22689 research outputs found
Sort by
Intelligent Patient Management and Control
Get PDFThis thesis aims to develop a multi-agent Reinforcement Learning (RL) framework to assist clinicians in optimizing Pressure Support Ventilation (PSV) settings by aligning them with patient-specific respiratory patterns under cases healthy, obese, and ARDS. Traditional approaches rely heavily on manual adjustments, which are time-consuming, prone to human error, and lack the adaptability to dynamic patient conditions. To address these limitations, we constructed a Markov Decision Process (MDP) to model the dynamics of the respiratory-ventilation system over time, using LT Spice simulations to capture the nonlinear mechanical behavior of lung airway interactions. This involved calculating the work of breathing (WOB) across a comprehensive range of state configurations defined by key respiratory parameters, including inspiratory pressure (Pinsp), positive end-expiratory pressure (PEEP), variable airway resistances, and multiple variable lung compliances. Algorithms were developed to process simulation outputs and calculate WOB for each state, enabling the creation of policies and reward functions, thus adjusting ventilation parameters to better assist patient breathing. We also constructed a customized AI Gym environment to simulate PSV management and validated the correctness of state transitions and rewards. After integrating with a deep reinforcement learning framework, this approach has the potential to automate PSV adjustments in real time together with clinician decisions, improving efficiency, accuracy, and patient outcomes in conditions such as acute respiratory distress syndrome (ARDS) and obesity-related respiratory challenges
SEX DIFFERENCES ASSOCIATED WITH MIP-3α-ANTIGEN FUSION DNA VACCINE IN TUBERCULOSIS MODEL
Get PDFBackground: Tuberculosis (TB) remains a major cause of global mortality. A
therapeutic DNA vaccine targeting RelMtb has been shown to increase the efficacy of
antitubercular drugs, and fusing macrophage-inflammatory protein 3α (MIP-3α to RelMtb)
further increases the vaccine’s therapeutic efficacy. A previous study in the laboratory
found that overall MIP-3α enhanced vaccine uptake and cell activation by iDCs. Female
mice showed greater levels of antigen presentation, especially in DCs able to cross
present antigen, explaining why they had the best outcomes. The current study
hypothesizes that females compared to males will show enhancement of genes associated
with CD8+ T cell recruitment and cross-presentation that are more pronounced with MIP
3α-antigen fusion vaccines and that females will have higher levels of immune cell
infiltration and Yae+ dendritic cells that are more pronounced with the vaccine construct
including MIP-3α.
Methods: To determine which genes might be contributing to the sex difference
phenotype qRT-PCR analysis was performed. Additionally, fluorescence microscopy of
the muscle was utilized to determine the amounts of GFP, Yae vaccine antigen, and
Cd11c+ dendritic cells at the site of vaccination.
Results: Most of the genes upregulated in MIP-3α females were chemokines involved in
immune cell motility and attracting cells to the node, while many of the genes that are
upregulated in males are important in monocyte-derived DC-driven T cell responses or
signal DC maturation. At the site of vaccination, Mip-EαGFP females had higher
ii
amounts of Cd11c+ signal, DAPI+ signal, and a higher percentage of DAPI+ signal that
is also Cd11c+ when compared to EαGFP females.
Conclusions: The Rel vaccine enhances TB bacterial clearance by antibiotics with a
female bias, and that fusion of MIP-3α enhances both the overall response and the sex
bias. The data supports the hypotheses that fusing MIP-3α to the antigen leads to better
targeting and activation of APCs at the vaccination site, and more dendritic cell and
immune cell infiltration in females. More research studying the mechanisms involved in
the sex differences is pivotal to fully understanding how vaccination elicits a more robust
immune response and how to best treat persistent TB
CO2 Utilization via Microbial Electrosynthesis Using Earth-abundant Metal as Anode
Get PDFReducing carbon emissions from the water sector via microbial activities is highly feasible. Carbon dioxide can be converted to methane and other valuable organic products in microbial electrosynthesis cells (MEC) for industrial purposes. However, cubic MECs have high internal resistance due to the space between two electrodes, hindering methane production, and are ideal only for biocathode acclimation. The previous study has addressed high internal resistance by using zero-gap flow MECs with cation exchange membrane, yet the overall reaction is dependent on oxygen evolution reaction (OER) on the anode and the methane production rate still needs to be improved by using anode materials that are resilient to OER. Precious metals such as Platinum and Iridium have shown high resilience to OER, but industrial-scale implementation with precious metals is too expensive. Therefore, we are testing earth-abundant titanium with surface deposition of γ-MnO2 as anodes for suppressed manganese ion dissolution and enhanced protection on titanium against OER, and acclimated carbon felt from cubic MEC as a biocathode. γ-MnO2/Ti with high loading rates demonstrated superior performance than of Pt/Ti in the cubic MEC. Our current average methane production rate in zero-gap flow MEC duplicates is 0.17 ± 0.103 L/L-d at a current density of 0.205 ± 0.2038 A/m2. We expect the discrepancies to be greatly reduced by addressing the limitations of poor internal connections and gas leakage from the system
A VERSATILE IMMERSIVE VIRTUAL REALITY PLATFORM FOR THE ANALYSIS OF VISION-DRIVEN BEHAVIORS IN ZEBRAFISH LARVAE
No full textUnderstanding how animals interact with environments through their behavior is one of fundamental goals of systems neuroscience. In this study, we present a low-cost, flexible, and immersive virtual reality (VR) system designed to investigate visuomotor behaviors in larval zebrafish. The platform utilizes a single projector, a 3D-printed curved screen, and real-time tracking algorithms to deliver panoramic, closed-loop visual stimulation. Compared to existing VR systems that often rely on multiple projectors and complex rendering engines, our system is easy to assemble and supports seamless switching between free-swimming and head-fixed experiments.
We first evaluated the system using looming-evoked escape assays in zebrafish at 5 and 7 days post-fertilization (dpf). Consistent with prior findings, fish in planar projection conditions exhibited a conserved escape threshold at a visual angle of approximately 73°, independent of developmental stage. However, fish in the immersive VR environment demonstrated significantly lower visual angle thresholds for escape, indicating higher stimulus saliency in immersive settings.
To enable real-time interaction, we developed a deep learning–based translation algorithm using Transformer architecture to predict fish velocity and angular displacement from tail motion. Our model achieved high predictive accuracy (angle MSE: 1.7°, displacement MSE: 0.17 mm), outperforming prior machine learning methods. Additionally, we introduced a geometric correction and panorama update algorithm to generate 360° stimuli responsive to the fish's inferred virtual position and orientation.
Finally, the system is compatible with advanced optical imaging techniques, such as light-sheet microscopy. This configuration enables simultaneous behavioral and neural activity recordings. Taken together, our VR platform provides a powerful and adaptable tool for investigating the neural circuits underlying visually guided behavior in larval zebrafish
MOLECULAR CHARACTERIZATION OF HISTOLOGIC RISK FACTORS FOR BREAST CANCER
Get PDFBreast cancer is the second leading cause of cancer-related deaths of women in the United States. Breast cancer development is influenced by a complex interplay between genetic, environmental, and histological factors. Histological risk factors for breast cancer are currently limited in number, emphasizing the need for more positive characterization. Differences in race and genetic ancestry has been shown to influence breast cancer subtypes, risk, and overall survival. Additionally, long-term environmental exposures, including particulate matter 2.5 (PM2.5) and neighborhood deprivation, have prominent roles in increasing breast cancer risk and disproportionately affect different racial and ethnic populations. In this work, we studied the associations between two histological risk factors, crown-like structures of the breast (CLS-B) and terminal duct lobular units (TDLUs), and breast cancer risk based on differences in genetic ancestry and environmental exposures. This work aims to offer insights into disparities in breast cancer outcomes by expanding the scope of histological risk factors to genetic, molecular, and environmental factors, ultimately advancing early detection and personalized treatment strategies
Associations between Physically Demanding Jobs and Dementia Risk among Older Adults in the United States
Get PDFDementia represents a growing global health challenge among older populations, characterized by progressive cognitive decline and memory impairment. Understanding the risk factors associated with dementia is critical to developing preventative strategies and improving health outcomes in older adults. Existing studies suggest that cognitively demanding occupations may serve as protective factors against dementia by fostering cognitive resilience. However, the relationship between physically demanding jobs and cognition remains to be established. The study aims to examine whether exposure to physically demanding jobs is associated with dementia incidence among older adults in the United States (US), while considering other sociodemographic risk factors.
This study utilized data from the Health and Retirement Study (HRS). Study participants were enrolled between 1992 and 2016, with follow-up data collected biennially from 1992 to 2020. We specifically used data from the 15th wave of biennial visits from HRS between March 2020 to May 2021. A total of 7,132 participants age 50 and above were included. Occupations were categorized into physically demanding jobs and non-physically demanding jobs based on participants’ self-reported occupational information. Dementia cases were identified through self- or proxy (e.g., a family caregiver)- reported diagnoses. Univariable logistic regression was used to examine the associations between dementia risk and occupation type, as well as potential confounders such as age, total working time, sex, race, marital status, education level, life insurance, depression history, and region of birth. Multivariable logistic regression was used to determine whether a physically demanding job was associated with differences in dementia prevalence, adjusting for the aforementioned potential confounders.
Our study revealed that there was a significant association between physically demanding jobs and increased dementia risk (OR = 3.33; 95% CI = 1.60, 8.49; p = 0.004). However, after adjusting for covariates, the association was no longer significant (OR = 2.00; 95% CI = 0.93, 5.21; p = 0.108). Other significant predictors of increased dementia risk included older age, male sex, shorter working time, living alone, and a history of depression. This study utilized a small sample of dementia cases from the US, limiting the generalizability of the findings to other populations. It is imperative for future research to use larger, population-based occupational data of aging individuals to better establish significant associations
THE BIOSOLIDS WORKER: CHARACTERIZATION OF TASKS AND FACTORS THAT INFLUENCE EXPOSURE TO LAND-APPLIED BIOSOLIDS
No full textBackground: Biosolids, or treated sewage sludge, contain several pathogenic and chemical contaminants that remain after treatment and may be harmful to workers who regularly interact with the sludge during land application. Workers’ risks of exposure remain uncharacterized, in part, because we lack critical information about how, when, and how often workers come into contact with biosolids during land application tasks.
Objective: We aim to identify how workers come into contact with biosolids during land application and describe contextual factors that may influence their contact and refine exposure assessments for contaminants in biosolids.
Methods: We conducted 11 semi-structured, in-depth interviews with biosolids workers who apply, transport, and/or use biosolids across the United States and Canada. We transcribed the interviews verbatim and developed a priori codes based on our study objectives. We added new codes and reorganized a priori codes as patterns and themes emerged to aid in interpreting workers’ descriptions of land application and their contact with biosolids.
Results: Workers described six tasks regularly performed during land application: hauling, loading, spreading, post-application field work, cleaning, and maintenance. Workers’ contact depended on whether they performed these tasks inside or outside the enclosed cabs of land application equipment. The saturation of the biosolids, the weather, and the actions they took to reduce their contact were also important contextual factors that modified workers’ contact. Our findings revealed novel worker-specific exposure pathways and suggested modifications to quantitative exposure models that may enhance exposure estimates.
Significance: Our tailored approach to understanding workers’ contact with biosolids identifies novel exposure pathways and key considerations that modify workers contact that will be critical for refining the assessment of workers’ exposure to biosolids
Animality and Ecology in Post-war and Postcolonial German and Assamese Literatures
No full textThis dissertation ‘Animality and Ecology in Post-war and Post-colonial German and Assamese Literatures’ explores various aspects of the ‘animal turn’ in contemporary German and Assamese works. In Why look at Animals? (1980), John Berger argues that metaphors have gradually supplanted real animals heightening alienation between humans and animals in modernity. Centering on Berger’s disappearing animal, this study engages the multi-scalar ramifications of ecological issues to examine how they shape new eco-literary trends in the literary cultures of German and Assamese. Despite the vast cultural and geographical differences, in both German-speaking regions and Assam, violent geopolitical movements have endured in the twentieth century alongside efforts for territorial domination or indigenous sovereignty. Arguably, these processes have impacted how the non-human, forests, animals, and landscapes are viewed, culturally mediated, and managed. This research aims to situate these political and cultural manifestations in light of contemporary ecological discourse. Specifically, it sets out to explore how cultural and social structures and change in political systems have shaped the literary representation of the non-human animal against the historical background of post-war Germany and of post-colonial Assam. Thus, the juxtaposition of a wider range of thinking and writing about human-animal entangled relationships in the two literary traditions further highlights the larger questions of ecology and animality within the discourse of the global climate change and what is now known to us as the ‘sixth mass extinction’. Furthermore, this research seeks to investigate to what extent or whether Nietzsche’s modernist concern with animals is asserted within the post-war context of a divided Germany. Ultimately, by situating the literary works in German and Assamese in a conjoined frame, this project demonstrates how a discourse-based inter-literary reading makes a previously understudied dimension accessible. At a time when economic uncertainty and energy scarcity have fundamentally reshaped our relationship with the natural world, my research seeks to foster an inter-literary dialogue between three novels of German wildlife documentary filmmaker and author Horst Stern, and Assamese eco-novels by Jatindra Borgohain and Prabhat Goswami
JOB SATISFACTION, RETENTION AND PRODUCTIVITY LEVELS OF RESEARCH ADMINISTRATORS IN RELATION TO REMOTE WORK
Get PDFThis thesis has explored the perceived productivity, job satisfaction, and psychological well-being of employees who work remotely. Due to the pandemic, many employees, including Research Administrators, were forced to transition from working in the office to a remote environment. This unexpected shift disrupted many organizations and their employees. With the lifting of social distancing regulations, employers questioned whether they should allow employees to continue working remotely or require them to return to the office. The chosen studies for this meta-analysis conducted surveys where employees self-reported their perceptions of productivity while working at home and how it related to their job satisfaction and work-life balance. The results of this meta-analysis have suggested that there was no significant difference in perceived productivity between the different work settings. Since productivity levels remain unaffected and employees have reported many perceived benefits from working remotely, employers are incentivized to modernize work-from-home policies to allow more hybrid and fully-remote positions to increase employee retention, reduce burnout and turnover, which studies show does have a positive impact on productivity
Statistical Methods for Etiologic Associations and Individualized Prediction of Muscle Involvement in Systemic Sclerosis
Get PDFMyopathy in systemic sclerosis (SSc) significantly increases disability, reduces quality of life, and elevates mortality risk, yet remains understudied. Current knowledge derives primarily from cross-sectional studies, limiting our understanding of longitudinal patterns and hindering individualized risk assessment.
This thesis addresses these limitations through two longitudinal modeling approaches applied to electronic health record data from over 2,800 patients in the Johns Hopkins Scleroderma Center Research Registry. First, we implemented a transition-based framework examining clinical risk factors for both incidence and continuance of muscle involvement across three outcomes: abnormal Medsger Muscle Severity Score (MSS ≥ 1) indicating proximal muscle weakness, elevated creatine phosphokinase (CPK ≥ 200 U/L), and their co-occurrence as a myopathy surrogate. Our findings corroborated previous research while revealing important distinctions: diffuse cutaneous subtype, restrictive lung disease, tendon friction rubs, and anti-PM/Scl and anti-Fibrillarin antibodies increased odds of incident myopathy, while different predictors influenced disease persistence.
To support personalized clinical decision-making, we developed a dynamic prediction framework utilizing a Bayesian generalized linear mixed model with cross-validation and variable selection. Using proximal muscle weakness as a proof-of-concept outcome, the model incorporated baseline characteristics, time-varying covariates, and patient-specific medical history to generate individualized risk trajectories. This approach demonstrated superior predictive performance (AUC: 0.867 [0.864, 0.869]) and reasonable calibration, when compared to standard regression and random forest models, allowing for continuously updated risk predictions as new clinical data emerges.
Together, these complementary approaches advance precision medicine by combining population-level risk factor identification with dynamic, patient-specific prediction. This work introduces the first longitudinal, real-time, individualized prediction model for myopathy-related outcomes in SSc. The framework offers a generalizable approach for potentially modeling other outcomes in electronic health record data to improve risk stratification and more timely, patient-centered intervention