UW Madison UW School of Medicine and Public Health
UW Biiostatistics & Medical Informatics UW Biostatistics & Medical Informatics UW Madison UW Biostatistics & Medical Informatics Site Map


General Departmental Seminar Series

How To Solve Challenging Bayesian Problems in MR? A: Graph Cuts!

Ashish Raj
University of California San Francisco

Monday, May 19th, 2008
11:00 pm
Room 3310 CS

Everyone agrees that Bayesian methods (math-ese for "using priors") are desirable for many MR/CT problems. Examples include segmentation, tractography, reconstruction, super-resolution, distortion correction, ...,etc. A really powerful class of priors are expressed in terms of Markov Random Fields. MRFs sound scary, but are merely a tool for imposing spatial coherence and edge preservation in images. They have a huge fan following in statistics and vision circles, and in this talk I will attempt to convince you that they are great for MR too.

Unfortunately even those who love MRFs agree that they are fiendishly difficult to actually solve (due to non-convexity), sometimes taking hours even days to get anywhere. I will present a new algorithm borrowed from graph theory, called Graph Cuts, which can efficiently and speedily solve Bayesian problems in MR involving MRFs. In particular I will show our latest results for MR parallel imaging. While not quite lightening fast, processing takes minutes rather than days.

Joint work with Ramin Zabih and Gurmeet Singh (Cornell CS and Radiology).

Bio: Dr. Raj earned a Bachelors in Electrical Engineering from the University of Auckland, New Zealand, and a Ph.D. in Electrical and Computer Engineering from Cornell University, NY, USA. At Cornell he specialized in signal processing and wrote his dissertation on new computational algorithms for MR imaging, working closely with the department of radiology at Weill-Cornell Medical College in New York. This work led to the development of new techniques for reconstructing and processing MR images using prior or redundant information. After graduating in December 2004, he joined the Centre for Imaging of Neurodegenerative Diseases and the Department of Radiology at University of California San Francisco.

Return to seminar list