Q-Space Undersampled Diffusional Kurtosis Imaging

Ali Tabesh¹, Jens H. Jensen¹, Els Fieremans¹, Joseph A. Helpern^1,2

We assess the feasibility of reconstructing DKI parametric maps from q-space undersampled brain DKI scans. We present two constrained tensor estimation methods that extend our previous convex quadratic programming-based formulation, and enable estimation of DKI maps from undersampled DKI acquisitions. We compare the performance of these methods on highly undersampled scans. Our results indicate that mean and radial kurtosis maps can be estimated from undersampled DKI scans with acceptable fidelity. As a consequence, useful DKI maps are obtainable at 3T from scans as short as 90 seconds.