AI-based applications for the diagnosis of skin cancer can perform as well as physicians, according to a recently published prospective study in The Lancet Digital Health that assessed two different diagnostic apps. One app was shown to be equivalent to experts and superior to non-expert physicians in terms of diagnosis. However, when it comes to treatment decisions, expert physicians excel.
For the study, researchers assessed two different scenarios. The first included 172 suspicious lesions in 124 patients, of which 84 were malignant. The second included 5,696 pigmented lesions in 66 patients, of which 18 were malignant. In each scenario, lesions were assessed by each AI, a group of expert physicians, and a group of non-expert physicians.
One app, using a 7-class AI algorithm, showed equivalent diagnostic accuracy compared to the experts and was significantly superior to the less experienced physicians. The other app, using an ISIC algorithm, performed significantly worse compared to experts but better than the inexperienced users.
Experts were superior to AI for treatment decision-making. According to lead researcher, dermatologist Harald Kittler from MedUni Vienna, “The AI application tends to remove more benign lesions in the treatment recommendation than experts would. If you take this into account, the AI application can certainly be used. It should also be borne in mind that if it is used uncritically, too many false-positive findings would have to be clarified.”