Generalisability of epileptiform patterns across time and patients

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Technology

Background:

• Accurate localization of the epileptogenic zone (EZ) is crucial for successful epilepsy surgery.
• Current methods using electroencephalography (EEG) often yield suboptimal EZ localization due to variable interictal and ictal activity patterns.
• Identifying generalizable neural signal features could improve EZ localization accuracy.

Purpose of the Study:

• To determine if neural signals reflecting epileptogenicity generalize from interictal to ictal periods within patients.
• To investigate whether epileptiform patterns generalize across different patients.
• To identify specific signal features that are generalizable for improved EZ localization.

Main Methods:

• Utilized an intracranial EEG dataset from 55 epilepsy patients.
• Extracted 34 features from stereo-EEG (SEEG) and electrocorticography (ECoG) signals during interictal and ictal phases.
• Employed decision tree classifiers to assess generalizability across time windows and patients.

Main Results:

• Strong evidence for generalizability of epileptogenic signals from interictal to ictal periods across patients.
• Consistent epileptogenicity patterns observed within patients across time windows.
• Signal features, particularly complexity and high-frequency network attributes, showed cross-patient generalizability, especially during the ictal phase.

Conclusions:

• Epileptogenic neural activity patterns exhibit significant generalizability across time and patients.
• Signal complexity and network features are key contributors to cross-patient generalization.
• These findings have implications for developing automated EZ localization tools to aid surgical planning.