Digital Twin for EEG seizure prediction using time reassigned Multisynchrosqueezing transform-based...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epileptic seizure prediction is a challenging problem in brain disorder analysis.
• Digital twins (DT) offer personalized healthcare solutions, requiring substantial patient data and advanced algorithms like Machine Learning (ML) and Deep Learning (DL).
• Electroencephalogram (EEG) recordings are non-stationary, necessitating advanced time-frequency methods for accurate seizure prediction.

Purpose of the Study:

• To develop a digital twin-based system for predicting epileptic seizures.
• To leverage Time-Reassigned MultiSynchroSqueezing Transform (TMSST) for extracting patient-specific EEG features.
• To utilize a Deep Learning model (CNN-BiLSTM-Attention) for classifying these features and enabling seizure prediction.

Main Methods:

• Applied Time-Reassigned MultiSynchroSqueezing Transform (TMSST) to EEG data for advanced time-frequency analysis.
• Extracted patient-specific impulse features from EEG signals using TMSST.
• Developed and employed a CNN-BiLSTM-Attention deep learning model, termed 'Digital Twin-Net', to learn and classify extracted time-frequency signatures.
• Validated the system on 22 patients from the CHB-MIT dataset.

Main Results:

• The proposed 'Digital Twin-Net' achieved a high accuracy of 99.70% in predicting epileptic seizures.
• The method demonstrated superior performance compared to existing approaches.
• TMSST effectively captured temporal EEG behavior, and the CNN-BiLSTM-Attention model successfully learned these signatures.

Conclusions:

• The developed digital twin system, integrating TMSST and a CNN-BiLSTM-Attention model, is an efficient and accurate method for EEG-based epileptic seizure prediction.
• This approach holds promise for personalized healthcare in managing epilepsy.
• The study highlights the potential of advanced signal processing and deep learning in neurological disorder research.