DenseNet-Based Classification of EEG Abnormalities Using Spectrograms

Background

EEG interpretation is essential for diagnosing neurological disorders but requires significant expert time and experience.

Research

This study developed a DenseNet-based deep learning model to classify normal and abnormal EEG recordings using different EEG representations, including signal images, spectrograms, and scalograms. LIME and Grad-CAM were applied to explain model predictions.

Potential Impact

The proposed approach provides an interpretable AI solution for automated EEG screening and may support more efficient neurological assessment.