Automatic seizure detection based on kernel robust probabilistic collaborative representation.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Machine Learning for Healthcare

Background:

• Visual inspection of electroencephalogram (EEG) recordings for epilepsy diagnosis is time-consuming and labor-intensive.
• Developing automated computer-assisted diagnostic systems is crucial to alleviate the workload of neurologists.
• Existing methods may struggle with the non-linear separability of EEG data.

Purpose of the Study:

• To propose a kernel version of the robust probabilistic collaborative representation-based classifier (R-ProCRC) for accurate epileptic EEG signal detection.
• To enhance the classification performance by mapping EEG signals into a higher-dimensional space.
• To provide a reliable and efficient automated system for epilepsy diagnosis.

Main Methods:

• Wavelet transform with five scales was used for initial EEG signal processing.
• Kernel R-ProCRC was employed to collaboratively represent test EEG samples on training sets.
• Classification was performed by maximizing the likelihood of a sample belonging to seizure or non-seizure classes.
• Post-processing techniques were applied to refine results and ensure stability.

Main Results:

• The proposed method achieved 99.3% accuracy for interictal and ictal EEGs on the Bonn database.
• An average sensitivity of 97.48% and specificity of 96.81% were obtained from the Freiburg database.
• The kernel R-ProCRC demonstrated superior performance in detecting epileptic EEG signals compared to existing methods.

Conclusions:

• The kernel R-ProCRC is a highly effective method for the automated detection of epileptic EEG signals.
• This approach offers a promising solution for computer-assisted diagnosis, reducing the burden on clinical experts.
• The method's high accuracy and robustness make it suitable for real-world clinical applications.