Estimating Effective Residue Functions in Model-Free ASL: Comparison of Deconvolution Algorithms

Andr Ahlgren¹, Ronnie Wirestam¹, Freddy Sthlberg¹, ², Linda Knutsson¹

To improve estimated residue function characteristics and cerebral blood flow (CBF) quantification in arterial spin labeling (ASL), we recently proposed the implementation of nonlinear stochastic regularization (NSR) in model-free ASL. Here we present quantitative results from an improved implementation and compare it to the more common singular value decomposition (SVD) methods. SVD based methods yielded unrealistic residue functions and reasonable (cSVD) or low (oSVD) CBF estimates. NSR yielded realistic residue functions and robust perfusion maps but somewhat low CBF values.