Effects of denoising on diffusion-weighted MRS data

Diffusion-weighted (DW) MRS is a useful tool for detecting microstructural changes linked to neurological diseases. However, DW-MRS suffers from low signal-to-noise ratio, which makes human application extremely difficult. To overcome this issue, a denoising algorithm based on low-rank approximation was recently utilized. Here, effects of the application of low-rank based denoising algorithm to DW-MRS data are described based on synthetic data. Severely artifactually low group SDs for the apparent diffusion coefficients (ADCs) of all metabolites and biased mean ADC values were observed after denoising. These findings suggest that low-rank approximations are detrimental to a reliable quantification of ADCs of metabolites.