NeuroAttention-Net: Deep Learning with Channel Attention for Robust Epileptic Seizure Detection from EEG Spectrograms

Inhalt:
Epileptic seizures are one of the most common neurological disorders in the world, and their timely detection is extremely critical for effective management and treatment for patients who suffer from it. In this study, a deep learning-based approach using the channel attention network NeuroAttention- Net is proposed for the accurate detection of epileptic seizures from EEG signals. The EEG signals are first converted into a spectrogram image, using the Short-Time Fourier Transform (STFT), which is then passed to a deep convolutional neural network architecture with an integrated channel attention mechanism. This attention mechanism selectively identifies and prioritizes significant features in the spectrograms, improving the ability of the model to distinguish between seizure and non-seizure events. Furthermore, the channel attention mechanism significantly reduced false positive rates, enhancing diagnostic reliability. The final trained model achieved 99.3% accuracy in detecting seizure activity, with optimized feature extraction and a balanced representation of the dataset. The proposed approach shows significant potential for enhancing the automated detection of seizures, providing an efficient and accurate diagnostic tool for real-world clinical applications and improving patient outcomes.