Improved Detection and Classification of Convulsive Epileptic and Psychogenic Non-epileptic Seizures Using FLDA and...

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epileptic seizures (ES) are frequently misdiagnosed due to similar-presenting conditions, notably psychogenic non-epileptic seizures (PNES).
• Current gold-standard diagnosis for PNES, video-electroencephalography monitoring (VEM), is resource-intensive.
• There is a clinical need for objective, accessible methods for PNES diagnosis.

Purpose of the Study:

• To develop and evaluate a novel approach for detecting and classifying PNES using accelerometry data.
• To assess the feasibility of a wrist-worn device for differentiating PNES from ES.

Main Methods:

• Extraction of time, frequency, and wavelet features from wrist-worn accelerometer signals.
• Application of Fisher linear discriminant analysis (FLDA) for feature compression.
• Training a Bayesian classifier with a kernel estimator for seizure detection and classification.

Main Results:

• The algorithm achieved 100% classification accuracy for PNES and 75% for ES among detected seizures.
• All epileptic seizures (ES) were detected, with 20 false alarms.
• The system was trained and validated on data from 16 patients undergoing VEM.

Conclusions:

• A wrist-worn accelerometer system shows promise for objective PNES detection and classification.
• This technology could offer a more accessible and less resource-demanding alternative to VEM.
• Further validation is warranted to refine diagnostic accuracy for both PNES and ES.