Recent Computational Advances in Denoising for Magnetic Resonance Diffusional Kurtosis Imaging (DKI)

Abstract

Magnetic resonance imaging (MRI) is widely used in clinical practice and medical research for the assessment of disease. Magnetic resonance diffusional kurtosis imaging (DKI) is a specific MRI technique that is useful for quantifying microstructural properties of biological tissues, particularly in brain. However, images derived with DKI can be sensitive to noise, as the MRI sequences needed for DKI strongly attenuate the signal. To mitigate this inherent noise sensitivity of DKI, advanced denoising methods maybe applied. Although a variety of denoising approaches have been considered in the broad context of MRI, the specific performance of these methods for DKI has not yet been thoroughly investigated. In this review, we examine three different denoising strategies for DKI—Gaussian filtering, non-local means filtering, and a local principal components analysis technique. These three denoising methods are compared qualitatively in terms of their abilities to increase image fidelity and to remove noise bias for the DKI-derived parametric maps.