Related Experiment Video
Updated: Aug 6, 2026
12:21
Probing the Brain in Autism Using fMRI and Diffusion Tensor Imaging
Published on: September 12, 2011
Multimodal neuroimaging fusion with hierarchical structure-function coupling for autism spectrum disorder diagnosis
Jianping Qiao1, Zhongchen Zhou1, Houyuan Zhu1
1School of Communication and Electronic Engineering, Shandong Normal University, Jinan, China.
Neuroscience
|July 20, 2026
Summary
A new deep learning framework, DMDC, effectively integrates brain functional and structural data for improved autism spectrum disorder (ASD) diagnosis. This approach enhances diagnostic accuracy and identifies key brain regions involved in ASD.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Biomedical Engineering
Background:
- Multimodal neuroimaging fusion is crucial for diagnosing autism spectrum disorder (ASD).
- Existing methods often fail to effectively couple functional and structural brain information, limiting diagnostic capabilities.
- There is a need for advanced frameworks that capture intrinsic cross-modal relationships for better ASD identification.
Purpose of the Study:
- To develop a deep multimodal dual-level coupling (DMDC) framework for enhanced autism spectrum disorder (ASD) diagnosis.
- To improve model discrimination and interpretability by integrating brain functional and structural information.
- To identify key brain regions and connectivity patterns associated with ASD.
Main Methods:
- Developed the deep multimodal dual-level coupling (DMDC) framework.
- Employed a pyramid-inverted graph-attention network with Top-k node filtering for function-structure coupling.
- Utilized skeleton-based white matter projection of fMRI signals onto DTI white matter skeletons, followed by a hierarchical convolutional network.
- Integrated features using a weighted mechanism and optimized with cross-entropy loss for ASD identification.
Main Results:
- The DMDC framework demonstrated superior performance compared to state-of-the-art methods in ASD identification.
- Identified key discriminative brain regions including the hippocampus, anterior cingulate cortex, amygdala, and frontal gyri.
- Observed both hyperconnectivity and hypoconnectivity in ASD, particularly in regions critical for social and emotional processing.
Conclusions:
- The DMDC framework offers robust diagnostic performance for autism spectrum disorder (ASD).
- The identified brain regions and connectivity patterns serve as reliable biomarkers for neurological assessment.
- The study provides valuable insights into the neural mechanisms underlying ASD.