Meeting Banner
Abstract #4198

Local Temporal Point Spread Function for CS Reconstructions Exploiting X-F-Sparsity

Tobias Wech1, Daniel Stb1, Andre Fischer1, Dietbert Hahn1, Herbert Kstler1

1Institute of Radiology, University of Wrzburg, Wrzburg, Germany

Compressed Sensing reconstructions exploiting spatio-temporal sparsity have successfully been applied to accelerate dynamic MRI. However, the non-linear and non-stationary CS algorithms prohibit the straightforward evaluation of the temporal resolution through a single global temporal point spread function. In this work a pixelwise perturbation strategy was utilized to assess local temporal point spread functions for every image pixel. The method therefore allows an appropriate assessment of the temporal resolution and can thus improve the choice of sampling patterns and algorithm settings.

Keywords

accelerated acceleration according acquisition additionally adjacent affected affects agent amount amplitude amplitudes analogically applied arrows artifacts beating behavior blurring broadening cardiac choice cine clear color compressed concept correlation corresponds created dataset degradation density depending depicted described detectable determine determined determining discussed domain double dynamic elite enable enables encoded especially every examination examine examined exhibit exploiting features frames frequency fully function functional funded hardly healthy heart identical improvement in vivo indicating induced influence institute introduced investigated iterative lateral linear lobe local location locations loss maps matrix measure mere minor model motion myocardium neither network normalized objective parts pattern patterns perfusion perturbation perturbations pixel preserved previously prior procedure proved radiology random recently reconstructed reconstruction reconstructions recovering recovery regularization repeating representing resolution respective retrospectively sampled sampling saturation sensing series since slice soft space sparse sparsity spatial spread stationary straightforward subtracting successfully suffer systolic temporal thresholding throughout transform trio underestimated usually utilized variation visible volunteers wall width