Classical accelerated dynamic MRI schemes rely on the sparsity or banded structure of the data in specied transform domains (eg. Fourier space). Clearly, the utility of these schemes depend on the specic data and the transform. For example, these methods only provide modest accelerations in free-breathing myocardial perfusion MRI. In this paper, we discuss a novel blind linear model to recover the data when the optimal transform is not known a-priori. Specically, we pose the simultaneous recovery of the optimal linear model/transform and its coecients from the measurements as a non-convex optimization problem. We also introduce an ecient majroize-minimize algorithm to minimize the cost function. We demonstrate the utility of the algorithm in considerably accelerating free breathing myocardial perfusion MRI data.

Date Published: 27 September 2011
PDF: 8 pages
Proc. SPIE 8138, Wavelets and Sparsity XIV, 81381V (27 September 2011); doi: 10.1117/12.893060

Published in SPIE Proceedings Vol. 8138:
Wavelets and Sparsity XIV
Manos Papadakis; Dimitri Van De Ville; Vivek K. Goyal, Editor(s)