AUROC
Also called: AUC, C-statistic, area under the ROC curve
A single number summarizing discrimination across all thresholds: the probability the model ranks a random positive case above a random negative one.
AUROC summarizes how well a model separates cases from non-cases across every possible decision threshold. It ranges from 0.5 (no better than chance) to 1.0 (perfect separation), and it is the most quoted metric in medical AI papers.
Where This Gets Misread
It is also the least clinically actionable one. AUROC ignores calibration, is insensitive to class imbalance, and describes performance at thresholds nobody would ever use. Two models with identical AUROC can behave completely differently at the threshold you actually deploy. Ask for sensitivity and specificity at the operating point.