Ensemble Hierarchical High-Order Functional Connectivity Networks for MCI Classification.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Imaging

Background:

• Conventional functional connectivity (FC) analyzes pairwise correlations between brain regions.
• High-order functional connectivity (HOFC) models more complex relationships involving multiple regions, offering unique insights for brain disease classification.
• Existing HOFC methods often use clustering, but single networks may miss complementary information.

Purpose of the Study:

• To develop a novel framework for brain disease diagnosis using high-order functional connectivity (HOFC).
• To address the limitation of information loss in single HOFC networks by generating and integrating multiple networks.
• To enhance the discriminative power of HOFC for classifying brain diseases like mild cognitive impairment (MCI).

Main Methods:

• Proposed a multi-layer HOFC network construction strategy with hierarchical clustering.
• Implemented a selective feature fusion method combining sequential forward selection and sparse regression.
• Ensembled features from multiple HOFC networks generated at different hierarchical layers.

Main Results:

• The novel HOFC framework successfully generated multiple, complementary HOFC networks.
• Selective feature fusion effectively identified discriminative features for classification.
• The proposed method significantly outperformed single HOFC networks in diagnosing mild cognitive impairment (MCI).

Conclusions:

• Hierarchically generated and ensembled HOFC networks provide a more comprehensive characterization of brain connectivity.
• The selective feature fusion method enhances the diagnostic accuracy for brain diseases.
• This HOFC-based framework shows significant promise for clinical applications in neurological disease diagnosis.