EEG-based stroke severity classification using higher-order topological features and graph convolutional networks.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Electroencephalography (EEG)-based stroke analysis traditionally uses signal and network descriptors.
• Higher-order brain network structures in stroke patients are not fully understood.

Purpose of the Study:

• To apply persistent homology for extracting higher-order topological features from EEG functional networks.
• To integrate these topological features with conventional EEG data for stroke severity classification.
• To investigate the role of higher-order brain network organization in post-stroke alterations.

Main Methods:

• Persistent homology was employed to derive cycle-based topological features from EEG functional networks.
• These features were combined with standard EEG representations.
• A graph convolutional network was utilized for stroke severity classification.

Main Results:

• The framework achieved 86% accuracy in differentiating mild from moderate stroke.
• Cycle ratio analysis identified the prefrontal cortex as having significant higher-order structures.
• This suggests prominent involvement of the prefrontal cortex in post-stroke brain network organization.

Conclusions:

• Higher-order topological features from EEG can improve stroke severity classification.
• This method provides deeper insights into post-stroke brain network changes.
• The findings highlight the importance of topological network analysis in understanding stroke pathophysiology.