Denoising of DKI images: effect on feasibility and accuracy of kurtosis parameters

Publication Type:

Conference Abstract


ESMRMB, European Society for Magnetic Resonance in Medicine and Biology, 28th Annual Scientific Meeting, Leipzig, Germany, p.236 (2011)


denoising, diffusion kurtosis imaging, maximum likelihood estimation


Diffusion Kurtosis Imaging (DKI) is a new MRI technique, which quantifies non-Gaussianity of water diffusion. DKI requires high b-values, resulting in a low SNR of acquired diffusion weighted images (DWIs). Generally, kurtosis tensors are estimated from these DWIs using Weighted Least Squares (WLS) or Maximum Likelihood (ML) methods. In this work, we apply the Restricted Local Maximum Likelihood (RLML) denoising algorithm to remove noise from DWIs prior to WLS estimation. Then, we compare the RLML+WLS approach to the WLS and ML methods in terms of feasibility, that is the percentage of estimated tensors that satisfy certain DKI model constraints; and in terms of accuracy of calculated values of Mean Kurtosis.