The Automatic Detection of Seizure Based on Tensor Distance And Bayesian Linear Discriminant Analysis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) is crucial for epilepsy diagnosis but manual analysis is time-consuming and costly.
• Automated seizure detection from EEG is essential for efficient epilepsy management.

Purpose of the Study:

• To develop and evaluate a novel automated method for seizure detection using tensor distance (TD) on EEG signals.
• To assess the performance of the TD method in terms of sensitivity, specificity, and accuracy.

Main Methods:

• EEG signals were transformed into time-frequency representations using wavelet transformation.
• Tensor representation of EEG signals was created, and Tucker decomposition was applied to extract principal components.
• Seizure detection was performed by calculating tensor distances and classifying features using Bayesian Linear Discriminant Analysis (Bayesian LDA).

Main Results:

• The proposed TD method achieved high performance metrics: 95.12% sensitivity, 97.60% specificity, and 97.60% recognition accuracy.
• A low false detection rate of 0.76 per hour was recorded on an invasive EEG dataset.
• The method demonstrated high computational speed, indicating potential for real-time applications.

Conclusions:

• Tensor distance (TD) provides an effective approach for seizure classification in EEG.
• This novel method shows significant promise for rapid and accurate real-time epilepsy diagnosis.