TECH

AI Is Now Helping Doctors Diagnose Skin Cancer Faster

There are 15,000 new cases in the UK every year.

23/08/2017 11:20

Doctors are hoping that artificial intelligence could be the key to detecting signs of melanoma skin cancer far earlier than the current methods of diagnosis allow.

The machine-learning software, developed by the University of Waterloo, Canada, would hopefully shorten the current process which relies entirely on patients presenting lesions (such as moles) and doctors then judging them on their appearance alone.

If they deem them to be potentially hazardous, patients than require a biopsy to get more information.

damiangretka via Getty Images

Professor Alexander Wong, who worked on the study, said: “There can be a huge lag time before doctors even figure out what is going on with the patient.”

Instead, the AI would anaylse images of the lesions, and look for telltale biomarkers of cancer that it has been taught through studying tens of thousands of images, and then it could provide doctors with objective data to make a decision.

The signs it is looking out for would include changes in the concentration and distribution of eumelanin (a chemical that gives skin its colour), and hemoglobin, both strong indicators that a melanoma is present.

This system, which could be available as early as next year, could drastically reduce the number (and cost) of unnecessary biopsies, according to the team.

“This could be a very powerful tool for skin cancer clinical decision support. The more interpretable information there is, the better the decisions are,” said Wong.

According to Cancer Research UK malignant melanoma is the second most common form of cancer among people under 50. It also happens to be the fifth most common cancer in the UK.

There are 15,419 new cases of melanoma skin cancer in the UK each year, and 16% of these (2,459) are deadly.

Cancer Research UK say that although 90% of patients go on to survive melanoma skin cancer for ten or more years, 86% of these could have been prevented in the first place.

Suggest a correction
Comments

CONVERSATIONS