A real-time epilepsy seizure detection approach based on EEG using short-time Fourier transform and Google-Net...

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Epilepsy is a prevalent neurological disorder characterized by recurrent seizures, significantly impacting patient quality of life.
• Accurate and timely seizure detection is crucial for effective epilepsy management, diagnosis, and treatment strategies.
• Electroencephalography (EEG) signals offer a valuable, non-invasive modality for monitoring brain activity and detecting epileptic seizures.

Purpose of the Study:

• To develop and evaluate a real-time system for detecting epilepsy seizures using EEG signals.
• To leverage advanced signal processing and machine learning techniques for enhanced seizure detection accuracy and efficiency.
• To provide a practical tool for improving the diagnosis and clinical management of epilepsy.

Main Methods:

• Utilized the Short-Time Fourier Transform (STFT) for feature extraction from EEG signals.
• Employed a Google-net convolutional neural network (CNN) architecture for seizure classification.
• Implemented a sliding window technique for real-time processing of EEG data.

Main Results:

• Achieved a high classification accuracy of 97.74% and sensitivity of 98.90% on the CHB-MIT database.
• Demonstrated a low false positive rate of 1.94%, indicating robust seizure detection.
• The real-time implementation processed EEG episodes in 0.02 seconds with an average delay of 9.85 seconds from seizure onset.

Conclusions:

• The proposed STFT and Google-net CNN approach offers a highly accurate and efficient method for real-time epilepsy seizure detection.
• This system has the potential to significantly aid in the diagnosis, monitoring, and treatment of individuals with epilepsy.
• The real-time processing capabilities and high performance metrics suggest clinical applicability for improving patient care.