AUROC
Area under the ROC curve, a common measure of how well a model separates two outcomes (for example disease vs no disease). 1.0 is perfect, 0.5 is a coin toss. A high AUROC does not guarantee the model is useful in a real clinic.
Two candidate tools for flagging pulmonary embolism on CT reach a radiology committee with AUROCs of 0.91 and 0.89, and the room leans toward the higher number. Each figure summarizes performance across every possible threshold, most of which the department would never use, and two ROC curves can cross. At the 95% sensitivity the department actually wants, the lower-scoring tool could produce fewer false flags. Ask both suppliers for specificity at that fixed sensitivity on comparable data, and compare those figures instead.
Terms like this come up in real clinical scenarios across the HelloAI courses: bite-sized modules with verifiable certificates. An account takes one minute, no password needed.
Sign in →