Cornell CS Parallel Imaging Project


Since its introduction in the late 90's, parallel imaging has proved to be an extremely powerful method for accelerating MR imaging. Current techniques, such as SENSE, GRAPPA, etc. rely on a least squares approach to the problem. In this research project we are addressing two major technical shortcomings of these approaches. They do not take advantage of the piecewise spatial smoothness of images They assume that the system matrix is error-free. Our research focuses on the use of graph cuts to address the first shortcoming, and a TLS-inspired technique to address the second.

People	Ramin Zabih, Ph.D Ashish Raj, Ph.D Bryan Kressler Gurmeet Singh Yi Wang, Ph.D.

Papers	MRF's for MRI's: Bayesian Reconstruction of MR Images via Graph Cuts. Ashish Raj, Gurmeet Singh and Ramin Zabih. To appear CVPR 2006. Bayesian Parallel Imaging with Edge-Preserving Priors. Ashish Raj, Gurmeet Singh, Ramin Zabih, Bryan Kressler, Yi Wang, Norbert Schuff, Mike Weiner. Submitted to Magnetic Resonance in Medicine, 2005. A Graph Cut Algorithm for Generalized Image Deconvolution. Ashish Raj and Ramin Zabih. In: International Conference on Computer Vision, 2005. A Maximum Likelihood approach to Parallel Imaging with Coil Sensitivity Noise. Ashish Raj, Yi Wang, Ramin Zabih. Submitted to IEEE Transactions on Medical Imaging. Improvements in Magnetic Resonance Imaging using Information Redundancy. Ashish Raj. PhD thesis, Cornell University, May 2005.

Talks	MRF's for MRI's: Bayesian Reconstruction of MR Images via Graph Cuts (Cornell AI Seminar, October 21, 2005)