Epileptic Focus Localization via Brain Network Analysis on Riemannian Manifolds

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Accurate seizure onset zone (SOZ) localization from intracranial electroencephalography (iEEG) is crucial for epilepsy treatment.
• Existing methods for analyzing brain network dynamics in iEEG data often suffer from information loss.
• The dynamic nature of ictal processes requires sophisticated methods to capture evolving connectivity patterns.

Purpose of the Study:

• To develop an effective measure for brain network dynamics to improve SOZ localization.
• To address the limitations of existing methods in accurately computing and comparing connectivity matrices over time.
• To enhance the robustness and accuracy of computer-aided SOZ localization.

Main Methods:

• Represented time-varying connectivity matrices as points on a Riemannian manifold.
• Utilized geodesic distance on the manifold to measure similarity between connectivity matrices, preserving more information.
• Applied clustering methods to brain network dynamics derived from the Riemannian approach, followed by temporal segmentation for SOZ localization.

Main Results:

• The proposed Riemannian method demonstrated superior performance in SOZ localization compared to existing approaches.
• Achieved an average area under the curve (AUC) score of 0.875 across six epilepsy patients.
• The method effectively captures information in brain connectivity descriptors, leading to more robust localization.

Conclusions:

• The Riemannian manifold approach offers a more robust and informative method for analyzing brain network dynamics in iEEG.
• This technique significantly improves the accuracy of seizure onset zone localization.
• The method shows considerable potential for clinical applications in epilepsy treatment.