Optimized Seizure Detection in EEG Using Dual-Branch Feature Fusion and Machine Learning Technique

Epilepsy is a neurological disorder of the brain that generates seizures due to abnormal electrical activity. The diagnosis and management of the disease primarily depend on recordings of the EEG. A multistage methodology for seizure detection with enhanced accuracy and reliability is proposed in this work. Isolation of key components of the EEG signal is performed to carry out robust segmentation using CNN and RNN. Features will be extracted through DBFFN and Hjorth parameters, which obtain the optimal features after optimization through salp swarm optimization and firefly optimization algorithm. Data classification has been done using Naive Bayes and Random Forest with an accuracy of 99.47%, sensitivity of 99.78%, specificity of 99.70%, and an F1 score of 99.51%. Performances are higher in comparison with other techniques. DBFFN feature extraction methodology with firefly optimization was identified to be effective in enhancing seizure detection performance. Unlike existing single-path or handcrafted feature-based methods, the proposed DBFFN with dual metaheuristic optimization (SSO-FOA) jointly captures complementary spatial-temporal and spectral EEG features while reducing redundancy, resulting in superior accuracy, robustness, and computational efficiency. This computationally efficient system emerges as a reliable real-world seizure detection tool in epilepsy management.