computing the scalar projection of the mean feature vector for each patch onto the FCD average unit vector, which was computed from the remaining FCD lesion masks, after leaving the selected patient out (Figure 1c). We then applied a threshold to identify the top 1% of vertices and retained clusters with 5 or more adjacent remaining vertices. For each surviving cluster, we computed the mean FCD similarity.
Our automated FCD detection method is based on the FCD similarity maps, in which the mean feature vector for every cortical patch is projected onto the FCD average unit vector. The top 1% were identified, and clusters with 5 or more adjacent remaining vertices were retained. The final detection threshold was selected based on the optimal trade-off between sensitivity and specificity. We initially evaluated the 11 MRI+ patients (5 initially MRI-) included in the leave-one-out analysis, resulting in an AUC of 0.91. At the optimal threshold determined from this analysis, our classifier correctly identified the lesions in 11/11 MRI+ patients (100% sensitivity) and 12/15 patients overall (80% sensitivity), with a true positive (TP) lesion in one additional patient that did not reach the threshold. In each patient, 1-3 clusters were identified; of 13 patients with TP clusters, FCDs had the highest mean cluster weight in 11. Both patients without TP lesions had no visually identifiable lesion, even retrospectively. Lesions were identified in 9/30 HVs (70% specificity); 7 HVs had 1 false positive cluster, 1 HV had 3 false positive clusters, and 1 HV had 4 false positive clusters. Table 2 summarizes the results individually for each patient; data for 3 exemplar patients are shown in Figure 5.