2.7. Machine learning-based analysis

Finally, the 68 ROI features extracted were assessed using a machine learning classifier to differentiate FCD from HC. The support vector machines (SVM) classifier was employed to complete the classification task, with the post-surgical confirmed EZ used as the ground truth. For the prediction tasks, 10-fold cross-validation is used to assess classification performance. Here, k-fold cross-validation has been considered one comprehensive way for predictive model evaluation [35]. The goal is to make the best use of our training data to estimate the achievement of a model on unseen data. The model's accuracy and sensitivity were also estimated.