Multiway analysis of epilepsy tensors

Area of Science:

• Neuroscience
• Biomedical Engineering
• Data Science

Background:

• Epilepsy surgery success hinges on precise seizure origin localization.
• Ictal electroencephalogram (EEG) analysis is a standard for identifying the epileptic focus.

Purpose of the Study:

• To develop an automated and robust method for analyzing large EEG datasets.
• To model epilepsy seizure structure using multiway analysis.

Main Methods:

• Constructed an Epilepsy Tensor from multi-channel ictal EEG using wavelet analysis.
• Applied parallel factor analysis (PARAFAC) for seizure structure modeling and artifact extraction.
• Utilized multilinear subspace analysis for artifact removal.

Main Results:

• Successfully localized seizure origins in 8 out of 10 cases using initial PARAFAC modeling.
• Improved localization for the remaining 2 cases by applying artifact removal before PARAFAC analysis.
• Demonstrated PARAFAC's effectiveness in modeling seizure structure and extracting artifacts.

Conclusions:

• Multiway analysis, specifically PARAFAC, offers a promising approach for epilepsy seizure analysis.
• Artifact removal is crucial for robust seizure localization in complex EEG data.
• The proposed method provides an automated and effective tool for EEG analysis in epilepsy.