A Method for Explainable Epileptic Seizure Detection Through Wavelet Transforms Obtained by...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Signal Processing

Background:

• Accurate classification of electroencephalogram (EEG) signals is crucial for diagnosing neurological disorders like epilepsy.
• Existing methods may lack transparency, hindering clinical trust and understanding.

Purpose of the Study:

• To propose an explainable deep learning (DL) method for detecting epileptic seizures from EEG signals.
• To enhance the transparency of DL models used in neurological disorder diagnosis.

Main Methods:

• EEG signals were converted into audio waveforms.
• Two continuous wavelet transforms (Morlet and Mexican Hat) were applied to create time-frequency representations (spectrograms).
• Convolutional Neural Network (CNN) models processed these spectrograms for seizure detection, incorporating Class Activation Mapping (CAM) techniques for explainability.

Main Results:

• Wavelet-based spectrograms effectively captured temporal and spectral characteristics of EEG data.
• The proposed explainable DL method achieved a high accuracy of 0.922 in epileptic seizure detection.
• Class Activation Mapping techniques successfully visualized salient regions influencing model predictions.

Conclusions:

• Wavelet-based preprocessing is effective for EEG signal analysis in epileptic seizure detection.
• The developed explainable DL approach offers a promising tool for prompt and transparent seizure detection.
• Integrating explainability techniques enhances the reliability and interpretability of DL models in clinical neuroscience.