Epileptic Seizure Detection Based on Variational Mode Decomposition and Deep Forest Using EEG Signals

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Manual analysis of electroencephalography (EEG) for epilepsy detection is labor-intensive.
• Automated seizure detection is crucial for effective computer-assisted treatment.

Purpose of the Study:

• To develop and evaluate a novel algorithm for automatic epileptic seizure detection using EEG signals.
• To improve the accuracy and efficiency of epilepsy diagnosis through computational methods.

Main Methods:

• EEG signals were processed using variational modal decomposition (VMD) to extract key variational modal functions (VMFs).
• Time-frequency distributions were constructed, and log-Euclidean covariance matrices (LECMs) were computed to derive EEG features.
• A deep forest (DF) model, a non-neural network deep learning approach, was employed for EEG signal classification.
• Postprocessing techniques, including moving average filtering and adaptive collar expansion, were applied to enhance classification accuracy.

Main Results:

• The algorithm achieved high sensitivity (99.32%) and specificity (99.31%) on the Bonn EEG dataset.
• On the Freiburg long-term EEG dataset, the method demonstrated a mean sensitivity of 95.2% and specificity of 98.56% with a low false detection rate (0.36/h).

Conclusions:

• The proposed VMD and DF-based algorithm shows superior performance for epileptic seizure detection.
• This method holds significant research potential for advancing automated epilepsy diagnosis and treatment.