Diffusion MRI Transformer with a Diffusion Space Rotary Positional Embedding (D-RoPE)

arXiv cs.CV / 3/30/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes a dMRI-specific transformer framework that can jointly model spatial structure, diffusion-weighting, and direction-dependent characteristics of diffusion MRI signals.
It introduces a diffusion space rotary positional embedding (D-RoPE) to better encode diffusion-direction information and to remain robust when acquisition protocols vary (e.g., different numbers of diffusion directions).
The method uses self-supervised masked autoencoding pretraining, then evaluates the learned representations on multiple downstream tasks.
Reported results indicate competitive or superior performance versus several baselines, including a reported 6% accuracy improvement for classifying mild cognitive impairment and a 0.05 increase in correlation for predicting cognitive scores.
The authors provide code via a public GitHub repository, supporting reproducibility and potential adoption for dMRI representation learning.

Abstract

Diffusion Magnetic Resonance Imaging (dMRI) plays a critical role in studying microstructural changes in the brain. It is, therefore, widely used in clinical practice; yet progress in learning general-purpose representations from dMRI has been limited. A key challenge is that existing deep learning approaches are not well-suited to capture the unique properties of diffusion signals. Brain dMRI is normally composed of several brain volumes, each with different attenuation characteristics dependent on the direction and strength of the diffusion-sensitized gradients. Thus, there is a need to jointly model spatial, diffusion-weighting, and directional dependencies in dMRI. Furthermore, varying acquisition protocols (e.g., differing numbers of directions) further limit traditional models. To address these gaps, we introduce a diffusion space rotatory positional embedding (D-RoPE) plugged into our dMRI transformer to capture both the spatial structure and directional characteristics of diffusion data, enabling robust and transferable representations across diverse acquisition settings and an arbitrary number of diffusion directions. After self-supervised masked autoencoding pretraining, tests on several downstream tasks show that the learned representations and the pretrained model can provide competitive or superior performance compared to several baselines in these downstream tasks (even compared to a fully trained baseline); the finetuned features from our pretrained encoder resulted in a 6% higher accuracy in classifying mild cognitive impairment and a 0.05 increase in the correlation coefficient when predicting cognitive scores. Code is available at: github.com/gustavochau/D-RoPE.