The use of artificial intelligence (AI) for the early detection of skin cancer is showing impressive progress, as demonstrated by a recent study presented at the European Academy of Dermatology and Venereology Congress in 2023. The accuracy of the diagnoses reaches almost 100 percent.
Skin cancer is the most common type of cancer worldwide. The good news is that skin cancer is usually very treatable as long as it is diagnosed early. This is why skin cancer prevention plays an important role and should be carried out regularly. Early detection saves lives. Regular skin examinations at home and at the dermatologist are particularly important.
For some time now, scientists have been trying to increase the effectiveness of early detection of skin cancer through the use of AI. There are now encouraging results. An AI-based diagnostic program analyzed the data of over 20,000 patients who were suspected of having skin cancer within two and a half years. The result of this study is remarkable: of 190 skin cancer cases diagnosed in this patient group, 189 were correctly identified, which corresponds to a detection rate of 99.5 percent. In addition, 541 of 585 precancerous lesions (92.5 percent), which are considered precancerous lesions, and all 59 melanoma cases were successfully detected.
Results of the study in detail
In October 2023, the study was carried out on the Congress of the European Academy of Dermatology and Venereology (PDF-Doc) presented. It shows that AI can detect potentially fatal cancers in early stages. As part of the study by Dr. Kashini Andrew, a consultant at University Hospital Birmingham NHS Foundation Trust, used artificial intelligence to study 22,356 patients with suspected skin cancer over a period of two and a half years. The program detected 189 of 190 skin cancer cases (99.5%) that occurred among study participants, as well as 541 of 585 precancerous lesions (92.5%) and all 59 melanoma cases. This is a big improvement over previous models. The first AI model tested in 2021 only detected 85.9% of melanoma cases, 83.8% of skin cancer cases, and 54.1% of precancerous lesions.
“This study has shown how AI is rapidly improving and learning, with high accuracy directly attributable to improvements in AI training techniques and the quality of data used to train AI,” said Dr. Andrew in a statement. Dr. Andrew explained that the latest version of the software saved over 1,000 face-to-face consultations in secondary care between April 2022 and January 2023. This meant more time was available for patients who needed urgent action.
The data is impressive. Nevertheless, the researchers point out that AI cannot and should not replace a good dermatologist. The study missed one case of basal cell carcinoma, which was later identified by a dermatologist. The team had a safety net of specialists in place as a backup. “We would like to emphasize that AI should not be used as the sole tool for detecting skin cancer and that AI is not a replacement for a dermatologist,” said Dr. Irshad Zaki, consultant dermatologist at UHB NHS Foundation Trust and co-author of the study.
Current research data shows that it is a promising technology for future healthcare. Artificial intelligence cannot yet replace going to the dermatologist, but several online doctors are also working on AI for the initial diagnosis of skin cancer. “We are currently still training our AI with image, anamnesis and metadata,” says Dr. Christian Drerup from Hamburger Online dermatologist doctorderma. Birthmarks are matched using pattern recognition. However, the AI does not yet make a diagnosis or even recommend treatments, says Drerup.
Skin examination using the ABCDE rule
Skin cancer is the cancer that you can see. In contrast to other types of cancer that develop inside the body, skin cancers arise on the skin and are usually visible and can therefore be detected and treated early with appropriate attention. The so-called ABCDE rule can be used for self-examination. It is a method to check the skin for possible malignant changes. Each letter represents a characteristic that can occur in a mole or birthmark. Specific changes can indicate black skin cancer:
A = asymmetry: Normally the halves of a mole are (almost) symmetrical. An asymmetrical shape can indicate a potential risk.
B = Limit: Blurred or unclear edges of a mole are also a sign of an existing risk of skin cancer.
C = Color: A mole is usually brown or black. If the color of a mole or skin tag changes or the coloring is inconsistent, it may be a sign of skin cancer.
D = diameter: The diameter of skin tags should be observed. Significant growth and a size of more than 5 millimeters may indicate an abnormality.
E = sublimity or development: A raised or newly developed skin lesion may also indicate malignancy.
Sources:
Andrew, K. et al. Continued Improvement of Artificial Intelligence in Identifying Skin Cancer (abstract). Presented at the EADV Congress 2023; 12 October 2023; Berlin, Germany.
February 8, 2024