Adaptive sampling design for compressed sensing MRI

Compressed Sensing (CS) takes advantage of the sparsity of MR images in certain bases or dictionaries to obtain accurate reconstructions from undersampled k-space data. The (pseudo) random sampling schemes used most often for CS may have good theoretical asymptotic properties; however, with limited data they may be far from optimal. In this paper, we propose a novel framework for improved adaptive sampling schemes for highly undersampled CS MRI. While the proposed framework is general, we apply it with a recently proposed MRI reconstruction algorithm employing adaptive image-patch based sparsifying dictionaries. Numerical experiments demonstrate up to 7 dB improvements in reconstruction PSNR using the adapted sampling scheme, on top of the large improvements reported in our previous work for the adaptive patch-based reconstruction scheme over analytical sparsifying transforms.