Even highly skilled medical professionals may find it challenging to distinguish melanoma from other skin abnormalities. But since melanoma is a dangerous kind of skin cancer, it’s critical to catch it early.
In Swedish healthcare, there has yet to be any recognized AI-based tool for evaluating skin lesions.
Primary care doctors see many skin lesions every day. Because they have limited resources, they must decide how to treat patients who may have skin melanoma. This frequently leads to a large number of specialist referrals or removals of skin lesions, most of which prove to be harmless.Â
A study led by Linköping University in Sweden reports about a new mobile app that uses AI to analyze images of suspected skin lesions. It can diagnose melanoma with very high precision.Â
Panos Papachristou, a researcher affiliated with Karolinska Institutet and specialist in general practice, the main author of the study, and co-founder of the company that developed the app, said, “We wanted to see if the AI support tool in the app could perform better than primary care physicians when it comes to identifying pigmented skin lesions as dangerous or not, in comparison with the final diagnosis.”
And the results are promising.
Primary care physicians in the research diagnosed suspicious skin tumors using a standard protocol. In cases where the doctors suspected melanoma, they would either remove the skin lesion for tissue analysis and diagnosis or refer the patient to a dermatologist.
The doctor used the AI-based tool only after determining how to treat the suspicious melanoma. Here, the doctor uses a smartphone that has a dermatoscope—a type of enlargement lens—to snap a photo of the skin lesion. Upon analyzing the image, the app makes recommendations regarding whether or not the lesion on the skin looks to be melanoma.
The researchers compared the AI-based app’s answer to the diagnoses produced by the standard diagnostic technique to assess the app’s effectiveness as a decision-support tool.
Out of the more than 250 skin lesions that were investigated, doctors discovered 10 in situ melanoma precursors and 11 melanomas. The application detected every melanoma and only overlooked one precursor. There was a 99.5 percent chance that the app was right when it said that a suspected lesion—including in situ melanoma—was not a melanoma.
Magnus Falk, senior associate professor at the Department of Health, Medicine and Caring Sciences at Linköping University, said, “This method could be useful. However, in this study, physicians weren’t allowed to let their decisions be influenced by the app’s response, so we don’t know what happens in practice if you use an AI-based decision support tool. So even if this is a very positive result, there is uncertainty, and we need to continue to evaluate the usefulness of this tool through scientific studies.”
Researchers are now planning to proceed with a large follow-up primary care study in several countries, in which the app’s use as an active decision-support tool will be compared to not using it at all.
Journal Reference:
- Panagiotis Papachristou, My Soderholm, Jon Pallon, Marina Taloyan, Sam Polesie, John Paoli, Chris D Anderson and Magnus Falk. Evaluation of an artificial intelligence-based decision support for the detection of cutaneous melanoma in primary care: a prospective real-life clinical trial. British Journal of Dermatology. DOI: 10.1093/bjd/ljae021