Automated EEG pathology detection based on different convolutional neural network models: Deep learning approach.

Area of Science:

• Neuroscience and Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Manual electroencephalogram (EEG) analysis for brain pathology diagnosis is time-consuming, subjective, and requires specialized expertise.
• Automated systems can enhance diagnostic speed and accuracy, reducing errors in clinical practice.

Purpose of the Study:

• To develop and evaluate an automated system for detecting brain pathology from EEG signals.
• To leverage time-frequency spectrum analysis and Convolutional Neural Networks (CNNs) for robust feature extraction.

Main Methods:

• EEG signals were converted into time-frequency spectrum images.
• Three CNN architectures (DenseNet, Inception-ResNet v2, SeizureNet) were employed for feature learning.
• Extracted features were classified using a Support Vector Machine (SVM) classifier.
• The system was validated on the Temple University Hospital (TUH) abnormal EEG corpus.

Main Results:

• The SeizureNet-SVM system achieved state-of-the-art performance.
• Achieved accuracy of 96.65%, sensitivity of 90.48%, and specificity of 100% on the TUH dataset.
• Demonstrated the effectiveness of deep learning on EEG spectrum images for pathology detection.

Conclusions:

• The proposed framework shows significant potential as a diagnostic tool for clinicians.
• Assisting in the early detection of EEG pathology can lead to timely treatment interventions.
• Automated analysis of EEG signals offers a viable alternative to manual interpretation.