Masters Theses

This study investigates the molecular and evolutionary mechanisms underlying copper tolerance in Mimulus guttatus. I focused on the role of multi-copper oxidases (MCOs), specifically the T-MCO gene located in the Tol1 locus in Mimulus guttatus, to provide direct evidence of its contribution to the plant’s copper tolerance mechanism. The T-MCO gene was overexpressed in non-tolerant backgrounds of M. guttatus and Arabidopsis thaliana, with results indicating that T-MCO is central to copper tolerance in M. guttatus. While MCO-centered mechanisms were previously associated primarily with bacteria and fungi, this finding suggests a novel copper tolerance mechanism in plants. Transgenic A. thaliana lines exhibited enhanced germination rates and root growth across various copper concentrations, confirming the significant impact of T-MCO overexpression. I used generalized linear mixed models (GLMMs) to validate these findings, highlighting the importance of T-MCO in copper homeostasis. This research advances our understanding of plant copper tolerance mechanisms and suggests potential applications in crop development. Future work should explore T-MCO’s interactions within the copper homeostasis network and its applicability in other plant species.

The study aims to explain the factors leading to changes in tax bureaucrats' effort levels and how closely these factors are associated with the anti-corruption campaign launched by President Xi Jinping after he took office in 2012. I constructed an original dataset and employed a fixed-effect counterfactual estimator to decompose the effects of routine anti-corruption actions and unique measures in the anti-corruption campaign on tax bureaucrats' effort levels, proposing different mechanisms to explain the differences between these effects. I argue that, first, based on the formal relationship between municipal party secretaries and local tax bureaucrats in Chinese prefecture-level cities, the turnover of party secretaries triggers a "new broom sweeps clean" effect, increasing tax bureaucrats' effort levels; second, for tax bureaucrats at any level, the changes in external conditions brought by the anti-corruption campaign are not sufficient to reduce tax bureaucrats' effort levels in the long term.

Regression results validate my theory: for general political turnover of municipal party secretaries, tax bureaucrats' effort levels increase in the second year after the party secretary takes office; while for cities where party secretary turnover occurred due to corruption, local tax bureaucrats' effort levels are lower than other types of political turnover, but the effect remains positive. Furthermore, the negative impact on bureaucratic efficiency brought by central inspections since the 18th National Congress of the Communist Party of China is short-lived.

This paper investigates the relationship between the prevalence of female combatants and armed rebellion with post-conflict outcomes. Particularly, it probes whether the increased employment of rebel women on the frontlines of battle generates greater gains for women upon the conflict’s conclusion. It utilizes recently published cross-sectional data from the Women in Armed Rebellion (WAAR) dataset which disaggregates women’s contributions to armed rebellion into 22 measures and then rates the prevalence of qualifying variables on a four-point ordinal scale, where 1 = low levels of women’s involvement and 4 = high. I utilize an indicator measuring the prevalence of women’s employment to test whether variation in the quantity of a rebel group’s female combatants impacts the status of women in post-conflict society. Ultimately, I find no relationship of significant nor substantive significance between women’s prevalence on the frontline of rebel conflict and the status of women post-conflict, but I do uncover a potential relationship between post-conflict outcomes and whether a state legally imposes a given religion on its citizens.

This study analyzes the inherent limitations of traditional filtering methods in the filtered back projection (FBP) algorithm for CT image reconstruction. The main goal is to explore the feasibility of using convolutional neural networks (CNN) to replace traditional filters, thereby improving the quality of CT image reconstruction. This study introduces a novel deep learning back projection (DLBP) framework that combines a CNN with a back projection operator. This framework takes the advantage of CNN as deep learning filter in image processing. And keep the conventional back projection operator for transformation from sinogram domain to image domain. Unlike the other study, the back projection operator also involved the training process. While the CNN filter and the fixed back projection operator together mapping the sinogram and target image, the CNN part automatically becoming a data driven neural network filter. The materials used in this study include a public dataset from Shanghai Medical College of Fudan University, which contains 27,800 high-resolution chest and abdominal CT slices from 30 patients. These image data are preprocessed and the corresponding sinograms are generated, then sinograms used as inputs, while the original images are used as targets to form dataset pairs. In the study, the DLBP method was compared with the conventional FBP algorithm using image evaluation metrics such as mean square error (MSE), structural similarity index measure (SSIM), and peak signal-to-noise ratio (PSNR). The results show that the DLBP method performs significantly better than the FBP algorithm in noisy environments, achieving lower MSE, higher SSIM, and better PSNR values. The results highlight the potential of the DLBP framework in enhancing the quality of CT image reconstruction and suggest future research directions, namely validating the framework on larger datasets, exploring complex geometries, and leveraging advanced hardware to improve performance.

Surface guided radiation therapy (SGRT) is an emerging technology that uses non-ionizing methods for patient positioning and motion tracking during radiotherapy delivery. However, the use of radiotherapy boluses, which are tissue-equivalent materials placed on the skin to increase surface dose, has been shown to interfere with SGRT systems due to reflections from the bolus surface. This thesis presents the development and validation of an opaque and non-glossy radiotherapy bolus called the "Surface Guidance Optimized" (SGO), which is a variation of the previously developed transparent Clearsight bolus.The Surface Guidance Optimized bolus was rendered opaque by adding 0.6% titanium dioxide and given a matte finish using matte release paper. Spectroscopy measurements confirmed optimal opaqueness, while gloss meter readings verified a non-glossy surface. The bolus density was quantified to be 0.853 g/cm3 using water displacement and CT methods. Dosimetric characterization through direct surface dose measurements and Monte Carlo simulations demonstrated the SGO bolus mimics the dose deposition of water-equivalent materials when accounting for density differences. Compatibility testing with the AlignRT SGRT system showed the bolus allowed accurate surface reconstruction and submillimeter tracking (within 0.4 mm) under different lighting conditions. Overall, the SGO bolus mitigates issues of transparency and glossiness that interfered with SGRT systems, while maintaining desirable dosimetric properties for clinical use as a radiotherapy bolus compatible with modern surface guided techniques.

Introduction: The current head-and-neck (HN) fluence map generator tends to producehighly modulated fluence maps and therefore high monitor units (MUs) for each beam, which leads to more delivery uncertainty and leakage dose. This project implements diffu- sion into the training process and modifies the loss functions to mitigate this effect.

Methods: The dataset consists of 200 head-and-neck (HN) patients receiving intensity mod-ulated radiation therapy (IMRT) for training, 16 for validation, and 15 for testing. Two models were trained, one with-diffusion and one without. The original model was a con- ditional generative adversarial network (GAN) written in TensorFlow, the model without diffusion was written to be the PyTorch equivalent of the original model. After confirming the model was properly converted to PyTorch by comparing outputs, both new models were modified to use binary cross entropy for the GAN loss and mean absolute error as a third loss function for the generator. Hyperparameters were carefully selected based on the training script for the original model, and further tuned with trial and error. The diffusion was implemented based on Diffusion-GAN and the associated GitHub repository. The two new models were compared by plotting training loss vs epoch over 500 epochs. The two models were compared to the original model by comparing the output fluence maps to the ground truth using similarity index and comparing DVH statistics among the three models.

Results: The with-diffusion model and no-diffusion model achieved similar training loss.The diffusion model and no-diffusion model consistently delivered better parotid sparing than the original model and delivered less dose to four of the six tested OAR. The with- diffusion model delivered less dose to five of the six tested OAR. The diffusion model had the least MUs: 23% less than the original model and 3% less than the no-diffusion model. The diffusion model had lower D2cc: 4% less than the original model and 1% less than the no-diffusion model on average. All three plans deliver 95% of the prescription dose to nearly the same percentage of PTV volume.

Conclusion: Implementing diffusion does not provide a significant impact on training timeand training loss. However, it does enable comparable dose performance to both the no- diffusion and original models, while significantly reducing the total MU’s and 3D max 2cc relative to the original model and slightly reducing these metrics relative to the no-diffusion model, indicating smoother fluence modulation. In addition, both new models reduced dose to the right and left parotids relative to the original model, and to four of six tested OAR total, while the with-diffusion model consistently delivers less dose to OAR than the no- diffusion model. This indicates that both the new loss functions and diffusion reduce the overall dose to the OARs while preserving dose conformity around the target.

This thesis presents a comprehensive study on nonparametric Bayesian density estimation using Gaussian processes (GP). We explore the logistic Gaussian Process (LGP) and introduce an innovative approach termed the tree-logistic-link Gaussian process (TLLGP). This method aims to improve computational efficiency while maintaining modeling flexibility. We address the computational challenges traditionally associated with LGP by implementing a novel tree-based strategy, thereby reducing the complexity of posterior computations. Through a series of numerical experiments, we demonstrate the effectiveness of TLLGP in various scenarios, comparing its performance with other methods. The results highlight the advantages of our approach in terms of computational speed and accuracy in density estimation tasks. This work contributes to the fields of Bayesian statistics and machine learning by providing a more efficient tool for density estimation, especially beneficial for large high-dimensional data where traditional methods fall short due to their computational demands.

This paper seeks to contribute to the study of early postwar Japanese women’s history by focusing on rural women, a group that has received relatively less attention in recent scholarship. It aims to understand the changes in the lives and worldviews of Japanese farm women from 1945 to 1950 as shaped by the ambitious initiatives of the Supreme Commander for the Allied Powers (SCAP), the Japanese government, and the local reception and internalization of new ideologies. Through the lens of women-and-lifestyle-related content in Ie no hikari 家の光 (Light of the Home), the most influential rural family magazine in prewar and postwar Japan, this paper intends to explore how the magazine tailored official campaigns to the rural context with the help of local activists and farm women themselves, leaving both tangible and intangible impacts on the daily lives of women and their families. It also investigates the various ways in which local women responded to and interacted with the official new life campaigns that promised them concrete improvements in material lives and social status. As the magazine served as a middle ground where top-down initiatives intersected with local efforts to internalize official languages in the late 1940s, it also provides access to the local voices of farm women at the time. These precious voices, however limited, allow us to better situate rural women within the tabulating social milieu of early postwar Japan and to delve deeper into their daily lives.

Purpose: Currently, the Magnetization Prepared Rapid Gradient Echo (MPRAGE) Magnetic Resonance Imaging (MRI) sequence is frequently used for brain tissue segmentation in the clinic due to its high image contrast. However, one of the limitations of the MPRAGE sequence lies in its susceptibility to metal artifacts, while the Turbo Spin Echo (TSE) sequences, can resist metal artifacts. Previous studies have shown that for patients with metal implants, metal-artifact-reduced MPRAGE images can be generated from TSE images. Conventional brain segmentation methods on MPRAGE images, such as FreeSurfer, are time-consuming. Therefore, the purpose of this study was to investigate a fast brain segmentation method via deep learning-based frameworks for patients with metal implants, using TSE images as input.Materials and Methods: A dataset consisting of 369 patients in total was used. Each patient contained 160 two-dimensional slices of T1-weighted (T1WI), T2-weighted (T2WI), and PD-weighted (PDWI) TSE brain MR images, respectively. The matrix size of the original images was 240 × 240. Two types of MPRAGE as intermediate steps were synthesized from T1WI, T2WI, and PDWI using mathematical calculations or Conditional Generative Adversarial Network (cGAN) algorithms. FreeSurfer software was used to generate brain segmentations on the MPRAGE, which were considered as the ground truth for deep-learning network training and eventual evaluation. Two research aims were investigated. Aim 1 was to utilize three-channel TSE images (T1WI, T2WI, and PDWI) to first mathematically synthesize MPRAGE images, and then perform segmentation via deep learning-based models. Aim 2 was to use single-channel TSE images as input directly or indirectly to achieve brain segmentation using deep learning-based models. Both UNet and UNet++ models were examined. The Dice coefficient was used to evaluate the performance of the above-mentioned segmentation aims. Results: For Aim 1, the Dice coefficient between the ground truth and the cortex segmentations generated by the UNet++ network using three-channel TSE images as original input and mathematically synthesized MPRAGE as direct input was 0.919 ± 0.03. For Aim 2, the Dice coefficient between the ground truth and the cortex segmentations generated by the UNet network using single-channel TSE images directly as input was 0.602 ± 0.06. The Dice coefficient between the ground truth and the cortex segmentations generated by the single-channel TSE images as original input and cGAN-synthesized MPRAGE as direct input using the UNet++ network was 0.766 ± 0.07. Conclusion: Two aims using three-channel or single-channel TSE images as original input and brain segmentation as output were investigated in this study. Three-channel TSE images as original input, and mathematically synthesized MPRAGE as direct input to the UNet++ network showed superior results. Single-channel TSE images as original input and cGAN-synthesized MPRAGE as direct input to the UNet++ network showed relatively lower performance. Further research is warranted to improve the performance of single-channel TSE-based deep-learning segmentation methods. Keywords: UNet++, MRI, Brain Image, Segmentation, TSE, MPRAGE

AbstractPurpose: DW-MRI and their derived apparent diffusion coefficient (ADC) maps have been shown to be beneficial in the diagnosis and treatment of various cancer types. This work determines the potential role of DW-MRI and ADC maps in GTV delineation for gynecological cancer patients undergoing external beam radiation therapy (EBRT) and brachytherapy. Our study also looked at the longitudinal changes in DWI/ADC values during the course of external beam treatments, as well as during the five brachytherapy fractions. Methods: The first aspect of this study involved validating the console-derived DW image sets and ADC maps using an in-house Matlab code designed for this purpose. Next, the b-value, which describes the sensitivity of the imaging sequence to diffusion, was optimized through a quantitative and qualitative analysis. The quantitative analysis involved maximizing the contrast-to-noise ratios between the tumor and various structures, including the endocervical canal, endometrium, myometrium, and gluteal subcutaneous fat. The qualitative analysis had two radiation oncologists ranking different DWI sets at various b-values based on tumor conspicuity and total image quality on a scale of 1-5, 1 being the best and 5 being the worst. After determining the optimal b-value for DW image calculation, an analysis of GTV contouring was performed in Medical Image Merge (MIM). This involved a radiation oncologist contouring GTVs on three image sets; axial T2 MRI, axial T2 MRI fused with DWI at b=1300 s/mm2, and axial T2 MRI fused with ADC at b=0, 1000 s/mm2. This was done for 16 patients, 5 of whom had pre-EBRT and pre-brachytherapy scans and 11 of whom had only pre-brachytherapy scans. The contours between the three sets were compared on each scan using the Hausdorff distance, Jaccard index, DICE coefficient, and mean pixel value, all of which were calculated in MIM. The final portion of this study was a longitudinal look at the CTVHRs throughout the course of brachytherapy. The CTVHRs were analyzed on the axial T2 MRI, DWI, and ADC maps. Results: The contrast-to-noise ratios of the endocervical canal, endometrium, myometrium, and gluteal subcutaneous fat all compared to tumor were optimized at either b=1300, 1600, or 1800 s/mm2. DW images at b=1300s/mm2 were ranked the best by both physicians in terms of total image quality and tumor conspicuity for the qualitative analysis for b-value optimization. For GTN analysis, the volumes of the GTVs contoured with the help of the DW images and with the help of the ADC maps were not significantly different (p=0.23404). The Dice coefficients, Hausdorff distances, and Jaccard indices calculated with respect to the reference GTVs were not significantly different between the GTVs contoured with the help of the DW images and the GTVs contoured with the help of the ADC maps. The p-values were 0.84148, 0.56868, and 0.95216, respectively. The volumes of the reference GTVs compared to the volumes of the GTVs contoured with the help of DWI and ADC maps were not statistically significant with p-values of 0.6672 and 0.42372, respectively. The mean pixel values in the reference GTVs compared to the mean pixel values in the GTVs contoured with the help of DWI and ADC maps were not statistically significant with p-values of 0.17384 and 0.68916. The mean pixel values within the GTV contoured with the help of DWI were almost significantly lower than the mean pixel values within the GTV contoured with the help of the ADC maps (p=0.0536). Looking to artifact quantification, no significant artifacts were seen in the axial T2 MRI, DWI, or ADC map outside of the tandem contour in the ice water phantom experiment. In the longitudinal analysis of the CTVHRs, the percent difference between the largest and smallest average of the mean pixel values in Figure 7 is 17.4% with no apparent trend along fractions. The percent difference between the largest and smallest average of the mean pixel values in Figure 8 and Figure 9 are 27.3% and 6.5%, respectively. There is no apparent trend for the average of the mean pixel values when looking at the ADC maps, but the average of mean pixel values decreases throughout the brachytherapy fractions when looking at the DW images. The average standard deviation of the pixel values within each CTVHR on the axial T2 MR and DW image sets generally decreases throughout the course of brachytherapy, while the average standard deviation on ADC generally increases along fractions. However, the percent difference between the largest and smallest average standard deviation on the axial T2 MR image set, the DW image set, and the ADC maps is 18.5%, 43.4%, and 4.9%, respectively. Therefore, while the standard deviation on the ADC maps is generally increasing, it is to a small extent. Conclusion: The results of this study indicate the potential of using ADC maps in tandem with axial T2 MRI to increase the accuracy of GTV delineation in cervical cancer patients undergoing EBRT and brachytherapy. However, a larger sample size is needed to provide more insight into their use during the contouring workflow.

Due to a high burden of disease of pneumonia in Vietnam, the country not including the pneumococcal conjugate vaccine (PCV) in its National Expanded Programme of Immunization (EPI), and the scarce data on PCV vaccine coverage or caregivers’ behavior within the country, it is imperative to assess the Knowledge, Attitude and Practice (KAP) of the caregivers’ community, to further explore ways to increase PCV uptake. The purpose of this study is to understand the KAP of caregivers towards PCV inoculation for children in Hanoi, VietnamMethodology: 338 respondents fulfilled the Qualtrics questionnaire and 26 respondents (16 caregivers and 10 health workers) were interviewed in Hanoi, Vietnam, using semi-structured interviews in June-December 2023. Materials and data were transcribed between Vietnamese and English, and analyzed according to selected themes. Discussion/Conclusions: Although the findings suggest that caregivers in Hanoi have limited knowledge on PCV, support for attitude and practice on accepting PCV exists, especially from caregivers with high socio-economic status. This study wished to contribute to a better understanding of the KAP factors regarding childhood vaccines, which may support decision-making about vaccine policies, and be utilized for creating suitable vaccine promotion materials for child caregivers.

Background: Cervical cancer is a common gynecological malignancy among women worldwide. Among the primary modalities for treating cervical cancer, radiation therapy occupies a central role. Using Cone-Beam Computed Tomography (CBCT) scans obtained prior to treatment for target registration and alignment holds critical significance for precision radiation therapy. Accurately contouring targets and critical-organs-at risk (OARs) is the most time-consuming task for radiation oncologists. The OAR contouring in CBCT plays a crucial role in the radiotherapy of cervical cancer. Specifically, the location and volume of the rectum and bladder can significantly impact the precision of cervical cancer treatment, as the patients need to drink certain amount of water to fill the bladder prior to the treatment for target localization. The resulting change in position of rectum and bladder may lead to alterations in the target dose. Further, changes in radiation dose to these two OARs can directly affect the severity of the acute and late radiation induced damage. Therefore, the OAR contouring not only allows for better localization before each radiotherapy session, but also provides valuable reference for clinicians when they need to adjust the treatment plan.Purpose: The objective of this study is to evaluate the capabilities of four deep-learning models for contouring OARs in CBCT images of cervical cancer patients. Materials and Methods: The study dataset comprising 40 sets of CBCT images were collected from the Fujian Provincial Cancer Hospital in China. Two experienced radiation oncologists meticulously delineated 10 groups of OARs (Body, Bladder, Bone Marrow, Bowel Bag, Femoral Head L, Femoral Head R, Femoral Head and Neck L, Femoral Head and Neck R, Rectum, Spinal Canal) on the CBCT images as reference/ground truth. Subsequently, the 24 sets of CBCT reference were used to train the CBCT model, and the unedited CBCT images of the remaining 16 sets were used for comparing with their reference to test the four models. The only difference between these four models is the adoption of different neural network structures. They are classic U-Net, Flex U-Net, Attention U-Net (ATT), and SegResNet respectively. The evaluation of contouring quality for the four models was performed using the metrics such as 95 percentile Hausdorff Distance (HD95), Dice Similarity Coefficient (DICE), Average Symmetric Surface Distance (ASSD), Maximum Symmetric Surface Distance (MSSD), and Relative Absolute Volume Difference (RAVD), respectively. Results: The average DICE was 0.86 for bladder contouring among four models. The average DICE for rectum on CBCT image was 0.84 for four models. Conclusion: According to the quantitative analysis, classic U-Net neural network architecture with minor adjustments can obtain competitive segmentation on CBCT images.