Biomedical Informatics Commons^™

Ai-Driven Biomarker Discovery & Progression Modeling For Precision Diagnosis Of Glaucoma, Cheng Huang

Computer Science and Engineering Theses and Dissertations

This dissertation presents a comprehensive study on the integration of artificial intelligence (AI) for glaucoma diagnosis and retinal image analysis. Leveraging multimodal imaging data including fundus photography, Optical Coherence Tomography Optical Coherence Tomography (OCT) and Optical Coherence Tomography Angiography (OCTA), the research develops a suite of deep learning frameworks designed to detect early glaucomatous changes with high precision, robustness, and interpretability. A series of novel architectures are introduced, spanning vessel segmentation networks, biomarker discovery pipelines, and multimodal fusion models, all designed to enhance diagnostic accuracy and generalizability across diverse populations. To facilitate reproducible and scalable ophthalmic AI research, this work …

Bridging Data Gaps In Retinal Imaging: From Structural Domain Adaptation To Topology-Aware Synthesis, Gözde Merve Demirci

Dissertations, Theses, and Capstone Projects

Comprehensive visualization of the retina is essential for diagnosing and monitoring blinding diseases such as Diabetic Retinopathy and Retinopathy of Prematurity (ROP), where pathological changes often extend beyond a single field of view. Despite significant advances in automated retinal image analysis, clinical deployment remains limited by two fundamental data gaps: a structural learning gap, arising from scarce expert annotations and poor generalization across imaging domains, and a spatial coverage gap, caused by the difficulty of acquiring multi-view retinal images in fragile populations. Although these challenges are often addressed independently, this dissertation argues that they are tightly coupled: accurate, …

Agent-Based Modeling Of Cellular Dynamics In Adoptive Cell Therapies, Yujia Wang

Dissertations and Theses (Open Access)

Adoptive cell therapies (ACT) have shown promising progress in combating cancer, but clinical responses remain heterogeneous. Key determinants, including the functional states of infused cells, tumor–immune interactions, and dosing strategies, are difficult to resolve experimentally in the dynamic and heterogeneous tumor microenvironment, limiting optimization of ACT products and treatment regimens. To address the gap, we developed ABMACT (Agent-Based Model for Adoptive Cell Therapies), a computational framework that reconstruct key cellular functions, interactions, and molecular signatures to recapitulate cell population dynamics associated with differential treatment responses. ABMACT encodes autonomous “virtual cells” using literature-informed rules calibrated with experimental data, enabling simulation of …

Essays On Ai-Driven Analytics For Enhanced Healthcare Delivery: From Onset To Outcomes In Neurological Diseases Management, Gabriel Owusu

Theses and Dissertations

Chronic neurological diseases, particularly Alzheimer's disease and related dementias (ADRDs), pose mounting global healthcare burdens. Despite advances in AI and health information technologies, early detection, precise diagnosis, and effective caregiver support remain elusive. This dissertation proposes innovative AI-enabled analytical frameworks to address these challenges across the full ADRD continuum.

The first essay designs an explainable AI-enabled clinical decision support system (CDSS) using a graph neural network for early AD identification, evaluated through usability studies demonstrating improved risk awareness and clinical engagement. The second essay introduces MADT-RNN, a multi-level attention-based deep transfer recurrent neural network that fuses convolutional and recurrent architectures …

Network Analysis: An Application Of Graph Theory In Biology, Kiana Dunbar

Honors Theses

Graphs are simple, visual representations of entities as nodes connected by edges. They are used to convey relationships within a system. Networks are comprised of interrelated entities that are not easily separable, producing complex, structured data. These can be seen in social groups, highway systems, and even in biological systems. This paper surveys concepts in graph theory for application to the analysis of network data to understand disease. The features of graphs provide a useful framework for thinking about relationships between different genes or proteins. The structure of graphs is also compatible with a variety of machine learning tools. Especially …

Development And Application Of A Multiplex Bead-Based Platform For Quantifying Host-Pathogen Protein-Protein Interactions, Larah Gorayeb

Electronic Theses and Dissertations 2020 - Present

Host-pathogen protein-protein interactions are the key drivers of infectious diseases, where events such as pathogen adhesion and signaling are directly associated with disease severity. One example of this is Malaria, in which the Plasmodium falciparum infected erythrocytes adhere to endothelial receptors to avoid splenic clearance. This study investigated interactions between PfEMP1 CIDR1 field domains and host receptor CD36 using a multiplex platform to quantitatively determine binding affinities. Variants RP18 and RP27 exhibited strong affinity, suggesting a role in cytoadherence and disease severity. To evaluate broader applicability, the platform was expanded to the interaction between host protein SIVA1 and Helicobacter pylori …

Deep Learning-Based Automated Pneumonia Detection From Chest X-Rays: A Comparative Study Of Custom Cnn And Transfer Learning Architectures, Ahmed Sajim

Honors Theses

Pneumonia is a leading global cause of mortality, claiming approximately 2.5 million lives an-nually and placing exceptional diagnostic pressure on radiologists in resource-limited settings. Manual interpretation of chest X-ray (CXR) images is time-consuming, subject to inter-observer variability, and limited by radiologist availability. This thesis presents a systematic investiga-tion into deep learning-based automated pneumonia detection comparing five convolutional neural network (CNN) architectures: a custom-designed 2D CNN and four pretrained transfer learning models—ResNet, DenseNet, MobileNet, and VGG19.

A targeted data augmentation pipeline addresses the severe class imbalance in the Kag-gle Chest X-Ray Pneumonia dataset, expanding the Normal class from 1,583 to 9,495 …

Deep Learning Based Approaches For Low Cost Defense Detection, Adele J. Noel-Rickert

All NMU Master's Theses

Pulmonary fibrosis is a progressive interstitial lung disease characterized by the accumulation of fibrotic tissue within the lungs, leading to impaired respiratory function and reduced quality of life. Early detection is important for disease management; however, accurate diagnosis often relies on high-resolution computed tomography (CT), which may not be accessible in all clinical settings. Chest radiography provides a lower-cost and widely available imaging modality, but interpretation of chest X-rays for fibrotic disease can be challenging due to subtle radiographic patterns and overlapping anatomical structures. This thesis investigates the use of multimodal deep learning techniques to assist in pul- monary fibrosis …

Uncovering Potential Therapeutic Targets For Spinal Cord Injury; Insights From Single-Cell And Cross-Species Transcriptomics, Sara Samy Abouzekry

Theses and Dissertations

Central nervous system (CNS) injury remains a devastating healthcare problem. Despite years of research, it remains almost impossible to achieve meaningful functional repair following traumatic brain injury (TBI) and spinal cord injury (SCI). Spinal cord injury (SCI), a highly devastating subset of CNS trauma, is considered a challenge due to the extremely limited regenerative capacity of the nervous system in adult mammals. In contrast, certain species display remarkable repair abilities followed by moderate regenerative ability in neonatal mammals, indicating that regeneration is shaped by the evolutionary context as well as the developmental age.

To uncover the core molecular mediators of …

Metabolic Syndrome Beyond Diagnostic Criteria: Population-Scale Integrative Metabolomics Characterization, Marwa Talal

Theses and Dissertations

Background: Metabolic syndrome (MetS) is a complex cluster of interrelated metabolic abnormalities associated with elevated cardiometabolic risk. While diagnosis is based on well-established five clinical criteria, these may overlook early or atypical metabolic alterations. Large-scale metabolomic profiling offers an opportunity to identify biochemical signatures of MetS beyond diagnostic bias and to evaluate their relative importance across different presentations of the syndrome.

Methods: Data from 117,147 UK Biobank participants were analyzed in a cross-sectional design. High-throughput NMR quantified 75 circulating metabolites, for. Univariate analyses, MetS subtype stratification, and elastic net models with SHAP interpretation were applied to assess feature …

Previsit Ai: A Retrieval-Augmented Generation For Patient Readiness In Clinical Encounters, Rolande Umuhoza

All Graduate Theses, Dissertations, and Other Capstone Projects

With healthcare systems under growing pressure from rising patient volumes and shrinking consultation windows, improving how patients communicate with physicians has become essential to delivering quality care. Yet patients routinely arrive at appointments unable to clearly describe their symptoms, recall their medical history, or articulate concerns, contributing to miscommunication, diagnostic inefficiency, and pre-visit anxiety. This study introduces PreVisit AI, a conversational system designed to address this gap through structured, knowledge-based patient preparation. The system is built on a Retrieval-Augmented Generation (RAG) architecture combining HuggingFace sentence embeddings (all-MiniLM-L6-v2), a Chroma vector store, and Google’s Gemini language model over a curated seven-document …

Integrative Approaches And Data Analysis For Single-Cell Rna Sequencing Data, Teng Long

Computer Science and Engineering Dissertations

The rapid growth of single-cell RNA sequencing and transcriptomic datasets has created major computational challenges in causal discovery, representation learning, and biologically faithful data generation. To address these challenges, this dissertation presents three complementary deep learning frameworks for the analysis and modeling of transcriptomic data. Together, these methods form an integrative computational toolkit for understanding complex biological systems from high-dimensional and heterogeneous gene expression data.

First, this dissertation introduces DAG-VAERL, a causal discovery framework that integrates variational autoencoders, graph neural networks, reinforcement learning, and attention mechanisms to infer directed acyclic graphs for gene regulatory network analysis. DAG-VAERL improves causal structure …

A Specification-Driven Framework For Self-Supervised Learning In Specialized Vision Domains, Mahmut S. Gokmen

Theses and Dissertations--Computer Science

Self-supervised learning (SSL) has emerged as a principled approach to visual representation learning that derives supervisory signal directly from unlabeled data, enabling foundation models to be trained at scale without manual annotation. Deployments in medical imaging and biometric recognition have demonstrated the potential of this paradigm, yet the assumptions that make SSL effective on natural image benchmarks fail systematically in specialized domains. Generic SSL pipelines encode a tacit assumption that the most informative correspondence is spatial proximity within a single acquisition. In specialized domains this assumption breaks at the level of the data-generating process: the signal that carries domain-specific information …

Adversarial Robustness In Biomedical Time-Series Models, Rohan Tiwari

Bioengineering Theses

This study investigates adversarial vulnerabilities in deep learning models for biomedical time-series classification across two clinically important modalities: electrocardiography (ECG) and electroencephalography (EEG). Using the MIT-BIH Arrhythmia and CHB-MIT seizure datasets, I evaluate time-domain attacks (FGSM, PGD), Fourier-domain constrained attacks, and learned spectral perturbations designed to reveal modality-specific sensitivity patterns. Across both tasks, a consistent trend emerges low-frequency components (0–5 Hz) constitute a dominant axis of adversarial vulnerability, with perturbations in this range producing the steepest degradation in classification performance. In ECG models, protecting the physiologically relevant QRS band (5–20 Hz) significantly improves robustness, whereas EEG models remain highly sensitive …

Robust And Interpretable Medical Image Segmentation Via Retrieval And Mechanistic Analysis, Salma Ahmed

Theses and Dissertations (Comprehensive)

Medical image segmentation is a critical component of clinical decision-making, yet deep learning based segmentation models face persistent challenges. These models typically require large volumes of densely annotated data, exhibit strong sensitivity to domain shifts across scanners and patient populations, and lack explicit anatomical priors. Furthermore, despite continual architectural advances beyond fully convolutional networks, most segmentation models remain opaque, where they lack the granularity required for component-level analysis. As a result, when segmentation models fail, it is difficult to identify which internal representations are responsible, limiting developers’ ability to diagnose errors, improve robustness, or establish trust in clinical settings. Existing …

Hyperglycemia Detection From Sigle - Lead Ecg Using A Hybrid Cnn & Transformer Model, Adam Ayomikun Ogunjembola

Graduate Theses, Dissertations, and Problem Reports (ETD)

Abstract
Hyperglycemia Detection from Single-Lead ECG using a Hybrid CNN & Transformer Model
Adam Ogunjembola

Diabetes Mellitus is known as high blood glucose. This high blood glucose level happens when the body has a problem with producing or using insulin. Insulin is a very important hormone that the pancreas makes to control how much glucose gets into the bloodstream and cells. Diabetes Mellitus has an effect on the body if it is not treated, such as damaging the blood vessels and nerves which can lead to stroke, kidney failure, heart attack and permanent loss of vision. Since people with diabetes …

Recurrence And Co-Occurrence Of Enhancer-Promoter Interactions Across Human Samples, Satvik Gunjala

Honors Undergraduate Theses

Experimental mapping of enhancer-promoter interactions (EPIs) is resource-intensive, and current computational prediction methods struggle with intrinsic genomic complexity and reliance on limited training data. To address this bottleneck and provide insights for improved computational methods, this study systematically analyzed chromatin contact datasets to investigate the recurrence and co-occurrence of enhancer-promoter interactions across human samples. Putative interactions were evaluated across two HiChIP datasets comprising 218 total samples and one Hi-C dataset comprising 266 samples to assess recurrence across samples and assess sequencing depth related to unique EPIs. Additionally, a preliminary item-based collaborative filtering recommender model was developed to assess co-occurrence patterns …

A Picture Tells A Thousand Words—, But Ecg Signals Have More To Say, Ashley N. Gomez

Theses and Dissertations

With the increasing adoption of deep learning classification models in the medical domain, a critical challenge remains: achieving high predictive accuracy while maintaining clinical Inter-pretability. This study examines how model architecture, dataset origin, and the use of full versus subset data affect both classification performance and Interpretability in Electrocardiogram (ECG) signal analysis. ResNet18 is evaluated using an open-source ECG Image Dataset, thus a custom dataset derived from digitized ECG images. Post-hoc explainability methods, such as Integrated Gradients, are applied to determine which time steps have the most significant influence on model decisions. The findings demonstrate that model architecture and dataset …

Ai-Powered Multi-Omics Integration For Predictive Modeling Of Genotype-Environment-Phenotype Relationships, You Wu

Dissertations, Theses, and Capstone Projects

This dissertation presents a series of machine learning frameworks for modeling genotype–environment–phenotype relationships through integrative predictive modeling of multi-omics data. The work addresses three major axes of biological complexity: modeling biological information transmission cross-levels from genes to proteins to phenotypes, predicting molecular features cross-scale from cells to tissues to organisms, and translating phenotypes cross-species from model systems to humans. Each proposed method also tackles key machine learning (ML) challenges in the biomedical domain, including data scarcity, domain shift, out-of-distribution (OOD) generalization, and hierarchical modeling. Specifically, this dissertation introduces five novel deep learning algorithms: MultiDCP predicts drug-induced transcriptomic and viability responses …

Modification Of Epic System’S Simple Visit Coding (Svc) Evaluation Rule To Reduce Medical Claim Denials, Edith Ballard

Translational Projects (Open Access)

Background: The recruitment, development, and retention of skilled medical coding staff has become increasingly challenging for healthcare providers. To alleviate staffing burdens, Epic Systems’ electronic health record (EHR) introduced the Simple Visit Coding (SVC) work queue, designed to evaluate hospital patient encounters and automatically assign International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes. However, the current SVC work queue evaluation rule at a prominent academic Oncology Care Organization (OCO) in the South does not utilize autonomous, generative, or computer-assisted coding (CAC) technologies. The work queue evaluation rule uses a simple rule-based program logic that relies on …

Enhancing Coagulase-Negative Staphylococcus Interpretation: Leveraging Data Visualization To Improve Diagnostic Accuracy And Patient Outcomes, Miriam Booker Barrett

Translational Projects (Open Access)

When implemented effectively, data displays can enhance interpretive decisions by helping physicians recognize patterns and relationships within various datasets. Data visualization has been widely studied for its role in converting raw data into actionable insights and supporting data-driven decision-making. However, few studies have been published that specifically explored using data displays to aid in interpreting coagulasenegative Staphylococcus (CoNS) in blood cultures.

One of the primary challenges infectious disease physicians face when interpreting CoNS in blood cultures is the disorganized nature of clinical data within electronic health records. The information necessary for accurate interpretation is often scattered across various sections of …

Translating The Hba1c Lab Test Name And Results From English To Arabic For Enhanced Patient Access And Workflow Efficiency, Aiman M. Alrawabdeh

Translational Projects (Open Access)

In Jordan, where the healthcare system primarily operates in English, Arabic-speaking patients encounter significant challenges in understanding laboratory test names and results, including Hemoglobin A1c (HbA1c), an essential diagnostic indicator for diabetes management. The lack of multilingual access to medical information limits health literacy and patient autonomy, increasing dependence on healthcare staff for translation assistance. This translational informatics project addressed these challenges by developing and implementing an Arabic translation process for the HbA1c lab test name and results. HbA1c was selected as it is a standard lab test among patients with diabetes in Jordan; nearly 30% of Jordanians had type …

Topodino: Self-Supervised Topological Representation Learning For Neuronal Morphologies, Yasser Binbisher

Master's Theses

Neuronal cell types are categorized by transcriptomic identity, yet their morphological heterogeneity defies this classification. In response, researchers have adopted unsupervised graph representation learning as a tool to reveal morphological variation within single-class transcriptomic types. However, the complex geometry of neuronal morphology—especially long axons and dense dendrites—challenges graph neural networks, which struggle with message propagation across extended structures. To mitigate this, current approaches enforce sub-sampling on neuronal graphs and omit axons entirely, sacrificing critical biological features for computational efficiency. To overcome this trade-off, this thesis introduces TopoDINO, a self-supervised, topology-aware representation learning model designed to preserve the full hierarchical organization …

Explaining Patient Data Towards Improving Patient Understanding And Medical Efficiency, Lupe K. Hernandez Jr.

Masters Theses

Patients frequently access medical test results through electronic health record (EHR) portals, often before receiving input from healthcare providers. These results typically lack interpretation or are presented with complex clinical terminology, leading to confusion, anxiety, and misinformed decision-making. This thesis examines using large language models (LLMs), such as ChatGPT, to generate test result summaries in detailed and simplified formats automatically. A Python-based middleware was developed to interface between OpenEMR and OpenAI’s API, enabling automated summary generation. Summaries included a clinically detailed version and a simplified version written at a third-grade reading level. Two surveys were conducted: one with undergraduate and …

Elucidating The Multi-Omics Of Early-Onset Colorectal Cancer, Jumanah Alshenaifi

Dissertations and Theses (Open Access)

The incidence and mortality rates of sporadic early-onset colorectal cancer have increased in recent decades, but there is no clear etiological basis for this trend. EOCRC is commonly defined as colon and rectal cancers diagnosed before the age of 50 years. The rising incidence of EOCRC has made it the second most common cancer and the third leading cause of cancer death in this age group. The rising incidence of EOCRC is also documented internationally in more than 20 countries across different continents. Clinically, EOCRC has a distinct, more aggressive clinical profile than LOCRC. While approximately 15% of EOCRC cases …

Toward The Application Of Natural Language Processing In Electronic Health Record Analysis For Taxonomy Development, Latoya Mcdonald

All Dissertations

Electronic health records (EHRs) are pivotal resources for nurse practice because they increase the timeliness and reliability of patient information at the point of care and support access by multiple healthcare providers and the individual patients themselves. However, it is widely recognized that data extraction from EHRs is challenging due to the variability in the language used in clinical care notes and the lack of standardized terminology across healthcare systems. The broad objective of this dissertation is to develop taxonomy-based classification models for nursing care by applying feature engineering approaches to EHRs that include nursing care of ostomy patients following …

Improving The Oral Pathology Referral Process In An Academic Dental Institution, Gregory Olson

Translational Projects (Open Access)

Introduction: Inefficiencies in the oral cancer referral process pose serious risks, including delayed diagnoses and poor patient outcomes. This project aimed to improve the reliability and efficiency of referrals between dental students and oral pathology residents for patients with suspected oral cancer. Although early detection is widely acknowledged as critical, significant process breakdowns were identified, including the absence of a standardized workflow, inconsistent follow-up practices, and a lack of accountability and tracking mechanisms. As a result, approximately half of the patients recommended for biopsy did not undergo the procedure, leading to delayed diagnoses and potentially worse prognoses.

Methodology: …

Leveraging Observational And Rct Data For Understanding Interventions’ Efficacy: Applications In Progressive And Acute Neurological Diseases, Yaobin Ling

Dissertations and Theses (Open Access)

The advancement of drug repurposing for progressive and acute neurological diseases is hampered by the limitations of randomized clinical trials (RCTs) and observational data. This dissertation presents a comprehensive framework to integrate data-driven insights from multiple sources, including observational studies, RCTs, and synthetic data generation, to overcome these challenges.

The first study focuses on estimating treatment effects on the population level, which investigates the effects of routine and high-dose influenza vaccines on the risk of Alzheimer’s Disease and Related Dementias (ADRD) through a trial emulation framework applied to health claims data, addressing biases inherent in observational studies. The second study …

Standardizing Social Determinants Of Health Factors From Heterogeneous Sources To Improve Data Fairness, Yifang Dang

Dissertations and Theses (Open Access)

Social Determinants of Health (SDoH) significantly influence health outcomes, yet their representation in computational models remains fragmented. This dissertation addresses this gap by constructing an SDoH ontology (SDoHO), leveraging it to improve large language model (LLM) performance, and using LLMs to extract new knowledge and enrich the ontology. The overarching goal is to establish a feedback loop where ontology development enhances LLM-based extraction, and LLM-derived insights refine and expand the ontology. The methodology is structured across three aims: (1) ontology construction, (2) ontology-assisted LLM enhancement, and (3) LLM-driven ontology enrichment, with a focus on Alzheimer’s Disease and Related Dementias (ADRD). …

Computer-Aided Integrated Rehabilitation For Post-Stroke Patients Using Deep Learning Techniques, Kaichen Tang

Dissertations and Theses (Open Access)

Stroke is a leading cause of long-term disability, often requiring intensive rehabilitation and frequent clinical assessments such as the Fugl-Meyer Assessment. However, traditional in-person evaluations are resource-heavy and difficult to scale, limiting access for many patients. Additionally, monitoring vital signs like blood pressure and oxygen saturation typically depends on specialized equipment, leading to fragmented care and incomplete recovery insights. Even when assessments and monitoring are available, patients often struggle to maintain high-quality, intensive exercise routines without supervision, further limiting rehabilitation outcomes.

To address these challenges, this dissertation proposes an AI-driven, smartphone-based framework that integrates automated motor function assessment, non-invasive vital …