Automatic Seizure Detection Using Logarithmic Euclidean-Gaussian Mixture Models (LE-GMMs) and Improved Deep Forest...

Area of Science:

• Biomedical Engineering
• Machine Learning
• Neuroscience

Background:

• Epileptic seizure detection from EEG is crucial for patient care and research.
• Existing methods face challenges in accuracy and efficiency.
• Automatic detection can improve treatment planning and reduce clinical workload.

Purpose of the Study:

• To develop and validate a novel automated system for epileptic seizure detection using EEG signals.
• To enhance the accuracy and efficiency of seizure detection algorithms.
• To explore the potential of Logarithmic Euclidean-Gaussian Mixture Models (LE-GMMs) and Deep Forest for EEG analysis.

Main Methods:

• Utilized variational modal decomposition (VMD) to process EEG signals.
• Constructed Logarithmic Euclidean-Gaussian Mixture Models (LE-GMMs) for feature extraction.
• Employed a Multi-Pooling and error Screening Forest (MPSForest) learning algorithm for classification.
• Applied post-processing techniques including moving average filtering and adaptive collar.

Main Results:

• Achieved high average sensitivity (98.22%) and specificity (98.99%) on UPenn and Mayo Clinic datasets.
• Demonstrated excellent performance on a long-term Freiburg EEG dataset with 98.47% sensitivity and 98.57% specificity.
• Reported a low false detection rate of 0.24/h, indicating robust performance.

Conclusions:

• The proposed LE-GMMs and MPSForest method offers highly accurate epileptic seizure detection from EEG.
• The approach effectively distinguishes between seizure and non-seizure EEG signals.
• This method shows significant potential for clinical applications and diagnostics in epilepsy management.