UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
Machine Learning Based Decision Support Systems for Timely Transfusion in Trauma
No full textTrauma is a leading cause of mortality worldwide, with hemorrhage representing a substantial proportion of preventable deaths. In emergency settings, waiting for laboratory results can delay critical blood product administration. Addressing this challenge, this work leverages machine learning and deep learning techniques to predict whether a patient will require transfusion using data obtained at initial hospital presentation.
The Trauma Quality Improvement Program (TQIP) serves as the primary data source, offering a robust dataset for training and validating the predictive models. The algorithms under investigation include Logistic Regression, Lasso Regression, Ridge Regression, Support Vector Machines, K-Nearest Neighbors, Random Forest, XGBoost, Multilayer Perceptrons, and Tab Transformers. Their performance will be evaluated using the Area Under the Receiver Operating Characteristic (AUROC) curve, providing a clear measure of classification accuracy. Shapley values will further elucidate each model’s predictions by highlighting the most influential features, thereby offering greater transparency for clinical decision-making.
Although results are not yet available, this approach aims to enhance early identification of transfusion needs, potentially reducing preventable deaths and improving outcomes for high-risk trauma patients.Artificial Intelligenc
Quantitative Detection of General Fecal Indicator Bacteria and Host-specific Source Tracking Markers to Determine Fecal Contamination along Galveston Island, Texas
No full textAccording to the “Safe for Swimming” report issued by the Environment America Research & Policy Center and Frontier Group, 55 out of 61 beaches along the Texas Gulf Coast had potentially unsafe levels of fecal indicator bacteria, at least for one day in 2020. While the coastal waterways experience fecal contamination from sewage leakage from malfunctioned septic systems, domestic and wildlife feces, it gets exacerbated by rising sea levels and increased rainfall. This condition poses a serious risk to humans since swimming in contaminated waters can cause gastrointestinal illness, respiratory illness, and other health problems. The preliminary objective of our investigation is to determine the distribution and abundance of fecal indicator bacteria (Enterococci) and specific host-associated fecal genetic markers (human-, dog-, avian- associated Bacteroidales) using qPCR assays in the Texas coastal area, focused on Galveston Island. Three sampling transects have been chosen based on distinguished land-use characteristics and well water, surface water, and pore water were collected from each transect from November 2021 to March 2021. Comparatively, both FIB and MST markers were detected more frequently in high concentration in surface water than well and pore water. Statistical analyses revealed significant differences in marker concentrations among sites (Kruskal-Wallis Test, p < 0.05). Detection frequency of culturable enterococci in surface water samples, often exceeding safety thresholds, indicating widespread fecal contamination. Entero1 was the most prevalent marker, showing significant differences among sites. HF183 was less frequently detected (4%), suggesting limited human fecal pollution. BacCan was more common in three well samples collected from the Heritage preserve and Jamaica Becah RV park suggesting fecal pollution through infiltration from wild species (cayote) and domestic dogs with vacationers. GFD was consistently present in all type of samples, reflecting the significant impact of avian sources. Multivariate Statistical Analyses were conducted on the to explain the clustering patterns and the variation of the sites based on FIB and MST marker concentrations. Cluster Analysis classified well, surface and pore sites separately with slight variation while PCoA explained about 70% variation in the dataset with close clusters for different types of sitesCivil Engineerin
Design and Implementation of a Medium Voltage Converter for an Ultra-Fast Electric Vehicle Charging System
Full text linkThe objective of this work is to design and implement a medium voltage converter for an ultra-fast charger for Medium- and Heavy-Duty Vehicles (MHDV). A two-stage conversion system is proposed: an AC-DC three-level converter and an isolated DC-DC converter with a high-frequency transformer. The charger’s AC input is a three-phase 4.16 kV, eliminating the necessity of low voltage service transformers and improving the overall efficiency.
This thesis presents detailed modeling, designing critical components for each conversion stage (LCL filter, DC-Link Capacitor, High-Frequency Transformer) and of the control strategies implementation. The proposed model is connected to an IEEE-13 bus and simulated in MATLAB/Simulink to verify the performance. There are three case scenarios considered: Case 1 is the system performance under nominal conditions, Case 2 is when the load is increased by adding multiple charging stations, and finally Case 3 is when the distribution system is overloaded.
Results show that proposed converter has an efficiency of 95 % and power factor of 0.98 in nominal conditions. However, the excessive increase of charging station’s load created a lot of harmonic distortions, voltage sags and a significant degradation in the power factor. It is concluded that, even though the coupling to a medium voltage grid is feasible and has multiple benefits, implementation of mitigation strategies such as better control and intelligent load management are required to avoid severe impacts to distribution grid.Electrical and Computer Engineerin
Development of Derivative-enhanced Surrogate Modelling Capabilities for Optimization and Design of Dynamical Structural Systems
No full textThe full text of this item is not available at this time because the author has placed this item under an embargo until August 26, 2026.Ensuring the safety and reliability of structures is paramount in modern societies that heavily rely on infrastructure. This dissertation presents the development of novel numerical differentiation techniques and computational tools to support the design process of safer and more reliable engineering systems. Central to this work is Hypercomplex Automatic Differentiation (HYPAD), a numerical method for computing arbitrary-order derivatives of analytic functions. HYPAD is integrated with finite element formulations and dynamic solvers to evaluate sensitivities in structural responses, enabling derivative-enhanced surrogate modeling and optimization.
The dissertation introduces multiple original contributions across sensitivity analysis, high-fidelity simulation, surrogate modelling, and structural optimization, with applications in vibration attenuation and structural health monitoring (SHM) systems design. These developments provide a foundation for accurate sensitivity computation in dynamic systems and support the training of efficient, derivative-enhanced Gaussian Process surrogate models. The outcomes of this dissertation include:
• A novel method for computing arbitrary-order eigenvalue and eigenvector sensitivities in damped and undamped vibration problems using HYPAD and residual-based formulations.
• The first implementation of HYPAD to coupled piezoelectric guided wave simulations using high-order spectral finite elements, including the development of cohesive interfaces, piezoelectric elements, low-reflective boundaries, and Mindlin–Reissner shell elements, all extended with HYPAD differentiation.
• Integration of these tools into derivative-enhanced Gaussian Process surrogate models for structural optimization, with applications in designing vibration attenuation materials and Model-Assisted Probability of detection (MAPOD) assessment in SHM systems.
• A sensitivity-informed Bayesian calibration framework for MAPOD estimation in SHM, reducing experimental requirements while preserving accuracy.
This dissertation follows a compendium format comprising five publications along with supporting work. It contributes a unified methodology for derivative-enhanced surrogate modeling that combines advanced numerical differentiation with surrogate modeling techniques to improve the safety and reliability of structural systems.Mechanical Engineerin
Exploring Emotional Self-Efficacy as a Mediator of Positive Leisure Experience and Subjective Well-Being Among Elementary School-Age Children in a Marginalized Community
Full text link<b>Background</b>: Prior research has established a positive relationship between emotional self-efficacy and life satisfaction in elementary school-age children. However, less is known about the direct impact of positive leisure experience on subjective well-being and the potential mediating role of emotional self-efficacy. <b>Objectives</b>: This study examined whether emotional self-efficacy mediates the association between overall leisure enjoyment and life satisfaction among elementary schoolchildren. It was hypothesized that both direct and indirect effects are statistically significant. <b>Methods</b>: A quantitative, cross-sectional design was used with 100 fifth- and sixth-grade students from a U.S.&ndash;Mexico border community. Participants completed the Children&rsquo;s Assessment of Participation and Enjoyment (CAPE), the emotional subscale of the Self-Efficacy Questionnaire for Children (SEQ-C), and the Student Life Satisfaction Scale (SLSS). Mediation analysis was conducted in R with bootstrapping (500 simulations). <b>Results</b>: Overall leisure enjoyment was positively associated with life satisfaction (&beta; = 0.54, 95% CI [0.23, 0.90], <i>p</i> = 0.004). The direct effect remained significant after accounting for emotional self-efficacy (&beta; = 0.41, 95% CI [0.15, 0.73], <i>p</i> = 0.004). The indirect effect through emotional self-efficacy was also significant (&beta; = 0.13, 95% CI [0.03, 0.29], <i>p</i> = 0.016), accounting for approximately 25% of the total effect. <b>Conclusions</b>: Emotional self-efficacy partially mediated the relationship between overall leisure enjoyment and life satisfaction, suggesting that positive leisure experience enhances children&rsquo;s emotional coping confidence and subjective well-being. These findings underscore the importance of promoting accessible and enjoyable leisure opportunities within marginalized communities that simultaneously foster children&rsquo;s emotional self-efficacy and well-being
Toward a Multimodal Fusion Framework For Polar Sea Ice Classification using Image and Point Cloud Satellite Datasets
Full text linkRapid climate-induced changes in polar regions have significant global implications, requiring accurate monitoring and classification of sea ice. This work proposes a multimodal deep learning framework for classifying polar sea ice thickness by fusing Sentinel-2 multispectral satellite imagery with ICESat-2 altimetry point cloud data. The objective is to distinguish between thick ice, thin ice, and open water, which are crucial categories for understanding sea ice dynamics and predicting future changes. Our methodology employs deep learning to extract meaningful features from optical imagery (Sentinel-2) and integrates them with complementary laser altimetry measurements (ICESat-2), enhancing classification accuracy. We designed three fusion strategies and conducted experiments on a dataset consisting of 30,728 Sentinel-2 image patches and corresponding ICESat-2 ATL03 measurements. Results showed that Gated Fusion (GF) consistently performs better than other strategies, achieving 97.70% overall accuracy. The advantages of fusion remain most evident under missing-modality conditions, which are common in polar regions due to cloud interference, emphasizing the resilience and adaptability of the Gated Fusion framework.Computer Scienc
Twain Goes Native and Huck Plays Indian in the Manifest Destiny of Vanishing Indians
Full text linkNot only the surface story of Twain’s humor and narrator Huck’s picaresque adventures, but also more veiled, complex, subsurface presumptions of Manifest Destiny and vanishing or vanished Indians generated the ultramegacanonization of Twain “going native” and Huck “playing Indian” in Twain’s two most famous novels, The Adventures of Tom Sawyer (1876) and the sequel Adventures of Huckleberry Finn (1885). Twain not only absorbed and portrayed the EuroAmerican cultural zeitgeist of vanishing or vanished Indians, but Twain also subtly perpetuated and reinforced vanishing Indian mythologies and self-righteous Manifest Destiny that resonated with his mostly white readers.Englis
A Course-based Undergraduate Research Experience (CURE) Focused Broadly on Research Methods in Computer Science
Full text linkThis experience report introduces a novel, adaptable course-based undergraduate research (CURE) model designed to nurture research interest in undergraduate CS majors. Traditional CURE models often target specific disciplines. Our model allows students to explore research questions within their chosen CS subfields. A pilot study assessed the model's effectiveness. Undergraduate CS majors participated, completing pre- and post-course surveys to gauge research motivation. The course culminated in a final research project requiring students to delve into a specific CS area, propose a novel approach, and outline a research plan. Findings from this pilot will inform the model's refinement for wider implementation. Further, the developed materials are shared openly at https://github.com/amandanko/UTSA-CS-CURE.Computer Scienc
Trade Regulation of the Cites Convention: An Examination of Compliance in Brazil
Full text linkWildlife trafficking is one of the leading causes of biodiversity loss across the globe. The Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) is a Multilateral Environmental Agreements (MEA) that was formed to stop the international trade of wildlife. This research will us a qualitative document analysis to understand the national compliance of Brazil based on the structure of the CITES convention. The CITES convention text is written as a trade regulation structure rooted in deterrence theory to prevent and stop the international trade of listed species. The deterrence behind the trade regulation structure of CITES is undermined by state capacity and poaching motivation. CITES has set up Brazil to fail in compliance by not considering these factors. A scoring measurement shows how Brazil is within textual legislative compliance with the convention but does not live up to enforcement compliance based on the factors of state capacity and poaching motivation. The broad approach of the trade regulation structure in CITES overlooks factors undermining deterrence and the research will discuss the alternative approach Normative Compliance.Political Science and Geograph
Uncertainty Informed Safety Frameworks for Disturbance Analysis of Structures in Extreme Environments
Full text linkThis work presents the development of two computational frameworks aimed at enhancing the safety and reliability of systems operating under harsh and uncertain environmental conditions, such as micrometeorite impacts and harsh solar radiation. Reduced-order modeling of lunar habitats is employed to enable efficient analysis under the concept of good-enough approximations-models that are sufficiently accurate to support timely and defensible decision-making. The frameworks are demonstrated through smart habitat applications on the lunar surface. The first framework models the degradation of a structural protective layer subjected to periodic micrometeorite impacts, enabling the determination of optimal protective thickness and scheduling of inspections and repairs to mitigate radiation exposure and penetration risks. Building on the first framework, the second framework is a control-oriented computational tool that implements a disturbance-fault-response model, where stochastic environmental disturbances lead to subsystem faults and trigger agent-level responses. Reduced-order models, informed by probabilistic representations of environmental conditions, support both predictive maintenance and controller optimization for subsystems responding to disturbances. For the long term framework, lunar regolith was shown to be reliable up to 3.25 years, after which repair must be initiated to protect the inhabitants from solar radiation. When analyzing the short term framework, settling time comparisons demonstrated that softening the designated controller response decreased state of charge consumption by as much as 43% during a 48 hour pressure leak simulation. Together, these frameworks demonstrate how uncertainty-informed modeling and control strategies can guide safety-critical decisions and system-of-systems reliability in extreme environments.Mechanical Engineerin