Automatic seizure detection using correlation integral with nonlinear adaptive denoising and Kalman filter

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Automatic seizure detection from electroencephalogram (EEG) is crucial for epilepsy management.
• Traditional methods like correlation integral have limitations in performance and computational cost.
• Need for efficient and accurate seizure detection algorithms for clinical applications and wearable devices.

Purpose of the Study:

• To re-evaluate the efficacy of correlation integral for automatic seizure detection.
• To improve the performance of correlation integral by incorporating advanced signal processing techniques.
• To assess the clinical utility and potential of the developed algorithm for epilepsy diagnosis and monitoring.

Main Methods:

• Re-examination of the correlation integral method for seizure detection using scalp EEG data.
• Implementation of nonlinear adaptive denoising for pre-processing EEG signals.
• Application of a Kalman filter for post-processing to enhance detection accuracy.
• Development of a three-stage algorithm combining these techniques.

Main Results:

• The developed three-stage algorithm achieved 84.6% sensitivity for seizure detection.
• A low false detection rate of 0.087 per hour was recorded.
• The algorithm demonstrated performance comparable to machine learning methods at a significantly lower computational cost.

Conclusions:

• The enhanced correlation integral algorithm offers a computationally efficient and effective approach for automatic seizure detection.
• The algorithm shows potential for improved performance with intracranial EEG data.
• Clinical applications include assisting diagnosis and serving as a reliable warning system in wearable devices for epilepsy patients.