36214 research outputs found
Sort by
Cellular Evolution in PEI: Mapping Bubble Dynamics through Solid-State CO_2 Foaming Processes
No full textThesis (Master's)--University of Washington, 2025Future applications of polyetherimide (PEI) extend beyond structural panels and insulation to include vibration-damping layers, impact-resistant aerospace skins, thermal-acoustic insulation, and multifunctional components in satellites and aircraft cabins. By using dynamic mechanical analysis (DMA), a constitutive model for bubble growth under CO_2 saturation is developed, capturing glass transition effects and growth of nanofoam cells. Key properties such as cell growth rate, CO_2 concentration, porosity, and stress are extracted, supporting simulation-driven design of next-generation aerospace structures with enhanced durability, thermal stability, and energy absorptio
Evaluating Multi-Modal Data Fusion Approaches for Predictive Clinical Models Using Multiple Medical Data Domains
No full textThesis (Ph.D.)--University of Washington, 2025Disease outcome prediction is a central research focus in biomedical informatics, as it facilitates precision health related interventions and scientific discovery by enabling digital clinical trials and multiple other benefits. Multimodal deep learning models have emerged as powerful tools in biomedical research, offering the ability to integrate diverse data sources such as clinical records, multi-omics data, imaging, survey responses, and wearable data to enhance predictive accuracy and deepen understanding of medical phenomena. Central to multimodal modeling is the process of data fusion, where information from different modalities is integrated into a unified model. Three primary fusion strategies exist in deep learning: early fusion (feature-level), intermediate fusion and late fusion (decision-level). While widely adopted in other domains, their comparative performance and implementation considerations remain underexplored in biomedical applications, where data heterogeneity, missingness, and varying dimensionality present additional challenges.This dissertation aims to evaluate the implications of data fusion strategies for developing multimodal predictive models in medicine. Across three distinct aims, I assess the impact of early, intermediate, and late fusion techniques on predictive performance, implementation complexity, and generalizability using diverse combinations of data types, outcomes, and modeling strategies. These studies span multiple datasets and outcome types (binary categorial variables vs continuous ratio variables) providing a broad view of fusion strategy utility in real-world biomedical settings.
In Aim 1—Evaluation and comparison of early, intermediate, and late fusion techniques for combining exposures, clinical and genomics data for disease risk prediction task using All of Us: Risk of CKD in patients with type 2 diabetes—I evaluated and compared early, intermediate, and late fusion strategies for integrating longitudinal EHR, genomic, and survey data to predict chronic kidney disease (CKD) progression in patients with type 2 diabetes using a novel transformer-based multimodal architecture. Using data from the NIH’s All of Us initiative, I trained models on a cohort of approximately 40,000 patients. While the best performing unimodal model achieved a baseline performance with an AUROC of 0.73 (0.71 - 0.75), the inclusion of multimodal data offered only marginal improvement with an AUROC of 0.74 (0.72 – 0.76), with the benefit limited to the early fusion approach and lacking statistical significance. This aim highlighted the challenges of integrating multimodal data with different dimensions using transformer models and emphasized the role of modality-specific relative predictive strength.
In Aim 2—Development and assessment of the incremental value of combining a deep convolutional neural network feature extractor on imaging data and clinical data on a binary prediction task: Predict post-surgical margin status in soft tissue sarcoma—I extended the fusion analysis to imaging data by combining a convolutional neural network (CNN) trained on longitudinal cross-sectional imaging with a shallow neural network trained on clinical and pathology variables to predict post-surgical margin status in patients with soft tissue sarcoma (n=202). Here, the intermediate fusion strategy significantly outperformed other approaches, achieving an AUROC of 0.80 (0.66–0.95), suggesting that cross-modal interactions between histologic features and imaging embeddings may be best captured through intermediate fusion. This result demonstrated the potential value of intermediate fusion when complementary signals exist across modalities.
In Aim 3—Evaluation and comparison of early, intermediate, and late fusion techniques for combining imaging and clinical data on a regression prediction task: Estimation of CT-based body composition metrics from chest radiographs—I explored fusion strategies for estimating continuous CT-derived body composition metrics (e.g., visceral, and subcutaneous fat volumes) using only chest radiographs and clinical variables in a dataset of 1,088 patients. A multitask multimodal model was developed and evaluated across early, intermediate, and late fusion strategies. Late fusion consistently delivered the best performance across most body composition metrics, closely followed by intermediate fusion. These results suggest that when individual modalities offer high independent predictive power, decision-level integration may be optimal for regression tasks.
Collectively, these aims provide a broad evaluation of data fusion strategies in multimodal biomedical modeling, highlighting their strengths, limitations, and practical considerations. Findings suggest that no single fusion strategy universally outperforms the others; rather, optimal fusion depends on data characteristics, model architecture, and task-specific objectives. This dissertation lays the groundwork for future research aimed at developing adaptive fusion strategies tailored to the complexities of real-world biomedical data
Physical Activity Patterns and Kidney Function Changes Among Hispanic/Latino Adults: An Analysis from the HCHS/SOL Cohort
No full textThesis (Master's)--University of Washington, 2025Background: Physical activity (PA) may be a modifiable factor for CKD prevention, but evidence specific to Hispanic/Latino populations is limited. Objective: To examine the association between PA patterns and longitudinal changes in kidney function among Hispanic/Latino adults. Methods: We conducted a secondary analysis of the Hispanic Community Health Study/Study of Latinos, a prospective cohort of 16,415 self-identified Hispanic/Latino adults aged 18–74 years, recruited from Chicago, Miami, the Bronx, and San Diego (2008–2011). Participants with complete accelerometer data, kidney function measurements at baseline and follow-up, and relevant covariate data were included. Of 8,073 participants in the main analysis, 7,177 were included in the incident CKD analyses after excluding those with baseline CKD. PA was assessed via 7-day accelerometry (light PA: 10–1534 counts/min; moderate-to-vigorous PA [MVPA]: ≥1535 counts/min) and self-reported measures using the Global Physical Activity Questionnaire, categorized as total MVPA and recreational MVPA. Primary outcomes were annual percent change in estimated glomerular filtration rate (eGFR) and annual change in urine albumin-to-creatinine ratio (UACR). The secondary outcome was incident CKD, defined as eGFR 1 mL/min/year decline or UACR ≥30 mg/g at follow-up. Results: Among 8,073 participants (mean age 42.5 years; 55.7% women), the average annual percent change in eGFR was –0.61% (95% CI: –0.68% to –0.54%), and the average annual change in UACR was +2.03 mg/g (95% CI: 0.98 to 3.07). In fully adjusted models, accelerometer-measured light PA and MVPA were not significantly associated with eGFR change (light PA: –0.009% per 15 min/day; 95% CI: –0.024% to 0.006%; MVPA: –0.012% per 15 min/day; 95% CI: –0.046% to 0.023%). However, self-reported recreational MVPA was associated with significantly lower UACR (–0.208 mg/g per year per 15 min/day; 95% CI: –0.390 to –0.026). No significant associations were found between PA and incident CKD. Conclusions: Among Hispanic/Latino adults, higher recreational MVPA was associated with reduced albuminuria, suggesting potential kidney health benefits. However, no significant associations were observed between overall physical activity and kidney function decline or incident CKD. These findings highlight the need for further research to explore why recreational PA, but not total or accelerometer-measured PA, may be linked to kidney health in this population
A Paradigm Shift in Precipitation Modeling: Moving Beyond Numerical Models
No full textThesis (Master's)--University of Washington, 2025Accurately representing surface precipitation in weather and climate modeling is crucial to practical operational use of these models. Presently, global numerical weather prediction (NWP) models struggle to recreate the precipitation variable due to unresolved physical processes at the subgrid level as well as use of poorly constrained microphysical parameterizations. The advent of machine learning in the field of weather and climate prediction has proven to be beneficial to advance modeling efforts and has the potential to also target weak points in NWP such as estimating the precipitation field. Training a deep learning model using satellite data, we can bypass the parameterizations traditionally used by NWP to produce precipitation, achieving a field that more closely matches observations than the widely used ERA5 reanalysis dataset. The resulting model can compute precipitation from only ten ERA5 input fields and is able to better capture extremes while also improving the issue of overproduction of light precipitation in the ERA5 product when evaluated against the IMERG satellite dataset. The machine learning model is also used to produce precipitation from NWP forecast fields, improving on the forecasted precipitation of the NWP model. This work supports future development of deep learning models that meet the needs of current weather and climate modeling
Students’ Engagement with Mental Disability Knowledge and Theory in the Writing Classroom: Implications for Transfer Research
No full textThesis (Ph.D.)--University of Washington, 2025Drawing on data from a 10-week, classroom-based, qualitative study, this dissertation investigates students’ micro-level processes of transferring knowledge about mental disability from a critical disability studies perspective in a general writing course. The study was conducted at the University of Washington within my self-designed and taught Intermediate Expository Writing course which was meant to build students’ critical disability literacy. This study builds on the work of disability studies, writing studies, uptake theory, and knowledge transfer scholars, including the ongoing and prominent conversations about knowledge transfer in writing studies as well as theories of students’ boundary-work. The results demonstrate that students’ engagement with mental disability knowledge and theory in the writing classroom is deeply informed by their incoming relationships with the subject matter and the innumerable interactions that occur between students’ experience, each other, concepts, readings, writing tasks, and material space/place. Student writing, survey, and interview data were analyzed using thematic and discourse analysis and within the framework of Dylan Medina’s (2017) concept of micro-transfers or moment-to-moment interactions which inform how students define people, places, and things. Tracing such micro-transfers revealed how students uniquely take up and adapt knowledge about mental disability within a single course and across writing tasks. Through this analysis, I present five themes that help complicate traditional understandings of knowledge transfer as broad, easy-to-see translations of knowledge across contexts. The analysis illuminates the unique shifts in students’ relationships to disability, their dynamic process of adapting knowledge about disability, the impact of class environments on knowledge transfer, students’ need for alternative pathways for engaging with difficult concepts, and the potential of courses centered around critical disability studies to support student advocacy stances. These findings illustrate the use of micro-transfer as a framework for the analysis of knowledge transfer and course design and the benefits of incorporating critical disability studies discourse into general writing courses. To close, I offer implications for transfer, writing, and disability studies research; teaching; and writing program administration
Gốc Rễ as Craft: Notes on Survival and Knowing
No full textThesis (Master's)--University of Washington, 2025There is one thing I hold dear is the notion that narrative desire, in any circumstances, arises most strongly once historical erasure takes place, a local, personal reckoning naturally forming against the global, sweeping forces. After the American War in Vietnam, after generations lost to and were displaced by violence, the search for an identity in each individual becomes more sacred, symbolic and urgent, but as quiet as it can be. This critical essay is an amalgamation of autobiographical and critical writing on craft, survival, and language through a transnational lens informed by Vietnamese and Vietnamese American experiences. I look at the notion of mất gốc as a personal trauma and a generative space for craft. I look at the gaps and silences within the literary and historical canon. The works by Ocean Vuong, Nguyễn Phan Quế Mai, Aimee Phan are relevant here as they facilitate my inquiries into the queer, diasporic, and transnational magnitudes of Vietnamese and Vietnamese American narrative works. I imagine a future for a novel that resists the assimilationist gaze and places the fragmented, haunted lives of Vietnamese Amerasians at the center stage. Ultimately, I argue for a literature and storytelling as an embodied act that sees the "art of staying afloat" as craft
Atrial Cardiomyopathy and Longitudinal Outcomes in the UK Biobank
No full textThesis (Master's)--University of Washington, 2025Background: Atrial cardiopathy often precedes atrial fibrillation (AF) and has emerged as an independent risk factor for cardiovascular outcomes. Previous studies assessing atrial function in relation to clinical outcomes have been small, and the importance of right-sided function is largely unknown. Methods: In 51,693 UK Biobank participants with no previous history of AF, we assessed left and right atrial volumes and ejection fraction from cardiac magnetic resonance imaging (CMR) using deep learning segmentation. We evaluated associations with new-onset AF, ischemic stroke, heart failure, and dementia and conducted a genome-wide association study of each atrial measure. Potential causal associations with downstream outcomes were evaluated using Mendelian randomization.
Results: During a median follow-up time of 4.0 (IQR 2.9-5.4) years, 964 (1.9%) developed AF, 266 (0.5%) developed ischemic stroke, 365 (0.7%) developed heart failure, and 72 (0.1%) developed dementia. After adjustment for clinical risk factors, left and right atrial measures were independently associated with risk of new-onset AF, ischemic stroke, and heart failure, with stronger associations in women. Left atrial minimal volume was also associated with increased risk of dementia. We identified 51 genetic loci associated with left and right atrial measures (p <5Ã 10-8), including 27 novel trait-locus associations, many of which do not overlap with established AF loci. Genetic correlations showed that both chambers had similar correlations with AF but varying correlations with blood pressure, body mass index, and type II diabetes. In Mendelian randomization analyses, left atrial measures had direct causal effects on AF, ischemic stroke, and cardioembolic stroke, though stroke associations were no longer significant after accounting for AF variants.
Conclusion: In this largest assessment of atrial structure and function to date, both left and right atrial cardiopathy were associated with clinical outcomes attributed to AF. We identified several novel genetic loci for left and right atrial traits and observed unique genetic correlations between left and right atrial traits and cardiovascular phenotypes, providing insight into chamber-specific remodeling. Several of these measures are likely to be causal determinants of cardiovascular complications previously attributed to AF, which may be mediated by intercurrent AF. These findings highlight atrial cardiopathy’s potential as a screening tool to identify individuals at high risk for cardiovascular complications
Morphological Comparisons between the subtidal clam Glycymeris septentrionalis and the intertidal clam Leukoma staminea using micro-CT scanning
No full textThis study utilized micro-CT scanning to compare the morphologies of the subtidal clam Glycymeris septentrionalis and the intertidal clam Leukoma staminea. Three subtidal clams (Glycymeris septentrionalis) were collected from Van Veen grabs on the Kittiwake in the San Juan Channel (48 32.491' N, 122 58.940' W) on July 3rd, 2025. Three intertidal clams (Leukoma staminea) were collected from Argyle Lagoon (48°31'15.2"N, 123°00'48.7"W) on June 26th and July 10th, 2025. Clam species were identified based on their size and location compared to iNaturalist to preserve morphology for micro-CT scanning. All six clams were relaxed in magnesium chloride before being preserved in formalin overnight. The clams were stained in iodine, with small holes drilled in ones that did not relax open under the magnesium chloride. Staining the clams with iodine allowed the soft tissues to be imaged by the scanner. To this end, the clams were soaked in iodine for at least 18 hours. All six clams were scanned using a micro-CT scanner at a resolution of 45 μm under a continuous scan. Based on a previous literature search, this was the first time these two clam species were scanned using a micro-CT scanner. The software 3D slicer was used to examine the six clams. Each clam was segmented into its own file. Threshold values 89.25 and 255.0 were used to only render the clam shells. This study had two aims for the micro-CT scans: a qualitative review of what morphology features were modeled by the scans, and a quantitative study of shell thickness with the hypothesis that subtidal clams would have a thicker shell than intertidal clams, since the intertidal is a harsher environment that could wear down the shells. Radial ribs, hinge ligaments, hinge teeth, lateral teeth, and some of an adductor muscle for an intertidal sample were observable
in the micro-CT scans. Shell thickness was measured in 3D Slicer for all six clams. An F-test was conducted to determine if the T-test for mean shell thickness should assume equal or unequal variance. The p-value was 0.17. A T-test of two samples assuming unequal variance was conducted in excel to determine if the difference in average shell thickness for the two species was statistically significant. The p-value was greater than 0.05 (p=0.60), indicating a failure to reject the null hypothesis that the average shell thickness between the two species was different. As such, the hypothesis that the subtidal clams would have thicker shells than the intertidal clams was not supported. Still, this study demonstrates how micro-CT scanning allows for measurements to be conducted without sample destruction
Mapping the Interior: Land Offices, Technology, and Bureaucratic Development in 19th Century America
No full textThesis (Ph.D.)--University of Washington, 2025Founded in 1812, the General Land Office (GLO) was the bureaucratic engine of American settler expansion. Beginning in the early 1800s as a loosely affiliated network of land offices, the GLO would develop throughout the 19th century into the primary institution which determined the pattern and process of settlement in the American west. In this dissertation, I argue that the GLO developed this power through a process of bureaucratic consolidation. As the agency consolidated, the GLO utilized technologies of mapping and contract to claim autonomous power. These technologies of land management were bound to the racial and social order of American political development. Political elites created early land offices to establish control over settler populations. Rather than seeking to dominate these distant populations through force, elites used the public provisions of the land office to encourage settlers to recognize the American state. I demonstrate this process using the case of Mississippi Territory, where land offices gave formal recognition to settler property rights, providing economic security for settlers and forming the foundation for the growth of enslavement in the region. As settlement expanded under President Andrew Jackson many of the processes undergirding the structure of early land offices became increasingly untenable under the GLO. In 1836, this would become a crisis for the agency, creating a critical juncture and leading to the first major consolidation of the GLO after Congressional action. However, in passing new legislation, Congress failed to change the fundamental structure of GLO technologies. In 1849, the GLO would consolidate again, becoming part of the newly created Department of Interior. In this new department, the office would seek to exercise autonomous power through its technologies of mapping and contract. Examining the case of Oregon Territory, I show that, collaborating with the Office of Indian Affairs, the GLO would claim the power to map Indian Reservations, despite lacking the formal authority. This project would lay the foundation for allotment, leading to the seizure of Indigenous land and the forced assimilation of Native nations. By the end of the 19th century, the consolidation and autonomy of the GLO had reached their peak. However, with new modes of land management emerging, the outdated technologies of the GLO would struggle to adapt. As a result, the 20th century would lead to the decline of the GLO as a center of bureaucratic power
Metabolite-predicted biological age acceleration and associations with oncologic, neurodegenerative, and adiposity-related endpoints in multi-ancestry populations
No full textThesis (Master's)--University of Washington, 2025Chronologic age is widely understood as a significant risk factor for many chronic diseases and is frequently used in risk stratification to guide selective screening. However, for age-related diseases, chronological age alone may not accurately reflect the underlying heterogeneity in biological aging that contributes to disease risk. Using prospectively collected data from the Women’s Health Initiative (WHI), the Multiethnic Cohort (MEC), and the Wisconsin Registry for Alzheimer’s Prevention (WRAP), we developed statistical models to estimate biological age acceleration from untargeted LC/MS metabolomics data and evaluated associations with, primarily, breast and prostate cancer, cognitive function scores, and BMI. Additionally, we conducted pathway analyses to identify metabolic pathways significantly associated with chronologic age. Model performance was consistent across cohorts, with RMSE ranging from [5.36, 6.41] years. We did not find evidence of a statistically significant association between a 10-year increase in metabolite-predicted biological age acceleration and disease risk, including for any oncologic outcomes, cognitive scores, or BMI. We detected several enriched metabolic pathways across cohorts, including C21-steroid hormone biosynthesis and metabolism, urea cycle/amino acid group metabolism, and carnitine shuttle