AI software achieves 100% melanoma detection rate
AI skin cancer detection has come on in leaps and bounds in the last 2 years. Presented at EADV 2023, the latest software could help to reduce the global burden of skin cancer.
Presented at this year’s European Academy of Dermatology and Venereology Congress (EADV) (11–14 October, Berlin, Germany), impressive results have emerged from a study using AI software to detect skin cancer.
AI isn’t new to the game when it comes to skin cancer detection, but few AI tools are actually approved for clinical application and even fewer have reached the melanoma detection rate that this latest tool claims to achieve: 100%. Yes, you are reading that number correctly, no rounding up.
Skin cancer is the most common cancer in the US, and worldwide, skin cancer makes up one in every three diagnosed cancers. The three main types of skin cancer are basal cell carcinoma, squamous cell carcinoma and melanoma. Melanoma is considered to be the most serious type as it tends to spread to other areas of the body and can be fatal if not treated promptly. The new AI tool could now help to reduce the global burden of skin cancer through early detection, particularly for melanoma.
The study presented at EADV took part over a 2.5-year period, involving 22,356 participants with suspected skin cancers. Not only did the software achieve an 100% detection rate for melanoma (59/59 identified cases), but it was also highly sensitive in detecting all skin cancers (99.5% -189/190 identified cases) as well as pre-cancerous lesions (92.5% – 541/585).
This is the third version of the AI software that has been tested. Previously, the first version of the software was tested in 2021 and achieved a detection rate of 85.9% (195/227 cases) for melanoma, 83.8% (903/1078 cases) for all skin cancers and 54.1% (496/917 cases) for precancerous lesions.
Kashini Andrew, Specialist Registrar at University Hospitals Birmingham NHS Foundation Trust (UK), remarked, “This study has demonstrated how AI is rapidly improving and learning, with the high accuracy directly attributable to improvements in AI training techniques and the quality of data used to train the AI. The latest version of the software has saved over 1,000 face-to-face consultations in the secondary care setting between April 2022 and January 2023, freeing up more time for patients that need urgent attention.”
Although these results are enormously promising, the study authors stress that the AI software should only be used with the clinical oversight and support of a Consultant Dermatologist, not as a stand-alone tool. This point is illustrated through the fact that among the basal cell carcinoma cases, one was overlooked by the AI out of 190. This case was later identified by a dermatologist acting as a ‘safety net’. It is important to emphasize that whilst AI can be immensely helpful in dermatology, it is not a replacement for a dermatologist.
