Brain functional connectivity analysis based on multi-graph fusion.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Diagnostics

Background:

• Resting-state functional magnetic resonance imaging (rs-fMRI) is crucial for brain disease diagnosis.
• Existing functional connectivity network (FCN) analysis methods face challenges with noise, inter-subject variability, and heterogeneity.
• Current approaches often rely on single FCNs, limiting the exploration of comprehensive brain network information.

Purpose of the Study:

• To propose a novel framework for brain disease diagnosis using rs-fMRI data.
• To reduce the impact of noise, inter-subject variability, and heterogeneity in FCN analysis.
• To improve diagnostic accuracy by integrating information from multiple FCNs.

Main Methods:

• A multi-graph fusion method is proposed to combine information from a fully-connected FCN and a 1 nearest neighbor (1NN) FCN.
• Graph fusion generates a robust representation of rs-fMRI data with high discriminative ability.
• L1-support vector machine (L1SVM) is employed for joint brain region selection and disease diagnosis.

Main Results:

• The proposed framework was evaluated on datasets for Fronto-Temporal Dementia (FTD), Obsessive-Compulsive Disorder (OCD), and Alzheimer's Disease (AD).
• The framework achieved superior diagnostic performance compared to existing state-of-the-art FCN analysis methods.
• Effective brain region selection was demonstrated for classification tasks.

Conclusions:

• The multi-graph fusion framework offers a significant advancement in FCN-based brain disease diagnosis.
• This approach effectively addresses limitations of single-FCN analyses by leveraging complementary network information.
• The method shows promise for accurate and reliable diagnosis of various neurological disorders using rs-fMRI.