Implementation AI to detect melanoma in primary care – a randomised multicentre clinical trial

Purpose and goal

The project aims to evaluate the effects of implementing an app-based AI-driven decision support to detect melanoma in primary care, as compared to standard clinical care. This will be done by randomisation; in half of the patients the doctor may use the app when assesing the participating patient´s skin lesions, and in the other half not (= standard clinical care). The project goal is to enable and secure safe and reliable use of AI for the assessment of skin lesions.

Expected effects and result

In a previous, non-randomized clinical trial, the results indicated that using the app could reduce the number of harmless skin lesions subjected to surgery or referral to dermatologist. Therefore, the hypothesis is that a corresponding difference between the experimental group and the control group will show, i.e. that the number of harmless changes that are surgically removed or referred unnecessarily will decrease when the app is used. It is also possible that fewer melanomas will be missed.

Planned approach and implementation

A total of 3,000 participants are intended (approx. 600 in Sweden), consisting of patients undergoing skin lesion assessment by primary care physicians. Half of the primary care centers are randomized to the possibility to use the app for the assessment (=intervention), while the other half follow standard care (=control). After inclusion of half of the intended sample size, the centers switch from intervention to control, and vice versa. Clinical management of the patients is then compared.