Meeting Banner
Abstract #1230

Motion-Guided Temporally-Constrained Compressed Sensing for Dynamic MRI

Xiao Chen1, Michael Salerno2, 3, Patrick F. Antkowiak1, Frederick H. Epstein1, 2

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

Many MR image series present temporal sparsity, in which image signal intensity changes smoothly through time, and such images are inherently suitable for acceleration using compressed sensing (CS) reconstruction. However, object motion between images violates temporal smoothness constraints and significantly degrades the quality of the CS-reconstructed images. To overcome this problem, we propose a general motion-guided CS algorithm which tracks object motion and guides the CS sparsity transform along the direction of motion. Improved image quality was observed using the proposed algorithm in dynamic contrast-enhanced images compared to non-motion guided reconstruction at an acceleration rate of 4.

Abstract PDF Poster PDF

Keywords

acceleration accurately acquisitions affine another apply applying arrow arrows background biomedical capable cardiac cardiology cine closely column columns commonly compressed compute computed consecutive constrained constraint contrast correct corrected correction correctly correlation cost cross dashed degrades describes details developed domain dynamic early edges engineering enhanced evaluation expertise exploits fast feasible fidelity finer flow fourth frame frequency fully function funded general generalized global gradient guidance guided highlighted illustrated importantly improve improved includes incorrect increasing iteration iterations iteratively measured medical minimizing model motion movement nonlinear object obscured occurs operator operators optical optimization optimize parallel particular pass pattern perfusion presence preserving previously propose proposed providing quality quite radiology randomly rather reconstruct reconstructed reconstructing reconstruction regional registered registration require required resemble resolution respectively respiratory retrospective rigid sampled sampling sensing series several sharper significantly smoothly smoothness solid space sparsity spatial spline straight substantial systematic temporal temporally term thank theory third tracking tracks transform types unregistered updated violates