[论文解读] Unsupervised Denoising of Diffusion-Weighted Images with Bias and Variance Corrected Noise Modeling

Jine Xie, Zhicheng Zhang+3 more|arXiv (Cornell University)|Feb 22, 2026

Advanced Neuroimaging Techniques and Applications被引用 0

一句话总结

作者在 Deep Image Prior 框架内提出两种无监督损失函数，显式对 diffusion MRI 中的瑞利噪声进行建模，校正偏差和异方差以改进去噪和扩散指标。

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.

研究动机与目标

• 在低信噪比条件下驱动扩散加权MRI 的去噪，其中瑞利偏差和方差显著
• 基于一阶和二阶矩统计的偏差-方差校正损失函数用于幅值 dMRI
• 在无监督的 Deep Image Prior 框架中验证所提出的损失函数，无需大规模训练数据集
• 在 simulated 与 in-vivo 数据中展示相较于现有基线方法的图像质量和扩散参数可靠性的提升

提出的方法

• 定义纳入自适应加权的噪声校正的一阶矩和二阶矩损失，以在体素层面归一化方差
• 以一阶矩及其方差在体素级加权的损失构造 DIP-M1-W1
• 以二阶矩及其方差在体素级加权的损失构造 DIP-M2-W2
• 在 DIP 框架内的 3D U-Net 上实现以去噪 4D dMRI 数据
• 与 MPPCA、Patch2Self、DDM2、Replace2Self（带瑞利偏倪校正）进行公平比较

实验结果

研究问题

• RQ1 在无监督的 DIP 框架中，瑞利偏差和幅度dMRI中的异方差噪声能被有效校正吗？
• RQ2 基于一阶和二阶矩的噪声校正损失是否比现有方法在去噪质量和扩散指标可靠性上有提升？
• RQ3 在非均匀、体素级变化噪声条件下，所提出的损失表现如何？
• RQ4 偏差-方差感知的损失是否能在不同分辨率下保留细微结构细节并使扩散指标更稳定？

主要发现

• DIP-M1-W1 和 DIP-M2-W2 相较基线在 simulated 数据中更有效地降低瑞利偏差并抑制噪声波动
• 所提出的损失在去噪图像上具有更高的 PSNR 和 SSIM，且 FA 与 MD 图在 RMSE 更低且 SSIM 更高
• 在非均匀噪声条件下，这些方法在图像质量和扩散指标精准度方面始终优于竞争对手
• in-vivo 结果显示在不同分辨率下的去噪鲁棒性更好，结构细节保留更好且扩散指标稳定性提升（FA 的变异系数更低）
• 该方法将偏差校正在去噪中整合，避免了多步骤基线方法中出现的误差积累现象