An artificial intelligence programme performs just trained dermatologists in diagnosing skin cancer, according to a new study published in Nature.
The study, performed by researchers at Stanford University in California, found that a computer trained to classify two specific types of skin cancer – keratinocyte carcinomas and malignant melanomas – was just as good as professional dermatologists at distinguishing between benign and malignant versions of each respective cancer.
Traditionally, skin cancers are diagnosed visually dependent on various factors, including their symmetry, borders, colour and size. Biopsies and histology tests then confirm the diagnosis.
The computer software used in this study used an algorithm known as a deep convolutional neural network (CNN) – a type of artificial intelligence designed to replicate animal visual perception and recognise visual patterns.
The software was initially ‘trained’ using 129,450 images of skin lesions with their respective diagnoses represented. A total of 2,032 different skin diseases were included in the image gallery.
A number of digital images were then presented to the computer and a set of 21 trained dermatologists. Diagnoses were then verified using biopsies.
The computer was found to be as effective at distinguishing between keratinocyte carcinomas (the most common form of skin cancer) and benign seborrheic keratoses, as well as between malignant melanomas (the deadliest form of skin cancer) and benign nevi (birthmarks).
“The CNN achieves performance on par with all tested experts across both tasks, demonstrating an artificial intelligence capable of classifying skin cancer with a level of competence comparable to dermatologists,” write the authors.
The authors also elude to the use of their technology in a smartphone app, claiming that “mobile devices can potentially extend the reach of dermatologists outside of the clinic.”
If possible, the technology could represent a universal, cost-effective means of preventing or diagnosing one of the most common forms of cancer worldwide.
It could lead to plenty of lives being saved thanks to early diagnosis of malignant melanomas which, in their later stages, have a five-year survival rate of around 14%. In contrast, five-year survival for early stage malignant melanoma is almost 100%.