Epilepsy faces diagnostic challenges due to the low sensitivity of scalp EEG and the subjectivity of its manual interpretation. This study analyzes multimodal physiological signals (EEG, ECG, EMG, and ACC/GYR) using statistical methods to identify features that can reliably discriminate between ictal and non-ictal events. Descriptive analysis, significance tests (Mann-Whitney U, Cliff’s Delta), and inter-signal correlation analyses were applied to extract interpretable and physiologically metrics. Additionally, these features were used to train a Random Forest classifier, achieving promising results across EEG, ECG, and EMG modalities (F1-score:0.79, 0.77, and 0.75 respectively). The model’s performance demonstrates that metrics such as amplitude range and standard deviation, selected through prior statistical analysis, provide sufficient discriminative power for automatic seizure detection. These findings highlight the importance of interpretable signal features as a foundation for explainable machine learning systems in clinical and portable contexts.