The use of computers in medical image analysis has seen tremendous growth following the development of imaging technologies that can capture image data in-vivo as well as ex-vivo. While the field of radiology has adopted computer aided image analysis in research as well as clinical settings, the use of similar techniques in histopathology is still in a nascent stage. The current gold standard in diagnosis involves labor-intensive tasks such as cell counting and quantification for disease diagnosis and characterization. This process can be subjective and affected by human factors suach as reader bias and fatigue. Computer based tools such as digital image analysis have the potential to help alleviate some of these problems while also offering insights that may not be readily apparent when viewing glass slides under an optical microscope. Commercially available high-resolution slide scanners now make it possible to obtain images of whole slides scanned at 40x microscope resolution. Additionally, advanced tools for scanning tissue images at 100x resolution are also available. Such scanning tools have led to a large amount of research focused on the development of image analysis techniques for histopathological images. While the availability of high-resolution image data presents innumerable research opportunities, it also leads to several challenges that must be addressed.
This dissertation explores some of the challenges associated with computer-aided analysis of histopathological images. Specifically, we develop a number of tools for Follicular Lymphoma and Lupus. We aim to develop algorithms for detection of salient features in tissue biopsies of follicular lymphoma tumors. We analyze the algorithms from a computational point of view and develop techniques for processing whole slide images efficiently using high performance computing resources. In the application of image analysis for Lupus, we analyze mouse renal biopsies for characterizing the distribution of infiltrates in tissue as well as develop algorithms for identification of tissue components such as the glomeruli, which play a significant role in the diagnosis of the disease. Finally, we explore the development of a web-based system for dissemination of high-resolution images of tissues with the goal of advancing collaboration, research and teaching. Through the use of web technologies and developments in the field of geospatial imaging, we demonstrate the efficacy of an online tissue repository that can enable pathologists, medical students and all researchers to explore these images as well as use high performance computing systems to leverage computer-aided diagnosis tools in their field.