Scientists have created a method utilizing artificial intelligence to detect skin cancer. In a recent study, this AI tool proved to be more effective than current techniques.
Researchers from the East of England have crafted a new approach employing artificial intelligence for skin cancer detection, which outperformed traditional methods according to a fresh study.
The collaborative effort involved experts from Anglia Ruskin University, Check4Cancer, University of Essex, and Addenbrooke’s Hospital, and was based on data from 53,601 skin lesions across 25,105 patients.
During the research, the team utilized machine learning alongside combination theory to narrow down 22 clinical features to the seven most crucial indicators that suggest whether a skin lesion might be concerning. These factors include: changes in size, color, or shape of the lesion, indications of pinkness or inflammation, and the individual’s hair color at age 15.
The researchers assigned proportional weights to these seven features to develop the C4C Risk Score, achieving an accuracy rate of 69%. This score notably surpassed existing methods like the 7PCL (62%) and the Williams score (60%).
Some of the newly identified risk factors, including the lesion’s age, degree of pinkness, and hair color, were significant across all skin cancer types, whereas older methods primarily targeted melanoma, a particular kind of skin cancer.
Professor Gordon Wishart, the Visiting Professor of Cancer Surgery at Anglia Ruskin University and Chief Medical Officer at Check4Cancer, remarked, “This research emphasizes the value of utilizing clinical data for classifying skin lesions, which should enhance skin cancer detection.”
“Our innovative AI model, which integrates the C4C risk score with images of skin lesions, could lead to fewer patient referrals for biopsies, reduced waiting times for skin cancer diagnosis, and improved outcomes for patients.”
Consultant Plastic Surgeon Per Hall, recently retired from Addenbrooke’s, added, “This paper offers significant insight by helping to identify patients whose skin lesions require referral for direct examination.”
“Historically, the focus has been on pigmented lesions and melanoma, but there are other skin conditions needing attention, such as basal cell and squamous cell carcinomas.
“The NHS faces an overwhelming number of referrals for skin lesion assessments, most of which are harmless. This initiative aims to filter out lesions that may be serious and quickly identify patients whose skin is more susceptible to cancer, ensuring they receive timely care.”
The study, partially funded by a Knowledge Transfer Partnership (KTP) Grant from Innovate UK, was published in the Nature journal Scientific Reports.
There are hopes that regulatory approval for the AI model will be achieved by 2025.
