University of Illinois at Chicago
University of Illinois at Chicago: UIC INDIGO (INtellectual property in DIGital form available online in an Open environment)Not a member yet
21439 research outputs found
Sort by
Transform Domain Deep Neural Network Layers and Their Applications
No full textThis dissertation presents a unified research framework that integrates orthogonal transform theory with deep neural network architectures to achieve efficient data compression, representation learning, and image correction. The motivation stems from the increasing demand for accurate and resource-efficient data processing in biomedical and industrial systems, where large-scale sensor data must be transmitted and reconstructed under strict computational and bandwidth constraints. Traditional deep learning methods provide strong modeling capability but often require excessive parameters and energy consumption, while classical transformbased approaches offer interpretability and sparsity yet lack adaptability to nonlinear and highdimensional data. To bridge this gap, the dissertation introduces a family of asymmetrical neural networks in which orthogonal transforms are embedded directly into trainable layers. These models exploit transform-domain sparsity and energy compaction while leveraging the benefits of deep network training, which adapts to the data
Computational Fluid Dynamics-Informed Analysis of Diaphragm-Free Shock Tubes and Insert Designs
No full textThe design and simulation of a novel, compact, diaphragm-free shock tube is presented using computational fluid dynamics (CFD) in ANSYS Fluent 2022 R1 to better understand shock tube behavior for improved control in high-pressure, high-temperature shock experiments. The study quantifies geometric and viscous effects by evolving from an idealized diaphragm-controlled inviscid straight shock tube to a viscous wrap-around configuration featuring a retracting plate valve in place of a diaphragm. Different valve opening speeds are assessed at matched P4/P1 pressure ratios and matched shock velocities, isolating the effects of valve motion on shock formation and test section conditions. A driver insert is designed and evaluated to control the rate of rise behind the reflected shock. The study assesses shock structure, rate of rise, and temporal homogeneity using computationally generated x-t diagrams, numerical Schlieren imaging, pressure contours, and test time pressure and temperature data. Valve-based and diaphragm-based configurations are compared to analyze the impact of valve opening speed, pressure ratio, and insert geometry on shock uniformity and test time characteristics.
The geometric evolution and viscous effects leading to the compact, wrap-around, diaphragm-free shock tube are examined, beginning with an inviscid straight shock tube. Viscosity and boundary layer development are assessed in a straight shock tube, followed by the addition of a wrap-around driver section and a valve housing (plenum). Upstream influences on shock structure, wave uniformity, and flow evolution in the test section are examined. The valve is then incorporated by simulating a retracting plate valve at an experimentally predicted opening speed of 10.16 m/s. The impact of the valve on shock uniformity, turbulence generation, compression wave attenuation, and the overall test time performance is analyzed. Additional simulations at valve opening speeds of 5.08 m/s and 20.32 m/s examine valve speed effects under matched P4/P1 pressure ratios and matched shock velocities. The trade-offs between valve speed, rate of rise, homogeneity, and test duration are identified. It is found that increasing valve speed reduces the rate of rise behind the reflected shock but introduces greater temporal inhomogeneity, revealing a critical balance between rapid shock formation and uniform test conditions. Driver insert effects are explored through parametric studies, varying the insert width (diameter), length, and head angle/shape to assess their impact on shock attenuation and rate of rise at a constant P4/P1 pressure ratio. An optimal driver insert configuration is designed based on its ability to sustain uniform test conditions across various P4/P1 pressure ratios.
This work advances the understanding of unsteady compressible flow physics by establishing a simulation framework for moving-valve systems using Chimera mesh techniques. It examines the role of valve speed on shock formation and test section performance. The impact of a driver insert on the flow uniformity and test time conditions is assessed, and an initial comparison between 2D CFD predictions and experimental results is presented
Barriers to HIV Care and Randomized Controlled Trial Retention among Black Men Who Have Sex with Men
No full textThe HIV epidemic remains a significant global health challenge, including in the United States, where disparities in HIV care continuum outcomes persist between priority populations for Ending the HIV Epidemic (EHE) initiative. Among these priority populations, young Black men who have sex with men (YBMSM) have increased HIV acquisition and are more likely to experience challenges in being retained in HIV care. Despite advances in HIV prevention and treatment over the past five decades, YBMSM face intersecting social, economic, and structural barriers, including stigma, limited access to culturally competent care, and inadequate support services. These factors contribute to delayed HIV diagnosis, missed HIV care appointments, and poor adherence to antiretroviral therapy (ART). Young Black men are also often underrepresented in behavioral research and have higher rates of attrition, leading to gaps in tailored interventions that this population may benefit from to improve HIV care continuum outcomes. Addressing these disparities requires dismantling systemic inequalities to improve access to HIV prevention and treatment services, including tailored behavioral interventions.
Mobile health (mHealth) interventions have emerged as promising tools to address disparities in HIV care continuum outcomes, including care retention, particularly among populations that experience multiple barriers to care, such as YBMSM. mHealth interventions provide innovative solutions to overcome logistical, social, and economic barriers that hinder traditional in-person behavioral interventions and healthcare engagement. For YBMSM, mHealth tools can offer discreet, convenient, and culturally competent support, which may facilitate improved HIV care continuum outcomes, such as ART adherence and HIV care retention. Previous research has shown that mHealth interventions can improve HIV self-management by offering personalized reminders, educational resources, and virtual peer support.
Additionally, mHealth platforms are private, easily accessible, and may be relational, which can improve engagement with hard-to-reach populations. By expanding accessibility, mHealth interventions have the potential to improve health outcomes for YBMSM and reduce disparities that continue to contribute to the ongoing HIV epidemic.
In the United States, we developed an mHealth app to improve HIV care continuum outcomes, including HIV care retention, among YBMSM. We conducted a randomized controlled trial (RCT) to explore the 6-Month intervention effect of the My Personal Health Guide mHealth app on HIV care continuum outcomes among YBMSM. We found that the app demonstrated an elevated but insignificant effect on self-reported HIV care retention after 6-months of follow-up. We believe My Personal Health Guide warrants further development to improve HIV care retention. In this sample, we observed a substantial proportion of research participants that were loss-to-follow-up (LTFU), so investigated factors associated with LTFU in an RCT among YBMSM. Loss-to-follow-up was associated with lower social support and YBMSM randomized to the attention control arm of the study, suggesting future behavioral interventions should consider whether the attention control is as engaging as the intervention. Additionally, determining the prevalence of lower self-perceived social support during RCT approach development might be considered to obtain accurate sample size estimates. Lastly, as the RCT began enrollment concurrent with the COVID-19 pandemic lockdowns, we conducted a qualitative analysis of barriers and facilitators of the HIV care experience among a convenience sample of YBMSM from the RCT. The most commonly reported barriers to HIV care retention were healthcare access factors, including administrative or technological issues, clinic hours, clinic location, and experiencing stigma in the clinic. Notably, positive interactions with provider and staff was commonly cited as a facilitator of HIV care retention, so future behavioral interventions might consider incorporating patient satisfaction with provider and staff interactions to improve care retention.
Effective and tailored interventions that retain YBMSM and improve their retention in HIV care is particularly important due to the higher burden of HIV diagnoses within this group and the disparities they face in healthcare access, social determinants of health, and HIV-related stigma. While there have been improvements over time in national metrics for HIV care continuum outcomes, disparities still exist between priority populations, with YBMSM suffering a disproportionate burden of diagnoses, non-optimal adherence, and poor retention in care. Ultimately, these findings point to a multifaceted approach to improve retention in HIV care and RCTs among YBMSM. These data provide support for continued development of mHealth delivered HIV behavioral interventions, with enhanced efforts to recruit and retain a larger sample of YBMSM with engaging control arms, and leveraging positive interactions with providers to improve care retention in this vulnerable population
Training and Inference in Early-Exit Deep Q-Networks for Efficient Reinforcement Learning
No full textNeural networks have become instrumental in various machine learning tasks, but their increasing complexity poses challenges in terms of computational resources and real-time decision-making. To address these challenges, this thesis explores the integration of early-exit neural networks (EENNs) with reinforcement learning (RL), a novel approach that has received limited attention in the literature. Early-exit strategies introduce intermediate exit points in neural networks, allowing for dynamic decision-making based on the model’s level of confidence. Reinforcement learning, on the other hand, enables autonomous systems to make intelligent decisions through sequential decision-making. We hypothesize that combining early exits with reinforcement learning can enhance learning efficiency and generalization. In this study, we design and implement Early Exit Deep Q-Network (EEDQN), a framework that integrates EENNs with RL algorithms. Through a series of experiments using benchmark datasets and complex environments, we assess the approach on two OpenAI Gym tasks with distinct input regimes: CartPole (vector state) and Atari Pong (pixels). Across both domains, EEDQN yields results that preserve baseline return while lowering expected per-decision FLOPs, with larger savings when many states are “easy” and confidently handled by the early exit. These results demonstrate that conditional computation via early exits can make RL policies adaptive to state difficulty without sacrificing control, minimizing overthinking, delays, and power utilization
Electrohydrodynamic Manipulation of Sessile Droplets by the In-Plane Electric Fields
No full textThis doctoral thesis is devoted to the exploration and understanding of the physical principles of electrohydrodynamics to develop and facilitate innovative processes of painting and coating of large dielectric surface areas with less material and a higher rate; aiming at optimizing these processes and reducing final production costs. From the beginning, this comprehensive study explores the effects of the in-plane electric fields (with DC voltages in the kV-range applied) on sessile droplets of multiple pure liquids, solvents, colloidal suspensions (with particles and pigments) and conducting polymers commonly used in painting and coating processes. This fundamental research aims at understanding the principles and mechanisms of interaction of the in-plane electric fields and sessile droplets through in-depth experimental, theoretical and numerical studies. Furthermore, this electric field-assisted approach aiming at an increase in the painting and coating efficiency, leads to multiple experimental results and provides a deeper understanding of sessile droplet manipulation by the in-plane electric fields on solid and porous dielectric surfaces. The present work attempts to fill significant knowledge gaps in this field
Development of Multielectrode Electroretinography (meERG) for Human Clinical Use
No full textDevelopment of Multielectrode Electroretinography (meERG) for Human Clinical use.
Introduction
Progressive retinal diseases such as macular degeneration, glaucoma, and retinitis pigmentosa begin with localized dysfunction that spreads, often remaining undetected until advanced stages. Existing clinical electroretinography (ERG) techniques either lack spatial sensitivity (full-field ERG) or are restricted to central vision (multi-focal ERG). Prior animal studies have demonstrated that multielectrode electroretinography (meERG), which records simultaneous ERG waveforms from multiple corneal locations, can reveal localized deficits in retinal function. Translating this technique to humans could provide a clinically feasible, objective method for functional retinal imaging.
Purpose
This work aimed to develop and evaluate a human-compatible meERG system using a Contact Lens Electrode Array (CLEAr) lens, with the overarching goal of advancing the technique toward clinical translation.
Methods
Three Specific Aims were pursued. Aim 1 focused on improving CLEAr lens design, incorporating wireless amplification, improved fit and optimized stimulus delivery; this Aim was the main focus of the overall study. Aim 2 involved pilot human recording to establish proof of concept for the improved designs. Aim 3 developed a computational bioelectric model to relate corneal potential maps to underlying retinal activity through forward and inverse modeling.
Results
The first human meERG system with potential for translation to a clinical setting was designed, fabricated and tested. Key developments included an improved CLEAr lens design to minimize channel crosstalk while accommodating inter-subject variability in corneal and scleral topography. Light transmission (for the ERG stimulus) was enhanced by optimizing electrode cable design, improving nasal–temporal trace symmetry for more uniform retinal illumination, and implementing transparent electrode cables (BioCLEAR), resulting in a 13% improvement in overall light transmittance. A full-field light stimulus source with excellent spatial uniformity (luminance CV <5%) was developed to minimize systemic bias in meERG measurement. Additional advances improved lens stability, biocompatibility, shelf life, packaging, and machining yield were achieved. Initial proof-of-concept human recordings demonstrated, for the first time, spatially distinct meERG signals with excellent signal to noise ratio. An analysis strategy for measuring small spatial differences in corneal potentials was established and demonstrated. Finally, a computational forward model based on a human atlas dataset was developed, capable of predicting field potentials across the anterior surface of the eye.
Conclusion
This work establishes the feasibility of human meERG as a clinically relevant functional imaging technique. By providing rapid, objective, and spatially resolved measures of retinal function, meERG has the potential to significantly improve early detection of retinal diseases and monitoring of localized therapies such as gene or protein delivery. Continued development will focus on optimizing clinical usability and validating diagnostic sensitivity in patient populations
Orthogonal Protein Synthesis by Ribosomes with an Integrated mRNA
No full textRibosomes drive protein synthesis by translating the mRNA into the amino acid sequence of proteins. Given the importance of proteins and peptides, many efforts have focused on engineering the ribosome to expand its functionality beyond its repertoire of only 20 naturally occurring proteinogenic amino acids. However, developing such systems faces the challenge of overcoming the detrimental effects that most ribosomal modifications impose upon the ability of the engineered ribosomes to synthesize cellular proteins. One way to circumvent this limitation is to functionally isolate the population of engineered ribosomes from the rest, thus generating two pools of ribosomes in the same cell: (I) wildtype ribosomes translating endogenous mRNAs and (II) specialized ribosomes translating their designated mRNAs. To develop a translation system with functional separation with respect to wildtype ribosomes, we have designed a new ribosome in which the mRNA is physically tethered to the rRNA of the small ribosomal subunit. We provide evidence that these engineered ribosomes are highly specialized, indicating that their attached mRNA is translated predominantly in cis. Furthermore, we successfully integrated this design into ribosomes with tethered subunits to expand functional isolation to the large ribosomal subunit. This ribosome-mRNA design demonstrates the viability of a fully integrated translation system whose specialization can theoretically allow the assignment of individual ribosomes to the synthesis of specific polypeptides, including those containing non-natural or difficult-to-translate amino acid sequences. The results of our work open new opportunities for synthetic biology
Role of Rostromedial Tegmental Nucleus Afferents in Symptoms of Withdrawal from Chronic Ethanol Exposure
No full textA significant number of individuals with alcohol use disorder struggle to maintain sobriety as a result of withdrawal symptoms including heightened pain sensitivity. Although common, the precise neural mechanisms underlying nociceptive symptoms of withdrawal are not well understood. The rostromedial tegmental nucleus (RMTg) is a GABAergic region that plays a critical role in avoidance behavior, aversive signaling, and pain modulation. Accumulating data provides evidence of RMTg hyperactivity during acute withdrawal and points to a role for the RMTg in regulating withdrawal symptoms. Yet, the role of this region in withdrawal-induced changes in pain sensitivity and the afferents that promote RMTg activity during acute withdrawal are unknown. In this thesis, I examined the role of RMTg afferents in regulating hyperalgesia during acute withdrawal. I found that thermal hyperalgesia emerged early after the cessation of ethanol exposure and persisted into protracted withdrawal. cFos expression, a marker of recent neuronal activity, was significantly increased in both the RMTg and RMTg-projecting lateral habenula (LHb) neurons during acute withdrawal. In vivo chemogenetic inhibition of this circuit attenuated withdrawal-induced thermal hyperalgesia in a sex-dependent manner. Collectively, these data suggest that RMTg-projecting LHb neurons play an important role in nociceptive symptoms of withdrawal from chronic ethanol exposure
Synthesis and Photophysics of Varied Substituted DPNDs for Potential PDT Applications
No full textUnderstanding how molecular architecture and solvent polarity govern excited-state dynamics is fundamental to developing functional chromophores for photo-energy conversion and manipulation for opto-electronic applications. Here, the photophysical properties of three dipyrrolonaphthyridinedione (DPND) derivatives, tBu-DPND–1 (β-Suzuki), tBu-DPND–1 (β-Sonogashira), and tBu-DPND–3 (α-Sonogashira), were prepared and characterized using advanced photophysical methods to unveil how π-conjugation and topology influence excited-state dynamics. Furthermore, we varied the polarity of the samples to shed light on the contributions of charge transfer (CT) dynamics and their interplay with triplet excited state formation of the three DPND chromophores. Steady-state and time-resolved absorption and emission studies suggest that the three DPND chromophores exhibit longer excited-state lifetimes and stronger triplet signatures in non-polar methylcyclohexane (MCH) solvent. Conversely, in polar 50% (v/v) DCM:EtOH, the excited states were stabilized and underwent equally non-radiative deactivation and ISC to populate the corresponding triplet manifold. Time-resolved Transient absorption spectroscopy (TAS) confirmed the ISC and triplet formation. Moreover, singlet-oxygen generation using these DPNDs as sensitizers yielded quantum yields (ΦΔ) ≈ 0.6. The interplay between conjugation length, positional effect, and solvent polarity dictates the ordering and mixing of excited singlet, CT, and triplet states, thereby controlling ISC efficiency. These findings establish the new π-conjugation DPNDs as polarity-responsive organic chromophores suitable for applications such as triplet sensitization, photodynamic therapy, and singlet-fission processes
Validating Computational Software With Exact Analytical Solutions: A review of the last three decades
No full textExact analytical solutions of canonical electromagnetic scattering problems that have been developed in the last 30 years are reviewed to emphasize that many geometrical shapes exist for the validation of computational electromagnetic (CEM) methods. The focus is on exact solutions that can be computed as the sum of series expansions, where the expansion coefficients are known analytically and do not require, for example, the solution of a system of equations. The introduction of a class of artificial materials, known as isorefractive, has enabled the development of new analytical solutions, some of which involve complications not previously found, such as involving different materials, sharp edges, and cavities. In addition, other complicated geometries have been examined to create new exact solutions, without requiring the use of isorefractive materials. Many numerical evaluations exist for the new exact solutions so that for the purposes of numerical validations, one could simply use the existing ones.</p
Data sources:
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇