Wavelet-based sparse functional linear model with applications to EEGs seizure detection and epilepsy diagnosis

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Epilepsy diagnosis and seizure detection rely on effective feature extraction and classification of electroencephalogram (EEG) signals.
• Current methods often involve complex algorithms, necessitating simpler yet effective approaches.

Purpose of the Study:

• To develop a wavelet-based sparse functional linear model for EEG signal representation.
• To capture discriminative random components of EEG signals using wavelet variances for improved diagnostic accuracy.

Main Methods:

• A sparse functional linear model utilizing wavelet variances for EEG signal analysis.
• A forward search algorithm to determine optimal wavelet decomposition levels.
• Application of the model with a 1-Nearest Neighbor (1-NN) classifier.

Main Results:

• Achieved 100% classification accuracy for epilepsy diagnosis and seizure detection using the University of Bonn EEG database.
• Obtained 99% overall classification accuracy for the University of Freiburg EEG data.
• Demonstrated superior performance compared to complex methods like Support Vector Machines (SVM).

Conclusions:

• The proposed wavelet-based sparse functional linear model offers a highly effective and accurate method for EEG-based epilepsy diagnosis and seizure detection.
• Simple classifiers like 1-NN combined with this model yield state-of-the-art results, simplifying diagnostic procedures.