Accurate Dynamic Susceptibility Contrast MR Perfusion Quantification using Spatiotemporal Noise Filtering Algorithms
Kosior J, Frayne R
University of Calgary, Foothills Medical Centre, Calgary Health Region
Quantitative cerebral blood flow (CBF) values obtained from dynamic susceptibility contrast MR perfusion data are subject to variation due to non-stationary noise artifacts, which cause deconvolution algorithms to become unstable. The purpose of this study was to investigate 4D noise filtering algorithms (space and time) to improve the accuracy of CBF estimates. We compared 4D-Gaussian, 4D-anisotropic diffusion and 4D-bilateral filtering algorithms using a novel digital brain perfusion phantom and a stroke patient. The bilateral filter provided the most accurate CBF estimates in heterogeneous tissue regions and performed comparably well to Gaussian filtering in homogenous tissue regions.