Unsupervised Denoising of Diffusion-Weighted Images with Bias and Variance Corrected Noise Modeling

2026-02-21

AI Summary

Diffusion MRI (dMRI) at high diffusion weighting suffers from low signal-to-noise ratio and non-Gaussian Rician noise, leading to degraded image quality and biased downstream analyses. Existing unsupervised denoising methods inadequately model the bias and heteroscedasticity inherent in Rician noise. This work proposes a Rician statistics–based unsupervised denoising approach that integrates first- and second-moment correction losses into the Deep Image Prior framework to explicitly remove mean and squared-signal biases, respectively, while employing an adaptive weighting strategy to account for variance heterogeneity. Notably, this method is the first to explicitly model Rician noise characteristics in unsupervised dMRI denoising without requiring architectural modifications to the underlying network. Experiments on both simulated and real data demonstrate substantial improvements over current state-of-the-art methods, effectively suppressing noise-induced bias and enhancing both image fidelity and the reliability of derived diffusion metrics.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Unsupervised & Self-Supervised Learning
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Diffusion magnetic resonance imaging (dMRI) plays a vital role in both clinical diagnostics and neuroscience research. However, its inherently low signal-to-noise ratio (SNR), especially under high diffusion weighting, significantly degrades image quality and impairs downstream analysis. Recent self-supervised and unsupervised denoising methods offer a practical solution by enhancing image quality without requiring clean references. However, most of these methods do not explicitly account for the non-Gaussian noise characteristics commonly present in dMRI magnitude data during the supervised learning process, potentially leading to systematic bias and heteroscedastic variance, particularly under low-SNR conditions. To overcome this limitation, we introduce noise-corrected training objectives that explicitly model Rician statistics. Specifically, we propose two alternative loss functions: one derived from the first-order moment to remove mean bias, and another from the second-order moment to correct squared-signal bias. Both losses include adaptive weighting to account for variance heterogeneity and can be used without changing the network architecture. These objectives are instantiated in an image-specific, unsupervised Deep Image Prior (DIP) framework. Comprehensive experiments on simulated and in-vivo dMRI show that the proposed losses effectively reduce Rician bias and suppress noise fluctuations, yielding higher image quality and more reliable diffusion metrics than state-of-the-art denoising baselines. These results underscore the importance of bias- and variance-aware noise modeling for robust dMRI analysis under low-SNR conditions.

Problem

• diffusion MRI
• denoising
• Rician noise
• bias correction
• variance heterogeneity

Innovation

• Rician noise modeling
• unsupervised denoising
• bias correction
• variance heterogeneity
• Deep Image Prior