Compressed Sensing in MRI playing around

A project that I taken at the University of Queensland. We proposed a new method to apply Compressed Sensing (CS) in Magnetic Resonance Imaging (MRI). Basically, the novel method we proposed has the similar idea as original CS but the algorithm to reconstruct the image is modified. Our algorithm finds the compressible image closest to the (under)sample in k-space data with the constraints are set in image-domain instead of in k-space domain as it was set in original CS. This is the demonstration video to show the reconstruction procedure with a pretty T2-weighted image which can be downloaded from http://www.eecs.berkeley.edu/~mlustig/CS.html. Enjoy a little messy video with a nice background music.