Related Experiment Video
Updated: Jul 26, 2025
08:43
Application of Granger Causality Analysis of the Directed Functional Connection in Alzheimer's Disease and Mild Cognitive Impairment
Published on: August 7, 2017
7.9K
Hierarchical Graph Convolutional Network Built by Multiscale Atlases for Brain Disorder Diagnosis Using Functional
Summary
This study introduces a new deep learning framework for diagnosing brain disorders using multiscale functional connectivity networks from fMRI. The method accurately identifies Alzheimer's disease, mild cognitive impairment, and autism spectrum disorder.
Area of Science:
- Neuroscience
- Medical Imaging
- Artificial Intelligence
Background:
- Functional connectivity networks (FCNs) from fMRI are vital for brain disorder diagnosis.
- Current methods often overlook hierarchical functional interactions across different spatial scales.
Purpose of the Study:
- To develop a novel framework for multiscale FCN analysis to improve brain disorder diagnosis.
- To integrate hierarchical brain organization into deep learning models for enhanced diagnostic accuracy.
Main Methods:
- Utilized multiscale brain atlases to compute multiscale FCNs.
- Introduced 'Atlas-guided Pooling' (AP) to leverage hierarchical relationships.
- Developed a multiscale-atlases-based hierarchical graph convolutional network (MAHGCN) for diagnostic information extraction.
Main Results:
- Achieved high diagnostic accuracies: 88.9% for Alzheimer's disease (AD), 78.6% for mild cognitive impairment (MCI), and 72.7% for autism spectrum disorder (ASD).
- Demonstrated significant advantages over competing methods.
- Validated on a large dataset of 1792 subjects.
Conclusions:
- The proposed MAHGCN framework effectively diagnoses brain disorders using resting-state fMRI and deep learning.
- Exploring multiscale functional interactions within brain hierarchies is crucial for understanding neuropathology and advancing diagnostic tools.
