Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DDCSR: A Novel End-to-End Deep Learning Framework for Cortical Surface Reconstruction from Diffusion MRI

Chengjin Li, Yuqian Chen, Nir A. Sochen, Wei Zhang, Carl-Fredrik Westin, Rathi Yogesh, Lauren J. O'Donnell, Ofer Pasternak, Fan Zhang

TL;DR

The paper tackles the challenge of cortical surface reconstruction directly from diffusion MRI, removing reliance on anatomical T1-weighted data. It introduces DDCSR, a two-module framework where SDFNet predicts voxel-wise surface representations and DiffCoSeg converts these into 3D meshes, enabling end-to-end CSR from dMRI. The approach demonstrates superior accuracy and efficiency compared to T1w-based methods and traditional dMRI pipelines, and shows strong generalization across datasets with different acquisition protocols. This work has significant practical implications for rapid, scalable CSR in multi-site diffusion MRI studies.

Abstract

Diffusion MRI (dMRI) plays a crucial role in studying brain white matter connectivity. Cortical surface reconstruction (CSR), including the inner whiter matter (WM) and outer pial surfaces, is one of the key tasks in dMRI analyses such as fiber tractography and multimodal MRI analysis. Existing CSR methods rely on anatomical T1-weighted data and map them into the dMRI space through inter-modality registration. However, due to the low resolution and image distortions of dMRI data, inter-modality registration faces significant challenges. This work proposes a novel end-to-end learning framework, DDCSR, which for the first time enables CSR directly from dMRI data. DDCSR consists of two major components, including: (1) an implicit learning module to predict a voxel-wise intermediate surface representation, and (2) an explicit learning module to predict the 3D mesh surfaces. Compared to several baseline and advanced CSR methods, we show that the proposed DDCSR can largely increase both accuracy and efficiency. Furthermore, we demonstrate a high generalization ability of DDCSR to data from different sources, despite the differences in dMRI acquisitions and populations.

DDCSR: A Novel End-to-End Deep Learning Framework for Cortical Surface Reconstruction from Diffusion MRI

TL;DR

The paper tackles the challenge of cortical surface reconstruction directly from diffusion MRI, removing reliance on anatomical T1-weighted data. It introduces DDCSR, a two-module framework where SDFNet predicts voxel-wise surface representations and DiffCoSeg converts these into 3D meshes, enabling end-to-end CSR from dMRI. The approach demonstrates superior accuracy and efficiency compared to T1w-based methods and traditional dMRI pipelines, and shows strong generalization across datasets with different acquisition protocols. This work has significant practical implications for rapid, scalable CSR in multi-site diffusion MRI studies.

Abstract

Diffusion MRI (dMRI) plays a crucial role in studying brain white matter connectivity. Cortical surface reconstruction (CSR), including the inner whiter matter (WM) and outer pial surfaces, is one of the key tasks in dMRI analyses such as fiber tractography and multimodal MRI analysis. Existing CSR methods rely on anatomical T1-weighted data and map them into the dMRI space through inter-modality registration. However, due to the low resolution and image distortions of dMRI data, inter-modality registration faces significant challenges. This work proposes a novel end-to-end learning framework, DDCSR, which for the first time enables CSR directly from dMRI data. DDCSR consists of two major components, including: (1) an implicit learning module to predict a voxel-wise intermediate surface representation, and (2) an explicit learning module to predict the 3D mesh surfaces. Compared to several baseline and advanced CSR methods, we show that the proposed DDCSR can largely increase both accuracy and efficiency. Furthermore, we demonstrate a high generalization ability of DDCSR to data from different sources, despite the differences in dMRI acquisitions and populations.

Paper Structure

This paper contains 11 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Datasets and Preprocessing
  4. SDFNet: SDFs Prediction
  5. DiffCoSeg: Surface Extraction from dMRI
  6. Implementation
  7. Experiments
  8. Experimental Design
  9. Results on HCP-YA data
  10. Results on ABIDE and In-House Data
  11. Conclusion

Figures (3)

  • Figure 1: Graphic overview of the DDCSR Framework.
  • Figure 2: Surface reconstruction comparison across the different methods. From left to right: WM surface, pial surface, slice view (yellow contour: WM surface; red contour: pial surface). The background is based on FA.
  • Figure 3: Surface reconstruction on example ABIDE and In-House datasets. From left to right: WM surface (up), pial surface (down), slice view (yellow contour: WM surface; red contour: pial surface). The background is based on FA.