Automatic seizure detection in long-term scalp EEG using an adaptive thresholding technique: a validation study for...

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Scalp electroencephalogram (EEG) is crucial for epilepsy diagnosis.
• Automatic seizure detection aims to improve efficiency and accuracy in analyzing long-term EEG recordings.
• Previous methods showed promise but required validation in larger, diverse patient cohorts.

Purpose of the Study:

• To validate an improved automatic seizure detection algorithm in a large, unselected patient group.
• To assess the general applicability of the algorithm in routine clinical practice.
• To demonstrate the algorithm's effectiveness across different epilepsy types.

Main Methods:

• Algorithm development based on Short Time Fourier Transform (STFT).
• Calculation of integrated power in the 2.5-12 Hz frequency band.
• Application of adaptive thresholding for seizure identification using multi-channel EEG data.

Main Results:

• 159 patients with 25,278 hours of EEG data analyzed.
• Overall sensitivity of 87.3% with 0.22 false detections per hour (FpH).
• High performance in temporal-lobe epilepsies (TLE) (89.9% sensitivity, 0.19 FpH) and extra-temporal lobe epilepsies (ETLE) (77.4% sensitivity, 0.25 FpH).

Conclusions:

• The validated algorithm offers high sensitivity and selectivity for seizure detection in large EEG datasets.
• It effectively analyzes EEG data without requiring prior assumptions about seizure patterns.
• The tool is valuable for rapid and efficient screening of long-term scalp EEG recordings in clinical settings.