker Saffet Guleryuz, E.T. McKinley, P. Zhao, A. Fu, J.R. Buck, H.C. Manning Introduction

Translocator protein (TSPO) is an 18kDa outer-mitochondrial membrane protein that participates in regulation of numerous cellular processes including cholesterol metabolism, steroid biosynthesis, proliferation and apoptosis. Elevated TSPO expression is well documented in oncology (1), where recent clinical studies have linked TSPO protein levels in tumors with diminished survival (2). Moreover, emerging data from our laboratory suggests that TSPO expression predicts response to molecularly targeted therapy in colorectal, breast, and lung cancer. Collectively, these data provide rationale for exploration of TSPO ligands as cancer imaging biomarkers. Complementary to our novel probe development activities, our group seeks a molecular-level understanding of how TSPO participates in cancer progression, survival, and response to therapy, as well as elucidation of mechanisms by which TSPO expression is affected by oncogenes and tumor suppressor genes.

Methods

Molecular studies being initiated in this project feature a variety of innovative proteomic and genomic approaches and include analysis of human tumors collected from patients with advanced colorectal cancer, mouse modeling, and human tumor cell lines. Our goals are to identify coordinately regulated genes (by microarray, validated by q-RT-PCR) and proteins (by shotgun MS, validated by MRM) associated with TSPO expression and utilization.

Results

To establish model systems representative of tumors expressing low, moderate, and high levels of TSPO expression, we surveyed TSPO ligand uptake in 22 human and rodent tumor cell lines in preliminary studies. We found that nearly all lines assayed exhibited ligand uptake measurable over background, with several lines showing uptake 2-3 fold above displaceable levels, indicative of target activity (Fig. 1). We have extended these observations to in vivo models and further molecular and functional analysis is underway.

Discussion

We have previously developed fluorescence-based TSPO ligands for imaging tumors in mouse models. More recently our probe development efforts have focused on PET ligands suitable for translation. We anticipate that the molecular studies presented here will lay the biological groundwork for TSPO as a potentially important cancer imaging biomarker.

References

1. Papadopoulos, V., et al., TRENDS in Pharmacological Sciences. 27, 402 (2006) 2. Maaser, K., et al., Clin Cancer Res. 10, 3205-9 (2002)

Example 4

Figure 1: TSPO is elevated in human and rodent tumor cell lines. TSPO can be visualized/quantified by IHC and western blot methods from in vivo specimens.

Development of Materials for TSPO-Directed HTS Matthew R. Hight, Jason R. Buck, Dewei Tang, Eliot T. McKinley, Saffet

Guleryuz, and H. Charles Manning Introduction

The outer mitochondrial membrane receptor translocator protein (TSPO) has in recent years been investigated as a target for cancer imaging due to its roles in apoptosis, cholesterol metabolism, and immunomodulation. We are developing libraries of TSPO ligands for development as small molecule positron emission tomography (PET) probes for use in cancer imaging. To discover and characterize compounds being synthesized in our group, we are also developing high throughput screening (HTS) assays capable of measuring the affinity and potency of these potential ligands. There are currently no available HTS assays suitable for the discovery of novel TSPO ligands; thus, the goal of this project is to develop novel reagents that will serve as the foundation for high throughput plate reader bioassays

Methods We are preparing TSPO ligands that are coupled with organic dyes for fluorimetric assays [1], lanthanide complexes for time-resolved fluorescence energy transfer (TR-FRET) assays [2,3], and novel multi-valent nanomaterials to facilitate a broad range of assays and detection motifs (Fig 1).

Results We have already prepared a number of dye coupled reagents which incorporate a variety of TSPO ligands labeled with a dye via diamine linkers possessing various alkyl chain lengths. Such reagents include the commercially available dye lissamineTM rhodamine B sulfonyl chloride coupled to the aryloxyanilide scaffold using 1,8-octanediamine. These reagents are currently being evaluated for their TSPO-affinity using radioligand displacement of [3H]-PK11195.

Discussion Our study aims to develop a multitude of spectroscopically active materials-based reagents for use in high throughput micro plate reader bioassays. Such tools will be used to screen novel TSPO ligands in terms of binding affinity and potency. Potentially, these studies could lead to the development of a high throughput smart assay technique that would permit the screening of TSPO ligands on a multi-facetted level using a variety of biologically relevant criteria.

References 1. H. C. Manning et al., Bioconjugate Chem. 17 (2006). 2. J. M. M. Griffin et al., Tetrahedron Letters. 42 (2001). 3. H. C. Manning et al., Organic Letters. 4, 7 (2002).

Fig 1: General reaction scheme (left) outlining the attachment of spectroscopically active materials (green) to TSPO ligands (blue) through organic linkers (red).

An Optimized Composite Refocusing Pulse for 7T MRI Marcin Jankiewicz, Jay Moore, Adam W. Anderson, John C. Gore

Introduction

A refocusing pulse is often needed in various imaging scenarios. Solutions currently used in the imaging community involve adiabatic pulses (BIR-4), or sequences of block pulses susceptible to field inhomogeneities. A composite refocusing pulse design is presented here. The assumption is that the refocusing pulse follows an excitation pulse [1,2]. The refocusing solution is immune to inhomogeneities within a predefined space of B1

+ and ΔB0 values.

Methods In our preliminary work, we used a composite, amplitude and phase modulated pulse consisting of a train of block sub-pulses. The amplitude and phase of each sub-pulse was subject to a numerical optimization. A set of custom Matlab optimization routines was used to minimize the expression:

where i is the i-th B1+ index, and j is the

j-th ΔB0 index. Mxinit, My

init correspond to the initial value of the transverse components of magnetization after execution of the slice selective pulse executed directly before the refocusing pulse. The optimization is performed on an orthogonal basis of magnetization vectors Minit

target=(1,0,0), Minittarget=(0,1,0) and

Minittarget=(0,0,1). In the example presented here, respective B1

+ and ΔB0 ranges of 0.3-1.0 as measured in units of nominal B1

+ and 0.5kHz bandwidth were selected to represent typical variations throughout the human head at 7T.

Results Figure 1 shows affected components of the final magnetization vector for the three mutually orthogonal initial magnetization values obtained through a conventional block sequence, BIR-4 pulse, and optimized composite pulse. The refocusing of the optimized pulse is superior to that of the block and BIR-4 sequences.

Discussion

The solution presented here is independent of subject specific and geometric complexity of B1

+ patterns, and hence represents a robust alternative to existing refocusing scenarios. Future work includes adaptation of this concept for slice-selective refocusing.

References 1. M. Jankiewicz et al., J. Magn. Reson. 203, 294 (2010). 2. J. Moore et al., Proc. Intl. Soc. Mag. Reson. Med. 18, 2856 (2010).

!

Mx

init "Mx

final + My

init + My

final[ ]j=1

n

#i=1

m

#

Figure 1: Components of the magnetization vector as a function of B1

+ and ΔB0 after execution of three different refocusing sequences.

Figure 1. (A) in vivo Ge3D image atinjection site.

A

Probing demyelination using high resolution 3D quantitative magnetization transfer (qMT) and diffusion tensor imaging (DTI) in a lipopolysaccharide (LPS) model of Multiple sclerosis (MS)

Vaibhav Janve, Song-Yi Yao, Ke Li, Zhongliang Zu, Subramaniam Sriram, Mark Does, Daniel Gochberg

Introduction

MS is a white matter disease, with a complex combination of pathologies including demyelination,  inflammation, axonal loss, and gliosis. Distinguishing these pathologies is difficult using convential imaging. In this study, we examine the sensitivity and specificity of qMT and DTI imaging to demyelination in an animal model of MS. The animal model is LPS injection, the first model for type III MS lesions, which are characterized by oligodendrocyte dystrophy [1, 2].

Methods

Rats were injected intracerebrally with LPS or an equal amount of saline into the corpus callosum (CC). The injection site was 1mm posterior and 1mm lateral to bregma. 28 days post injection imaging was performed on a 9.4T Varian scanner. High resolution 3D gradient echo structural scans (Ge3D) were used to locate the injection site (Fig 1) and 2D qMT and DTI scans were taken. After this in vivo imaging ex vivo 3D, high resolution (167μm isotropic) qMT and DTI scans were performed on the perfusion fixed brains. 4mm tissue sections containing the site of injection were excised and used for 3D co-registered histological analysis.

Results

We have measurements from 1 rat. Pool size Ratio (PSR) maps obtained from qMT analysis show a reduction in CC near the injection site. Reduction in PSR maps from qMT analysis may indicate demyelination in the CC.

Discussion

Ex vivo measurements on an experimental set of 9 rats consisting of 2 controls and 7 LPS rat brains are planned. High resolution 3D ex vivo scans will hopefully minimize the partial voluming effects that can be problematic in vivo. qMT and DTI parameters will be correlated quantitatively with histological volumes constructed from LFB stained slides.

References 1. Felts PA et al, Brain 128, 1649-1666 (2005). 2. Lucchinetti C et al, Ann Neurol 47, 707–17(2000).

Radiation Dose-Based Comparison of PET and SPECT for Bone Imaging

Lindsay Johnson, Julie Sterling, Michael Stabin, Todd Peterson Introduction

Both PET and SPECT are commonly used in a variety of imaging applications. While it is possible to image similar processes with both modalities, the question of which modality is better cannot be answered simply based on standard properties such as spatial resolution, sensitivity, or SNR. An application-specific comparison will give more insight into the practical strengths and weaknesses of using either PET or SPECT. Metastatic bone lesions have frequently been imaged in both PET and SPECT with the radiotracers Flouride-18 (18F) and technetium-99m-methylene diphosphonate (99mTc-MDP), respectively1. Radiation dose is a concern in preclinical imaging studies as it can lead to alterations in what is being studied2. Because of this, a comparison method based on an equal radiation dose to the subject from each radiotracer was chosen.

Methods

Nude mice were injected intracardially with an osteolytic breast cancer cell line, MDA-MD-231. Three weeks after cancer cell injection mice were retro-orbitally injected with 2.83mCi of 99mTc-MDP and SPECT images were acquired beginning 2.5 hours post injection. After SPECT imaging was completed, mice were injected with 250µCi of 18F and PET images were acquired 1 hour post injection. Each injected radiotracer quantity gives an estimated absorbed radiation dose of 592mGy to the skeleton.

Results

Three mice were imaged with the same-day PET and SPECT protocol, and representative images are shown in Figure 1. Two mice had lesions only in one knee region while the third had lesions in both knees. Preliminary quantification methods based on regions of interest around the knee area have shown non-significant differences between areas of bone lesion and healthy bone.

Figure 1: 18F PET and 99mTc-MDP SPECT images of one mouse are shown in (A) and (B) respectively. Both knees of the hind limbs have an osteolytic lesion.

A B

Discussion

Cardiac-injection models of bone tumor metastases are advantageous because they allow formation of bone lesions in a variety of locations. In the small sample of mice imaged so far, lesions were only present in the knee-joint area. We plan to image more mice at additional time points so as to sample a range of lesion sizes and locations. In addition to quantitative analyses we will also consider employing a lesion-detection methodology to compare the performance of the two modalities.

References

1. E. Even-Sapir et al.,J Nucl Med. 47, 287 (2006). 2. D. Bhattacharjee et al, In vivo. 15, 87 (2001).

Figure 1: Preliminary ADC Maps for frequencies of 50-500 Hz.

50 Hz 100 Hz 200 Hz

300 Hz 400 Hz 500 Hz

µm2/ms

Parametric Mapping of Biological Tissues Using Temporal Diffusion Spectroscopy

Susan D. Kost, Junzhong Xu, Daniel C. Colvin, Mark D. Does, John C. Gore

Introduction Diffusion-weighted MRI reflects the sensitivity of water diffusion rates to the cellular and molecular composition of tissues. Diffusion MR is often characterized by the apparent diffusion coefficient (ADC) which measures the effective root mean square displacement of water in a specific time. The use of an oscillating gradient spin-echo (OGSE) pulse sequence reduces the diffusion time compared to conventional pulsed gradient spin-echo (PGSE) techniques, potentially providing a method to study intra-cellular structure. Diffusion measurements at different frequencies create a temporal diffusion spectrum that can be fitted to analytical models to reveal cellular parameters of interest [1,2]. We aim to create parametric maps of biological tissues based on the OGSE method, which may elucidate how the micro-structural properties of tissues effect ADC measurements.

Methods Diffusion-weighted coronal images of abdominal mouse tissue were made using a Varian 4.7T horizontal bore imaging system and OGSE pulse sequence [3]. Gradient oscillation frequencies were varied between 50 and 500 Hz, and ADC maps were generated using a two-point fit to the Stejskal-Tanner diffusion equation [4], with b-values of 0 and 400 s/mm2. Data were also obtained using PGSE methods with the same diffusion parameters (Δ=24.02ms, δ=20ms) as the OGSE scans.

Results Figure 1 shows some preliminary OGSE ADC maps which include liver and kidney tissues. As expected, ADC values increase with increasing frequency for both tissue types. However, vibrational motion artifacts degrade the data at frequencies above 300 Hz. These artifacts must be corrected before the data are fit to analytical models.

Discussion ADC maps of mouse tissues and corresponding diffusion spectra have been obtained using OGSE measurements up to 500 Hz. It is our aim to determine cellular parameters, including cell size, intracellular diffusion coefficient and surface-to-volume ratio of healthy tissue by fitting such OGSE data with equations relating ADC to structure dependent constants [1,2]. Parametric maps based on diffusion MR may also be applied to tissue pathologies, serving as a noninvasive tool for cancer detection and to probe intra-cellular structure changes caused by treatment.

References 1. J. Xu et al., J Magn. Reson. 200, 189 (2009). 2. E.C. Parsons et al., Magn. Reson. Med. 55, 75 (2006). 3. D.C. Colvin et al., Cancer Res. 68, 5941 (2008). 4. E.O. Stejskal, J.E. Tanner, J Chem. Phys. 42, 288 (1965).

An Information Theory Approach to Parameter Estimates from Multi-Compartment Models of MRI Contrast

Christopher L. Lankford, Mark D. Does Introduction

The magnetization of many biological tissues has been shown to relax according to a multicompartmental model rather than mono exponential functions dictated by the single-compartment Bloch equations [1]. In the case of white matter, it is well established that the fast relaxing signal component is associated with water within or interacting with myelin. Using appropriate pulse sequences [2], this fast-relaxing signal amplitude can be measured and, in-turn, related to the relative volume of myelin within each imaging voxel, but such measurements are slow and require high signal-to-noise ratios (SNR). Also, these estimates are biased by water exchange between myelin and surrounding tissue during the measurement. Deoni, et al have proposed a combination spoiled gradient-recalled echo (SPGR) and fully-balanced steady-state free precession (bSSFP) experiment to characterize myelin content [3]. This approach is attractive because of its purported insensitivity to water exchange and because both SPGR and bSSFP are standard sequences capable of fast, multi-slice or 3D image acquisitions. This study aims to use an information-theory approach to determine the potential precision of estimates of myelin content and other multi-compartment model parameters derived from Deoni’s method.

Methods

Computational models have been constructed which provide phantom magnetization values for a given set of parameter values under both the SPGR and bSSFP scenarios. Using these in conjunction with the fitting algorithm described by Deoni, et al [3], numerical approximations of the Jacobian matrix for the model were calculated at the appropriate parameter values. The Cramer-Rao Lower Bound (CRLB) of the variance of each parameter was calculated via the Fisher Information matrix, assuming Gaussian noise. We intend to repeat the process assuming Rician noise.

Results

The Fisher Information matrix calculated was nearly singular. Initial calculations demonstrated a weak dependence of magnetization (low values in the Jacobian) on the exchange rates at reasonable parameter values. The CRLB of the variance of the exchange rate was significantly larger than those of the relaxation rates and intrinsic myelin water fraction.

Discussion

A singular Fisher Information matrix often indicates that one or more parameters are underdetermined. This is supported by the high CRLB of the exchange rate’s variance. Each of these points provides evidence against the ability of the SPGR/bSSFP combination to determine magnetization exchange rates with precision.

References

1. S. Deoni et al., J. Mag. Res. Imaging. 27, 1421 (2008). 2. CS Poon, RM Henkelman, J. Mag. Res. Imaging. 2, 541 (1992).

3. S. Deoni et al., Mag. Res. in Medicine. 60, 1372 (2008).

Comparison of qMT Techniques with Optimal schemes Ke Li, Richard D. Dortch, Zhongliang Zu, Vaibhav A Janve,

John C. Gore, Mark D. Does, Seth A. Smith, Daniel F. Gochberg Introduction

Magnetization transfer refers to spin exchange between free water and macromolecular pools in biological tissues. There is continuing interest in rapid quantitative measurements of magnetization transfer (MT) parameters. Several techniques have been developed. Selective inversion recovery fast-spin-echo (SIR-FSE) method (1) is based on observing the transient response of a RF pulse that selectively inverts the free water protons. Pulsed-MT (2) technique applies MT pulses at different powers and offset frequencies and fits the observed signal to specific models. In this study, a comparison of these two techniques with optimal schemes was performed.

Methods The optimization was based on Cramer-Rao lower bounds theory. Measurements were performed on Varian magnets. For SIR-FSE, a five-point optimal acquisition scheme (1) with maximum precision efficiency was used. For pulsed-MT, a ten-point optimal scheme with maximum precision was used (2). Comparisons were based on data from bovine serum albumin samples and in vivo rat brain.

Results

The BSA data shows that the SIR-FSE method has higher precision efficiencies. The in vivo data shows that they have roughly similar precision efficiencies. Figure 1 shows a comparison of measured pool size ratio maps of an in vivo rat brain with optimal SIR-FSE and pulsed-MT methods. Ramani’s model was employed to process pulsed-MT data. Both methods yield similar pool size ratios for white matter and gray matter, by analyzing the statistics in the region of interest.

0

0.1

0.2

0.3

0

0.1

0.2

0.3

Figure 1: Measured pool size

ratio maps of an in vivo rat brain with optimal SIR-FSE (top) and pulsed-MT (bottom) methods.

Discussion

Pulsed-MT data have to be combined with additional B0, B1 and R1 mappings to fit for qMT parameters. The fitting process is more complex and has a higher computational cost. Additionally, investigations show that the fitted qMT parameters have a strong dependence on acquisition parameters, such as the excitation flip angle, calling the accuracy of the method into question. SIR-FSE is a simpler method, both for data acquisition and processing. However, any ghosting artifacts in the FSE readout adversely affect the accuracy of the sample quantification. Further investigations will be performed on other sequence parameters. Comparison investigations will be performed on human magnets as well.

References

1. K. Li et al. Magn Reson Med. 2010 (in press). 2. Cercignani et al. Magn Reson Med. 56, 803 (2006).

Resting State Functional Connectivity Analysis: a Potential Model for Human fMRI Studies

Arabinda Mishra, Baxter P. Rogers, John C. Gore, Li Min Chen Introduction

Resting state functional connectivity has been shown in several different brain systems. The inter-regional temporal correlations of low frequency fluctuations of BOLD signals have been hypothesized to indicate the intrinsic functional architecture or functional connectivity of the brain, and potentially to reflect underlying anatomical connections [1][2]. To test this hypothesis, we are investigating the intrinsic resting state functional connectivity in a well-established model system: somatosensory areas 3a, 3b, 1 and 2 of non-human primates. Analysis of high-resolution resting state BOLD signals reveals differential inter-areal functional connectivity.

0 2 4 6 8 10 12 14 16 18 20-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

Results Four squirrel monkey fMRI data (20 sessions) were analyzed for connectivity. We observe a similar correlation pattern across runs within the same imaging session (across area 3a, 3b, area 1 and 2). Connectivity correlation coefficients computed using a linear regression analysis (using area1 time course as seed) are shown in figure (a). Figure (b) shows the profile of the connectivity index variation in 3D. The functional connectivity variation in both x and y directions are analyzed separately. Figures (c) and (d) show the mean connectivity index in x and y directions.

Discussion Correlations between area 3b-area 1and 2 are found to be stronger in comparison to other pairs, which is consistent with their known anatomical connections. The connectivity patterns in the neighborhood of and the selected ROIs and control regions are significantly different. The point spread function for functional connectivity can be reconstructed using this approach. This study provides a reference for comparable studies in humans at high field and suggests future directions for the analysis of the functional heterogeneity of various anatomic structures responding to similar stimulation at a finer scale.

References 1. Morcom AM & Fletcher, PC. 2007 NeuroImage. 37:1073-1082. 2. Vincent et al. 2007 Nature 447:83-86.

he null hypothesis both data of equal mean and variance (p=407e-6)

0.6

0 2 4 6 8 10 12 14 16 18 20-0.1

0

0.1

0.2

0.3

0.4

0.5

(p = 276e-14)

mean connectivity x axis Roi1polynomial fit, n = 8control pointspolyfit

mean connectivity y axis Roipolynomial fit, n = 8control pointpolyfit

Connect iv i ty map (a) over la id on anatomic image and 3-D v iew of area1and 3b (b) . Mean connect iv i ty prof i le in x-y d i rect ion (c-d) .

(c) (d)

(a) (b)

Methods All scans were performed on a 9.4T 21-cm bore Varian INOVA magnetic, using a 3cm surface transmit-receive coil positioned over the somatosensory cortices. Piezoelectric stimulation was performed on the monkey’s distal finger pad (30/30s stimulus on and off, TR: 1.5s) after resting state acquisition of EPI images were done. Seed voxels in targeted brain areas (3a, 3b, 1, 2 and control) were defined by stimulus driven activation and intracortical electrophysiology maps. Standard pre-processing, and fMRI connectivity analysis software generated functional connectivity maps.

Identifying EEG Correlates of Activity in the Working Memory Network Measured with fMRI

Allen T. Newton, Victoria L. Morgan, John C. Gore

Introduction EEG theta power (3-8Hz) has been used as a marker of working memory activity because it is modulated during performance of working memory tasks, but source modeling studies have raised questions as to whether it should be attributed to working memory or so called “default mode” regions [1,2]. One purpose of this study is to investigate whether theta power is actually a marker of working memory activity, and how other bands of the EEG signal vary with working memory activity.

Methods

Simultaneous EEG and fMRI data were acquired in six subjects in the steady state (multiple working memory loads, N-back task, N=0,1,2,3), and during a block designed N-back task (N=0 vs. N=2). Analysis of the block designed data identified the working memory network, while steady state data were used to measure the correlation between theta power and signal variations in BOLD fMRI. All data were acquired using a 64 channel Neuroscan system and a 3T Philips Achieva MR spectrometer. Images were preprocessed, spatially normalized, and analyzed using SPM5. Maps of correlation between theta power and BOLD signal were calculated for all subjects and loads. Initial analyses identified those voxels whose BOLD signal were significantly correlated with frontal theta power across subjects and loads (p<0.005 unc).

Results

Our results suggest that theta power is predominantly correlated with activity in premotor and motor areas (Figure 1). No significant correlations were detected to the working memory network, as located by analysis of data from the block designed task.

Discussion Simultaneously recorded EEG and fMRI data have provided preliminary evidence that the sources of theta power may be different from those that have been previously described. Going forward, we will now ask the question of whether features aside from theta power are better predictors of activity in working memory regions of the brain, potentially yielding a better EEG marker of working memory activity.

References 1. Onton J et.al., Neuroimage 27(2):341-56. 2. Scheeringa et al. International Journal of Psychophysiology 67 (2008) 242–251

Figure1:‐(top)Meangroupactivitymapsduring block designed working memorytask. red=activation, blue=deactivation .(bottom)GroupcorrelationofsteadystateBOLDsignalstothetapower.Red=positiveblue=negative.

Synthesis and Evaluation of COX-2 PET Imaging Probes Donald D. Nolting, Mike Nickels, Brenda Crews, Larry Marnett, Ning Guo,

Ronald Baldwin and Wellington Pham Introduction

Currently, two COX isoforms, namely COX-1 and COX-2 have been identified and scrutinized regarding their roles in biological systems. Several lines of research have demonstrated the implication of COX-2 in the neurodegeneration seen in Alzheimer’s disease (AD). Elevated expession of COX-2 was found in the brains of AD patients compared to age matched groups. Epidemiological studies have shown that nonsteroidal anti-inflammatory drugs reduce the risk and delay the onset of AD. One of the most significant obstacles encountered in studying COX-2 in AD has been relying on in vitro analysis of postmortem brain tissue. This impedes researcher’s ability to obtain molecular information repeatedly, in real time, and in an intact environment. We believe it is of paramount importance to develop a robust technique with which to image COX-2 activity in vivo, since this would facilitate the ability to decipher the mechanism of this enzyme and assess the efficacy of therapy.

Methods Utilizing our experience in azulene chemistry, we synthesized a small library of COX-2 PET probes. We started by transforming tropolone into a lactone which was subjected to an [8+2] cycloaddition reaction to afford the 2-methylazulene core of the molecule (1). Initial synthesis schemes to derivatize the azulene ring structure with thiazole moieties proved unsuccessful. By exploring alternative routes the final target molecule (Fig. 1, inset) and precursor PET compounds were synthesized successfully.

Results The COX-2 precursor PET probes were synthesized with an overall yield of 6-10% after undergoing an 11-step synthesis. The purity of the final product was confirmed by HPLC. IC50 studies performed with one of the products showed an 11-fold greater selectivity for COX-2 than COX-1 with an IC50 value of 185 nM (Fig. 1). Preliminary PET imaging studies have shown that in a wild-type mouse, the material can cross the blood brain barrier and accumulate briefly in the brain.

Discussion Our current focus is on the chemical modification of the lead compounds to improve specificity harboring innovative synthesis methodology. This approach explores the untapped bioisosteres associated within the available chemical structure of the synthesized compounds. To this end, we have recently completed large-scale synthesis of one such PET precursor. Additional work has been planned to validate the probes in a transgenic mouse model of AD. Detailed results of that effort will be reported in a timely manner.

Figure 1: IC50 study resultsfor C23H20N2OS.

References

1. D.D. Nolting et al. Synthesis of bicyclo[5.3.0]azulene derivatives, Nat. Protoc. 4, 1113 (2009).

MR Measurement of Cerebral Blood Volume in the Hippocampus Swati D. Rane, Samet A. Kose, Malcolm J. Avison, Stephan C. Heckers and

John C. Gore Introduction

Cerebral blood volume (CBV) measurement is important for characterizing both pathological and physiological changes in the brain. We employed contrast enhanced steady state T1 (SS) and dynamic susceptibility contrast (DSC) T2* imaging at 3T to compare CBV in the hippocampus (HF) of healthy controls and patients with Schizophrenia.

Methods

SS imaging [1] compares the local fractional increase in tissue signal after an intravenous contrast agent has equilibrated in the blood and thoroughly perfused in the microvasculature. CBV is calculated by subtracting the high resolution, post- and pre-contrast images and normalizing by the difference in voxels of pure blood. DSC, on the other hand, tracks a rapidly administered bolus of contrast on its first pass through the tissue before equilibration, at a very high temporal resolution. The concentration time curve for the passage of the contrast agent is fitted by a gamma-variate function (Figure 1).

Results Baseline CBV in the HF of 4 healthy controls using SS imaging was 3.45 ± 1.12% and 11.04 ± 0.76 ml/100g with DSC. The mean MTT and CBF were 10.07 ± 1.01 s and 63.18 ± 16.96 ml/100g/min respectively.

Discussion Baseline CBV was successfully measured in healthy subjects using SS and DSC methods. However, the CBV values obtained using SS imaging and DSC imaging differ significantly. One possible reason is that while the contrast agent is uniformly mixed in the blood, it is compartmentalized in the tissue vasculature. Signal intensity changes associated with susceptibility agents are due to T2* relaxation, and the transverse relaxivity of the contrast agent is likely quite different in pure blood than when confined to micro-vessels within tissue, thus overestimating the CBV in DSC. In future work the CBV in hippocampus will be evaluated in a group of subjects with schizophrenia.

References 1. W. Lin et al., J. Magn Reson Imaging, 9(1), 44 (1999)

Normalizing the area under the gamma curve for the tissue, with a similar area that is obtained in an artery, estimates the CBV. Blood flow (CBF) and the mean transit time (MTT) of the bolus passag may also be calculated from these curves. 115 volumes were acquired between the pre- and post- contrast image acquisitions for SS imaging. (Contrast: Magnevist, 10% of body wt, time to equilibrate ~ 5 mins, Rate: 5ml/s ).

Figure 1: Perfusion curves (light) and their gamma-variate fit (dark). HF: hippocampus (blue), MCA: Middle Cerebral Artery (red)

Magnetic Resonance Spectroscopic Imaging of Brain at 7 T Indrajit Saha, James M. Joers, Saikat Sengupta, Subechhya Pradhan, Jason E.

Moore, and John C. Gore

Introduction Proton MR spectroscopic imaging (MRSI) at 7 T offers potential for increased SNR and higher spectral resolution for mapping of a large number of neurometabolites [1]. However, the shorter T2 relaxation times of metabolites at ultra-high magnetic field, B0 inhomogeneities of the 7 T magnet, and limitations on achievable B1 field strength at 7 T are challenging [1,2]. Here we present our ongoing work on developing 2-D/single slice MRSI methodology at 7 T using an ultra short echo time STEAM sequence which will lay the foundation for achieving 3-D/whole brain MRSI capabilities at 7 T in the near future.

Methods

3-D inversion prepared T1 images were obtained and the reconstructed images served as anatomic guides for MRSI matrix placement along the midline of brain (FOV = 110 x 110 mm, VOI = 90 x 80 mm, voxel size = 10 x 10 mm, slice thickness = 10 mm). A SENSE accelerated STEAM sequence was employed for 7 T MRSI with echo time of 9.5 ms; mixing time of 32 ms; and repetition time of 2.5 s. Prior to MRSI acquisition, global second order field-map based shimming was performed using an in-house Matlab shimming tool. Additional RF power calibration, outer volume suppression and water suppression (using a MOIST scheme) were performed using standard Philips routines.

Results

Our preliminary MRSI results at 7 T (figure 1) show promises for generating and quantifying concentration maps of metabolites such as NAAG, Glu, Gln, and myo-Inositol in addition to the metabolites commonly reported at 3 T such as NAA, Creatine and Choline.

Discussion

We are now able to obtain in vivo MR spectra with increased number of quantifiable metabolites from MRSI voxels near the brain midline at 7 T by using the ultra short echo time STEAM sequence and implementing image based shimming to efficiently overcome the problems at 7 T. However, we have noticed a catastrophic deterioration of spectral quality near the skull due to infiltration of skull-lipid signal which currently limits our capabilities of acquiring MRSI data from clinically more relevant areas like pre-frontal cortex. We are currently working on incorporating a B1 insensitive pulse-train into the sequence to suppress lipid signal efficiently around the skull. Additionally, we are developing our data processing protocol to quantify metabolites using LCModel and to generate metabolite distribution maps.

References 1. Henning et al., NMR Biomed. 22, 683 (2009). 2. Tkac et al., Mag. Reson. Med. 46, 451 (2001).

Figure 1: Typical spectral result from a MRSI voxel near the brain midline (8 transients, 24 minutes)

Synthesis of Metabolic Tracers for Real Time Hyperpolarized MRI of Breast Cancer

Roman V. Shchepin, Kevin W. Waddell, Eduard Y. Chekmenev Introduction

Using a revolutionary hyperpolarization technique PASADENA (Parahydrogen and Synthesis Allow Dramatically Enhanced Nuclear Alignment) we can routinely increase Magnetic Resonance (MR) sensitivity of injectable tracer compounds by 10,000-1,000,000 fold, Fig. 1. Hyperpolarized MR permits real time metabolic imaging of malignant tumors and can potentially provide US population and oncologists with inexpensive and fast non-radioactive non-invasive (under 1 minute) exam rivaling the accuracy of FDG-PET tracer by reporting on real time metabolic status of tumors. Such exam would be useful for population screening as well monitoring response to treatment in breast cancer.

Methods, Results and Discussion

Choline metabolism is up-regulated in many cancers, and it is the primary target for organic synthesis of isotopically enriched unsaturated PASADENA precursor in our laboratory. Low 15N (<1%) natural abundance requires isotopic enrichment of the tracer compound. The requirement to simplify our spin system to three spins of two nascent protons and X nucleus (X = 13C, 15N) and the requirement to extend the lifetime of the hyper-polarized precursor demand the use of deuterium in positions adjacent to vinyl carbons, Fig. 1. Work is in progress to synthesize molecular precursor using the scheme below, Fig. 2.

References

1. C.R. Bowers & D.P. Weitekamp J. Am. Chem. Soc. 109, 5541-5542 (1987). 2. de Molina, A.R. et al., Cancer Res. 65, 5647-5653 (2005).

H

R1

R2

X

R3

R1, R2, R3 = D or other group except H

H

Catalyst

para hydrogen

Isotopically enriched precursor R1

R2

X

R3

H H

X = 15N and 13C enriched groups

Multistep organic synthesis Hyperpolarization

Hyperpolarized tracer

Purification

Injection in subject

Real time metabolic imaging of

breast cancer

Figure 1: Multistep organic synthesis yields precursors for PASADENA hyperpolarized tracers. After a reaction with parahydrogen molecule and polarization transfer and purification it can be injected in subject for real time metabolic imaging of breast cancer.

Figure 2: The diagram of synthetic strategy to yield molecular precursor for 15N-choline.

Parallel Image Reconstruction for 7T MRI Sepideh Shokouhi, Ha-Kyu Jeong, Marcin Jankiewicz, Adam Anderson, Brian

Welch, John Gore Introduction

Parallel acquisition techniques combine the signals of several coil elements to reconstruct the final image with the chief advantage being able to increase the spatial resolution, signal-to-noise ratio (for the same acquisition period) or to reduce scan time, to diminish distortion from echo planar imaging (EPI) readout trajectories, or to minimize the specific absorption rate, SAR, in scans using many RF pulses such as turbo spin echo, which is a problem in high-field MRI. In the past, several image reconstruction techniques were developed for parallel imaging with multiple coils. Among these techniques are SENSE [1] and GRAPPA. While it does not require signals from multiple receive coils, the emerging theory of Compressed Sensing (CS) [2] has also offered great insight into signal recovery at sampling frequency significantly below the Nyquist rate.

Methods We will implement, test and optimize different parallel and advanced data reconstruction techniques, such as SENSE and compressed sensing, for the 7T MRI scanner. We will use the 7 Tesla human MR scanner (Philips Healthcare) and a 16-channel receive-only coil. In the near future, a 32-channel coil will be available. Reconstruction algorithms developed for the 7T will be ported for use on the 3T human MR scanners as well.

Results We acquired 7T human brain EPI data (no ramp sampling) using 16 receive coils and a SENSE reduction factor of 3. After standard EPI phase correction is applied, a simple Fourier transform of the down-sampled k-space data yields aliased images (Fig 1) from 4 selected coils. The true image was recovered by implementing the SENSE reconstruction (Fig 2) using the data and sensitivity maps from all 16 coils.

Discus

Ourto ethe

Refere

1. 2.

Fig.1

Fig.2

sion current SENSE program is applicable to Cartesian k-space data. A future goal is xpand this to other non-Cartesian sampling trajectories (radial, spiral) as well as incorporation of random undersampling for compressed sensing reconstruction.

nces K. P. Pruessmann et al., Magnetic Resonance in Medicine. 42, 952, (1999) D. L. Donoho, IEEE Transactions on Information Theory . 52,1289, (2006)

Correlating 18FLT Uptake with Drug Delivery Using MALDI-IMS R. Adam Smith, Eliot T. McKinley, Allie Fu, Ping Zhao, Michelle Reyzer,

H. Charles Manning Introduction

Cellular proliferation is a critical biological process known to be disregulated in cancer cells; thus non-invasive, longitudinal imaging assessments could be of particular value in predicting and quantifying response to therapy. B-RAF, one important regulator of cell proliferation, is mutated in up to 15% of colorectal cancers (CRC) [1], with resulting tumors being particularly sensitive to small molecule B-RAF inhibitors. The goal of this project is to evaluate the effect of B-RAF inhibition on TK1 regulation and thus 18FLT uptake using multi-modality imaging coupled with MALDI IMS to observe spatially resolved drug distributions in tumors.

Methods

CRC cell line xenografts (Lim2405, Colo-205, and HT-29) were prepared in female athymic nude mice. Animals were treated with either vehicle control or PLX4720 by oral gavage and imaged by 18FLT PET/CT following drug treatment. Frozen sections were collected for both MALDI IMS and histology, with digital images of the remaining sample block face collected every 50µm. 18FLT PET/CT, MALDI IMS, and block face imaging were co-registered to provide a continuous mapping of in vivo images to the corresponding location in MALDI IMS and ultimately in histologic imaging sets.

Results

In general, 18FLT uptake was found to inversely correlate with drug delivery in all 3 CRC models. MALDI IMS illustrated drug distribution in collected sections, with highest signals originating from the stomach/bowel and tumor regions. From co-registered data, (Figure 1) elevated drug concentrations appeared to agree with reduced radiotracer uptake and in some cases potentially necrotic tissues following prolonged drug treatment.

Discussion

Correlation of MALDI IMS, histological, and 18FLT PET imaging data sets provides a powerful tool to study the utility of imaging metrics. In all 3 models employed, 18FLT uptake was inversely proportional to drug delivery, indicating a potential link between TK1 regulation and B-RAF inhibition in tumor xenografts. These feasibility studies suggest the potential to correlate drug delivery with imaging response and biological readouts on a voxel-by-voxel basis.

References

L. Rozek et al., Canc. Epidemiol. Biomarkers Prev.. 19, 838 (2010).

Figure 1 – (A) Block face (left), FLT (center) and MALDI IMS (right) of a Lim2405 xenograft following PLX4720 treatment. (B) Co-registration of CT, FLT, and MALDI IMS imaging sets. (C) H&E histological staining of sectioned sample.

B

A

C

Development of a Clinically Relevant 3 T Breast Protocol David S. Smith, E. Brian Welch, Lori R. Arlinghaus, Thomas E. Yankeelov

Introduction

High-field MRI provides theoretical gains in SNR and resolution, but in practice these gains have been difficult to realize in clinical breast imaging. Breast imaging at 3 T suffers from issues such as B0 and B1 inhomogeneity, increased specific absorption rate, relaxation rate changes, chemical shift artifacts, and susceptibility gradients (1). Thus, the standard field strength for breast imaging is still 1.5 T. We propose to exploit the recent availability of multichannel, dedicated breast coils and dual-channel transmit systems to develop a clinical 3 T breast protocol for use in breast cancer detection that is equivalent to or better than the existing 1.5 T protocols. The 3 T protocol will then be incorporated into an ongoing study to assess treatment response in breast cancer clinical trials.

Methods

Benchmark scans at 1.5 T were carried out at Vanderbilt 100 Oaks Breast Center on a breast phantom developed by Dr. Arlinghaus and consisting of two containers of lard with water inserts. Initial 3 T protocol development and testing is being done in the Philips Exam Card environment on a virtual machine using theoretical calculations to adjust for the change in tissue properties and artifacts as field strength is increased. These prototype exam cards are being tested at 3 T using the same lard phantom with water inserts. Image SNR is being estimated by comparing two independent acquisitions.

Results

We have developed an initial 3 T protocol and scanned the phantom. Initial 3 T image quality is excellent (see figure), but, without larger SENSE acceleration, scan times are proving to be longer than the 1.5 T equivalents, which is not clinically desirable.

Discussion

Through the rest of this year we will refine our 3 T protocol with experiments on the phantom and volunteers before ultimately performing a direct 1.5 T to 3 T comparison in patients. As few published 3 T clinical breast studies exist and only two studies have compared breast imaging at 1.5 T and 3 T (2,3), our study should have significant clinical relevance.

References

1. C. K. Kuhl, Mag. Res. Imaging. Clinics N. Am. 15, 315 (2007). 2. C. K. Kuhl et al., Radiology 239, 667 (2006). 3. A. Matsuoka, et al., Jap. J. Radiology 26, 15 (2008).

Figure 1: Comparison of T1-weighted axial images of the breast phantom. Upper is the 1.5 T protocol; lower is the 3 T scan. The 3 T images display ghosting due to excessive SENSE acceleration.

Imaging of Hyperpolarized 13C and 15N Tracers Diana E. Smith, Eduard Y. Chekmenev

Introduction

Cancer cells are known to exhibit alterations in metabolism, and these alterations are independent of the host tissue and of the molecular mechanism used to induce

tumor formation. Conventional MR with 13C and 15N labels is capable of detecting

steady-state metabolic flux in vivo, but at low physiological concentrations, the sensitivity is too low to be useful as a clinical diagnostic tool. Additionally,

hyperpolarized MR receptivity scales as !2 for spin-1/2 nuclei, which means that 13C-

and 15N-imaging is much less sensitive compared to proton imaging. However, recently developed techniques, namely PASADENA (Parahydrogen and Synthesis

Allow Dramatically Enhanced Nuclear Alignment) [1] and DNP (Dynamic Nuclear

Polarization) [2], have dramatically increased nuclear polarization. We refer to these

methods as hyperpolarized MR. These new techniques make direct imaging of metabolic flux in vivo at µM concentrations possible.

Methods

Proton polarization transfer experiments to characterize the relevant nuclear spin-

spin couplings have been conducted using a 12 T Bruker high-resolution NMR spectrometer using ~0.2 M solutions of 1-13C-pyruvate and 1-13C-lactate sodium

salts in D2O. The combination of 13C and 1H coupled and decoupled spectra were

collected. Development of polarization transfer sequences will eventually translate over to our MR imaging, and future work will involve developing and testing chemical

shift imaging pulse sequences on the 4.7 T and 9.4 T Varian MRI scanners.

Results

Based on preliminary results, detection sensitivity can be potentially enhanced by 47

fold for hyperpolarized 1-13C-pyruvate metabolism to 1-13C-lactate (a factor of 15.7 due to the (!1H/!13C)2 ratio combined with a factor of 3 from the three methyl protons

for each 13C label). This major enhancement in sensitivity will benefit many imaging-

and spectroscopy-based applications by decreasing the detection limit and/or increasing the spatial resolution.

Discussion Hyperpolarized MR allows imaging of changes in key metabolic pathways in cancer.

The methods employed can potentially be applied to many hyperpolarized 13C and

15N metabolic contrast agents in vivo, including hyperpolarized pyruvate, lactate, bicarbonate, glutamine, and choline. Hence, the hyperpolarized MR imaging of 13C

and 15N has potential to become a key tool in the study and treatment of cancer.

References

1. C. R. Bowers, D. P. Weitekamp, J. Am. Chem. Soc. 109, 5541 (1987).

2. A. Abragam, M. Goldman, Rep. Prog. Phys. 41, 395 (1978).

Increased Differentiation of the Lateral Pain Network by High Resolution fMRI at 7T

Elizabeth Ann Stringer, Li Min Chen, Robert M. Friedman, John C. Gore Introduction

Human neuroimaging studies have uncovered multiple cortical and sub-cortical regions involved in the perception of pain. Although the regions that comprise the lateral pain matrix are known to process the sensory components of pain, less is known about how these regions encode pain-specific attributes. In this study we take advantage of the higher sensitivity available for fMRI at high field (7T) to map cortical representations of pain perception at high spatial resolution.

Methods

We acquired high-resolution (1x1x2mm3) GE-EPI fMRI data at 7T (Philips Achieva with 16-channel NOVA head coil) from healthy subjects covering SI, SII, and insula. A Medoc CHEPS probe secured to fingers 2 and 3 generated innocuous warm or painful heat stimuli. Prior to scanning, subjects rated their pain level for a number of different temperatures. One warm and two painful temperatures that elicited a 5/6 and a 7/8 on a 0 (no pain) to 10 (strongest imaginable pain) pain numerical scale were presented twice per run in 30s on/off blocks. Subjects continuously rated their pain level in real-time during image data acquisition with visual feedback. To examine pain-specific activity that correlates with perception, subject’s ratings were utilized as regressors in a GLM analysis, implemented in BrainVoyagerQX. Activity maps were visualized at p(Bonf)<0.01.

Results

Robust bilateral activation was observed. Within primary somatosensory regions (SI) discrete activation clusters were localized to the posterior bank of the central sulcus, the crest of the post central gyrus, and anterior bank of the post central sulcus, corresponding to Brodmann’s areas 3b, 1, and 2, respectively. Non-human primate pain studies have indicated that areas 3b and 1 are responsive to innocuous cutaneous stimulation, while areas 3a and 1 can show responses to nociceptive stimuli. We did not find a robust nociceptive response in area 3a, but we did observe nociceptive processing in the other areas of SI. Within the SII region discrete activation clusters were localized to the posterior superior bank of the lateral sulcus, the medial superior bank of the lateral sulcus, and the anterior superior bank of the lateral sulcus, bordering the dorsal insula. These areas comprise the parietal operculum, a region involved in sensory processing and association, and a region critical to the ability to experience pain. Each of the areas highlighted show a robust differential response to pain vs. innocuous warm stimuli.

Discussion

Our data demonstrate that fMRI at 7T can be used to map fine-scale cortical representations of pain perception, and provide opportunities to identify pain specific encoding areas. While human imaging studies have indicated that areas SI and SII are involved in pain processing, this is the first human study to identify the sub-regions involved with millimeter resolution. These pain-specific foci will be used as seed regions in a functional connectivity analysis. We will investigate whether experiencing pain diminishes the correlations in BOLD signal fluctuations in the resting state between pain specific areas within the lateral pain network.

Table 2: Cut-off modes for cylindrical wave guide formed by 7T Philips magnet RF shield without any dielectric loading.

TE11   303  MHz  TM01   395  MHz  

TE21   502  MHz  TE01   630  MHZ  

TM11   630  MHz  Table 2 Cut-off mode for cylindrical wave guide formed by 7T Philips magnet RF shield without any dielectric with 3 mm water dielectric bore.

TE   260  MHz  

TM   304  MHz  TE   352  MHz  

TE   420  MHz  

Hybrid   440  MHz    

(a) (b) (c)

Figure 1 TM mode Field distribution a) in XY plane b) YZ plane c) XZ plane

 

Partially loaded travelling wave MRI Sasidhar Tadanki

Introduction

Radiofrequency (RF) field inhomogeneity has been a major challenge in today’s high-field magnetic resonance imaging (MRI) mainly due to the shortened RF wavelength in human tissue at higher frequencies. To overcome this problem use of travelling waves guided by the RF shield of the MR scanner has been suggested[1]. In this work we are exploring the use of other modes of cylindrical wave guides and their potential advantages and disadvantages in high field MR imaging.

Methods

Table 1 shows first 5 primary modes of cylindrical wave guide formed by 7T Philips magnet RF shield without any dielectric. The first cut-off mode for cylindrical wave guide is TE11. Primary characteristic of TE11 mode is presence of z component of magnetic field which will result in reduced probe efficiency. Reduced probe efficiency can be compensated by higher drive power which will directly affect SAR and sampling heating. The above mentioned problems can be reduced by partially loading the wave guide with dielectric material and bringing down cutoff frequency for other modes. Table 2 shows CST simulation results for cylindrical wave guide loaded with 3 mm water dielectric. It clearly demonstrates that the required mode can be brought down below cutoff frequency by partial dielectric loading.

Results and Discussion

Our CST simulations show that TM mode can be achieved on 7T Philips scanner by partial dielectric loading, but simulations also show that TM mode may not be the optimized mode as it has non uniform field distribution as shown in images below. Currently we are exploring other hybrid modes which will give uniform homogenous field. This is being achieved by various combinations of non-uniform dielectric loading.

References 1. Brunner, D. O., et al., Nature 457, 994–998 (2009).

Subject-specific CFD modeling of the vertebro-basilar system

Amanda K. Wake, John C. Gore, J. Christopher Gatenby Introduction

Intimal hyperplasia (IH) is a common cause of failure of arterial bypass grafts, and studies have shown a correlation between IH and hemodynamic factors such as wall shear stress (WSS), a parameter used in graft design [e.g., 1-4]. In the vertebro-basilar system (VBS) the vertebral arteries merge into the basilar artery (BAS). Understanding the hemodynamic features of the VBS in healthy subjects can potentially be used to improve vascular graft design. Our study uses high resolution quantitative MR angiography to characterize VBS hemodynamics in healthy subjects for the goal of identifying flow patterns in native artery confluences for use in design.

Methods Geometry and velocity data were obtained using a Philips Achieva 3T MR scanner (Philips Healthcare). Three-dimensional TOF angiography was used to determine cerebrovascular geometry of the VBS and to define acquisition planes for quantitative measurements by phase contrast (PCMR) imaging. Measurement planes were oriented perpendicular to flow, and the through-plane velocity was measured across the vessel using retrospectively ECG-triggered PCMR. Time-varying velocity distributions across the right (RV) and left vertebral (LV) arteries were measured at 21 time points throughout the cardiac cycle. Lumen geometry was segmented from the TOF data and reconstructed. The resulting arterial geometry was discretized, and FLUENT (Ansys, Inc.), was used to solve for steady flow just prior to peak systole. This time point was chosen because it is expected to exhibit upper limits of WSS values due to velocity profiles at higher Reynolds number flow.

Results Regions of high WSS (WSS > 2.8 Pa) are seen in the distal LV and on the anterior surface of the proximal BAS. The intersection point of the two arteries is an area of locally low WSS (WSS < 1.3 Pa), as is the posterior surface of the BAS.

Discussion

Locally high WSS values on the anterior face and the low WSS region on the posterior surface of the BAS are due to the curve in the BAS. The VBS curves in the anterior-posterior plane, which would cause the velocity profile to skew towards the anterior; thus, the WSS is higher on the outer curve of the vessel than on the inner curve of the vessel. The flow field of the VBS has been investigated for pathologies; however, physiologic flow in this system is not well characterized. These preliminary results demonstrate the feasibility of our approach to modeling hemodynamics in the VBS. Understanding the flow field in the normal VBS may yield useful insight for improving graft design. The pulsatility (and fluctuating flow divisions in the LV and RV) of physiologic flow likely will influence WSS distributions; therefore, we are conducting subject-specific, pulsatile CFD simulations of additional subjects.

References 1. A. Imparato and A. Bracco, Surg. 72, 1007 (1972). 2. H.S. Bassiouny et al., J. Vasc. Surg. 15, 708 (1992). 3. C.F. Ethier, et al., J. Biomech. 31, 609 (1998). 4. A.S. Anayiotos, et al., Ann. Biomed. Eng. 30, 917 (2002).

Comparison of Reduced-FOV Techniques at 7T Christopher J. Wargo, John C. Gore

Introduction

Improvements in spatial resolution for human brain imaging are possible at ultra-high field strength (7T) due to the increased signal obtainable. However, very high resolution imaging (≈ 100 microns) of a complete FOV would require acquisition of a very large number of voxels, increasing the overall scan duration. Beyond the impracticality of long scans, various artifacts are more severe for such long scan times. Reduced-FOV techniques take less imaging time and constrain data acquisition to smaller object regions using selective excitation methods such as STEAM, PRESS, OVS, or spectral spatial pulses [1,2]. Each involves a balance of resolution, SNR, efficiency, SAR, and produced artifacts. To date, a subset of selective excitation approaches have been compared in phantoms at 7T.

Methods

Experiments were performed on a Philips 7T Achieva using a 16 channel SENSE head coil and FBIRN phantom. Images were collected with STEAM, PRESS, OVS, SE, and SPOKE reduced-FOV preparations, where: TR/TE = 1200/32 ms, 128x128 points, 210x210x3 mm, and a 42 mm slab target FOV. Peak signal, SAR, and suppression were recorded.

Results

Overall, OVS produced the highest peak signal and SAR at 87% and 92%, but the lowest suppression at 67%. The other approaches were consistent, with 90-100% suppression, 11-22% SAR, and 24-40% peak signal. OVS best resembled the original object, with shape, width, and signal artifacts for other methods.

Discussion

Future work will investigate alternate pulse types to improve SNR and suppression while reducing SAR in the presence of 7T B1 inhomogeneity. Method comparisons will be repeated as described, with high-resolution (100-500 µm) human brain images collected. The optimal approach will be based on that enabling distinction of the highest resolution anatomical features, matching or exceeding traditional scans, with reduction of observable artifacts, and greater efficiency. This analysis will be made in both anatomical and functional based MR brain studies.

References 1. S.F. Keevil, Phys. Med. Biol. 51, R579 (2006) 2. H. Hardy, J Magn. Reson. 82, 647 (1989)

Figure 1 - A.) Object, B.) OVS, C.) STEAM, D.) PRESS, E.) SE, and F.) SPOKE.

A B C

D E F

Figure 3 - Measured peak signal, suppression, and SAR.

Figure 2 – Central object profile for each reduced-fov preparation.

Detecting Tumor Early Response to Chemotherapy Using Temporal Diffusion Spectroscopy: How Early Can We Get?

Junzhong Xu, Susan D. Kost, Saffet Guleryuz, R. Adam Smith, Ping Zhao, Xia Li, H. Charles Manning, Mark D. Does, John C. Gore

Introduction

Diffusion-weighted magnetic resonance imaging (DWI) has been suggested as a biomarker to detect tumor early response to treatment. Conventional DWI uses the pulsed gradient spin echo (PGSE) method, which has been reported to be sensitive to cell density (1). In the current study, an oscillating gradient spin echo (OGSE) method is used to detect tumor intracellular changes due to treatment that precede cellular variations. The detection of early cancer treatment response provides an opportunity to optimize individual patient treatment and reduce the cost of therapy.

Methods

Two groups of nude mice (seven for each) will be injected with 1×105 SW620 cancer cells into the hind limb. After two weeks of the injection, the control group will receive 0.1ml of drug vehicle (10% ethanol) and the treatment group 24mg/kg AZD1152. All treatments will be administered by a single intraperitoneal injection. Both OGSE and PGSE methods will be performed in the MRI scans with multiple axial slices. 1D navigator will be acquired to correct translational motion artifacts. Scans will be performed at 24 and 48 hours after the initial treatment and then all DWI images will be registered to high resolution T2-weighted baseline images before and after the treatment. After the scans, mice will be sacrificed and tumor histological images will be obtained with H&E staining.

Results

Figure 1 shows some preliminary OGSE images of different slices at 200Hz. The tumor ADC histograms before the treatment were also compared with those of after 24 and 48 hours of treatment, and the OGSE method has been found to detect ADC histogram shift to higher values at 24 hours while the PGSE method found a slight change of ADC. However, rotational motion artifacts that cannot be corrected by the 1D navigator have been found in the experiments.

Figure 1: Some preliminary OGSE images of different slices with gradient frequency at 200Hz.

Discussion

The conventional spin echo method was used in the experiments. However, it will be helpful to perform echo planar imaging (EPI) technique to achieve a much faster acquisition in order to achieve an ADC dispersion curve over a broad range of frequencies, and such curve will be fitted to theoretical models (2) to provide detailed tumor structural information, such as mean cell size and surface-to-volume ratio, which assist quantitative cancer treatment evaluation.

References

1. J. Xu et al., Magn. Reson. Med. 61, 828 (2009). 2. J. Xu et al., J. Magn. Reson. 200, 189 (2009).

High Resolution FMRI Mapping of Cortical Plasticity Following Spinal Cord Injury in Non-Human Primates

Xiang Ye, Feng Wang, Barbara Dillenburger, Calum Avison, Li Min Chen

Introduction Substantial cortical plastic changes occur in somatosensory cortices following spinal cord injury (SCI). The spatiotemporal dynamics of these plastic changes and their functional significance for behavioral deficit and recovery remain very poorly understood. An understanding of the relationship between plastic change and behavioral/functional recovery is essential to the development of mechanism-driven therapeutic interventions in spinal cord injury patients. As a first step toward this goal, this study focuses on longitudinal mapping of normal hand representations in primary (SI) and secondary somatosensory (SII) cortices, and aims to establish the baseline maps in these two cortical areas by using high resolution fMRI at 9.4T.

Methods All scans were performed on a 9.4T 21-cm bore Varian INOVA magnetic, using a 3cm surface transmit-receive coil positioned over the somatosensory cortices of anesthetized squirrel monkeys. Tactile stimulation (30/30s on/off) of individual distal finger pad was used to map hand responses. Functional gradient echo EPI images

(TR/TE = 1500/16 ms) were acquired with 250x250 m2 in-plane resolution. Areas of significant activation were identified by voxelwise correlation of the BOLD signal timecourse with the stimulus presentation boxcar function; time shifted to account for the hemodynamic delay. Electrophysiological mapping in the same animal validated Somatotopic maps obtained from BOLD mapping.

Results Multiple BOLD activation foci were detected in cortical areas around central (SI cortex) and lateral sulci (SII and surrounding areas). BOLD signal timecourses showed robust signal changes associated with stimulus presentation. Across imaging sessions of the same animal, activation consistency varies across cortical areas. Electrophysiology mapping confirmed the locations of activation foci in somatosensory areas 3b and 1, and classical SII. The underlying sources for within and across session activation variation are under investigation.

Discussion Our preliminary data not only confirmed the fine hand representation in SI of humans by measuring the same BOLD signals, but also resolved multiple sites of activation within SI and SII. In SI these multiple sites are consistent with the multiple somatotopically-organized subregions of areas 1, 3a, 3b and 2, which are differentially sensitive to benign and painful stimuli, while SII subregions have been implicated in higher order pain processing. The ability to resolve fine-scale plastic reorganization leading to inappropriate activation of these cortical areas is likely to provide novel insights into the role of plasticity in the emergence of central pain syndromes following SCI and peripheral nerve injuries.

References 1. Chen et al., J. Neurosci. 27:9181-9191 (2007). 2. Zhang et al, Neuroscience, 165:252-264 (2010).