48440 research outputs found
Sort by
Student Mental Health: Screening for Stress, Anxiety and Depression using Fitbit Data
No full textCollege students experience many stressors, resulting in high levels of anxiety and depression. Wearable technology provides unobtrusive sensor data that can be used for the early detection of mental illness. However, current research is limited concerning the variety of psychological instruments administered, physiological modalities, and time series parameters. In this research, we collect the Student Mental and Environmental Health (StudentMEH) Fitbit dataset from students at our institution during the pandemic. We provide a comprehensive assessment of the ability of predictive machine learning models to screen for depression, anxiety, and stress using different Fitbit modalities. Our findings indicate potential in physiological modalities such as heart rate and sleep to screen for mental illness with the F1 scores as high as 0.79 for anxiety, the former modality reaching 0.77 for stress screening, and the latter modality achieving 0.78 for depression. This research highlights the value and potential of wearable devices to support continuous mental health monitoring, the importance of identifying optimal data aggregation levels, and selecting modalities for mental illness and stress screening
Microscale gradient-driven behaviors in Staphylococcus epidermidis biofilms and colloidal dispersions
No full textMicroscale chemical gradients can modulate the properties, structures, and behaviors of colloidal dispersions and colloidal composites. Understanding the relationship between microscale gradients and biocolloidal and colloidal systems is important in applications ranging from the development of novel antimicrobials for bacterial biofilms to particulate removal and separation. My dissertation research investigates how subtle variations of chemical species at the microscale create gradients (pH, oxygen, solute) that significantly impact how bacteria in biofilms or colloidal particles interact with their environments. First, I simultaneously measure Staphylococcus epidermidis biofilm creep compliance, structure, and pH gradients during biofilm development, across bacterial strains, and under antibiotic treatment and construct multiple linear regression models to identify how pH gradients and structural features relate to biofilm mechanics. I find pH is a significant predictor of biofilm mechanics throughout S. epidermidis biofilm development and across bacterial strains. Second, I develop a solution-based oxygen sensor that enables in situ measurement of three-dimensional oxygen gradients in undisrupted biofilms. I utilize the solution-based sensor to reveal changes in microscale oxygen gradients throughout S. epidermidis biofilm development and across vancomycin treatments ranging from the minimum inhibitory concentration (MIC) to clinical dosages. I show depletion of oxygen availability at the microscale during S. epidermidis biofilm development and increases in microscale oxygen availability within biofilms treated with high doses of vancomycin. Finally, I combine experiments, dimensionless analysis, and theoretical modeling to describe how solute gradients can be harnessed to drive colloidal particle separation in a diffusiophoresis-based system, providing design rules for tuning separation performance based on system geometry, operating conditions, and particle properties. I find separation performance is most substantially affected by radial gap size which controls the solute gradient strength in a tube-in-tube-in-tube diffusiophoresis-based separation system. Characterization and analysis of chemical gradients in biocolloidal and colloidal systems through this work integrates structural and chemical measurements to predict biofilm mechanical behavior, introduces a new method to non-invasively map 3D oxygen gradients in intact biofilms, and provides a scalable and tunable framework for controlling gradient-driven particle separation. Together, these contributions advance experimental and theoretical approaches for studying microscale gradients and their impact on biocolloidal and colloidal systems. This work will enable future investigations of biofilm mechanics for guiding biofilm removal strategies, metabolic mapping of bacterial biofilms for understanding the response of biofilms to antimicrobials, and gradient-driven transport for designing low-energy, scalable separation technologies
Engineering Phase Behavior for Municipal Waste Circularity to Enhance Carbon Utilization
No full textOver the past two centuries, humanity’s industrial advancements have led to the greatest recorded increase in atmospheric carbon dioxide spanning back 800,000 years. This change in our atmospheric concentration has been correlated to the increase in global temperature, rapidly approaching a deviation of +2 °C, which is responsible for the harsher and dramatic weather patterns and loss of biodiversity around the world. The traditional response to this crisis has been to sequester carbon directly from the atmosphere or to scrub industrial flue gases to mitigate new carbon from entering the environment. However, at diluted concentrations current solvents and sequestration materials are less effective at capturing CO2 and are sensitive to degradation in the presence of oxygen and moisture, adding extraneous engineering costs. Organic wastes are a preconcentrated source of carbon and energy and currently make up 84% of the municipal solids waste produced in the United States. Landfills currently manage half of all waste produced in the states due to the low energy barrier for disposal, but this method wastes both the carbon and energy potential within waste. Biomass conversion technologies such as digestion, pyrolysis, and gasification exist, but the high moisture content and contamination of mixed wastes reduce the viability of these methods. Instead, hydrothermal methods operate at moderate temperatures and high pressures to convert wet organic waste into a mixture of carbon dense products. This work focuses on expanding the capabilities of hydrothermal systems to unlock wet organic waste as the future of carbon removal technologies. We go on to (1) build, design, and operate supercritical water systems to fractionate inorganic contaminants from organic carbon, allowing for effective catalytic upgrading to useable fuels, (2) leverage high-resolution molecular insight to optimize liquid-liquid extraction of traditionally wasted chemical byproducts, and (3) perform mass and energy balances for a new class of hydrothermal technologies, driving organic carbon through thermodynamically favorable oxidation reactions to inorganic carbonates within functional construction material alternatives
Molecular Mechanisms of Gating and Selectivity in the Channelrhodopsin-1,2 Chimera
No full textChannelrhodopsins (ChRs) are light-activated semi-selective cation channels formed by seven transmembrane helices and a covalently bound all-trans retinal at its center. Native to the green alga Chlamydomonas reinhardtii, channelrhodopsin-1 (ChR1) and channelrhodopsin-2 (ChR2) serve as photosensors and mediate phototaxis by depolarizing the plasma membrane. Neuroscientific research has enjoyed widespread application of ChR2 as a research tool in the field of optogenetics to control excitable cells with light. The channelrhodopsin-1,2 chimera (C1C2) is a synthetic construct composed of helices 1-5 from ChR1 and helices 6-7 from ChR2. C1C2 has served as a structural model for ChR2 and other ChRs since its crystal structure was solved in 2012. The ability to create custom-designed ChRs with fine-tuned properties to meet the needs of cell-specific and specialized applications is a long-standing goal of optogenetic research. Although much progress has been made in recent years, the molecular mechanisms governing channel properties such as ion conductance, selectivity, and gating kinetics in C1C2 have not been fully resolved. This knowledge gap has significantly hindered the rational design of new variants with altered properties. Gaining a deeper understanding of these mechanisms at the molecular level will enable the generation of new optogenetic research tools to address a broader range of pressing research questions in neuroscience and related fields. Here I use a combination of electrophysiological and computational methods to identify specific structural elements that govern ion selectivity and photocurrent kinetics and quantify their impact on channel function. This work helps to inform molecular engineering of new ChR optogenetic tools and makes significant contributions to basic science and the growing body of knowledge regarding the structure-function relationships of ion channels
Detection, Validation and Analysis of Deepfaking in Social Media
No full textAs AI continues to gain momentum and capture media attention, the prevalence of AI-generated voice, video, and image deepfakes on social media is rising dramatically. On some platforms, the AI attribution rate is estimated to be as high as 38.95\%, and continuing to increase month over month. Despite significant efforts in the literature to detect and mitigate these synthetic media, a substantial amount of deepfaked content still circulates undetected. This thesis investigates methods for identifying, classifying, validating, and analyzing deepfakes in the digital landscape. Furthermore, this project investigates the potentials of leveraging contextual insights through the analysis of user interactions with AI-generated content, and proposes a new technique to enhance the detection and mitigation of deepfakes across social media platforms. By outsourcing part of the overall detection computation to a distributed base of users, interaction-based detection could prove to be a successful tool in reducing the harmful effects of AI-generated deepfakes online
Procedural Generation in Houdini
No full textSide FX Houdini has advanced procedural generation technology in the field of film and television special effects and game development. It can generate effects that other software cannot replace: explosions, oceans, foam, shattering, cracks, etc. The VDB node in the software is an important part of achieving these effects. I use VDB nodes in procedural generation technology to convert geometry into rough pixel blocks and re-simulate them in detail to create realistic stone shattering texture effects. Convert the mesh to VDB SDF, and then use VDB to merge and subtract the broken shapes to simulate cracks and damage
Enhancing Language Models for Classification on Text-Attributed Graphs through Multi-Model Data Augmentation
No full textGraph representation learning has become a critical task across various domains such as social networks and recommender systems. Recently, the rise of large language models (LLMs) has opened up new possibilities for processing text-attributed graphs (TAGs), where nodes are associated with textual information. Despite promising progress, applying LLMs to TAGs faces significant challenges, including input window size limitations and the computational overhead of handling large-scale graphs with millions of nodes. To address these challenges, we propose a novel approach that leverages a multi-model profiling approach for data augmentation, thereby increasing the diversity and quantity of the training samples. The data generated by each model is then combined with the graph structure, prompts and ground-truth labels to create a comprehensive and varied fine-tuning dataset. By strategically selecting profiling models, an appropriate number of neighboring nodes and constructing concise yet informative fine-tuning prompts, our proposed method enables LLMs to process more complex graphs while operating within limited computational resources. Notably, our experiments demonstrate that it is unnecessary to construct intricate graph structures for fine-tuning to achieve strong performance. Our approach outperforms nine state-of-the-art baselines, showcasing its effectiveness. Furthermore, we have made our model publicly available on Hugging Face for reference and use
Characterization and Compensation of Temperature Effects of a Luminescence-Based Sensor Film in a Transcutaneous Oxygen Monitor
No full textTranscutaneous blood gas monitoring provides a noninvasive method for assessing oxygen levels in clinical settings. Traditional transcutaneous oxygen measurement techniques typically rely on heating the skin to enhance capillary blood flow and oxygen diffusion. However, this heating process can cause skin irritation and increase power consumption. To address these issues, a novel luminescence-based transcutaneous blood oxygen monitor was developed. This nonthermal device is designed to track blood oxygen levels noninvasively, eliminating the need for skin heating. Despite this advantage, the device remains susceptible to variations in both body temperature and ambient conditions. To evaluate the performance of this device, its precision was first assessed across a broad range of oxygen partial pressures. The sensor demonstrated high precision in the physiologically relevant low oxygen range (0–100 mmHg), with measurement uncertainties ranging from ±0.3 mmHg to ±5 mmHg. At higher oxygen concentrations, however, the precision decreased, with uncertainties reaching up to ±2.4% at 22% oxygen. These results underscore the device's suitability for clinical applications focused on measuring human blood oxygen levels. To investigate and mitigate the device's temperature sensitivity, both theoretical modeling and experimental analysis were performed. Controlled experiments were conducted at three discrete temperatures—22°C, 30°C, and 40°C—and revealed a consistent inverse relationship between temperature and phosphorescence lifetime ( τ), with an average temperature sensitivity of –0.14 µs/°C. In a dynamic temperature sweep from 22°C to 42°C, a maximum deviation of 17% from baseline lifetime values (at 22°C) was observed. To compensate for this temperature dependency, multiple correction strategies were implemented. These included a 2°C step-size lookup table, a high-resolution 0.1°C lookup table, and a universal scaling factor model. The final compensation algorithm, embedded within the device’s microcontroller, successfully reduced the lifetime error to below 1% and achieved a corrected standard deviation of less than 0.2 µs
Is This Real? The Integration of Blockchain Technology for Digital Content Authenticity
No full textIn today’s digital age, manipulated content such as deepfakes, miscaptioned media, and falsified material spreads rapidly online, eroding public trust in digital media. In collaboration with Dr. Emma Rawlings, we investigated whether blockchain technology could help restore that trust by enabling digital provenance tracking. We assessed the desirability, viability, and feasibility of this approach through interviews, surveys, and focus groups, focusing on journalism as a powerful use case. These insights guided the development of our blockchain-based image certification prototype. We conclude that blockchain can help verify media authenticity and recommend further exploration of its integration in contexts where digital asset integrity is critical
Meissa Microgrid | Multi-Tenant Renewable Energy Monitoring Platform with Solar PV Tracker Controls
No full textThis project aims to expand EV charging infrastructure in Indonesia with a technologically driven solution in collaboration with 360energy, an energy-as-a-service startup. Our solution features an automated reduced-scale dual-axis solar tracking system that captures sunlight as well as the controls and design for the full-scale system. Static analyses and mathematical models of the system were enlisted to select various components and enable controls tracking. Sensors were integrated into the system to detect solar and weather conditions. Tests were conducted for the controls of the full-scale system to ensure functionality. The system is connected to a centralized web application dashboard to monitor and control the solar tracking system and simulated battery energy storage system (BESS) for 360energy employees. The dashboard provides real-time data visualizations, including energy intake collected from the solar panel control monitor. This allows users to track system performance effectively. Additionally, the dashboard allows users to view different project locations on an interactive map showcasing all sites located in Indonesia. The dashboard also displays a battery energy storage system interface, where users can view certain information like battery health, charging schedules, locking status, and availability. A full day test with the reduced-scale model was conducted, displaying energy intake on the dashboard