A Integrated Deep Learning Approach Using Resnet And Efficientnet-B0 With Optimized Data Sampling For Non- Invasive Eeg-Based Epileptic Seizure Detection

Epileptic seizure detection is an important task in healthcare monitoring systems. Epileptic seizures occur due to abnormal electrical activity in the brain and may vary in severity, duration, and type. Accurate and early detection of seizures using electroencephalogram (EEG) signals can help doctors provide timely treatment and improve patient safety. In recent years, artificial intelligence and deep learning techniques have been widely used to automate seizure detection. This study proposes a deep learning-based framework for epileptic seizure detection using ResNet and EfficientNet-B0 architectures. The proposed model analyzes EEG signals to automatically learn complex patterns associated with seizure activity. ResNet helps extract deep hierarchical features from EEG data through residual learning, enabling efficient training of deeper networks. EfficientNet-B0 further improves feature extraction and classification performance by using a balanced scaling approach for network depth, width, and resolution. The combination of these architectures enhances the model's ability to accurately classify seizure and non-seizure EEG signals. Experimental results demonstrate that the proposed approach provides reliable and efficient seizure detection, making it suitable for real-time medical monitoring and clinical decision support systems