Motivation: Gaussian noise assumptions in commonly-used MPPCA dMRI denoising methods are violated due to accelerating acquisitions, combining coil signals, and using magnitude data, affecting denoising performance. Goal(s): To develop single-channel (i.e. closer to source/assumptions) complex denoising and compare performance against channel-combined denoising on noise variance and noise-floor suppression. Approach: We devised an offline reconstruction approach that incorporates denoising prior to channel combination and handles partial k-space data. We compared performance and the interaction of different reconstruction (single-channel/channel-combined, zero-fill/homodyne) with denoising (MPPCA/NORDIC) methods. Results: We found increased noise-floor suppression and signal dynamic range when denoising single-channel complex data compared to channel-combined data across reconstruction/denoising methods. Impact: Improved denoising from single-channel complex data highlights the importance of having access to data early in reconstruction. Less susceptibility of denoising performance to reconstruction choices, as observed for some methods, is important to consider when harmonising dMRI across scanners.
D’Antonio et al. (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: