Recent advancements in computer vision models have largely been driven by supervised training on labeled data. However, the process of labeling datasets remains both costly and time-intensive. This dissertation delves into enhancing the performance of deep neural networks when faced with limited or no labeling information. I address this challenge through four primary methodologies: domain adaptation, self-supervision, input regularization, and label regularization. In situations where labeled data is unavailable but a similar dataset exists, domain adaptation emerges as a valuable strategy for transferring knowledge from the labeled dataset to the target dataset. This dissertation introduces three innovative domain adaptation methods that operate at pixel, feature, and output levels.Another approach to tackle the absence of labels involves a novel self-supervision technique tailored to train Vision Transformers in extracting rich features. The third and fourth approaches focus on scenarios where only a limited amount of labeled data is available. In such cases, I present novel regularization techniques designed to mitigate overfitting by modifying the input data and the target labels, respectively.

With the exponential growth of multi-modal data in the field of computer vision, the ability to do inference effectively among multiple modalities—such as visual, textual, and auditory data—shows significant opportunities. The rapid development of cross-modal applications such as retrieval and association is primarily attributed to their ability to bridge the gap between different modalities of data. However, the current mainstream cross-modal methods always heavily rely on the availability of fully annotated paired data, presenting a significant challenge due to the scarcity of precisely matched datasets in real-world scenarios. In response to this bottleneck, several sophisticated deep learning algorithms are designed to substantially improve the inference capabilities across a broad spectrum of cross-modal applications. This dissertation introduces novel deep learning algorithms aimed at enhancing inference capabilities in cross-modal applications, which take four primary aspects. Firstly, it introduces the algorithm for image retrieval by learning hashing codes. This algorithm only utilizes the other modality data in weakly supervised tags format rather than the supervised label. Secondly, it designs a novel framework for learning the joint embeddings of images and texts for the cross-modal retrieval tasks. It efficiently learns the binary codes from the continuous CLIP feature space and can even deliver competitive performance compared with the results from non-hashing methods. Thirdly, it conducts a method to learn the fragment-level embeddings that capture fine-grained cross-modal association in images and texts. This method uses the fragment proposals in an unsupervised manner. Lastly, this dissertation also outlines the algorithm to enhance the mask-text association ability of pre-trained semantic segmentation models with zero examples provided. Extensive future plans to further improve this algorithm for semantic segmentation tasks will be discussed.

Alzheimer's disease (AD) and Alzheimer's Related Dementias (ADRD) is projected to affect 50 million people globally in the coming decades. Clinical research suggests that Mild Cognitive Impairment (MCI), a precursor to dementia, offers a critical window for lifestyle interventions to delay or prevent the progression of AD/ADRD. Previous research indicates that lifestyle changes, including increased physical exercise, reduced caloric intake, and mentally stimulating exercises, can reduce the risk of MCI. Early detection of MCI is challenging due to subtle and often unnoticed cognitive decline, traditionally monitored through infrequent clinical tests. As part of this research, the Smart Driving System was proposed, a novel, unobtrusive, and economical technology to detect early stages of neurodegenerative diseases. This system, leveraging a multi-modal biosensing array (MMS) and AI algorithms, assesses daily driving behavior, offering insights into a driver's cognitive function. The ultimate goal is to develop the Smart Driving Device and App, integrating it into vehicles, and validating its effectiveness in detecting MCI through comprehensive pilot studies. The Smart Driving System represents a breakthrough in AD/ADRD management, promising significant improvements in early detection and offering a scalable, cost-effective solution for monitoring cognitive health in real-world settings.

This thesis is developed in the context of biomanufacturing of modern products that have the following properties: require short design to manufacturing time, they have high variability due to a high desired level of patient personalization, and, as a result, may be manufactured in low volumes. This area at the intersection of therapeutics and biomanufacturing has become increasingly important: (i) a huge push toward the design of new RNA nanoparticles has revolutionized the science of vaccines due to the COVID-19 pandemic; (ii) while the technology to produce personalized cancer medications is available, efficient design and operation of manufacturing systems is not yet agreed upon. In this work, the focus is on operations research methodologies that can support faster design of novel products, specifically RNA; and methods for the enabling of personalization in biomanufacturing, and will specifically look at the problem of cancer therapy manufacturing. Across both areas, methods are presented attempting to embed pre-existing knowledge (e.g., constraints characterizing good molecules, comparison between structures) as well as learn problem structure (e.g., the landscape of the rewards function while synthesizing the control for a single use bioreactor). This thesis produced three key outcomes: (i) ExpertRNA for the prediction of the structure of an RNA molecule given a sequence. RNA structure is fundamental in determining its function. Therefore, having efficient tools for such prediction can make all the difference for a scientist trying to understand optimal molecule configuration. For the first time, the algorithm allows expert evaluation in the loop to judge the partial predictions that the tool produces; (ii) BioMAN, a discrete event simulation tool for the study of single-use biomanufacturing of personalized cancer therapies. The discrete event simulation engine was designed tailored to handle the efficient scheduling of many parallel events which is cause by the presence of single use resources. This is the first simulator of this type for individual therapies; (iii) Part-MCTS, a novel sequential decision-making algorithm to support the control of single use systems. This tool integrates for the first-time simulation, monte-carlo tree search and optimal computing budget allocation for managing the computational effort.

Over the past few decades, medical imaging is becoming important in medicine for disease diagnosis, prognosis, treatment assessment and health monitoring. As medical imaging has progressed, imaging biomarkers are being rapidly developed for early diagnosis and staging of disease. Detecting and segmenting objects from images are often the first steps in quantitative measurement of these biomarkers. While large objects can often be automatically or semi-automatically delineated, segmenting small objects (blobs) is challenging. The small object of particular interest in this dissertation are glomeruli from kidney magnetic resonance (MR) images. This problem has its unique challenges. First of all, the size of glomeruli is extremely small and very similar with noises from images. Second, there are massive of glomeruli in kidney, e.g. over 1 million glomeruli in human kidney, and the intensity distribution is heterogenous. A third recognized issue is that a large portion of glomeruli are overlapping and touched in images. The goal of this dissertation is to develop computational algorithms to identify and discover glomeruli related imaging biomarkers. The first phase is to develop a U-net joint with Hessian based Difference of Gaussians (UH-DoG) blob detector. Joining effort from deep learning alleviates the over-detection issue from Hessian analysis. Next, as extension of UH-DoG, a small blob detector using Bi-Threshold Constrained Adaptive Scales (BTCAS) is proposed. Deep learning is treated as prior of Difference of Gaussian (DoG) to improve its efficiency. By adopting BTCAS, under-segmentation issue of deep learning is addressed. The second phase is to develop a denoising convexity-consistent Blob Generative Adversarial Network (BlobGAN). BlobGAN could achieve high denoising performance and selectively denoise the image without affecting the blobs. These detectors are validated on datasets of 2D fluorescent images, 3D synthetic images, 3D MR (18 mice, 3 humans) images and proved to be outperforming the competing detectors. In the last phase, a Fréchet Descriptors Distance based Coreset approach (FDD-Coreset) is proposed for accelerating BlobGAN’s training. Experiments have shown that BlobGAN trained on FDD-Coreset not only significantly reduces the training time, but also achieves higher denoising performance and maintains approximate performance of blob identification compared with training on entire dataset.

Beta-Amyloid(Aβ) plaques and tau protein tangles in the brain are now widely recognized as the defining hallmarks of Alzheimer’s disease (AD), followed by structural atrophy detectable on brain magnetic resonance imaging (MRI) scans. However, current methods to detect Aβ/tau pathology are either invasive (lumbar puncture) or quite costly and not widely available (positron emission tomography (PET)). And one of the particular neurodegenerative regions is the hippocampus to which the influence of Aβ/tau on has been one of the research projects focuses in the AD pathophysiological progress. In this dissertation, I proposed three novel machine learning and statistical models to examine subtle aspects of the hippocampal morphometry from MRI that are associated with Aβ /tau burden in the brain, measured using PET images. The first model is a novel unsupervised feature reduction model to generate a low-dimensional representation of hippocampal morphometry for each individual subject, which has superior performance in predicting Aβ/tau burden in the brain. The second one is an efficient federated group lasso model to identify the hippocampal subregions where atrophy is strongly associated with abnormal Aβ/Tau. The last one is a federated model for imaging genetics, which can identify genetic and transcriptomic influences on hippocampal morphometry. Finally, I stated the results of these three models that have been published or submitted to peer-reviewed conferences and journals.

The COVID-19 outbreak that started in 2020, brought the world to its knees and is still a menace after three years. Over eighty-five million cases and over a million deaths have occurred due to COVID-19 during that time in the United States alone. A great deal of research has gone into making epidemic models to show the impact of the virus by plotting the cases, deaths, and hospitalization due to COVID-19. However, there is very less research that has anything to do with mapping different variants of COVID-19. SARS-CoV-2, the virus that causes COVID-19, constantly mutates and multiple variants have emerged over time. The major variants include Beta, Gamma, Delta and the recent one, Omicron. The purpose of the research done in this thesis is to modify one of the epidemic models i.e., the Spatially Informed Rapid Testing for Epidemic Model (SIRTEM), in such a way that various variants of the virus will be modelled at the same time. The model will be assessed by adding the Omicron and the Delta variants and in doing so, the effects of different variants can be studied by looking at the positive cases, hospitalizations, and deaths from both the variants for the Arizona Population. The focus will be to find the best infection rate and testing rate by using Random numbers so that the published positive cases and the positive cases derived from the model have the least mean square error.

Additive manufacturing consists of successive fabrication of materials layer upon layer to manufacture three-dimensional items. Several key problems such as poor quality of finished products and excessive operational costs are yet to be addressed before it becomes widely applicable in the industry. Retroactive/offline actions such as post-manufacturing inspections for defect detection in finished products are not only extremely expensive and ineffective but are also incapable of issuing corrective action signals during the building span. In-situ monitoring and optimal control methods, on the other hand, can provide viable alternatives to aid with the online detection of anomalies and control the process. Nevertheless, the complexity of process assumptions, unique structure of collected data, and high-frequency data acquisition rate severely deteriorates the performance of traditional and parametric control and process monitoring approaches. Out of diverse categories of additive manufacturing, Large-Scale Additive Manufacturing (LSAM) by material extrusion and Laser Powder Bed Fusion (LPBF) suffer the most due to their more advanced technologies and are therefore the subjects of study in this work. In LSAM, the geometry of large parts can impact the heat dissipation and lead to large thermal gradients between distance locations on the surface. The surface's temperature profile is captured by an infrared thermal camera and translated to a non-linear regression model to formulate the surface cooling dynamics. The surface temperature prediction methodology is then combined into an optimization model with probabilistic constraints for real-time layer time and material flow control. On-axis optical high-speed cameras can capture streams of melt pool images of laser-powder interaction in real-time during the process. Model-agnostic deep learning methods offer a great deal of flexibility when facing such unstructured big data and thus are appealing alternatives to their physical-related and regression-based modeling counterparts. A configuration of Convolutional Long-Short Term Memory (ConvLSTM) auto-encoder is proposed to learn a deep spatio-temporal representation from sequences of melt pool images collected from experimental builds. The unfolded bottleneck tensors are then further mined to construct a high accuracy and low false alarm rate anomaly detection and monitoring procedure.

In this thesis, the applications of deep learning in the analysis, detection and classification of medical imaging datasets were studied, with a focus on datasets having a limited sample size. A combined machine learning-deep learning model was designed to classify one small dataset, prostate cancer provided by Mayo Clinic, Arizona. Deep learning model was implemented to extract imaging features followed by machine learning classifier for prostate cancer diagnosis. The results were compared against models trained on texture-based features, namely gray level co-occurrence matrix (GLCM) and Gabor. Some of the challenges of performing diagnosis on medical imaging datasets with limited sample sizes, have been identified. Lastly, a set of future works have been proposed. Keywords: Deep learning, radiology, transfer learning, convolutional neural network.

High-dimensional data is omnipresent in modern industrial systems. An imaging sensor in a manufacturing plant a can take images of millions of pixels or a sensor may collect months of data at very granular time steps. Dimensionality reduction techniques are commonly used for dealing with such data. In addition, outliers typically exist in such data, which may be of direct or indirect interest given the nature of the problem that is being solved. Current research does not address the interdependent nature of dimensionality reduction and outliers. Some works ignore the existence of outliers altogether—which discredits the robustness of these methods in real life—while others provide suboptimal, often band-aid solutions. In this dissertation, I propose novel methods to achieve outlier-awareness in various dimensionality reduction methods. The problem is considered from many different angles depend- ing on the dimensionality reduction technique used (e.g., deep autoencoder, tensors), the nature of the application (e.g., manufacturing, transportation) and the outlier structure (e.g., sparse point anomalies, novelties).