Cutting-edge mobile phone-powered artificial intelligence (AI) is revolutionizing the battle against pigmented skin cancer, as highlighted in a recent Lancet Digital Health study. While conventional disease diagnosis typically relied on human experts, advanced machine learning has given rise to AI-based algorithms that hold the potential to surpass human capabilities. An essential advantage of these AI systems lies in their widespread accessibility.
In the world of diagnosing pigmented skin cancer, AI seems to be taking the lead. Studies have shown that computer algorithms designed for this purpose can often outdo human experts. In fact, they even outperformed top experts in the International Skin Imaging Collaboration (ISIC) 2018 Challenge.
However, there is a catch. Most of these studies compared AI and human experts in controlled and simulated environments, which don’t fully mimic real clinical situations. But there are scenarios where AI shines and especially when physical contact with the patient isn’t necessary such as in radiology.
Researchers aimed to bridge the gap between AI technology and real-world clinical practice. They conducted a clinical trial involving medical experts from the Sydney Melanoma Diagnostic Centre in Australia and the Department of Dermatology at the Medical University of Vienna in Austria. The trial included patients aged between 18 and 99 with suspicious pigmented skin lesions.
Both specialist doctors and junior dermatologists conducted non-verbal examinations of patients’ skin, documenting their diagnoses for subsequent comparison with AI assessments. Comprehensive photographic records of all patients’ skin were also established. This process involved utilizing a specialized dermoscopic tool attached to a mobile phone featuring DermEngine software of MetaOptima Technology.
To verify the accuracy, reference tests were conducted using histopathological examination. Lesions that remained unchanged in the total-body photographs were deemed benign, while altered lesions underwent digital dermoscopy monitoring.
