Extended Data Fig. 7. Estimated feature importance (mean and stdev) of the associated fragility heatmap used estimated using permutation.

Estimated feature importance (mean and stdev) of the associated fragility heatmap used estimated using permutation - The metric of interest was the concordance statistic (i.e. AUC) of the ROC curve. The original feature map is transformed into a 20dimensional set of time-varying statistics of its SOZ and SOZ^C electrodes describing the quantiles of the spatiotemporal heatmap (10% - 100% quantiles). This time-varying summary allows these heatmaps to be pooled together across subjects when training a Random Forest classifier as described in Methods.