Restoring Marčenko-Pastur noise estimation and denoising of diffusion MRI under common image reconstruction
Year: 2026
Type: preprint
Abstract: Diffusion magnetic resonance imaging can reveal microstructure in the living brain, but the measurements are inherently noisy. Marčenko–Pastur denoising is a widely used remedy, helping to separate signal from noise under a simplified noise covariance model. However, we find that the model breaks under common image reconstructions across 28 public in vivo datasets, resulting in weaker denoising than intended. We restore the intended denoising behavior with a noise covariance model estimated from the reconstructed images, and validate the estimated model on repeated ex vivo and in vivo scans. The method requires no additional measurements or knowledge of the reconstruction, opening headroom for microstructure analyses in both existing and newly acquired data.
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7234402166
