Meeting Banner
Abstract #4555

Block LOw-Rank Sparsity with Motion Guidance (BLOSM) for Accelerated Dynamic MRI

Xiao Chen1, Michael Salerno2, 3, Craig H. Meyer1, Frederick H. Epstein1

1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States; 2Medicine, University of Virginia, Charlottesville, VA, United States; 3Radiology, University of Virginia, Charlottesville, VA, United States

Several accelerated imaging techniques utilizing k-t undersampling have been proposed to model dynamic CMR behavior with a few spatiotemporal basis functions to reconstruct images. These algorithms are sensitive to respiratory motion and perform poorly when both signal intensity and object position and shape change during image acquisition. We propose a novel method that divides the images into blocks and tracks the blocks motions to exploit increased sparsity (Block LOw-rank Sparsity with Motion guidance). The simplified dynamics in the smaller, motion-compensated blocks can be better described by a limited number of basis functions, making the method insensitive to complex dynamics.

Abstract PDF Presentation Video

Keywords

accelerated acceleration accurately achieved acquisition adopted affine applications applied association avoid award background basis beginning behavior biomedical block blocking blocks cardiac chosen cluster coarse compensated complex concepts consist contents contributed correlation correlations datasets decomposition decreased degraded described displacement divided divides dynamic dynamics engineering error exploit exploits feature finishing form frame fully functions funded gaps gathered greatly guidance guided handle happened heart human improved improvement initiated insensitive intensity interpolation iterates iteration iterative leads limited making many maps matrix medicine model motion next norm note novel object objects occurred others outperformed overlapped patient patterns perfusion pixels poorly position predoctoral problem prominent proposed quality quantitative radiology rank rearranged reconstruct reconstructed reduced refined regional registration related represents respiratory retrospectively rigid sampled several severe severely shape significantly simplified since singular smaller smoothing soft solutions solve sparsity spatial specifically square strategy structure substantial substantially suffered takes temporally thresholding together track tracked tracking tracks trajectory utilizing variations whole