A comparative study of CNN-capsule-net, CNN-transformer encoder, and Traditional machine learning algorithms to...

Area of Science:

• Neurology
• Artificial Intelligence
• Signal Processing

Background:

• Epilepsy affects millions globally, characterized by neuronal discharge and convulsions.
• Current diagnosis relies on lengthy electroencephalogram (EEG) analysis by neurologists.
• Optimizing EEG analysis is crucial for efficient and timely epilepsy diagnosis.

Purpose of the Study:

• To compare traditional and advanced AI models for classifying EEG signals in epilepsy detection.
• To identify the most accurate AI model for improving the speed and efficiency of neurological diagnosis.
• To evaluate the performance of Capsule-Net and Transformer Encoder architectures against conventional methods.

Main Methods:

• Utilized artificial intelligence methods, specifically Capsule-Net and Transformer Encoder architectures.
• Compared the performance of these cutting-edge models with traditional machine learning and deep learning approaches.
• Applied models to a database for binary and multiclass classification of epileptic seizure detection.

Main Results:

• Capsule-Net achieved a 99.92% accuracy for binary classification of epileptic seizure detection.
• Transformer Encoder demonstrated an 87.30% accuracy for multiclass classification.
• State-of-the-art AI models significantly outperformed conventional methods in accuracy.

Conclusions:

• AI is indispensable for accurate epilepsy diagnosis and pathology identification.
• Model comparison is vital for selecting efficient diagnostic tools.
• Advanced AI models, coupled with effective data processing, enhance diagnostic accuracy and speed.