[Detection of epileptic waves in EEG based on wavelet transform].

Area of Science:

• Biomedical Signal Processing
• Neurology
• Artificial Intelligence in Medicine

Background:

• Epileptic wave detection in electroencephalograms (EEGs) is crucial for understanding seizure processes.
• Existing methods may require significant manual interpretation, increasing workload and potential for error.

Purpose of the Study:

• To develop an automated method for detecting epileptic waves, specifically spikes and spike-waves, in EEG signals.
• To utilize wavelet transform for enhanced time-frequency localization of high-energy epileptic waveform components.
• To establish criteria for automatic classification by comparing EEG signal characteristics between normal and epileptic states.

Main Methods:

• Applied discrete wavelet transform to EEG data for time-frequency analysis.
• Identified local maximal positions across dyadic scales to pinpoint sharp transitions indicative of epileptic activity.
• Calculated subwave periods and analyzed their distribution to differentiate between normal and epileptic EEGs, selecting optimal scales for detection.

Main Results:

• Wavelet analysis effectively localized high-frequency energy characteristic of spikes and spike-waves.
• Scale 3 of the wavelet transform demonstrated the highest accuracy in detecting epileptic waveforms.
• The developed system achieved significant workload reduction and enhanced detection accuracy compared to traditional methods.

Conclusions:

• The proposed wavelet-based system offers an accurate and efficient method for continuous automatic detection of epileptic waves in EEGs.
• The system's ability to select optimal scales and criteria enhances its reliability and applicability to other biomedical signals.
• Initial clinical results are encouraging, suggesting the system's potential for practical diagnostic use.