Automatic Seizure Detection Using Hierarchical Spectral-Temporal Feature Learning with an Imbalance-Aware Transformer

Epilepsy manifests as a chronic neurological condition marked by recurrent seizure. Recent advances in computational analysis of Electroencephalography (EEG) signals have enabled new possibilities for identifying ictal events in extended recordings. The current work develops a novel deep learning architecture that simultaneously resolves two fundamental challenges in automated seizure detection: comprehensive feature representation and class distribution imbalance. First, a multi-branch neural network structure is proposed to process EEG signals across varying spectral and temporal resolutions. Then, an attention-based feature refinement mechanism is utilized to automatically emphasizes clinically relevant signal characteristics. Finally, a modified loss function is leveraged to incorporate class-specific margin adjustments to handle data imbalance scenarios. Analysis of scalp EEG recordings yields detection accuracy with 96.06% sensitivity and 98.50% specificity, and false detection rate (FDR) is maintained at low level of 0.34 events per hour. When applied to intracranial EEG data, the algorithm demonstrates similar efficacy (95.90% sensitivity, 98.65% specificity) with further reduced false detections (0.18/hour). The consistent efficacy validated on diverse EEG modalities (scalp and intracranial) supports its clinical utility as a practical diagnostic tool.