MSBI Thesis & Abstracts 2022-2023

Shixian Du

Thesis Title: Evaluation of the feasibility of in vivo Ac-225 imaging

Advisor: Youngho Seo, PhD

Abstract: Targeted alpha therapy (TAT) is showing promise in the treatment of solid and liquid tumors. TAT uses alpha particles that have small effective range and high linear energy transfer (LET) to achieve high killing in tumor cells and spare normal tissues around the tumor cells. Currently, Actinium-225 (Ac-225) is the most sought-after alpha-emitter. To study the biodistribution of any Ac-225 labeled radiopharmaceuticals, we need to image Ac-225 in vivo in small animals at a sub-Ci activity level. With this research question, we investigate the feasibility of using a commercial preclinical single photon emission computed tomography scanner (SPECT) to quantitatively image a mouse-like phantom at low Ac-225 activity. We first developed an Ac-225 imaging protocol with the SPECT scanner. The protocol consisted of calibrating the scanner for two imageable photon emissions from the Ac-225 decay to Fr-221 and Bi-213 (218 keV and 440 keV). Then, we selected regions of interests (ROIs) from phantom images and characterized the quantitative accuracy of the resulting images as a function of activity in Ci equivalent to 1-hour exposure. For both energy windows, with three ROI methods, the recovery coefficients (RCs) showed consistent quantitative accuracy of SPECT images at an activity level as low as 3.36 Ci with 1-hour exposure. 20% or more deviation of RCs from the ground truth as well as increasing variations of RC values measured between phantom cavities were found at activities below 3.36 Ci. Additional phantom studies, at lower activity levels, and animal studies are being prepared for further investigation on the preclinical scanner’s limit of detection of Ac-225.

Eric Park

Thesis Title: Improving Mouse Brain Hyperpolarized 13C MR data through User-Independent tMPPCA Denoising

Advisors: Myriam M. Chaumeil, PhD & Jeremy Gordon, PhD

Abstract: This study aims to enhance the quality of metabolic data from mouse brain through the application of tMPPCA denoising and Whittacker baseline correction. The goal is to develop a robust pipeline for detecting spectral signal changes using MATLAB, with applicability to various diseases and pre-clinical spectra data. The study employed data collected from Alzheimer's disease mouse models, analyzing different groups with varied drug injections and time points. By comparing key ratios like Lactate/Pyruvate, Bicarbonate/Lactate, and Bicarbonate/Pyruvate obtained from the pipeline with MestreNova software, the pipeline's efficacy was validated. Furthermore, the application of denoising and baseline correction methods was crucial for reliable results. Notably, the pipeline demonstrated remarkable similarity in results with MestreNova software, validating its usability. This pipeline's importance lies in its potential to analyze low signals in diverse slab data. This study's contribution lies in its role in addressing low signal challenges through denoising and baseline correction in the realm of hyperpolarized carbon 13 studies. The conclusion emphasizes the pipeline's robustness and its validation through comparisons. A key focus is the observation of distinctive metabolic shifts in APOE 4 mutation mice, signifying enhanced glycolysis, anaerobic metabolism, mitochondrial dysfunction, and impaired oxidative metabolism. This thesis not only refines metabolic analysis precision but also enriches the understanding of metabolic dynamics in various contexts, offering insights into potential therapeutic interventions targeting metabolic pathways.

Paramjot Singh

Thesis Title: Predicting Newly Diagnosed Glioma Pathology with MRI and Deep Learning

Advisor: Janine Lupo, PhD

Abstract: Current methods of glioma pathology assessment using the tumor score metric rely on the extraction of a biopsy sample for evaluation by a pathologist. This method is limited by the fact that tumor score can vary within a glioma and that it only gives information regarding glioma pathology at one time point. An approach in which allows for the assessment of glioma pathology at various timepoints and in the entire brain is thus desirable. We explored such a method of glioma pathology prediction with machine learning, using both traditional and deep learning approaches. Using a dataset of patient information (MRI images and corresponding tumor scores, we performed several experiments with traditional machine learning models to explore the potential benefits of a deep learning based approach. We then developed, trained, and tuned a deep learning model that predicted tumor score from MRI data, and experimented with various forms of transfer learning to evaluate the impact of loading weights from different autoencoders. We determined the results of our traditional machine learning experiments showed a potential for a deep learning model’s ability to predict tumor score from MRI data. When evaluating our deep learning model we found that domain shift played a significant role in affecting our results in terms of testing accuracy, and we explored several methods to alleviate this issue. That said, our deep learning approach did outperform our traditional machine learning models, indicating the effectiveness of this approach.

Katelyn Vu

Thesis Title: Evaluating target-specific pre-to-post DBS effects on brain function in Parkinson’s disease using fMRI

Advisor: Melanie Morrison, PhD

Abstract: Parkinson's disease (PD) is a progressive neurodegenerative disorder affecting both motor and non-motor neural circuits. Common frontline treatment includes dopamine agonists, but long-term levodopa use can cause dyskinesias and lower quality of life. Deep brain stimulation (DBS) is a second-line therapy to treat motor symptoms caused by movement disorders. Overall, the project aimed to better understand how DBS affects brain networks to improve patient outcomes. The repeatability of resting state functional magnetic resonance imaging (rs-fMRI) data was evaluated in patients with DBS and stimulation-induced longitudinal and immediate changes in brain activity and connectivity were related to symptom improvement. It was hypothesized that functional connectivity (FC) and variability would decrease with sub-thalamic nucleus (STN) or globus pallidus (GPi) DBS over time, and that these changes would be associated with symptom improvement. The two oldest patients with the highest MDS-UPDRS raw scores exhibited superior repeatability, while those with the worst MDS-UPDRS raw scores displayed lower repeatability. The Wilcoxon sign rank test's rejection of the null hypothesis suggested challenges in achieving group-level repeatability due to motion artifacts. Adjusting degrees of freedom may mitigate this issue. The study found a decrease in connectivity within thalamic regions, supplementary motor area and cerebellum Crus I and Crus II from pre-op to post-op, indicating neuroplastic changes in the brain. The variability metrics examined the brain's network adaptability, with an overall decrease in variability in the post-op DBS off and on conditions.

Ningjing Zhang

Thesis Title: Improving Quantification of Prostate-Specific Membrane Antigen (PSMA) - Positron Emission Tomography (PET) Clinical Data

Advisor: Peder Larson, PhD

Abstract: Prostate cancer is a significant global health concern, ranking as the second most commonlydiagnosed cancer and fifth leading cause of cancer-related deaths in men worldwide. Prostatespecific membrane antigen (PSMA)-targeted positron emission tomography (PET) is used for staging, especially in intermediate to high-risk cases and biochemical recurrence, offering superior sensitivity and specificity compared to conventional methods. Despite advancements, the analysis of PSMA-PET data remains largely manual, prompting the need for computer-aided diagnosis using machine learning or deep learning to enhance efficiency and consistency. Some previous work have been done to develop a prediction model based on deep neural network (DNN), however it failed in prediction of high-volume disease cases. In this project, I hypothesized that refined lesion annotation and specialized batching strategies can enhance algorithm performance, leading to improved segmentation accuracy, and SUV measurements. The successful implantation of better contour of the lesion was done by a clinical fixed-threshold method. The results suggested the refined lesion annotations significantly impacted the SUVmean values in lesions and will further affected the result of DNN training.