36214 research outputs found
Sort by
The River at the End of the World: Architectural Coexistence Along the Klamath River
No full textThesis (Master's)--University of Washington, 2025This thesis explores the history of the Klamath River and the ever-changing landscape that it shapes and by which its path is determined. In late 2024, four out of the five dams blocking the Klamath River were removed allowing the river to return to its historical routes, opening pathways for migration and allowing the river to express natural behavior that can be observed through six different landscape sites: the spring, the floodplain, the ravine, the valley, the confluence, and the delta. Today, as the river begins to distinguish its course through the landscape once again unimpeded, this project proposes interventions within these six distinct landscapes to allow for the witnessing of these characteristics of the river to reconnect with the cycles of water and wetness, and to think of new ways of observing and engaging with the realities of changing landscape. This thesis focuses on two sites in particular: the ravine and the delta. At the ravine a bridge is proposed to span the river at the site of a former train bridge crossing. From the bridge a tower is constructed to mark the changes in water level during and after the presence of the dams. At the delta site a series of walls are proposed, which structures are built, to help visitors observe the uncertainty of a landscape that is in constant change. Together, along with the six other sites, these two proposals constitute a different kind of relationship architecture may have with water and landscape
Applying Compositional Data Analysis Methods To Complete Blood-Counts Data For Early COVID-19 Detection
No full textThesis (Master's)--University of Washington, 2025The investigation of using complete blood-count (CBC) data analysis for COVID-19 infection diagnosis has been a topic of interest in the last couple of years. It could be used as an affordable complementary tool to RT-qPCR and is particularly useful for developing areas struggling to test their population or suffering from massive COVID-19 infection. However, previous research of using CBC data for COVID-19 infection classification didn’t appreciate the compositional nature of white blood cell counts data. In this master’s thesis, we treat white blood cell counts data as compositional variables and apply compositional data visualization methods, using biplots based on log-ratio principal component analysis. Also, we apply compositional classification models to detect COVID-19 infection, using log-ratio linear discriminant analysis. We successfully illustrate the efficacy of compositional methods by building a compositional classification model superior to traditional models and highlight the benefits of analyzing CBC data from a compositional perspective. In a database of symptomatic individuals, we achieve a classification rate of 85% for the PCR-test result using the main CBC composition with some additional blood characteristics
Approximate Methods for Simulating Optical Properties of Plasmonic Nanoparticle Assemblies
No full textThesis (Master's)--University of Washington, 2025Simulating optical responses of plasmonic nanoparticle assemblies is a crucial milestone for advancing the design and optimization of optical metamaterials and photonic devices. Since optical responses are highly dependent on light frequencies sampled on particle structures, simulations often iterate its calculation on small increments of frequencies to preserve high resolution for the resulting spectra, leading to a high computational cost. In this thesis, we introduce the Kramers-Kronig resonance frequency (KK-res) approximation. KK-res only requires two particle dipole calculations at low- and high-frequency limit, rather than particle dipoles at each frequency, to estimate the full extinction spectra of large particle configurations (). This allows KK-res to gain an order-of-magnitude computation time improvement comparing to other coarse-grained mutual dipoles methods. Several experimental studies were also simulated with KK-res to gauge the accuracy and utility under conditions like various anisotropic structures and dielectric models
Evaluating and Enhancing Large Language Models (LLMs) in the Clinical Domain
No full textThesis (Ph.D.)--University of Washington, 2025Recent advancements in large language models (LLMs) have demonstrated human-level performance on many specialized medical tasks, even without annotated training data. However, three main challenges remain: (1) due to the sensitive and highly specialized nature of clinical narratives, as well as the high cost of human expert annotation, there is a lack of high-quality, well-structured, and clinically meaningful datasets for LLM training and evaluation; (2) current medical LLMs show limited generalization ability to interpret and extract complex clinical information on certain unseen natural language understanding (NLU) tasks; and (3) as LLMs are typically trained on vast amounts of data, there is a substantial risk of data contamination, where evaluation benchmarks unintentionally overlap with training data, leading to inflated test performance and potentially reduced performance on truly novel tasks. In this work, we address these limitations through three core aims: (1) develop benchmark datasets for clinical information extraction (IE), a key NLU subtask, across two critical medical domains, and evaluate the performance of multiple state-of-the-art (SOTA) transformer-based language models (LMs), under both fine-tuning and in-context learning settings; (2) develop a more generalizable medical NLU model via instruction tuning, demonstrating enhanced performance on previously unseen clinical NLU datasets; and (3) systematically review existing detection approaches for data contamination and evaluate those approaches on datasets used during pre-training and fine-tuning LLMs, with our own and three other widely used open-source LLMs. In summary, our work contributes to the development of both clinical benchmarks and robust LLMs, as well as highlighting the ongoing challenges in benchmarking LLMs' generalizability
Parameterizations and rectifiability via geometric functions and singular integral operators
No full textThesis (Ph.D.)--University of Washington, 2025Geometric measure theory provides the framework to examine the geometry of sets and measures that are not ``smooth'' enough to be studied using the classical methods of differential geometry. These rough sets arise naturally in many settings, for instance as the minimizers of certain geometric variational problems, e.g \cite{R60}. A relatively recent technique used to analyze the fine properties of sets of this type is through the study of geometric square functions, which capture simple geometric information about the set at each scale. This technique has gained popularity starting with the work of Jones \cite{J90} and David-Semmes \cite{DS91}. In \cite{Bi87} a discontinuous geometric function appeared, known as the Carleson -function. In \cite{JTV21} and \cite{FTV23} the authors prove that qualitative control on the Carleson -function, and higher-dimensional analogues, characterize tangent points of certain domains. In this thesis, we study higher regularity versions of Carleson's conjecture in the plane and in higher dimensions.
As for the study of the ``smoothness'' of a given measure on a commonly used tool in geometric measure theory are \textit{singular integral operators}. The fine properties of measures involve two things: the ``smoothness'' of the set on which the measure lives and the behavior of the measure on that set. In \cite{DS91}, the authors prove that uniformly rectifiable measures are characterized by the -boundedness of all Calder\'{o}n-Zygmund operators. In \cite{M95} and \cite{MP95}, the authors prove that the almost everywhere existence of principal values of the Riesz transform characterizes rectifiable measures. In this thesis we extend the work of \cite{M95, MP95} to a broader class of singular integral operators, in particular, a class of anisotropic Riesz kernels
An Experimental Study of Soil Health Score and Nutrient Density
No full textThesis (Master's)--University of Washington, 2025Regenerative agriculture has gained attention as a holistic approach to farming that prioritizes environmental and public health outcomes while reducing adverse outcomes associated with conventional, industrialized agriculture. A primary goal of regenerative agriculture is to improve soil health through practices such as minimizing soil disturbance and increasing soil organic matter. Although improving soil health is assumed to improve the nutritional content and health benefits of crops, such as those that come from phytochemicals and micronutrients, the relationships by which soil and environmental conditions impact food’s nutrient density are not fully understood. This preliminary study aims to contribute insights into how soil health impacts nutrient and phytochemical density of produce, and ultimately human nutrition. This study first investigates four plots farmed regeneratively for different lengths of time. This diverse time horizon produced a spectrum of soil health scores in which we grew two crops, Beta vulgaris (Swiss chard) and Phaseolus vulgaris (beans). These crops were harvested and analyzed for their mineral and phytochemical concentrations. These results were combined with soil health scores to test the hypothesized relationship between nutrient/phytochemical density and soil health. Results for both minerals and phytochemicals were mixed, with some non-associations found in Swiss chard and beans, and some positive or negative relationships with soil health score. These findings suggest four potential mechanisms worthy of further investigation
The Influence of Maternal Immune Events on the Establishment of Maternal Microchimerism
No full textThesis (Master's)--University of Washington, 2025Maternal microchimerism (MMc), or the transfer of maternal cells to the offspring, has been associated with immune modulation and improved health outcomes in newborns. Using data collected from a prospective cohort, we assessed how maternal immune events during pregnancy, including vaccinations and illnesses, influenced the odds of detecting MMc in cord blood. Additionally, we examined the dynamic changes that occurred in the maternal T cell repertoire across trimesters. We observed that vaccination was associated with higher odds of detectable cord blood T cell MMc (OR: 1.68, 95% CI: 0.42- 6.75) and that this effect was modified by gestational age, with the strongest effect with vaccination in first trimester (OR: 1.38, 95% CI: 0.36- 5.29). We also found that T cell repertoire varied greatly across individuals and by trimester. These findings suggest that the timing of immune exposures and the individual immune history of the mother are important factors influencing the transfer of maternal T cells to the fetus, with potential implications for offspring immune development and health
Healthcare resource utilization and costs of commercial and Medicare-enrolled adult patients with Multiple Sclerosis in Urban and Rural settings
No full textThesis (Master's)--University of Washington, 2025Background: Multiple sclerosis (MS) is a chronic progressive neurological disease that affects almost 1 million people in the United States, with an economic burden of over $85 billion USD. Previous studies have demonstrated that rural MS patients have reduced access to a neurologist compared to urban MS patients, but how this disparity impacted these patients was not explored. However, healthcare resource utilization and cost in MS patients in rural and urban areas has not been previously studied. Objective: The primary objective of this study was to describe the healthcare resource utilization (HCRU) and healthcare costs of commercial and Medicare-enrolled patients with MS in rural and urban settings. Methods: We conducted a retrospective cohort study using health insurance claims from the Merative MarketScan® Commercial and Medicare Supplemental Databases. MS cases were identified as having ≥2 MS outpatient (OP) claims, dated ≥30 days apart, within a 1-year period or ≥1 MS inpatient (IP) claim (ICD-10: G35) during the index/case identification period. The index date was defined as the first occurrence of meeting case criteria during December 31st, 2017, to December 31st, 2022. We examined all-cause HCRU by summarizing the number of IP admissions, OP services, emergency department (ED) visits, and OP pharmacy fills in the 6 months post-index. Additionally, we assessed all-cause health care costs, including IP, OP, ED, and pharmacy, over 1 year following index using the Kaplan-Meier sample average estimator to account for censoring. We reported total healthcare costs as the sum of IP, OP, ED and OP pharmacy costs. Analyses were conducted separately for the commercial and Medicare supplemental groups; all-cause HCRU and costs were summarized overall and by rural and urban status.
Results: Our cohort consisted of 47,636 commercially insured patients and 3,318 Medicare patients. During the 6 months following index date, rural commercial patients had a slightly elevated mean number of IP admissions (1.4) and length of stay (LOS) (4.4 days) than the urban patients (1.3, 4.4). This trend is not repeated in the Medicare Supplemental population. In that population, the rural cohort had a lower mean number of IP admissions (1.3) and LOS (5.00) compared to the urban cohort (1.4, 5.9). For ED visits, in both the commercial and Medicare cohorts, the rural patients had lower average ED visits than the urban patients (Commercial: 1.5 vs. 1.6 rural vs urban) (Medicare: 1.7 vs. 1.8). OP visits follow this trend with the rural cohort having fewer visits among those that did have OP visits 6 months post-index in both insurance cohorts (Commercial: 9.4 vs. 10.0 rural vs urban) (Medicare: 12.7 vs. 14.4). However, there is a difference seen in the OP pharmacy fills. For both commercial and Medicare insurance, the rural cohorts had higher mean and median OP pharmacy fills than the urban cohorts The commercial rural patients had a mean of 31.5 (SD 20.8) 30-day standardized pharmacy fills and a median of 27.3 (IQR 15.9 – 42.6) compared to the urban patients with a mean of 27.7 (SD 19.4) and median of 23.4 (IQR 13.4 – 37.3). The Medicare rural patients had a mean of 38.1 (SD 23.9) 30-day standardized pharmacy fills and a median of 35.9 (IQR 22.3 – 48.9), compared to the urban patients with a mean of 36.6 (SD 20.2) and a median of 34.2 (IQR 22.3 – 48.7).
Conclusion: Rural patients tended towards lower average healthcare resource utilization (HCRU), and thus, lower costs than the urban patients, with some notable exceptions. Rural patients had higher mean and median pharmacy fills and higher costs in both insurance categories. This may point to rural MS patients relying more on prescription drugs to compensate for reduced access to medical services. MS patients require complicated care and insight into their HCRU and costs, which can aid healthcare decision-makers in optimizing care for these patients
Spacecraft Rendezvous with Non-Linear Dual-Quaternion Controller Under Orbital Perturbations
No full textThesis (Master's)--University of Washington, 2025A non-linear control system based on several controllers is applied the satellite rendezvous problem with a dual quaternion framework. The spacecraft pose is simulated under the effects of typical orbital perturbations with some initial proximity already established in advance. The objective is to present an approach to the rendezvous and proximity operations problem with the addition of attitude constraints using control barrier functions and dual quaternion feedback. Such applications with dual quaternions may also be applied to planetary descent and landing, six degree of freedom simulation, video graphics, and robotic pose control
Redox-Active Metal–Organic Cages and Frameworks for Electrochemical Applications
No full textThesis (Ph.D.)--University of Washington, 2025With over thirty percent of industrial energy consumption in the United States stemming from chemical manufacturing, there is a pressing need to develop new methodologies to lower the energetic cost of chemical production. Industrial thermal catalysis strategies often require elevated temperatures and pressures, with the iron-catalyzed Haber–Bosch process operating at over 400 C and 200 atm. Electrocatalysis offers a greener, more sustainable, and critical alternative to traditional thermal catalysis, requiring less overall energy for chemical synthesis. This process is driven by redox–active molecules and materials that can reversibly donate or accept electrons to reduce or oxidize a substrate, triggering further chemical transformations into a desired product. While common electrocatalysts include redox–active organic molecules and supported metal nanoparticles, one particularly promising class comprises porous redox–active materials and molecules, such as coordination cages and metal–organic frameworks. This work describes the application of redox–active coordination cages and metal–organic frameworks in electrochemical applications, with a specific focus on their synthesis and characterization. Chapter 1 summarizes the current landscape of redox–active coordination cages and metal–organic frameworks, comparing these classes of materials with particular attention to their use in electrochemical methods. While coordination cages have been widely explored for traditional thermal catalysis, their loss of stability and solubility upon changes in charge state has limited their electrochemical applicability. Specific examples of coordination cage electrocatalysts, such as those used in carbon dioxide reduction, are discussed. Additionally, strategies for imparting conductivity to two-dimensional metal–organic frameworks are presented.
Chapter 2 introduces an [Fe4L6]8+ coordination cage electrocatalyst, capable of being readily recycled via precipitation triggered by an overall reduction in charge state. This perylene diimide cage exhibits a wide range of charge states, spanning from +18 to −16, with multiple reversible redox events across a potential window greater than 2 V in acetonitrile. It serves as an effective catalyst for the electrochemical reduction of a series of vicinal dihalides, producing the corresponding alkenes with Faradaic efficiencies near unity. Furthermore, this cage is easily recyclable, leveraging the loss of solubility upon reduction to simplify post-catalytic separation. This work motivates further investigation into the role of cage charge, cavity size, and host–guest interactions in electrocatalysis.
In Chapter 3, an analog of the perylene diimide cage based on a pyromellitic diimide ligand is introduced to modulate the redox properties of the resulting cage. Replacing the perylene diimide core with the pyromellitic diimide moiety shifts the cage’s reduction events to more strongly reducing potentials. The ability of this new cage variant to serve as an electrocatalyst for the reduction of a vicinal dibromide is evaluated. Additionally, evidence is presented for the role of ligand design in cage assembly: modification of the pyromellitic diimide ligand to include solubilizing methyl groups leads to the formation of the [Fe4L6]8+ cage as three discrete diastereomers, as opposed to the unmodified ligand, which yields both the tetrahedral cage and triple helicate species. This work expands the library of redox-active coordination cages and motivates the continued investigation into redox-active cage design.
Chapter 4 marks a departure from the electrochemistry of coordination cages to investigate the impact of ligand oxidation state on the synthesis of Cu3(HHTP)2 (HHTP = 2,3,6,7,10,11-hexahydroxytriphenylene), a two-dimensional conductive metal–organic framework (MOF). The morphology of conductive MOFs strongly impacts their performance in applications such as energy storage and electrocatalysis. However, identifying the appropriate conditions to achieve a specific nanocrystal size and shape can be a time-consuming, empirical process. This work demonstrates how partial ligand oxidation dictates the morphology of Cu3(HHTP)2, a prototypical 2D conductive metal–organic framework. Using organic quinones as the chemical oxidant, we show that partial oxidation of the ligand prior to metal binding alters the nanocrystal aspect ratio by over 60-fold. Systematically varying the extent of initial ligand oxidation leads to distinct rod, block, and flake-like morphologies. These results represent an important advance in the rational control of Cu3(HHTP)2 morphology and motivate future studies of how ligand oxidation impacts the nucleation and growth of 2D conductive metal–organic frameworks