An algorithm for diagnosing skin cancer performs similarly to dermatologists

A study has now been presented that will increase the evidence for the use of AI solutions in the diagnosis of skin cancer. With an algorithm designed by themselves, scientists at the University of Gothenburg are demonstrating the ability of technology to perform at the same level as dermatologists in assessing the severity of skin melanoma.

The study, published in the Journal of the American Theological Academy, and his results are the work of a research group at the Department of Dermatology and Venereology at the Sahlgrenska Academy, University of Gothenburg.

The study was conducted at Sahlgrenska University Hospital in Gothenburg. Its purpose, through machine learning (ML), was to train an algorithm to determine if skin melanoma is invasive and at risk of spreading (metastatizing), or if it is still at a growth stage in which it is confined to the epidermis, without risk of metastasis.

The algorithm was trained and tested on 937 dermatoscopic images of melanoma, followed by testing on 200 cases. All included cases were diagnosed by a dermatopathologist.

Most melanomas are diagnosed by patients rather than doctors. This suggests, for the most part, that diagnosis is relatively easy. Before surgery, however, it is much more difficult to determine the extent to which the melanoma has reached.

To make the classifications more accurate, dermatologists use dermatoscopes – instruments that combine a type of magnifying glass with clear illumination. In recent years, interest in the use of ML for the classification of skin tumors has increased, and several publications have shown that ML algorithms can perform equal to, or even better than, experienced dermatologists.

The current study is now giving further impetus to research in this area. When the same classification operation was performed with the algorithm on the one hand and seven independent dermatologists on the other, the result was a draw.

“None of the dermatologists did much better than the ML algorithm,” says Sam Polesie, a researcher at the University of Gothenburg and a specialist doctor at Sahlgrenska University Hospital, who is the corresponding author of the study.

In an improved form, the algorithm could be an aid in the task for assessing the severity of skin melanoma prior to surgery. The classification affects the breadth of activity required, so it is important for both the patient and the surgeon.

“The results of the study are interesting, and it is hoped that the algorithm can be used to support clinical decision making in the future. But it needs further refinement, and there is a need for prospective studies that monitor patients over time, “Polesie concludes.