Machine learning detection of epileptic seizure onset zone from iEEG

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Accurate seizure onset zone (SOZ) identification is critical for epilepsy surgical treatment.
• Intracranial electroencephalography (iEEG) is a key data source for SOZ localization.
• Machine learning (ML) offers advanced analytical capabilities for complex neurological data.

Purpose of the Study:

• To review recent advances in ML approaches for SOZ localization using iEEG data.
• To analyze current ML techniques, their effectiveness, limitations, and future directions in SOZ identification.
• To provide insights into the application and challenges of ML in epilepsy surgery.

Main Methods:

• Narrative review of peer-reviewed studies utilizing ML for SOZ localization with iEEG data.
• Analysis of trends in ML applications, performance metrics, benefits, and challenges.
• Identification of prevalent ML techniques, such as Support Vector Machine (SVM) with high-frequency oscillations (HFOs).

Main Results:

• Increasing adoption of supervised ML, particularly SVM with HFOs, for SOZ localization.
• High accuracy and sensitivity reported, especially in smaller sample size studies.
• Limited generalizability observed in patient-wise validation; ambiguity in SOZ definition and data scarcity are key limitations.

Conclusions:

• ML holds significant potential for advancing SOZ localization in epilepsy.
• Development of robust algorithms, multimodal data integration, and enhanced model interpretability are crucial for progress.
• Addressing current limitations can improve reliability, consistency, and real-world applicability of ML in epilepsy surgery.