Implicit Neural Representations for Accurate Estimation of the Standard Model of White Matter

Published in Communications Biology (2026)

DOI: https://doi.org/10.1038/s42003-025-09399-5

Diffusion magnetic resonance imaging (dMRI) enables non-invasive investigation of tissue microstructure. The Standard Model (SM) of white matter aims to disentangle dMRI signal contributions from intra- and extra-axonal water compartments. However, estimating its parameters remains challenging due to the model’s high dimensionality and noise sensitivity.

This work introduces an estimation framework based on implicit neural representations (INRs), which incorporate spatial regularization through sinusoidal encoding of spatial coordinates. The method is self-supervised and does not require labeled training data.

Evaluations on synthetic and in vivo datasets show that the INR approach outperforms existing methods, particularly in low signal-to-noise conditions. The framework also enables anatomically plausible spatial upsampling and supports joint estimation of model parameters and fiber orientation distributions.

Additional advantages include fast inference, robustness to noise, and compatibility with advanced diffusion MRI modelling techniques such as spherical harmonics up to high orders and gradient non-uniformity corrections.