UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
Experimental and Pattern Analysis Pipeline for Dynamic C. elegans Aggregates
Get PDFActive matter involves the motion of self-propelling particles whose local interactions result in the emergence of collective behaviors. Caenorhabditis elegans is a species of roundworm known to exhibit easily observable collective behaviors in the form of aggregation, swarming, synchronization and formation of dynamic networks, which, along with a few other characteristics, makes it a promising model for the study of active matter. While these behaviors are well documented in C.elegans, the principles governing these collective behaviors are largely unknown. To address this, I developed an experimental protocol and analysis framework to parametrize the patterns of the aggregation of the worms over time. The worms are imaged using time-lapse microscopy, and the resulting images are then segmented using ImageJ and Ilastik, and analyzed with Matlab. The characteristics of the patterns analyzed include the area, perimeter, perimeter-area ratio, bundle widths, branch lengths, and Euler number. Results show that this is a promising foundation to build upon for the continued study of the collective behavior of C. elegansBiomedical Engineerin
The Role of Emotion Regulation in the Early Life Stages of the Confluence Model of Sexual Aggression
Get PDFPrevalence estimates based on national samples suggest that 10% of college men have perpetrated rape since the age of 14 (Koss et al., 2022). Developed for the purpose of identifying life experiences and attitudes associated with an increased likelihood of sexual assault perpetration, the confluence model created by Malamuth et al. (1991) is one of the most commonly used theoretical models of sexual aggression. This thesis aimed to clarify mixed findings regarding some of the existing model variables through the inclusion of the additional variable of emotion regulation. As hypothesized, a significant indirect effect of childhood abuse on delinquency through emotion regulation difficulties was found, and in turn, delinquency had significant direct effects on both impersonal sex and hostile masculinity. Childhood abuse had a significant direct effect on hostile masculinity as originally hypothesized, as well as un-hypothesized direct effects on level of impersonal sex and number of sexually aggressive acts. While hostile masculinity had a significant direct effect on the number of sexually aggressive acts as hypothesized, impersonal sex did not. These findings contribute to existing theoretical models of sexual aggression and suggest several avenues for future research.Psycholog
Developing and Validating a Bioenergetic Model for Guadalupe Bass
Get PDFThe Guadalupe bass (Micropterus treculii, hereafter ‘GB’) is an endemic species found in central Texas streams draining the Edwards Plateau. GB are classified as a vulnerable species due to hybridization, habitat degradation, flow alteration, and climate change. Conservation of GB in response to threats has been hampered by a shortage of physiological knowledge. I addressed physiological knowledge gaps using a bioenergetic modeling (BEM) framework that predicts daily growth based on energy gains and losses. My objectives were to measure temperature and mass-dependent relationships, parameterize a BEM, and validate the BEM by comparing simulated growth to observed growth measured in the field and the lab. Temperature-dependent responses were measured from 10 to 35°C and mass-dependent responses at body sizes ranging from 45 to 150 mm total length. Temperature dependent consumption and respiration indicated an energetic thermal optimum of 30.1°C and 29.5°C, respectively, which is similar to other temperate warmwater fishes. Mass-dependent relationships followed a negative power function typical of allometric metabolic scaling theory. Lab growth was obtained from growth experiments at high and low temperatures and high and low rations. Field growth was estimated using modal progression of GB length frequencies across nine surveys from 12 April 2025 to 3 August 2025 in the San Antonio River. The BEM accurately predicted growth in the field but inaccurately in the lab, especially at cold temperatures and low rations. My study advances the goals of ecophysiology and conservation physiology by comprehensively characterizing the thermal niche dimensions of GB and providing a BEM for applied conservation goals.Environmental Scienc
MIA: Masked Inpainting-Based Image Augmentation with Diffusion Models for Enhanced Dermatology Image Classification
Get PDFRecent progress in few-shot learning and small-data representations in diffusion models has created new opportunities for data augmentation in machine learning. In this work, we introduce a novel data augmentation approach based on diffusion models to improve the performance of machine learning tasks on dermatological image datasets. By generating augmented images using diffusion models, we tackle the challenge of data scarcity in the healthcare domain. The augmented data, which exhibit diverse and realistic characteristics, enhance existing training datasets, thereby improving the model's ability to generalize to unseen real-world data. We integrate our augmentation strategy into the state-of-the-art machine learning model training pipeline and evaluate its impact on the performance of convolutional neural networks (CNNs) and vision transformers (ViTs) using the HAM10000 skin lesion dataset. Our experimental results show that this approach significantly improves the generalization ability and overall performance of downstream models, highlighting the potential of diffusion models in medical image analysis.Information Systems and Cyber Securit
Who will burn out, and who will leave? Demographic predictors of burnout and intent to quit in world language teachers
Get PDFTeacher burnout and attrition are significant concerns in the United States and globally, particularly in high-needs areas such as world language (WL) teaching. Despite extensive international research on teacher burnout and attrition, few studies have specifically examined how demographic characteristics may influence burnout and intent to quit among WL teachers. To address this gap, this study employed a cross-sectional research design utilizing a factorial multivariate analysis of variance (MANOVA) to explore the relationships between various demographic factors and teacher burnout and intent to quit in high school WL teachers. Analysis revealed statistically significant main, interactive, and between-participants effects for a range of personal characteristics (i.e. gender, age, race, and ethnicity), teacher characteristics (i.e. number of WLs taught, primary language status, years of experience, highest level of education completed, type of certification program, number of professional organizations, and number of additional certifications), and school characteristics (i.e. urbanicity of school, type of school, and region). Findings suggest a need to take these factors into consideration when addressing teacher attrition and burnout through research and practice.Modern Languages and LiteraturesInterdisciplinary Learning and Teachin
Parent-adolescent communication about sex and adolescent sexting behaviors
Get PDFThe term “sexting” refers to sexually suggestive and/or explicit messages exchanged in one-to-one or one-to-many forms over Internet-enabled devices. According to the multisystem perspective, parent-adolescent communication influences sexual socialization globally and sexting locally. In the present study, a model of the relationships among the breadth of sex topics discussed between parents and adolescent children, comfort in parent-adolescent sexual communication, and adolescent sexting is evaluated. U.S. nationally representative data were gathered from 592 parents and their adolescent children. The results demonstrated that for both adolescents and parents, a wider array of sex topics discussed is associated with increased comfort in parent-adolescent sexual communication. Additionally, increased comfort in parent-adolescent sexual communication is associated with lower incidence of adolescent sexting. Equivalent alternative models that rearrange the variables of interest did not fit the data. These findings contribute to the conceptualization of parent-adolescent sexual communication by delineating between content (i.e., “what is talked about”) and process (i.e., “how it is talked about”) components. Theoretical contributions involve expanding the multisystem perspective of adolescent sexual behaviors to include technologically mediated risks and introducing communication comfort as a regulatory mechanism between message exchange and effects.Communicatio
Machine Learning in Access Control: A Taxonomy [Systematization of Knowledge Paper]
Get PDFDeveloping and managing access control systems is challenging due to the dynamic nature of users, resources, and environments. Recent advancements in machine learning (ML) offer promising solutions for automating the extraction of access control attributes, policy mining, verification, and decision-making. Despite these advancements, the application of ML in access control remains fragmented, resulting in an incomplete understanding of best practices. This work aims to systematize the use of ML in access control by identifying key components where ML can address various access control challenges. We propose a novel taxonomy of ML applications within this domain, highlighting current limitations such as the scarcity of public real-world datasets, the complexities of administering ML-based systems, and the opacity of ML model decisions. Additionally, we outline potential future research directions to guide both new and experienced researchers in effectively integrating ML into access control practices.Information Systems and Cyber Securit
Designing Expert-Facing Surveys That Inform Generative Models for Health Applications
Get PDFHealthcare artificial intelligence (AI) faces a critical trust gap: patients prefer human physicians despite AI achieving superior diagnostic performance. This thesis addresses the underlying problem: AI systems lack transparent alignment with expert clinical reasoning. I developed a bidirectional framework where expert surveys inform AI development and AI refines survey design.Aim 1 designed quantitative surveys capturing expert trauma clinician reasoning to inform generative model development. Thirty-two trauma specialists across diverse institutions evaluated 17 emergency cases, providing freeform clinical rationales, relevance ratings, and blinded rankings of AI-generated recommendations. These responses optimized prompt engineering for Leah, a trauma decision-support system. Expert rankings demonstrated Leah's superior performance, receiving over double the first-place rankings compared to commercial large language models (LLMs), particularly excelling at clinically critical transfer decisions. Additionally, expert similarity surveys identified OpenAI embeddings as best matching human clinical similarity judgments, outperforming biomedical-specific alternatives.Aim 2 established frameworks for AI-informed survey refinement. Three validated cognitive tests (Stroop, Psychomotor Vigilance, N-Back) were integrated with survey administration to identify predictive cognitive metrics. Analysis of COVID-19 behavioral surveys revealed significant misalignments between self-reported activity and objective fitness tracker data. Three LLMs independently evaluated a problematic survey question, proposing refined versions with improved biomarker alignment through specific timeframes and quantifiable metrics.This work demonstrates that expert-facing surveys systematically capture clinical knowledge to create aligned AI systems, while AI analysis identifies and refines survey design flaws, both establishing methodology for trustworthy healthcare AI development.Biomedical Engineerin
A Dataset for the Prediction of Spanish Language Fluency by Quantification of Linguistic Components with Artificial Intelligence
Get PDFThe native language of an individual is the language acquired naturally in early childhood, typically from their family and immediate community. Native language acquisition intrinsically involves several linguistic components that help individuals develop their language skills, such as morphology, pragmatics, syntax, and semantics. In this work the goal is to predict Spanish language fluency by quantification of these linguistic components using an artificial intelligence (AI) pipeline. The pipeline includes a novel Spanish language question-answer dataset, automatic question text generation, data augmentation, preprocessing using Natural Language Processing (NLP) techniques, and a Transformer model that integrates the components to quantify and provide a prediction of fluency. We found that our model is able to predict language fluency with high accuracy using the components: morphology, syntax and pragmatics with higher scores for syntax. The results of this study show the possibility of the use of AI to verify if an individual is fluent in a particular language.Electrical and Computer Engineerin
Harnessing C. elegans as a Biosensor: Integrating Microfluidics, Image Analysis, and Machine Learning for Environmental Sensing
Get PDFEnvironmental contamination is becoming an increasingly evident risk to human health worldwide. The small, free-living nematode <i>Caenorhabditis elegans</i> (<i>C. elegans</i>) has become a compelling model organism for environmental toxicity studies in recent years, owing to its numerous advantages, including its transparent body, small size, well-characterized biology, genetic tractability, short lifespan, and ease of culture. Several assays have been developed using <i>C. elegans</i> to enable a better understanding of toxicant effects, from whole-animal to single-cell levels. While these methods can be extremely useful, they can be time-consuming and cumbersome to perform on a large scale. Recent advances in microfluidics have adapted many of these assays to enable high-throughput analysis of <i>C. elegans</i>, greatly reducing time and resource consumption while increasing efficiency and scalability. Further integration of these microfluidic platforms with machine learning expands their analytical capabilities and accuracy, revolutionizing what can be achieved with this model organism. This article will review the physiological basis of <i>C. elegans</i> as a model organism for environmental toxicity studies, and recent advances in integrating microfluidics and machine learning which could lead to using <i>C. elegans</i> as a promising living biosensor for environmental sensing