Doctor of Philosophy, Case Western Reserve University, 2018, Biomedical Engineering

Early detection and treatment of coronary artery disease (CAD) can improve prognosis and overall survival. However, current noninvasive assessment is highly inefficient: of patients referred to invasive angiography, >60% do not have obstructive CAD. Microvascular disease (MVD) accounts for a significant portion of these patients, particularly in patients with diabetes, smoking, hypertension, or other cardiomyopathies. Quantitative estimates of myocardial blood flow by myocardial perfusion imaging (MPI) can detect the physiologic impact of MVD and obstructive CAD. The combination of MPI with computed tomography (MPI-CT) and coronary CT angiography would enable rapid physiologic and anatomic evaluation of CAD and MVD in a single exam. Despite a number of promising MPI-CT reports, the lack of consensus in image acquisition and myocardial blood flow quantification methods, as well as concern regarding imaging artifacts and radiation dose, slows clinical adoption. Four projects are described in this dissertation. First, in a porcine model of flow-limiting stenosis scanned on a spectral detector CT, energy-sensitive reconstruction and dynamic imaging were shown to improve detection of myocardial ischemia as compared to conventional reconstruction and static imaging. Second, the role of imaging conditions and quantification methods was evaluated with regards to obtaining accurate and precise myocardial blood flow (MBF) estimates. Several methods from the literature, some implemented in commercial software, gave imprecise, biased MBF estimates. A proposed robust physiologic model was found to precisely and accurately quantify MBF. Third, a method to calculate MBF confidence intervals (MBFCI) was developed and used to select appropriate analysis models. Use of MBFCI and a goodness-of-fit metric, Akaike Information Criterion (AIC), selected a model with precise MBF estimates whereas AIC alone selected models with imprecise MBF estimates. Fourth, an advanced iterative reco (open full item for complete abstract)

Committee: David Wilson PhD (Advisor); Nicole Seiberlich PhD (Committee Chair); Hiram Bezerra MD, PhD (Committee Member); Raymond Muzic Jr., PhD (Committee Member); Steven Izen PhD (Committee Member) Subjects: Biomedical Engineering; Biomedical Research; Medical Imaging; Radiology

MS, University of Cincinnati, 2024, Engineering and Applied Science: Computer Science

Brain tumor segmentation is crucial for diagnosis, treatment planning, and patient monitoring. Traditional manual segmentation is labor-intensive and prone to variability, necessitating the development of automated, precise, and reproducible methods. This study enhances segmentation by integrating multi-modal MRI scans (T1-weighted, T2-weighted, FLAIR, and post-contrast T1-weighted) with clinical data using advanced deep learning techniques. Multi-modal MRI provides diverse tissue contrasts essential for identifying tumor regions like the enhancing core, peritumoral edema, and necrosis. Incorporating clinical data—such as patient age, survival days, and genetic markers—adds context influencing tumor appearance and growth patterns, refining the segmentation process. The proposed framework uses convolutional neural networks (CNNs), particularly U-Net architectures, trained on datasets like the BraTS Challenge, with data augmentation and cross-validation enhancing robustness and generalizability. A multi-input model processes imaging data and clinical features through parallel neural network branches, fusing them to form a comprehensive representation. This integration captures complex relationships between clinical variables and imaging features, improving segmentation outcomes. Performance is evaluated using metrics such as Dice Similarity Coefficient (DSC), Accuracy, and Intersection over Union. Initial results show that combining multimodal MRI with clinical data significantly improves segmentation accuracy and delineation of tumor boundaries. In conclusion, integrating multi-modal MRI and clinical data in brain tumor segmentation offers more accurate and clinically meaningful results. This approach harnesses the full spectrum of imaging information and contextual clinical insights, paving the way for more effective and personalized patient care. Future work will refine the model architecture, expand (open full item for complete abstract)

Committee: Vikram Ravindra Ph.D. (Committee Chair); FNU Nitin Ph.D. (Committee Member); Jun Bai Ph.D. (Committee Member) Subjects: Computer Science

Doctor of Philosophy, Case Western Reserve University, 2025, EECS - System and Control Engineering

In the context of artificial intelligence (AI) with a focus on machine learning (ML) and deep learning (DL), variability and generalizability are key challenges that significantly impact model performance and applicability across diverse real-world scenarios. Variability refers to the inconsistencies in data quality, feature extraction, and model outcomes that arise due to differences in data settings, populations, and other external factors. Generalizability, on the other hand, denotes a model ability to maintain robust performance when applied to new, unseen data beyond the specific conditions it was trained on. Tackling these challenges is critical to enable the broader adoption of ML and DL models, particularly in high-stakes fields such as medical imaging and signal processing, where reliable and consistent performance is essential. Addressing these issues requires systematically dividing the AI workflow into three primary stages: pre-analytical, analytical, and post-analytical. Each stage presents unique challenges, and this work introduces novel methods to address them, thereby improving the overall robustness, performance, and generalizability of AI models. In the pre-analytical stage, we propose new quality control mechanisms for medical imaging data, including the development of MRQy, a tool for automated MRI quality assessment, and DeepQC, a DL-based framework that leverages ResNet18 for detecting artifacts such as noise and motion. These innovations ensure that the data fed into AI models is clean, consistent, and suitable for analysis, minimizing biases that can degrade model performance. The analytical stage introduces several wavelet-based architectures to enhance feature extraction and model building. Key contributions include Sparse Wavelet Networks (SWN), Deep Hybrid Con-volutional Wavelet Networks (DHCWN), and Residual Wavelon Convolutional Networks (RWCN). These architectures improve the ability to capture multi-scale and complex patterns, part (open full item for complete abstract)

Committee: Satish Viswanath (Advisor); Wei Lin (Committee Member); Vira Chankong (Committee Member); Kenneth Loparo (Committee Chair) Subjects: Electrical Engineering

Master of Science, The Ohio State University, 2007, Graduate School

Committee: Not Provided (Other) Subjects:

Doctor of Philosophy, Case Western Reserve University, 2024, EECS - System and Control Engineering

This dissertation investigates the application of digital pathology for developing diagnostic and prognostic tools for 2 diseases: Biliary tract adenocarcinoma and Papillary Thyroid Carcinoma (PTC). We explore the realms of cytopathology, which studies exclusively the morphologies of epithelial cells, and histopathology, which includes the entire tissue region. Bile duct brush specimens are difficult to interpret as they often present inflammatory and reactive backgrounds due to the local effects of stricture, atypical reactive changes, or previously installed stents, and often have low to intermediate cellularity. As a result, diagnosis of biliary adenocarcinomas is challenging and often results in large interobserver variability and low sensitivity. In this dissertation, we first used computational image analysis to evaluate the role of nuclear morphological and texture features of epithelial cell clusters to predict the presence of biliary tract adenocarcinoma on digitized brush cytology specimens. We improved the sensitivity of diagnosis with a machine learning approach from 46% to 68% when atypical cases were included and treated as nonmalignant false negatives. The specificity of our model was 100% within the atypical category. PTC is the most prevalent form of thyroid cancer, with the classical form and the follicular variant representing the majority of cases. Despite generally favorable prognoses, approximately 10% of patients experience recurrence post- surgery and radioactive iodine therapy. Attempts to stratify risk of recurrence have relied on gene expression-based prognostic and predictive signatures with a focus on mutations of well-known driver genes, while hallmarks of tumor morphology have been ignored. In this dissertation, we introduce a new computational pathology approach to develop prognostic gene signatures for thyroid cancer that is informed by quantitative features of tumor and immune cell morphology. We show that integrating gene express (open full item for complete abstract)

Committee: Kenneth Loparo (Committee Chair); Anant Madabhushi (Advisor); Satish Viswanath (Committee Member); Sylvia Asa (Committee Member); Aparna Harbhajanka (Committee Member) Subjects: Artificial Intelligence; Biomedical Engineering; Biomedical Research; Biostatistics; Computer Engineering; Medical Imaging; Oncology; Systems Design

Doctor of Philosophy, Case Western Reserve University, 2024, Biomedical Engineering

Magnetic resonance fingerprinting (MRF) is a fast quantitative imaging technique that simultaneously maps multiple MR-related tissue properties from a single scan. While MRF has been applied to enhance diagnostic quality for various medical conditions, its extensive application in the clinical workflow has been limited by several factors. First, MRF scans still require relatively long scan times compared to conventional clinical MRI. Moreover, the performance of MRF scans can be influenced by real-world scan variations, including scan acceleration, system imperfections, and patient motion. The objective of this work is to improve the scan efficiency, robustness, and measurement accuracy of MRF scans against errors arising from these confounding factors. The following projects all contribute to this goal by optimizing the acquisition strategies of MRF scans. Firstly, we developed an optimization framework to design MRF pulse sequence parameters for accurate T1 and T2 measurements within short scan durations. This framework includes a fast MRF image simulation, significantly accelerating error estimation in the cost function calculation process and making the sequence optimization problem computationally feasible. To address the highly dimensional and nonconvex nature of the optimization problem, we employed a search space parameterization strategy and a fine-tuned stochastic algorithm. This study represents the initial step in establishing an optimization pipeline applicable to designing pulse sequences for any MRF implementation across various clinical applications. Secondly, we integrated the fast MRF image simulator with a new mathematical model, called the systematic error index (SEI). This addition further reduces the computational cost of error analysis in cost function evaluations. The SEI enables sequence optimization of high-dimensional MRF frameworks that quantify more than two tissue properties to achieve fast and accurate scans. This SEI-based optimizati (open full item for complete abstract)

Committee: Dan Ma (Advisor); Xin Yu (Committee Chair); Daniela Calvetti (Committee Member); Satish Viswanath (Committee Member); Mark Griswold (Committee Member) Subjects: Biomedical Engineering

PhD, University of Cincinnati, 2023, Engineering and Applied Science: Computer Science and Engineering

Deep learning techniques are being unified for decision support in various applications. However, it remains challenging to train robust deep learning models, due to the inherent insufficient labeled data that is usually time-consuming and labor-intensive. Self-supervised learning is a feature representation learning paradigm to learn robust features from insufficient annotated datasets. It contains two types of task stages, including the pretext task and the downstream task. The model is typically pre-trained with the pretext task in an unsupervised manner, where the data itself provides supervision. Afterward, the model is fine-tuned in a real downstream supervised task. Although self-supervised learning can effectively learn the robust latent feature representations and reduce human annotation efforts, it highly relies on designing efficient pretext tasks. Therefore, studying effective pretext tasks is desirable to learn more effective features and further improve the model prediction performance for decision support. In self-supervised learning, pretext tasks with deep metric/contrastive learning styles received more and more attention, as the learned distance representations are useful to capture the similarity relationship among samples and further improve the performance of various supervised or unsupervised learning tasks. In this dissertation proposal, we survey the recent state-of-the-art self-supervised learning methods and propose several new deep metric and contrastive learning strategies for learning effective features. Firstly, we propose a new deep metric learning method for image recognition. The proposed method learns an effective distance metric from both geometric and probabilistic space. Secondly, we develop a novel contrastive learning method using the Bregman divergence, extending the contrastive learning loss function into a more generalized divergence form, which improves the quality of self-supervised learned feature representation. (open full item for complete abstract)

Committee: Anca Ralescu Ph.D. (Committee Chair); Kenneth Berman Ph.D. (Committee Member); Dan Ralescu Ph.D. (Committee Member); Lili He Ph.D. (Committee Member); Yizong Cheng Ph.D. (Committee Member) Subjects: Computer Science

Master of Computer and Information Science, Cleveland State University, 2023, Washkewicz College of Engineering

Deep learning has revolutionized several fields including the medical image processing in the past decade. Convolutional Neural Networks can now perform many image processing tasks better than humans. As a result, Convolution Neural Networks (CNNs) are increasingly used in the automation of diagnosis of life-threatening diseases. CNNs perform complex image classification tasks with greater accuracy and output quality. However, recent discovery of adversarial attacks raises a significant threat against safety and accuracy of the CNNs. CNNs are vulnerable to perturbations in the input image that are imperceptible to human eyes, which leads to misclassification of the model output. This research work proposes a novel Super Resolution Generative Adversarial Network-based approach to improve classification robustness of CNN against adversarial attacks using MRI dataset as an example. Robustness of proposed novel network model is compared with existing state of the art models in the field. The experiment results demonstrate that proposed approach improves CNN model robustness by 95% against adversarial attacks when compared to state-of-the-art approaches such as context-aware-models and conventional CNN.

Committee: SATHISH KUMAR (Committee Chair); HONGKAI YU (Committee Member); JANCHE SANG (Committee Member) Subjects: Artificial Intelligence

Master of Science, The Ohio State University, 2023, Electrical and Computer Engineering

Diseases such as Alzheimer's and cancer cause alterations in the genome, the physiology and structure of afflicted organs. These variations can be studied for the development of powerful disease predictive tools. Machine learning algorithms for disease prediction and diagnosis have seen a rapid increase owing to the availability of large volumes of data. Vision based machine learning models have shown promising results in Alzheimer's disease and cancer detection using imaging data such as 3D MRI (Magnetic Resonance Imaging) and PET (Positron Emission Tomography) as input. Additionally, analysis of genomics such as study of changes in gene expression, identification of genetic variants and genome wide association studies have been used for designing successful models for Alzheimer's Disease risk assessment and cancer detection. Further, investigating the correlation between genomics and imaging features allows for the identification of radiogenomic biomarkers. These biomarkers can be used to design more reliable and robust predictive tools. In this work, we develop supervised and unsupervised machine learning based algorithms using imaging, genomics and radiogenomics, for the task of Alzheimer's disease prediction. Additionally, we employed genomics for the task of breast, pancreas, liver and colon cancer detection. With genomics data, we perform a study using unsupervised and reinforcement learning based algorithms for predicting an optimal empirical formula for multi omics (viz. copy number variation, DNA methylation and RNA gene expression) integration. We test the efficacy of the algorithm on multiple cancer datasets and report the prediction accuracies achieved. Following this, we extend this approach for Alzheimer's detection by integrating gene expression data corresponding to three genes namely, APOE, PSEN1 and PSEN2. Further, 3D structural magnetic resonance imaging is used as input to design a long term recurrent convolutional network (or LRCN) based frame (open full item for complete abstract)

Committee: Golrokh Mirzaei (Advisor); Wladimiro Villarroel (Committee Member); Asimina Kiourti (Committee Member) Subjects: Computer Science; Electrical Engineering

Doctor of Philosophy, Case Western Reserve University, 2023, Biomedical Engineering

In many chronic inflammatory diseases, the vascular endothelium becomes pathologically permeable due to conditions like angiogenesis and production of growth factors and inflammatory cytokines (e.g., histamine, bradykinin, etc.). In cancer, this process can be exploited for delivery of nanoparticles to tumors via the enhanced permeability and retention (EPR) effect. However, nanoparticle-based therapeutics reliant on the EPR effect have led to inconsistent results in patients. This is due to many factors, with a significant one being heterogeneous tumor vascular architecture and morphology both between patients and within a single tumor. Transport of the nanoparticle to the tumor and into the parenchyma is complicated by uptake by the immune system, ineffective margination, and inefficient extravasation. Guidance is needed to inform clinicians on what therapies may be most effective for each patient. Effective guidance could reduce health-care costs and negative side effects of medication. An inexpensive, safe, non-invasive, and real-time imaging method that has high temporal and spatial resolution may be capable of categorizing the extent of vascular permeability in tumors and once validated, personalize therapeutic regimens for patients. Such a tool could be used not only for tumors, but for all diseases involving pathologically permeable vasculature. With this goal in mind, the objective of this thesis is to work toward development of a real-time method for evaluating vascular permeability over the entire tumor using novel nanobubble (NB)-based contrast-enhanced ultrasound (CEUS). This work builds upon dynamic CEUS protocols used clinically with microbubbles (MBs). NBs, which are 100-400 nm in diameter, are approximately 10x smaller than MBs and have been shown to extravasate into the tumor interstitium. To reach the final objective of this work, NB dynamics from intravenous injection to retention in the tumor must be studied. To this aim, in vitro studies con (open full item for complete abstract)

Committee: Agata Exner (Advisor); Horst von Recum (Committee Chair); Anirban Sen Gupta (Committee Member); Aaron Proweller (Committee Member) Subjects: Biomedical Engineering; Medical Imaging; Medicine; Nanoscience; Nanotechnology; Oncology; Radiology

Doctor of Philosophy, Case Western Reserve University, 0, Molecular Medicine

Novel therapeutic regimens which spur the endogenous immune system to kill cancer cells, such as stereotactic radiosurgery (SRS) and immune checkpoint inhibition (ICI), are heterogeneously effective. Understanding causal factors of response is vital to guide risk assessment and treatment decisions. In this thesis, I evaluate the ability of three methods to prognosticate survival for brain metastases patients following SRS and ICI treatment. These include the clinically utilized response assessment in neuro-oncology for brain metastases (RANO-BM) protocol, as well as investigational computational methods such as radiomic feature analysis and convolutional neural network (CNN) image analysis. I find that easing the 10mm RANO-BM diameter threshold for measurable disease allows new lesions to be discovered as proof of progression in ICI-treated metastases. Further, I find that the trajectory of RANO-BM diameter can be more instructive for risk prediction than the ratio-change and that neither volume nor number of metastases, nor RANO-BM diameter can significantly predict survival until a year after treatment. Reproducing common radiomic methodology flaws observed in the published literature, I demonstrate that inconsistent partitioning, or the improper division of radiomic feature data into Training, Validation, Test, and External test sets, can provide a 1.4x performance boost to reported accuracy (AUROC) for predictive models. Additionally, I highlight how spurious correlations with biological variables can overstate the importance of radiomic features. Leveraging the conclusions from my radiomic reproduction study, I assess the ability of radiomic features and convolutional neural networks (CNNs) to predict overall survival in the largest ICI-treated brain metastases cohort assembled to date, comprising 175 patients from three institutions in two countries. I find that neither radiomic features nor any architecture of the survival AI model MetsSurv is capable of p (open full item for complete abstract)

Committee: Jacob Scott (Advisor); Brian Rubin (Committee Chair); Elizabeth Gerstner (Committee Member); Anant Madabhushi (Committee Member); Jayashree Kalpathy-Cramer (Advisor); Nathan Pennell (Committee Member) Subjects: Artificial Intelligence; Computer Science; Immunology; Medical Imaging; Molecular Biology; Neurology; Oncology; Radiology

Doctor of Philosophy, Case Western Reserve University, 2022, Biomedical Engineering

Patients with inadequate stent expansion are at high risk for adverse outcomes, including stent thrombosis as well as in-stent restenosis. Both are well-described complication usually causing acute coronary syndromes and, in the worst-case scenario, sudden cardiac death. Once a stent is deployed in atherosclerotic tissue that is highly resistant to dilation, it is often tricky to fully expand the implanted stent, even using a noncompliant balloon. Careful evaluation of these risks prior to the intervention will aid treatment planning, including potential application of a plaque modification strategy. Moderate to severe calcification in the treated vessel restrict full deployment of a stent. Intravascular optical coherence tomography (IVOCT) is a useful tool to identify calcification lesion severity, reference vessel size, lesion length, and extent of calcification for PCI planning. In addition, IVOCT imaging provides detailed evaluation of stent deployment including areas, stent expansion index, floating stent struts and stent dissection. In this dissertation, we developed a machine learning method to predict stent deployment with the intervention of creating software that one can use to identify lesions needing a plaque modification strategy. However, we confronted a challenge in that IVOCT single pullback generated more than 500 images in less than 2.5 second scan. The need for specialized training, uncertain interpretation, and image overload required automated analysis of IVOCT images. Therefore, we built on previous studies and used deep learning to perform semantic segmentation of the lumen and calcification within IVOCT images. We evaluated our segmentation model on manually annotated IVOCT volumes with calcifications, lipidous, or mixed lesions. We obtained sensitivities of 0.85 ± 0.04, 0.99 ± 0.01, and 0.97 ± 0.01 for calcified, lumen, and other tissue classes, respectively. Segmented lumen and calcifications labels were then used to develop the stent u (open full item for complete abstract)

Committee: David Wilson (Advisor); Andrew Rollins (Committee Chair); Umut Gurkan (Committee Member); Satish Viswanath (Committee Member); Sadeer Al-Kindi (Committee Member) Subjects: Biomedical Engineering

Doctor of Philosophy, The Ohio State University, 2021, Biomedical Engineering

Cardiovascular heart disease (CVD) is the leading cause of mortality in the U.S. and worldwide. Over the past several decades, the healthcare costs associated with CVD have steadily risen to more than 200 billion dollars per year and are expected to rise further with the aging population. Cardiovascular MRI (CMR) is a well-established imaging technique that provides the most comprehensive evaluation of the cardiovascular system. CMR is considered the gold standard for evaluating ventricular function and myocardial viability. Despite the growing evidence of its advantages over other imaging modalities and its potential as a “one-stop-shop” diagnostic tool, the role of CMR in clinical cardiology remains limited. One major impediment to its wider usage is the inefficient acquisition that makes CMR exams excessively long, often lasting for more than an hour; this diminishes its efficiency and cost-effectiveness relative to other imaging modalities. The current paradigm offers either a prolonged segmented acquisition that requires regular cardiac rhythm and multiple breath-holds or a fallback option of real-time, free-breathing acquisition with degraded spatial and temporal resolutions. Recently, 3D imaging has gained significant interest due to its volumetric coverage and isotropic resolution. In particular, 4D flow imaging has emerged as a powerful tool that provides temporally and spatially resolved velocity maps of the blood in the heart and great vessels. A major technical limitation of 4D flow imaging is the long acquisition, which makes the images susceptible to motion artifacts. In this work, we present a framework that provides a whole-heart coverage and enables a rapid, quantitative assessment of hemodynamics. In addition, the method employs self-gating and thus extracts and compensates the physiological motions from the information in the MRI data itself, obviating the need to utilize electrocardiogram or respiratory gating. Novel extensions of the method, whe (open full item for complete abstract)

Committee: Rizwan Ahmad (Advisor); Rengasayee Veeraraghavan (Committee Member); Orlando Simonetti (Committee Member); Jun Liu (Committee Member) Subjects: Biomedical Engineering; Medical Imaging

Doctor of Philosophy (Ph.D.), University of Dayton, 2021, Electrical Engineering

Purpose: To present and demonstrate a computationally efficient deep learning approach for computer-aided detection systems for medical imaging applications that include malaria, diabetic retinopathy, and tuberculosis. Approach: We propose a novel and a computationally efficient deep learning approach for medical image analysisusing convolutional neural networks (CNNs). We demonstrate the efficacy of our proposed method in the detection of malaria, diabetic retinopathy, and tuberculosis. We refer to our approach as Incremental Modular Network Synthesis (IMNS), and the resulting CNNs as Incremental Modular Networks (IMNets). Our IMNS approach is to use small network modules that we call SubNets that are capable of generating salient features for a particular problem. Then, we build up ever larger and more powerful networks by combining these SubNets in different configurations. At each stage, only one new SubNet module undergoes learning updates. This reduces the computational resource requirements for training and aids in network optimization. Results: We compare IMNets against classic and state-of-the-art deep learning architectures such as AlexNet, ResNet-50, Inception v3, DenseNet-201, and NasNet for the various experiments conducted in this study. Our proposed IMNS design leads to high average classification accuracies of 97.0%, 97.9%, and 88.6% for malaria, diabetic retinopathy, and tuberculosis, respectively. Conclusions: Our modular design for deep learning achieves the state-of-the-art performance in the scenarios tested. The IMNets produced here have a relatively low computational complexity compared to traditional deep learning architectures. The simpler IMNets train faster, have lower memory requirements, and process images faster than the benchmark methods tested

Committee: Russell Hardie Dr. (Committee Chair); Vijayan Asari Dr. (Committee Member); Youssef Raffoul Dr. (Committee Member); Temesguen Kebede Dr. (Committee Member); John Loomis Dr. (Committee Member) Subjects: Biomedical Engineering; Computer Engineering; Electrical Engineering

PhD, University of Cincinnati, 2021, Engineering and Applied Science: Computer Science and Engineering

Deep learning methods, especially convolutional neural networks (CNNs), have revolutionized many domains of artificial intelligence (AI) including natural image classification/segmentation, speech recognition, and natural language processing. It is slowly being acknowledged that deep learning methods have a huge potential to advance medical image analysis, medical diagnostics, and general healthcare. Traditionally, medical imaging interpretations have benefited from machine learning methods. In machine learning, feature extraction is significant for model design and experiment implementation. However, the exact etiologies behind many medical diseases are unknown. It is challenging to develop a feature extraction system when there is a lack of domain understanding. Compared to traditional machine learning methods, feature extraction is part of the learning process and discriminative features can be automatically extracted in deep learning. It has been demonstrated that deep learning methods can produce physiologically meaningful features and reveal new associations from high dimensional medical imaging data. Different from conventional methods, deep learning is an end-to-end solution for medical imaging problems without the feature extraction process. Data can be fed to deep learning models in their raw form, and high-level representations are automatically extracted.

Committee: H. Howard Fan Ph.D. (Committee Chair); Yizong Cheng Ph.D. (Committee Member); Carla Purdy (Committee Member); Ali Minai Ph.D. (Committee Member); Lili He (Committee Member) Subjects: Computer Science

Master of Science in Computer Engineering, University of Dayton, 2021, Electrical and Computer Engineering

Artificial intelligence (AI) based analysis is accelerating clinical diagnosis from pathological images and automating image analysis efficiently and accurately. Recently, Deep Learning (DL) algorithms have shown superior performance in pathological image analysis, such as tumor region identification, metastasis detection, and patient prognosis. As digital pathology becomes popular, it is crucial to evaluate the performance of DL approaches that show the best performance for the different color-space representations of pathological images. The main goal of this research is to analyze several supervised and unsupervised DL approaches in pathological image analysis. In this study, the Inception Residual Recurrent Convolutional Neural Network (IRRCNN) model has been examined in six different color spaces (RGB, CIE, HSB, YCrCb, Lab, and HSL) pathological images and evaluate the best color space for tissue classification tasks. In addition, the Recurrent Residual U-Net (R2U-Net) model is evaluated in six different color spaces images in nuclei segmentation tasks and selects the best color space. Also, R2U-Net based autoencoder models are examined for medical image denoising such as digital pathology, dermoscopy, Magnetic Resonance Imaging (MRI), and Computed Tomography (CT). The performance of the R2U-Net based auto-encoder model is also evaluated for the Transfer domain (TD) between MRI and CT scan images. Finally, as pathological images have higher dimensions, it is necessary to reduce the dimensionality for analyzing these samples by obtaining its original features representation in the lower dimensions. In this research, DL features have been extracted, and then the t-distributed Stochastic Non-linear Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) are applied for clustering and visualization of pathological images.

Committee: Tarek M Taha (Advisor) Subjects: Artificial Intelligence; Biomedical Research; Computer Engineering; Computer Science; Medical Imaging

Doctor of Philosophy, Case Western Reserve University, 2021, EECS - Electrical Engineering

For the last decade, robotic catheters have emerged as a promising technology for catheter ablation. The development of magnetic resonance image (MRI) guided robotic catheters is complicated by the need to track the position and orientation of these instruments within the MRI scanner, accurate localization of the desired target on cardiac surface, and precise control of the catheter. In order to accurately navigate the catheter to the desired location on the heart via MR images, it is necessary to register the robot space to the MR scanner's image space as well as track the catheter position and the cardiac surface motion from the MR images, while precisely controlling the catheter. This thesis details novel approaches to address the challenges in these topics. The first contribution of this work is to describe a framework to register robotic catheter to the MRI scanner, while taking into account the scanner related geometric distortions in the MR images. The geometric distortion is identified via a grid-based, custom-built 3D phantom, where morphological operations are applied to localize the control points in the phantom images, which in turn are used to determine the distortion map. The underlying distortion is modeled and corrected by employing thin plate splines. The catheter to scanner registration is performed via a differential, multi-slice image-based registration approach utilizing active fiducial coils. In the proposed scheme, the registration is performed with the help of a registration frame, which has a set of embedded electromagnetic coils designed to actively create MR image artifacts. These coils are detected in the MRI scanner's coordinate system by background subtraction. The detected coil locations in each slice are weighted by the artifact size and registered to known ground truth coil locations in the catheter's coordinate system via least-squares fitting. The proposed approach is validated by using a set of target coils placed within the wo (open full item for complete abstract)

Committee: Murat Cenk Cavusoglu PhD (Advisor); Murat Cenk Cavusoglu PhD (Committee Chair); Wyatt Newman PhD (Committee Member); Frank Merat PhD (Committee Member); Mark Griswold PhD (Committee Member); Nicole Seiberlich PhD (Committee Member) Subjects: Electrical Engineering; Engineering; Medical Imaging; Robotics; Surgery

Master of Science, Miami University, 2021, Chemical, Paper and Biomedical Engineering

Newts have exceptional capability of regenerating the lens throughout their lifetime. Since the 1890s, lens regeneration has been documented with ex vivo imaging techniques only. For the first time, we demonstrate that Optical Coherence Tomography (SD-OCT) can capture the in vivo essential morphological characteristics with abundant dynamic features. Monitoring lens regeneration using a single newt is now possible. The results show that the lens originates below the pupillary margin of the middle dorsal region and its early regenerating lens has an irregular elliptical shape. The lens volume expands quadratically, where its regenerating rate is linear from 14 to 60 days post-lentectomy. These findings warrant future research for tailoring OCT to study newt lens regeneration in vivo dynamically.

Committee: Hui Wang Dr (Advisor); Katia Del Rio-Tsonis Dr (Committee Member); Justin Saul Dr (Committee Member) Subjects: Biomedical Engineering

Doctor of Philosophy (PhD), Wright State University, 2020, Computer Science and Engineering PhD

As the population of older individuals increases worldwide, the number of people with cardiovascular issues and diseases is also increasing. The rate at which individuals in the United States of America and worldwide that succumb to Cardiovascular Disease (CVD) is rising as well. Approximately 2,303 Americans die to some form of CVD per day according to the American Heart Association. Furthermore, the Center for Disease Control and Prevention states that 647,000 Americans die yearly due to some form of CVD, which equates to one person every 37 seconds. Finally, the World Health Organization reports that the number one cause of death globally is from CVD in the form of either myocardial infarctions or strokes. The primary ways of assisting individuals affected with CVD are from either improved treatments, monitoring research, or primary and secondary prevention measures. In the form of cardiovascular structural monitoring, there are multiple ways of viewing the human heart. That is, Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET), Computed Tomography (CT), and Ultrasonography are the four fundamental imaging techniques. Though, continuous monitoring with these imaging techniques is far from currently possible. Large financial cost and size (MRI), radiation exposure (PET and CT), or necessary physician assistance (Ultrasonography) are the current primary problems. Though, of the four methodologies, Ultrasonography allows for multiple configurations, is the least expensive, and has no detrimental side effects to the patient. Therefore, in an effort to improve continuous monitoring capabilities for cardiovascular health, we design a novel wearable ultrasound vest to create a near 3D model of the heart in real-time. Specifically, we provide a structural modeling approach specific to this system's design via a Stereo Vision 3D modeling algorithm. Similarly, we introduce multiple Stochastic Petri Net (SPN) models of the heart for future functiona (open full item for complete abstract)

Committee: Nikolaos G. Bourbakis Ph.D. (Advisor); Soon M. Chung Ph.D. (Committee Member); Yong Pei Ph.D. (Committee Member); Iosif Papadakis Ktistakis Ph.D. (Committee Member); Konstantina Nikita Ph.D. (Committee Member); Anthony Pothoulakis M.D. (Other) Subjects: Biomedical Engineering; Biomedical Research; Computer Science; Medical Imaging

Doctor of Philosophy, Case Western Reserve University, 2020, Biomedical Engineering

The high morbidity associated with triple negative breast cancers (TNBCs) is directly related to its high risk of recurrence. TNBC recurrence is often invasive - leading to its metastasis (mTNBC) in visceral organs including the lungs, liver, and brain. With these phenotypic characteristics nearly all newly diagnosed patients with mTNBC will have a poor prognosis. The difficulty with metastatic disease is two-fold: 1) micrometastasis (> 1cm) cannot be reliably detected by conventional diagnostic techniques and 2) therapeutic windows are reduced as the metastatic lesions are not easily accessible to systemically administered agents. Thus, the difficulty in diagnosis and treatment lies to a great degree in adequately targeting imaging and therapeutic agents to metastasis. However, it is well documented that metastatic niches upregulate receptors that are not commonly found in healthy tissues such as nonendogenous matrix proteins (i.e. PTP-mu, fibronectin/fibrinogen), adhesion molecules (i.e. selectins, cadherins and integrins) and cell specific markers (i.e. EGFR and integrins). Nanotechnology offers a unique solution such by incorporating targeting ligands that can direct nanoparticles to these tumor-associated upregulated biomarkers. By decorating the surface of nanoparticles with targeting moieties, we can adequately administer nanoparticles loaded with either contrast agents, chemotherapeutics or immunotherapeutics. In this dissertation, we show that targeted nanoparticles can significantly improve diagnosis and treatment of metastatic breast cancer.

Committee: Efstathios Karathanasis Ph.D. (Advisor); Capadona Jeffery Ph.D. (Committee Chair); Yu Jennifer M.D., Ph.D. (Committee Member); Tiwari Pallavi Ph.D. (Committee Member); Samia Anna Ph.D. (Committee Member) Subjects: Biomedical Engineering