Doctoral Defense - Jian Mu
July 23, 2014, 11:30 am , 117-I Cushing Hall
Advisor: Dr. Danny Chen
Dr. Fang Liu Dr. Tijana Milenkovic Dr. Tim Weninger
New Methods for Biomedical Image Processing and Quantitative Analysis,
with Applications in Computer Vision
Image processing techniques have a wide range of applications in the processing and analysis of biomedical images for clinical and research purposes. They allow the fast and accurate processing of large quantities of biomedical images, and are extensively used by biologists and physicians to analyze experimental and clinical images, in order to get accurate diagnosis and develop effective therapeutic strategies to better treat patients. However, such image analysis tasks are usually quite challenging, due to the existence of various artifacts in images, such as image fuzzyness caused by movement of animals or diffusion of coloring substance; or the high complexity of the target objects to be analyzed, such as contours of multiple pieces of overlapping bones of the knee joint. In this dissertation, we present algorithms for the processing and quantitative analysis of several types of biomedical images. In several problems, we identify or segment target objects from biomedical images (e.g., blood clots in3-D microscopy images), quantitatively compute their features, propose hypotheses for biomedical mechanisms that are being studied by biologists based on our analysis, and study differences in the computed features in images from different study groups. Our analyses produce results that are of biomedical significance, and can assist biologists and physicians in the diagnosis and treatment of diseases. We also apply some techniques that we use to process biomedical images, such as the graph search algorithm, along with other graph algorithms, to solve problems in computer vision.