This paper presents a method to improve the efficiency and reliability of EEG-based seizure detection using a lightweight 1D CNN model and structural pruning. The model, trained using an early stopping technique to prevent overfitting, achieved 92.78% accuracy and a Macro-F1 score of 0.8686. Pruning 50% of the convolutional kernels reduced weights and memory usage by 50%, while maintaining the prediction performance. The accuracy increased slightly to 92.87% and the Macro-F1 score increased slightly to 0.8707. This demonstrates that structural pruning removes redundancy and improves generalization performance, and that, when combined with early stopping, it is a promising method for improving the efficiency and reliability of seizure detection in resource-constrained environments.
