Cloud based ensemble machine learning approach for smart detection of epileptic seizures using higher order spectral...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Epileptic seizures pose significant diagnostic challenges.
• Automated detection systems are crucial for timely intervention and patient care.
• Integrating IoT, cloud computing, and machine learning offers a promising approach for advanced EEG analysis.

Purpose of the Study:

• To propose a smart framework for the automated detection of epileptic seizures.
• To leverage IoT, cloud computing, and machine learning for enhanced EEG signal processing.
• To achieve real-time classification of EEG states with high accuracy.

Main Methods:

• Acquired scalp EEG signals processed using Fast Walsh Hadamard Transform.
• Extracted amplitude and entropy-based statistical features using higher-order spectral analysis.
• Selected features using correlation-based algorithm and classified using ensemble machine learning (AdaBoost M1 with C4.5 decision tree).

Main Results:

• The ensemble of AdaBoost M1 and C4.5 decision tree achieved 100% classification accuracy.
• Demonstrated 100% sensitivity and specificity for epileptic seizure detection.
• Achieved optimally small classification time, indicating real-time feasibility.

Conclusions:

• The proposed framework effectively detects epileptic seizures in real-time.
• Ensemble machine learning techniques, particularly AdaBoost M1 with C4.5, are highly effective for EEG-based seizure classification.
• The integration of advanced signal processing and machine learning offers a robust solution for automated epilepsy diagnosis.