Dynamic perfusion magnetic resonance (MR) imaging is a commonly used imaging technique that allows to measure the tissue perfusion in an organ of interest via assessment of various hemodynamic parameters such as blood flow, blood volume, and mean transit time. In this paper, we tackle the problem of recovering perfusion MR images from undersampled k-space data. We propose a novel reconstruction model that jointly penalizes spatial (local) incoherence on temporal differences obtained based on a reference image and the patch-wise (nonlocal) dissimilarities between spatio-temporal neighborhoods of MR sequence. Furthermore, we introduce an efficient iterative algorithm based on a proximal-splitting scheme that solves the joint minimization problem with fast convergence. We evaluate our method on dynamic susceptibility contrast (DSC)-MRI brain perfusion datasets as well as on a publicly available dataset of in-vivo breath-hold cardiac perfusion. Our proposed method demonstrates superior reconstruction performance over the state-of-the-art methods and yields highly accurate estimation of perfusion time profiles, which is very essential for the precise quantification of clinically relevant perfusion parameters.     «

Dynamic perfusion magnetic resonance (MR) imaging is a commonly used imaging technique that allows to measure the tissue perfusion in an organ of interest via assessment of various hemodynamic parameters such as blood flow, blood volume, and mean transit time. In this paper, we tackle the problem of recovering perfusion MR images from undersampled k-space data. We propose a novel reconstruction model that jointly penalizes spatial (local) incoherence on temporal differences obtained based on a r...     »