Improved Assessment of Regional Cerebral Blood Flow by Dynamic Susceptibility Contrast MRI Using a Kernel-based Deconvolution Approach

INTRODUCTION: In dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI), in order to estimate the Cerebral Blood Flow (CBF), a deconvolution operation between the arterial input function (AIF) and the voxel concentration of contrast agent (C(t)) must be performed to obtain the residue function (R(t)) [1]. Here, we propose a kernel based deconvolution approach that tackles the problem in a fully Bayesian framework and includes information on both R(t) continuity and on the system BIBO stability. The algorithm proposed by De Nicolao & Pillonetto (DNP) was validated on both simulated and clinical data and compared to the most common deconvolution techniques in DSC-MRI quantification, i.e. the singular value decomposition (SVD) [1] and the block-circulant SVD (cSVD) [2].