Naïve Bayes model results for the classification of patient outcome for DS patients. (A) The ROC curves computed from the leave-one-out Naïve Bayes validation set posterior probabilities for the model with the CReP30 feature based on HFO-RATE (grey) and the model with CReP30 of centralities and HFO-RATE (yellow). The statistical significance of the AUC of the ROC curve was evaluated for both curves. While the AUC of the HFO-RATE ROC was not significantly different (AUC = 0.72, 95% CI = 0.459, 0.974, P = 0.138 via bootstrap test, n = 17), the AUC of the centrality + HFO-RATE ROC was significantly different from a random classifier (AUC = 0.83, 95% CI = 0.636, 1.031, P = 0.008 via bootstrap testing, n = 17). (B) Confusion matrix computed from the chosen point ‘B’ in A with perfect specificity and an accuracy of 83%. (C) Violin plot of the leave-one-out Naïve Bayes posterior probabilities for the model with centralities features and HFO-RATE. Boxes represent the interquartile range with the horizontal line being the median and whiskers extending 1.5 times the interquartile range. Jittered data points are overlaid on top of the box plots. Results for DS-1 (n = 12) and DS-3+ (n = 5) patients are from leave-out cross-validation. DS-2 patient data are the mean and standard deviations of each patient (n = 6) to all 17 cross-validation models (see Supplementary Fig. 6). (D) Calibration curve of the centrality + HFO-RATE model. This plot demonstrates that the algorithm is very accurate when it is more certain (higher probability) of the outcome.