MSBI Thesis & Abstracts 2016-2017
Sivakami Avadiappan
Advisor: Dr. Janine Lupo, PhD
Thesis Title: Simultaneous Visualization of arteries, veins and cerebral microbleeds from 7T MRA-SWI images for assessing the effects of radiation therapy in brain
Abstract:
The use of radiation therapy for treatment of brain tumor is controversial due to its long-term effects on neurocognitive function, especially in pediatric patients. Developing objective criteria for evaluating the severity of radiation-related injury is critical for children, due to their longer survival times and impact on early cognitive development. Cerebral microbleeds (CMB) which are deposits of hemosiderin that initially accumulate around vessels and can appear as early as 6 months post radiation therapy and continue to increase in number over time, however their vascular etiology is unknown. The aim of this project is to develop a method for simultaneous visualization of arteries, veins, and CMBs in order to automatically calculate vascular metrics from the fusion of MRA and SWI images obtained from a multi echo sequence at 7 Tesla. A strategy to assess the distribution of CMB’s relative to surrounding arteries and veins would help establish a connection between CMB formation and underlying vascular pathology. The tools developed in this study will be evaluated in pediatric patients with CMBs who have been treated with uniform supratentorial cranial radiation therapy for childhood brain tumors in order to ultimately determine whether metrics describing vascular structure could serve as quantitative markers for radiation injury, help in identifying regions of the brain that are most susceptible to radiation, and predict the formation of new CMBs and subsequent cognitive impairment.
Jonathan Chan
Advisor: Dr. Steven Hetts, PhD
Thesis Title: In-Vitro Flow Characteristics of Chemotherapy Filter Device
Abstract:
Kevin Chang
Advisor: Dr. John Kurhanewicz, PhD
Thesis Title: Role of hyperpolarized Glutamate in Detecting Neuroendocrine Prostate Cancer
Abstract:
Hsu-Cheng Huang
Advisor: Dr. Benjamin Yeh, PhD
Thesis Title: Optimizing bone marrow lesion detection using dual energy CT
Abstract:
Introduction: Bone metastasis is the third most common metastasis in cancer patients, causing intractable pain and debilitating complications. However, studies have shown that conventional single-energy computed tomography (SECT) has limitation in detecting subtle bone metastases without obvious osteolysis/osteosclerosis. We hypothesize that dual-energy computed tomography (DECT) can distinguish different composition in tumors and in healthy bone marrow, facilitating the detection of bone marrow metastases.
Methods: We constructed 51 semi-anthropomorphic lumbar spine phantoms embedded with 75 simulated tumors (25 mild lytic, 25 isodense, and 25 mild sclerotic). These phantoms were initially scanned without outer torso phantom encasement, and then scanned again with outer torso phantom encasement under the same machine setting in a rapid-kilovoltage-switching DECT scanner. Two radiologists independently reviewed the virtual monochromatic reconstruction in 70 keV and material decomposition images (hydroxyapatite-water, water-hydroxyapatite, cortical bone-water, water-cortical bone). We recorded reviewer’s response regarding the presence of tumors, tumor conspicuity score and image quality using 3-point Likert scales. The sensitivity and specificity of different reconstruction algorithms were evaluated with McNemar test. Wilcoxon signed rank test was used to evaluate the tumor conspicuity and image quality of different algorithms. Then we validate our testing algorithms by retrospectively reviewing patients’ images in our institution.
Results: Hydroxyapatite-Water material decomposition algorithm achieved higher sensitivity in detecting isodense lesions as compared to the 70 keV reconstruction (without torso phantom encasement: 94% vs 82%, p = 0.031; with torso phantom encasement: 38% vs 18%, p = 0.013). Hydroxyapatite-Water material decomposition algorithms also possessed higher tumor conspicuity score (p < 0.0001) compared to 70 keV reconstruction, and was less affected by CT artifacts.
Conclusion: DECT with hydroxyapatite-water material decomposition may help detect spine marrow metastases, especially for subtle isodense tumors. Further study in prospective clinical scans is warranted.
Henry Li
Advisor: Dr. Robert Flavell, PhD
Thesis Title: Detection of prostate cancer through molecular imaging of Zn using hyperpolarized 13C MRI
Abstract:
Purpose: Hyperpolarized 13C MRI is a promising non-invasive imaging technique that can interrogate tumor microenvironments, and is amenable to clinical translation. The aim of this study was to synthesize candidate probes and evaluate their effectiveness in zinc-binding and hyperpolarized study. Methods: Candidate probes are synthesized, purified, and characterized via NMR, followed by testing in various zinc concentrations and pH buffers. Promising probes are enriched with a 13C label and evaluated with a hyperpolarizer and NMR. Results: Several probes show significant chemical shift upon binding to zinc, as well as negligible shift to different pH conditions. Conclusion: Results from this study show that molecular imaging of zinc is possible, but require further optimization of 13C compound synthesis routes and hyperpolarized experiments.
Cyrus Manuel
Advisor: Dr. Duygu Tosun, PhD
Thesis Title: Weakly supervised deep feature learning to predict amyloidosis from structural MR brain images
Abstract:
Many novel treatments for Alzheimer’s Disease (AD) are aimed to target Aβ, one of the pathological hallmarks of AD, but are hampered by potential non responders due to lack of target Aβ pathology in their brains. Specifically, about ∼25-40% of those clinically diagnosed with AD or mild cognitive impairment (MCI) would not have significant Aβ pathology. This study explores a deep learning framework to predict Aβ pathology positivity from baseline clinical assessments and structural MRI data routinely acquired from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Graph convolutional networks (GCNs) were trained on undirected graphs reconstructed from diffusion MRI and their performances was assessed to see their predictive value based on ground-truth Aβ-positivity estimates from AV45-PET scans. Anatomical brain parcellations with atrophy estimates from structural MRI constitute the vertices; tractography based connectivity estimates defined the edges of the graph model. A 10-fold cross validation on independent training and test sets were performed to assess the model performance in terms of classification accuracy, sensitivity, specificity, positive and negative predictive values. GCNs were able to learn from atrophy descriptors and network connectivity derived from MRI and predict Aβ-positivity. Atrophy was a significant predictor of Aβ-positivity in the AD model, but at a lesser degree in healthy and MCI models. The inclusion of other AD-related predictors showed: a significant improvement in test accuracy to 68±4%, sensitivity to 84±7%, specificity to 52±13%, negative predictive value to 77±5%, and positive predictive value to 64±4% in MCI models; and a significant improvement in test accuracy to 69±2% and specificity to 97±4% in HC models. Patterns of regional brain atrophy within large-scale brain networks might offer predictive value to whether or not a subject will test positive for an AV45-PET exam. Predictions are more accurate with the addition of well-established AD-related predictors, however more features may be necessary to increase the predictive ability in healthy and MCI subjects.
Michael Piel
Advisor: Dr. John Shepherd, PhD
Thesis Title: Predictive Modeling of Whole Body Dual Energy X-Ray Absorptiometry from 3D Optical Scans Using Shape and Appearance Modeling
Abstract:
Introduction: Malnutrition and lack of exercise have led to a steep increase in metabolic disorders worldwide. Even though diseases caused by malnutrition have become common, we still lack an accurate, inexpensive, and easily accessible method to assess a person’s risk of developing metabolic diseases. In this work, I test a novel method called 3D optical body composition that I hypothesized would be relatively accessible, accurate, and inexpensive.
Methods: The Shape Up! Adults Study is recruiting 720 adults for measures that include whole-body DXA and 3D optical scans. Like image types were spatially registered using 105 and 75 fiducial points for DXA and 3D optical scans respectively. Statistical appearance and shape modeling were then performed on each image type. The sex-specific population variances for shape (3D Optical) and bone, fat, and lean appearance (DXA) were captured as Principal Components (PCs) resulting in 8 PC models 4 for females and 4 for males. Stepwise linear regression was used to predict DXA PCs from 3D optical PCs and other anthropometric and demographic measurements. The predicted DXA PC coefficients of each participant were then inverted to create a pseudo-DXA image. Additionally, k-means cluster analysis was performed on the participants’ predicted DXA fat PC coefficients to determine different body phenotypes of males and females and corresponding health risks.
Results: A total of 72 men and 104 women were available at the time of the analysis. To describe 95% of the population variance in men, it required the following number of PC modes: 10 (optical), 32 (fat), 35 (lean), 35 (bone), with women having similar results. The pixel-to-pixel differences in mass between actual and predicted DXA values had no mean bias for all models. The difference in the pixel values had root mean square errors (RSMEs) of 0.015 g of fat, 0.023 g of lean, and 0.012 g of bone for the female data, and 0.013 g, 0.024 g, and 0.018 g for the male data respectively. These RMSE values were less than 5% of the maximum pixel value within the population. Lastly, I found 9 female and 5 male phenotypes of body fat that were related to unique metabolic characteristics and risk factors.
Conclusion: Whole body and regional distributions of fat, lean, and bone can be accurately predicted from 3D optical scans. With this accessible and accurate method, body composition and metabolic risk phenotype can now be defined in individuals. Our hope is that this will increase awareness of metabolic risks and motivate those at high risk to seek medical advice for risk-reducing strategies.
Niecholle Roco
Advisor: Dr. Henry Vanbrocklin, PhD
Thesis Title: Development and Imaging of Targeted Therapy Imaging Agents for BCBM
Abstract: Transforming growth factor-β (TGF-β) is a cytokine that acts as a tumor promoter in breast cancer brain metastasis. It is tumor promoting by activating immunosuppression and enhancing breast cancer cells’ ability to metastasize. Since TGF-β is overexpressed and plays a role in breast cancer brain metastasis progression, it can be exploited as a biomarker for immuno-PET imaging through radiolabelling of fresolimumab, a monoclonal antibody (mAb) that inhibits all active isoforms of TGF-β. Methods: An immuno-PET imaging agent was synthesized in the form of an 89Zr-DFO-fresolimumab targeting vector to investigate a breast cancer brain adapted mouse model. The number of DFO chelate sites per antibody were determined. Additionally, immunoreactivity experiments were done to ensure that the integrity of the antibody was maintained after radiolabeling. Lastly, the imaging probe was tested in-vivo in a mouse model to determine characteristics of the probe’s distribution. Results: DFO was chelated to the antibody fresolimumab. The number of DFOs per fresolimumab was determined to be 3.8 ± 0.8. The imaging probe, 89Zr-DFO-fresolimumab was prepared with high radiochemical purity (> 95%) and with good radiochemical yield (∼70%). Immunoreactivity was maintained, the radiolabeling process did not affect the properties of fresolimumab’s ability to bind to TGF-β. μPET/CT imaging was done to determine the distribution of the radiotracer. The targeting vector was able to cross the blood brain barrier in a specialized mouse model and be localized with adequate signal in the brain tumor. Conclusion: Results from this study showed the targeting vector’s ability to cross the compromised blood brain barrier and enough signal was seen to characterize the brain tumor. The tumor progression was imaged with high quality using diffusion weighted MRI. This study can be used to provide the baseline untreated model to be compared against mice with brain tumors after receiving treatment in future studies. These initial studies for 89Zr-DFO-Fresolimumab are promising however, further work is needed to assess the tracer characteristics.
Benjamin Speidel
Advisor: Dr. Robert Knowlton, MD
Thesis Title: Mapping EEG to Brain MRI for Epilepsy Localization
Abstract: Accurate localization of epilepsy foci is essential prior to resection of epileptogenic zones of the brain. EEG signatures indicative of epileptic seizures can be mapped to their location on a brain MRI by acquiring the MRI before implanting the electrodes in the brain, and localizing the electrodes within the head with a CT after implantation. However, distortion resulting from the craniotomy and the implantation procedure change the brain geometry causing the frame of reference to shift and this model to be inaccurate. Using B-Spline transformations and an Advanced Mattes Mutual Information metric, a nonrigid registration was performed between CT scans before and after the implantation procedure. This transformation was subsequently applied to the MRI in order to correct for the mismatch in registration in 5 subjects. After the application of this nonrigid transformation there was an average increase in accuracy of 12.5% ± 2.3% (p<0.001) in the ability of depth electrodes to localize gray matter. On average, the grid electrodes were 3mm ± 3mm (p= 0.06) closer to their locations given by an intraoperative photograph in one patient when the transformation was applied as well. This has been incorporated into a completely automated tool that can be easily implemented into a lab or clinical environment.
Radhika Tibrewala
Advisor: Dr. Viola Rieke, PhD
Thesis Title: Strategies in MR-Guided Focused Ultrasound
Abstract:
Purpose: To determine under-sampled k-space gradient echo trajectories for MR thermometry with PRF shift to increase acquisition speed while maintaining temperature accuracy.
Methods: A computer simulation was built and used to study the behaviour of temperature measurement errors as different lines of k-space were acquired. A fully acquired k-space was constructed from the information in the Bioheat Transfer Equation (BHTE), various k-space under-sampling schemes were employed and temperature maps were reconstructed using the PRF shift method as a basis. Temperature errors were calculated as the difference in the maximum temperature predicted by the BHTE and the maximum temperature measured by the under-sampled scheme.
Results: Under-sampled gradient echo keyhole trajectories showed promising results for making temperature measurements. Using 5 interleaves with a 63 keyhole size estimates temperature with only a 0.4% error while providing a 1.68x faster frame rate.
Conclusion: Gradient echo keyhole trajectories can be used to make temperature measurements for MR thermometry with PRF shift. This method can also be applied for shorter sonications and making temperature estimates for multiple slices.
Justin Yu
Advisor: Dr. Robert Flavell, PhD
Thesis Title: Optimization of pH imaging methodology for hyperpolarized 13C MRI
Abstract:
Background: The acidification of the tumor microenvironment is a result of extensive metabolic reprogramming in cancer cells and is linked with tumor metastasis. Hyperpolarized MRI is a method for imaging and quantifying this change in pH, but suffers from rapid signal loss from spin-lattice (T1) relaxation. We propose using hydrogen/deuterium exchange on hyperpolarized 13C probes in order to prolong hyperpolarized signal by reducing T1 relaxation. Methods: H/D exchange was performed on several amino acids and amino acid derivatives with utility in HP-MRI. Isotopic enrichment was evaluated using 1H NMR. The T1 relaxation constant was quantified by analyzing the decay of hyperpolarized signal of deuterated vs nondeuterated 13C compounds. Results: H/D exchange was successfully used to enrich compounds with deuterium with high isotopic enrichment and moderate to high chemical yield. The T1 relaxation constant of all fully analyzed 13C compounds exhibited a significant increase after deuteration at 3T: T1 of 13C Gly increased from 52.0±3.2 to 65.0±1.2s, 13C Ala from 52.9±2.2 to 66.4±1.7s, 13C Val from 38.1±1.1 to 49.2±0.4s. Conclusion: H/D exchange method described is a viable technique for inexpensive and direct deuterium labeling. Deuterium labeling be applied to hyperpolarized 13C MRI probes to prolong HP signal by lengthening T1.