User-guided interictal spike detection.

Area of Science:

• Neuroscience
• Medical Technology
• Signal Processing

Background:

• Epilepsy diagnosis and treatment necessitate long-term electroencephalogram (EEG) monitoring for interictal activities like interictal spikes.
• Manual visual inspection of EEG data by experts is time-intensive, leading to reliance on automated detection methods.
• Existing automated methods often fail to capture inter-individual variability in spike morphology.

Purpose of the Study:

• To develop a user-guided algorithm for efficient and accurate detection of interictal spikes in EEG.
• To create a system that balances the need for expert input with the limitations of fully automated detection.
• To improve the analysis of long-term EEG monitoring data for epilepsy research and clinical practice.

Main Methods:

• A user-guided algorithm was developed, requiring minimal user annotation to build spike templates.
• Mean Squared Error (MSE) testing was employed to detect potential spikes based on the generated templates.
• A multichannel, multi-template approach was utilized to account for variations in spike morphology during wake-sleep cycles.

Main Results:

• The algorithm achieved an average sensitivity of 96% in detecting interictal spikes.
• The system demonstrated an average of 4.8 false detections per hour.
• Detected events were rank-ordered, facilitating expert review and identification of true spikes.

Conclusions:

• The user-guided EEG spike detection algorithm offers an effective compromise between manual annotation and fully automated methods.
• This approach enhances the efficiency and accuracy of interictal spike detection in long-term EEG monitoring for epilepsy.
• The algorithm's ability to handle variable spike morphologies improves its applicability in clinical and research settings.