A British AI has been authorised to detect cancer without human oversight in a ‘world first’, its makers said.
It can analyse moles and lesions for skin cancer, and achieves a reported accuracy rate of 99.8% compared to 98.9% for a human dermatologist.
The model by Skin Analytics has been classed as a Class III CE marked medical device by EU regulators, and will be able to operate in Europe without oversight, makers said.
Deep Ensemble for Recognition of Malignancy (DERM) is already being used as a supporting tool in the NHS and has detected over 13,000 cancers.
The company says it will now be deployed widely in Europe where it will be able to autonomously discharge up to 40% patients given urgent referrals for suspected skin cancer.
How does the tech work?
If cancer is suspected, a smartphone can be used to take photos of skin lesions using a special dermoscopic lens attachment.
Images are uploaded to DERM’s online platform, which uses an algorithm to analyse them.
It can classify lesions as benign, pre-cancer, or cancer, with further subdivisions for the type of cancer.
Patients whose lesions are classed as benign can be discharged.
If the lesion is judged to be cancerous, a dermatologist will review the case and decide on a management plan.
As of last year, the cost was £30 per referral, with an extra cost of £8.20 per referral if NHS teledermatology staff virtually review a case to decide on the most appropriate outcome.
It costs an extra £17 to have a case reviewed by the company’s own second-read dermatologist.
It is hoped that the model could cut waiting times in future, in a field of medicine with ongoing shortages at consultant level.
There have long been hopes that AI is the future of medicine, and this could be good news for patients, with either earlier treatment, or an earlier end to anxiety.
The NHS also plans to use AI to increase breast cancer screening, and last year said it would trial to trial a ‘superhuman’ model that predicts a person’s risk of disease and early death.
Skin cancer cases are rising globally, in the last decade in the NHS there has been a 170% increase in the number of people referred urgently for assessment of suspicious skin lesions.
Most of these are benign, however, with only around one in 10 referrals being urgent skin cancer. It is hoped that AI can help with this initial triage, and filter out the cases which do not need treatment.
The ‘watershed moment in the history of medicine’ marks a ‘transformative era for global healthcare’, Skin Analytics said.
The tech has not yet been approved to operate autonomously in the NHS.
Last year, NICE draft guidelines said that more research was needed before it could be fully trusted as a diagnosis tool without a doctor checking too.
They raised particular concerns about how accurate AI technologies are at detecting cancer and non-cancer skin lesions in people with darker skin. This is because the data on accuracy mainly relates to white skin, while only around 3% of all data collected coming from people with black or brown skin.
NICE judged that ‘more data is needed to be certain that these technologies do not incorrectly detect or miss skin cancer in people with black or brown skin’.
Skin Analytics said this guidance was still in draft and didn’t include the most recent independent evaluation.
Medical devices in the EU must undergo testing to check they are up to standard, and Class III is for the most high risk products with the most stringent testing to be sure they are safe.
Neil Daly, founder and CEO of Skin Analytics, said: ‘This isn’t just an AI approval – it’s the first-time regulators have determined that artificial intelligence can match or exceed specialist doctors in making critical clinical assessments.
‘DERM can now independently clear patients it determines don’t have cancer, without requiring a doctor to check its work.
‘This is a real watershed moment in the history of medicine.’
Get in touch with our news team by emailing us at [email protected].
For more stories like this, check our news page.
MORE: Fears AI ‘could be made to suffer’ if it becomes conscious
MORE: Everything we know about King Charles’ cancer treatment one year since diagnosis
MORE: Lesbian couples ‘priced out of starting families by IVF rules’, says fertility expert
