A comparison of machine learning classifiers for pediatric epilepsy using resting‑state functional MRI latency data

Epilepsy affects 1 in 150 children under the age of 10 and is most common chronic pediatric neurological condition; poor seizure control can irreversibly disrupt normal brain development. The present study compared ability different machine learning algorithms trained with resting‑state functional MRI (rfMRI) latency data to detect epilepsy. Preoperative rfMRI anatomical scans were obtained for 63 patients epilepsy 259 healthy controls. distribution z‑scores from cohorts analyzed overlap 36 seed regions. In these regions, between ranged 0.44‑0.58. Machine features extracted z‑score maps using principal component analysis. Extreme Gradient Boosting (XGBoost), Support Vector Machines (SVM), Random Forest features. Area receiver operating characteristics curve (AUC), accuracy, sensitivity, specificity F1‑scores used evaluate model performance. XGBoost outperformed all other models a test AUC 0.79, accuracy 74%, 73%, sensitivity 77%. performed comparably across multiple metrics, but it had 31%. SVM did not perform >70% any metrics. highest detection Development could provide an adjunctive method diagnosis evaluation goal enabling timely appropriate care patients.