ECn-MultiBSTM: multiclass epileptic seizure classification using electro cetacean optimized bidirectional long...

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epileptic seizure classification from Electroencephalography (EEG) signals is crucial for diagnosis and treatment.
• Existing models struggle with EEG signal noise, high variability, and complex seizure patterns, limiting accuracy and reliability.
• Challenges include poor generalization and sensitivity to artifacts, hindering effective multiclass seizure detection.

Purpose of the Study:

• To develop an advanced model for accurate multiclass epileptic seizure classification.
• To address the limitations of existing models in handling complex EEG data and noise.
• To improve the reliability and performance of epileptic seizure detection systems.

Main Methods:

• Proposed the Electro Cetacean Optimization based Multi Bidirectional Long Short-Term Memory (ECn-MultiBSTM) model.
• Utilized Multi Bidirectional Long Short-Term Memory (BiLSTM) for robust feature extraction from sequential EEG data.
• Incorporated Electro Cetacean Optimization techniques for efficient hyperparameter tuning and improved model adaptability.

Main Results:

• The ECn-MultiBSTM model achieved 95.84% accuracy, 95.30% precision, and 95.54% F1-score.
• Demonstrated high sensitivity (95.79%) and specificity (95.88%) in classifying seizure types.
• Showcased superior performance on the CHB-MIT SCALP EEG dataset compared to existing methods.

Conclusions:

• The ECn-MultiBSTM model offers a significant advancement in multiclass epileptic seizure classification.
• The proposed approach effectively handles EEG signal complexity and noise, enhancing diagnostic reliability.
• This model shows promise for improving the clinical management of epilepsy through accurate seizure detection.