Congrats to Shawn for his successful depth exam
April 19, 2012
Congratulations to Shawn for his successful depth exam today! Further details below.
Title: UNCERTAINTY IN MEDICAL IMAGE SEGMENTATION: SOURCES, QUANTIFICATION, AND IMPORTNACE
Abstract: Segmentation is a key step in most medical image analysis tasks. Unfortunately, manual expert segmentations are time consuming to produce and still suffer from some variability, while fully automated segmentation methods do not achieve the accuracy required in sensitive clinical settings. The goal of automated segmentation methods should be to assist an expert user by accepting input from them and by conveying useful information to them. In particular, information regarding uncertainty in a segmentation result is helpful in a variety of ways. In this report, we discuss how uncertainty arises in medical image segmentation through limitations in the image acquisition tools and through probabilistic models used in the automated segmentation process. We describe how uncertainty quantification can be leveraged in both clincal and research applications. Finally, we review many popular segmentation frameworks and compare the various ways these frameworks quantify and encode uncertainty.
Ph.D. Depth Examining Committee:
Dr. Ghassan Hamarneh, Sr. Supervisor
Dr. Greg Mori, Supervisor
Dr. Torsten Moller, Examiner