Effectual seizure detection using MBBF-GPSO with CNN network.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Signal Processing

Background:

• Electroencephalography (EEG) is crucial for seizure diagnosis, reflecting brain's electrical activity.
• Conventional automatic seizure detection methods face challenges in feature selection, computational complexity, and accuracy.
• There is a need for advanced frameworks to improve seizure detection performance.

Purpose of the Study:

• To propose a novel framework, MBBF-GPSO with CNN, for effective and accurate automatic seizure detection.
• To address limitations of existing methods by optimizing feature selection and reducing computational burden.
• To enhance the efficacy of seizure detection using a hybrid approach.

Main Methods:

• Utilized a modified Blackman bandpass filter (MBBF) to eliminate noise and improve stopband attenuation.
• Employed greedy particle swarm optimization (GPSO) with time and frequency domain features for optimized feature selection.
• Integrated convolutional neural network (CNN) for productive classification and automatic feature learning.

Main Results:

• The MBBF-GPSO-CNN framework demonstrated superior performance in seizure detection.
• Optimized feature selection via MBBF-GPSO enhanced the efficiency of the detection process.
• CNN effectively learned distinct features for accurate seizure classification.

Conclusions:

• The proposed MBBF-GPSO-CNN framework offers a practical and effective solution for automatic seizure detection.
• This hybrid approach overcomes limitations of conventional methods, providing higher accuracy and efficiency.
• Further validation through performance and comparative analysis confirms the system's advantages.