Related Experiment Video
Updated: Aug 6, 2026
17:06
Co-analysis of Brain Structure and Function using fMRI and Diffusion-weighted Imaging
Published on: November 8, 2012
DDSurfer: A Weakly-Supervised Dual-Stream Deep Learning Framework for Cortical Surface Reconstruction From Diffusion
Chengjin Li1, Wei Zhang1, Xi Zhu1
1University of Electronic Science and Technology of China, Chengdu, China.
Advanced Science (Weinheim, Baden-Wurttemberg, Germany)
|July 23, 2026
Summary
DDSurfer, a deep learning framework, generates accurate cortical surfaces directly from diffusion MRI (dMRI) data. This T1-weighted-independent method enhances neuroimaging analyses like connectomics.
Area of Science:
- Neuroimaging
- Computational Neuroscience
- Medical Image Analysis
Background:
- Cortical surface reconstruction from diffusion MRI (dMRI) is vital for neuroimaging analyses.
- Current methods rely on T1-weighted images, facing challenges with registration accuracy due to dMRI's low resolution and contrast.
Purpose of the Study:
- Introduce DDSurfer, an end-to-end deep learning framework for direct cortical surface reconstruction from dMRI.
- Overcome limitations of traditional T1-weighted image-based surface reconstruction and registration.
Main Methods:
- DDSurfer employs a dual-stream architecture to fuse dMRI microstructural features.
- A weakly-supervised strategy with pseudo-ground-truth surfaces trains the model for diffeomorphic transformation and surface reconstruction.
Main Results:
- DDSurfer demonstrates superior geometric accuracy, morphological consistency, and generalization compared to traditional methods.
- Evaluations on diverse datasets confirm the framework's effectiveness.
Conclusions:
- DDSurfer offers a computationally efficient, T1-weighted-independent solution for high-fidelity cortical surface reconstruction from dMRI.
- This advances dMRI-centric connectomics and surface-based neuroimaging investigations.
