Mole selfies could help detect skin cancers early. But be ‘wary’ of AI apps


The hot days of summer are here, and so are the long evenings of regret: sunburnt shoulders, peeling noses and time spent anxiously inspecting suspicious lesions and changing moles.

Key points:

  • Australia has one of the highest rates of skin cancer in the world
  • A study has found using a smartphone to send photos of moles and lesions to a dermatologist could reduce the time to diagnosis
  • This telehealth model could partially replace routine skincare checks, researchers say

For years, a growing number of apps have promised to speed up this examination with software: just take a selfie of a mean-looking mole, and your phone will tell you whether it's a melanoma or some other kind of skin cancer.

Unfortunately, the technology isn't accurate enough yet, and experts strongly advise against relying on a machine's diagnosis over a trained professional's.

Researchers from Australia and New Zealand have been working on another approach: using phones to send photos of moles or lesions to a dermatologist, who then assesses whether they need to be followed up.

There's a national shortage of dermatologists in Australia, experts say.(AAP: Dan Himbrechts)

To test this, they've conducted a pilot study of 100 participants, with the results published today in the journal JAMA Dermatology.

The research concludes this approach appears to be a "safe, feasible and acceptable" way to monitor for melanomas.

"This is a proof of concept of whether using this type of [magnifying] device and these apps will lead to benefit," said senior author Katy Bell, an associate professor at the University of Sydney.

"There's a potential it could partially replace routine skincare checks."

How the study worked

Australia has one of the highest rates of skin cancer in the world, with about 1,300 people dying of melanomas of the skin each year.

it's hoped that smartphone technologies could help patients detect melanoma early and save lives.

Five common sunburn myths busted

Skin cancer rates remain high and sunburn is all too common in Australia — so why are these myths still believed?

Read more

For the study, 100 participants who had previously had a melanoma removed (and therefore were at higher risk of skin cancer) were evenly split into control and intervention groups.

The control group had little contact with researchers, while the second group received training and support in "patient-led surveillance".

This involved them being assigned a skin-check partner such as a relative or friend, shown how to perform skin self-examination, and taught how to use a dermascope to take highly magnified smartphone photos and then send them to a dermatologist via an app.

The participants were given a dermascope that fitted over the camera lens of their smartphone.
 (Getty: kali9)

Every two months they were reminded to examine themselves and upload images of suggestive lesions.

The control group did none of this.

After six months, all participants had an in-person check-up, where eight melanomas were found in the control group.

The patient-led surveillance group also had eight melanomas after six months, but five had already been detected, Dr Bell said.

"Five people in the patient-led surveillance group had a new melanoma diagnosed ahead of the [face-to-face] check-up, whereas no people in the usual-care group had a melanoma diagnosed before the check-up."

Participants were also assessed to see whether the task of checking themselves for cancerous lesions was making them anxious.

"What we found is the patient-led surveillance did seem to be safe and didn't cause adverse psychological outcomes," Dr Bell said.

More interest in telehealth with pandemic

Australasian College of Dermatologists president Dr Andrew Miller said the study was "interesting", but more research was needed before a similar model could be rolled out to the general community.

Who's the best person to check your skin spot?

When you've got a suspicious skin spot, where is the best place to get it checked out? A skin cancer clinic or just a regular GP?

Read more

The participants in the study were given much more support than patients would generally receive from a skincare clinic outside of the closely monitored environment of a trial, he said.

"Patients [in the self-surveillance group] felt happier and more supported, but that's because they were given more training and support," he said.

But he said a national shortage of dermatologists meant remote skincare checks could be a useful tool.

"In well-to-do areas in capital cities, you find there's a concentration of dermatologists, but if you move out to rural areas there's fewer of them," he said.

"You need a population base of 55,000 to support a dermatologist; that's where telemedicine helps."

The authors of today's paper, which included researchers from the universities of Sydney, Queensland, Otago, Monash and the Melanoma Institute Australia, are preparing a larger follow-up study.

"We're aiming to get a large enough sample to be able to definitively say whether this new model of follow-up care can detect melanomas early," Dr Bell said.

She added that the pandemic had created more interest in telehealth skincare checks and "faster access to dermatological opinion".

"It might be especially beneficial to people in Australia who live outside the major cities," she said.

"There's currently some problems around inequality of access to dermatology."

Are AI skin-check apps any good?

At a technology conference in the US in May, Google revealed there were almost 10 billion Google searches related to skin, nail and hair issues every year.

In response, it was developing an "AI-powered dermatology assist tool".

Once launched, users would be able to use their phone to take images of moles, lesions and other areas of skin that would then be checked against a database of hundreds of conditions.

"The tool is not intended to provide a diagnosis, nor be a substitute for medical advice, as many conditions require clinician review, in-person examination, or additional testing like a biopsy," the company wrote in a blog post about the proposed tool.

Loading

At the time, healthcare experts expressed concern the AI technology wasn't good enough and would either waste time and money through over-diagnosing skin cancers, or cost lives through under-diagnosing them.

Google planned to release a pilot of the app in 2021; more than six months later, it hasn't yet launched.

Other apps, however, claim to use AI to diagnose melanomas.

There's some evidence this could work, at least in theory. A 2017 study published in the journal Nature found an AI network was capable of classifying skin cancer about as well as dermatologists.

A 2018 Cochrane review, however, found that in practice, the available apps get it wrong a lot of the time.

One study cited in the review found four AI-diagnosis apps failed to detect between 27 per cent and 93 per cent of melanomas.

Dr Bell said AI skin-check apps should be avoided as they could give "false reassurance that everything is fine".

"A lot of people are working actively in that space to try and improve the AI systems," she said.

"They still wouldn't replace the dermatologist, but might be useful as a triage test, if they were sufficiently accurate."

Dr Miller agreed.

"I'd be very careful about using them," he said.

Want more science from across the ABC?

Science in your inbox

Get all the latest science stories from across the ABC.

Your information is being handled in accordance with the ABC Privacy Collection Statement.Email addressSubscribe

Source: https://www.abc.net.au/news




Recent Posts