By Liam Mannix
A huge and much-hyped government investment into 3D skin cancer scanners has hit an unexpected snag after early data showed the scanners performed no better than a simple skin check from a GP – and may lead to overdiagnosis.
The new data has stunned researchers, who are debating whether this represents a blip that will be ironed out as the tech improves or a cautionary tale about the promise and perils of shiny new medical technology.
The Australian Cancer Research Foundation spent about $10 million in 2018 to set up 15 3D full-body cameras across Australia. The Queensland-based research centre established to run the network received another $25 million in federal government research funding, as well as funding from the camera’s manufacturer.
A man has his skin scanned for melanomas in a Vectra machine at the Princess Alexandra Hospital in Queensland.
The scanners, each of which cost about $500,000, use dozens of cameras to generate a 3D image of a person, tracking the location of each mole and blemish.
When the first machines were installed in Australia in 2017 as part of a separate project, a glowing press release said the tech would “revolutionise melanoma detection”.
That revolution is not yet here.
In a study published earlier this year in JAMA Dermatology, researchers found adding the cameras to usual care led to a lot more lesions being cut out from volunteers’ skin – but no more melanoma being detected compared to standard skin checks. And the scanners added $945 per patient in healthcare costs.
“This study is like a cautionary tale,” said one leading melanoma researcher, working on a related project and granted anonymity to speak freely about the trial. “These are very costly devices. And they might not work if you don’t implement it properly. And you’re just wasting lots of money and potentially doing harm.”
“It does present a challenge for us going forward,” said Professor David Whiteman, a researcher at QIMR Berghofer and co-author of the study. “It does temper the enthusiasm a little for just how we go about dealing with skin cancer and its detection in Australia.”
Others disagree. Professor H. Peter Soyer heads the Australian Centre of Excellence in Melanoma Imaging & Diagnosis and led the study. “I still think our original vision, 3D total body imaging supported by AI … will basically lead to an improvement,” he said. “I have no doubt about it.”
A patient discusses her results from The Alfred hospital’s machine.Credit: Eddie Jim
Melanoma is Australia’s national cancer. We have the highest incidence rate in the world. More than 18,000 cases are diagnosed every year, and more than 1300 people die.
The cancer effects melanocytes, the pigment-producing cells responsible for skin’s colour, and occurs mainly in people with fair skin.
UV radiation from the sun is capable of directly altering the DNA code within these cells. Damage the code in genes crucial to controlling growth, and the cell can enter a frenzy of uncontrolled growth. If the new tumour grows deep enough, it can access the bloodstream and spread to our organs or brain.
That makes it a highly survivable cancer if it is detected early and cut out. Five-year survival rates for melanomas detected at stage 1 – when the cancer remains a single skin spot – are greater than 99 per cent.
If the cancer is detected at stage 2 – the cancer is at least a millimetre thick – that rate drops to 73 per cent. By stage 4, five-year survival rates plummet to 26 per cent.
Hence the interest in a screening program, which could theoretically pick up melanomas before they have a chance to spread. Late last year, the federal government committed $10.3 million to develop a road map.
But screening is harder than it would first appear. Diagnosis is somewhat subjective. Your skin is like a tapestry, painted with scars and freckles and moles – scientists call them naevi – that are not cancerous. And melanomas do not all look the same; there can even be disagreement among experts about which marks are benign and which cancerous. The accuracy of a visual test from a doctor ranges between 40 and 70 per cent.
“We are confronted with millions, really, of lesions on the skin that have to be assessed. And it’s very, very difficult for the practitioners to discriminate which are the nasty ones because there are no rewards for getting it wrong,” said Whiteman.
An effort to screen hundreds of thousands of Germans for melanoma led to no long-term reduction in cancer mortality; other screening studies have come to similar conclusions.
And then there’s the overdiagnosis problem. Melanoma was once a rare tumour but is now the third-most-commonly diagnosed in the US – an increase some refer to as an epidemic – yet there has been no actual increase in melanoma deaths.
This suggests, to some, we are cutting out way too many ‘could-be’ melanomas. “We are living in a fee-for-service society. If in doubt, cut it out,” said Soyer.
This is why the Australasian College of Dermatologists does not recommend melanoma screening.
“It is a waste of time. It’s not cost-effective. You stir up the worried well. They’ll have more procedures, so potentially there’s going to be more harm,” said incoming college president Dr Adrian Lim.
All of which brings us back to the 3D scanners.
The melanomas we really want to spot, and quickly, are the ones that are growing. What if you could quickly map out every spot on someone’s skin, and compare them, year-on-year, to spot malignancy? That’s the promise of Canfield Scientific’s VECTRA 360 system.
The patient steps inside the imager, where 92 cameras snap photos of every inch of exposed skin. A computer knits them together into a digital avatar, with each spot able to be analysed by a dermatologist – or an AI. And you can do it quickly, important if you’re going to screen millions of people.
In 2018, the Australian Cancer Research Foundation handed Soyer and his team $9.9 million to roll out the scanners – the only 3D cameras approved by Australia’s health regulator. “This is a significant game changer,” Soyer said in a 2021 video uploaded to Canfield’s YouTube channel. “This will allow us to detect your potential melanoma much, much earlier.”
Soyer has received consulting payments from Canfield Scientific since 2018, but the company was not involved in design or review of the study. The arrangement was “declared in all presentations and publications”, he said.
With all that promise, why are the early results so disappointing? Theories abound.
First, the study compared a 3D camera plus standard skin testing to just standard skin testing, so the intervention group was very heavily scrutinised. That could explain why so many extra moles were excised.
Why did the cameras not pick up more melanomas? Maybe because current skin checks work very well. “We have a very high bar,” said Lim.
The study also did not use the ability of the scanners to show a change in skin spots over time, a crucial melanoma symptom. “I do think that’s a key factor,” said Professor Anne Cust, who leads development of the skin cancer screening road map for Melanoma Institute Australia. And the machines may improve if AI is used to scan the data.
Soyer argues the results actually validated the technology because they “demonstrated that 3D imaging could identify skin lesions that should be reviewed by a dermatologist or clinician for appropriate treatment and diagnosis.”
But it may simply be the case machines are no better than a trained doctor. “That’s our default position – that this is the challenge we’re going to really struggle to overcome,” Whiteman said. “The computer has got to do better than that. And, at the moment, it does not seem like it can.”
The Examine newsletter explains and analyses science with a rigorous focus on the evidence. Sign up to get it each week.