In this work, we analyze the ability of an automated deep learning-based model to identify the three-dimensional spatial extent of pleural mesothelioma (PM) as presented on computed tomography (CT) scans, employ machine learning to classify an image-based biomarker for PM, and evaluate another model's generalizability in the task of classifying COVID-19 based on patients' chest radiographs (CXRs). PM is an aggressive form of cancer present in the pleural lining of the lung. It is usually the result of exposure to asbestos and has a very poor prognosis. Linear measurements are the clinical standard used in evaluating tumor response to therapy, but these measurements are only a surrogate for tumor volume. Tumor volume must be calculated to assess tumor burden completely and quantitatively. Determining the volume of tumor, however, is complicated and time consuming, since discerning PM tumor on a medical image is a challenge for human and computer observers alike due to its complex and irregular morphology. Tumor profiling and genetic testing can also be performed to identify the mutation status of the BRCA1-associated protein-1 (BAP1) gene, a prognostic factor in PM that can directly impact treatment options. Classifying the mutation status using machine learning can be informative as genetic testing is not current standard of care. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), an RNA virus that can impact mammals and birds, is the virus responsible for the COVID-19 global pandemic. The primary mode of transmission among humans is through exposure to respiratory fluids carrying infectious virus. Before widespread use of reverse transcription polymerase chain reaction (RT-PCR) tests, CXRs were recommended for triage, disease monitoring, and assessment of concomitant lung abnormalities. In recent years, there has been a substantial increase in the application of artificial intelligence techniques to medical imaging. Convolutional neural networks (CNNs), specifically, have been successfully employed for various objectives performed on medical images. CNNs are capable of learning both local and global patterns of an image, which is essential to identify nuanced disease presentations. The specific aims of this work were: (1) to study a deep learning model for the automatic segmentation of tumor volumetry in PM, (2) to investigate image texture analysis for differentiation of BAP1 mutation status, and (3) to evaluate a deep learning model for the generalizable classification of COVID-19. Aim 1 fully investigated the performance of the deep learning model used for PM segmentation, which better informed us of the generated outputs by the model. Aim 2 performed texture feature analysis to determine the somatic BAP1 mutation status based on the segmented region on a CT scan. Aim 3 evaluated the performance of a deep learning model in the task of COVID-19 classification in order to address and develop methods to achieve model generalizability. These results provided a novel pipeline with potential impact on the future treatment of patients presenting with mesothelioma or COVID-19, expediting many of the sequential steps a patient must undergo and improving the individualized prognostication process, which can also be implemented for other cancers and lung abnormalities.