Meeting Banner
Abstract #2174

Arterial Spin Labeling (ASL) Denoising with Markov Random Field (MRF) Optimization

Yongsheng Zhang1, Ze Wang1

1Department of Psychiatry, University of Pennsylvania, Philadelphia, PA, United States

ASL MRI has an intrinsic low signal-to-noise-ratio, requiring an efficient denoising method. Spatial smoothing is efficient for suppressing random noise but at the expense of a resolution loss. We proposed a resolution preserving ASL denoising method based on the Markov random field theory. Our evaluation results showed that the proposed method can gain up to 4 fold SNR increase while still keep the structural details intact. By contrast, spatial smoothing induced severe blurring effects when a similar level of SNR increase was achieved.

Keywords

adaptive adjusted adopted advances alternative anisotropic applied arterial averaging axial background basic blood blurred blurring blurs boundaries boundary brain cerebral clearly clinical coherence component continuous contrast control correlated database delay designed details develop deviation direct distributed evaluated even evenly existing field filtering fissure flow full function future gradient graph gray great greatly half homogeneous improvement included inevitably input intrinsic introduces inversely isotropic kernel label labeling local localized longitudinal make maps mater matrix measure measured mechanism medial model modeled needed neighborhood neighboring neurology noise noninvasive optimization output pairs perceivable perfusion post potential preserve preserved preserving principal procedure procedures process processed promising property proportional proposed pseudo psychiatry quantification random rather recode reduced regional regionally regular representative represents resolution respectively resultant routine separately severe similarly slice smaller smoothing smoothness spatial spatially specified spin strategies structure studies subject subjects successfully supported suppress tagged taking theory tissue undirected wavelet white widely width words worsened yield yielded