A Simple Retrospective Noise Correction for Diffusional Kurtosis Imaging

Russell Glenn¹, Ali Tabesh¹, Jens H. Jensen¹

Diffusion MRI (dMRI) measurements are positively biased by noise from use of magnitude reconstructed images. The effects of the noise bias increase with decreasing signal-to-noise ratio (SNR), which can be problematic in high resolution dMRI acquisitions. A simple, retrospective noise correction technique is described and a weighted linear least squares fitting algorithm is presented for diffusional kurtosis imaging (DKI). Noisy phantom data is analyzed in DKI datasets with variable SNR, and the results of the noise correction are compared to uncorrected and reference data. Noise correction substantially reduces the bias in kurtosis estimates.