Meeting Banner
Abstract #2242

Compressed Sensing Using Prior Rank, Intensity and Sparsity Model (PRISM): Applications in Cardiac Cine MRI

Hao Gao1, 2, Stanislas Rapacchi3, Da Wang3, John Moriarty3, Conor Meehan3, James Sayre3, Gerhard Laub4, Paul Finn3, Peng Hu3

1Mathematics, UCLA, Los Angeles, CA, United States; 2Mathematics, UCI, Irvine, CA, United States; 3Radiology, UCLA, Los Angeles, CA, United States; 4Siemens

We propose a novel CS method for dynamic MRI applications using Prior Rank, Intensity and Sparsity Model (PRISM) and evaluate this technique for cardiac cine MRI. PRISM differs from the previous CS techniques in the ability to apply the sparsifying transform (TF) after background suppression using rank minimization. PRISM was tested on cardiac cine MRI data sets acquired on 6 healthy subjects. The data was fully sampled and retrospectively undersampled. Results show dynamic 2D MRI could be greatly accelerated using PRISM. PRISM provides good-quality image series even from highly undersampled kspace data when state-of-the art traditional compressed sensing fails.

Abstract PDF Poster PDF

Keywords

ability abstract accelerated acceleration accuracy acquiring acquisition acquisitions additional although anatomy angiography another anterior applications applied apply approaches artifacts automatically avoid background blinded blood blurring captures cardiac changing cine coherent coils column combination commonly comparable complements complete component compressed considered considering contrast decomposed decrease denotes diastolic dimensions dividing drawn dynamic eddy enable evaluators even fails fair fidelity formulated frame frames full fully good greater healthy heartbeats hence highly impact important improve incoherent incoherently intensity iteratively john little mathematics matrix measuring minimization minimizing model myocardium norm novel nuclear optimization overall paired pattern pixel poor potential prior prism prob problem propose proposed providing quality quoted radiology random randomly rank recently reconstruct reconstructed reconstruction reconstructions recurring reduce represented required residual resp retrospectively reviewers sampled sampling scored scores sensing series sets sharp sharpness significantly simulated singular solved space sparsity started step structure subjects suppression temporal term theory tight traditional transform transformed true underlying undesirable vector