Stacked Topological Preserving Dynamic Brain Networks Representation and Classification.

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• Resting-state functional magnetic resonance imaging (rs-fMRI) is used to study brain networks.
• Dynamic functional connectivity networks (DFCNs) capture temporal brain activity better than static networks.
• Current DFCN methods often overlook brain topology and multi-dimensional features crucial for disease classification.

Purpose of the Study:

• To propose a novel method for constructing and representing DFCNs that integrates brain topology and temporal dynamics.
• To improve the accuracy of brain disease diagnosis, specifically for epilepsy and schizophrenia, by leveraging multi-dimensional DFCN features.

Main Methods:

• Fused blood oxygen level dependent (BOLD) signals and inter-region interactions to capture temporal and cross-temporal brain topology via block-structured adjacency matrices.
• Employed sparse tensor decomposition with sparse local structure preserving regularization for multi-dimensional feature extraction from DFCNs.
• Utilized kernel discriminant analysis for classification and decision-making.

Main Results:

• The proposed method effectively integrates topological structure and temporal variations in functional brain architecture.
• Demonstrated superior performance compared to existing state-of-the-art methods in identifying epilepsy and schizophrenia.
• Successfully extracted discriminative features from DFCNs using a multi-dimensional approach.

Conclusions:

• The novel DFCN construction and representation method offers a significant advancement in brain disease diagnosis.
• Integrating topological and temporal information in DFCNs is key to improving diagnostic accuracy.
• The method shows promise for clinical applications in neurological and psychiatric disorders.