Exploring Douglas-Peucker Algorithm in the Detection of Epileptic Seizure from Multicategory EEG Signals

Area of Science:

• Biomedical Signal Processing
• Machine Learning in Healthcare
• Neurology

Background:

• Accurate detection of epileptic seizures relies on identifying complex patterns within Electroencephalogram (EEG) signals.
• High dimensionality and correlation in multichannel EEG data pose challenges for efficient analysis and pattern recognition.

Purpose of the Study:

• To develop a novel scheme for extracting representative and discriminatory features from epileptic EEG data.
• To enhance the efficiency and accuracy of epileptic seizure detection using advanced signal processing and machine learning techniques.

Main Methods:

• Applied the Douglas-Peucker (DP) algorithm to extract salient data points from high-volume EEG signals, reducing sample size.
• Utilized Principal Component Analysis (PCA) to decorrelate EEG features and reduce dimensionality of DP-extracted samples.
• Evaluated feature extraction performance using Random Forest (RF), k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), and Decision Tree (DT) classifiers.

Main Results:

• The DP algorithm successfully compressed EEG signals by over 47%, retaining representative sample points.
• The proposed feature extraction method combined with the RF classifier achieved an Overall Classification Accuracy (OCA) of 99.85%.
• The developed method demonstrated superior performance compared to recently reported algorithms on the same epileptic EEG database.

Conclusions:

• The integration of DP and PCA provides an effective strategy for dimensionality reduction and feature extraction in EEG analysis.
• The proposed method significantly improves the accuracy of epileptic seizure detection.
• This approach offers a robust and highly accurate solution for automated epilepsy diagnosis from EEG data.