Nobody wants to be called in to see the doctor and find a robot wearing a white coat.
You may not have to worry your GP will be replaced by a droid anytime soon, but artificial intelligence is playing a growing role in health care, diagnostics and the delivery of services.
Skin care and cosmetics brand La Roche-Posay is among companies turning to AI as a diagnostic tool.
La Roche-Posay, a L’Oréal brand that specializes in products for sensitive skin, recently introduced the Effaclar Spotscan – a scanning tool that uses artificial intelligence to diagnose acne. It’s the latest in data-driven skin care that uses AI to develop personalized solutions for skin health. AI won’t take the place of your dermatologist, but it shows promise as a diagnostic tool, in everything from acne to skin cancer.
Othman Bennis, head of marketing for La Roche-Posay said a need for early intervention among acne patients was identified by dermatologists. That was the catalyst for a product that could help patients identify early on the level of their acne, measured on a scale from zero to 4+.
“They were telling us one of the main pain points they have in the management of acne patients was that most of the time acne patients were not going soon enough to see a dermatologist and when they were coming, many times it was too late,” said Bennis. “Even if they had the right treatment at that time, they would still have marks and scars because they hadn’t gone earlier.”
Dermatologists wanted a way for patients to get an early screening to diagnose the severity of their acne. They also wanted patients with moderate to high acne levels to be directed to a dermatologist for treatment.
That led to research into a digital tool that could provide an early and accurate diagnosis of acne, ranking it on the same scale used by dermatologists seeing patients in person.
It started with data collection, with some 6,000 photos, three for each patient, with varying degrees of acne. Photos had to be taken on both an iPhone and Android phones since the calibration of pictures isn’t the same across the different operating systems.
La Roche-Posay, a L’Oréal brand that specializes in products for sensitive skin, recently introduced the Effaclar Spotscan.The photos had to encompass skins types from all over the world. Dermatologists provided the photos and each photo was graded by three different dermatologists.
In the case of a disagreement on the numbers, something that’s not unusual in real life practice, the majority determined the deciding grade. So, for example, if two dermatologists said an acne was a grade three and one said it was a grade two, it would be counted as grade three.
The photo analysis focuses on pimples, blackheads and acne marks, like pigmentation of the skin.
Once that data collection and analysis was complete, the AI process began with the data fed into supercomputers. Machine learning resulted in a system that could define and grade acne with an accuracy on par with having two dermatologists deliver a diagnosis.
“Basically it is as precise as seeing a dermatologist face to face,” said Bennis.
Response to Spotscan, which launched three months ago, exceeded expectations, said Bennis. It is now available in 50 countries and two million people have used the online tool to evaluate their skin.
There is a professional version of the online tool as well as a consumer version.
Eighty per cent of people under the age of 25 have had acne at least once and 40 per cent of adult women have issues with acne.
“Acne is huge,” said Bennis. “It is a very widespread problem.”
The online tool has a number of functions, not all related to promoting La Roche-Posay skincare products. For ratings of zero to two, users are offered a skincare analysis with both generic suggestions and La Roche-Posay product suggestions for a skincare regime. At higher levels, it recommends a visit to a dermatologist.
Bennis said users don’t have to be La Roche-Posay customers to benefit from the online tool. Another function is a timeline simulation for people who have low to mild acne to show the effect of following a recommended skincare regime for eight weeks. For severe acne there is no such simulation because users are advised to see a dermatologist.
A second function is useful to all — the ability to track progress though a digital diary that compares early results to later ones as treatment is followed.
AI is also being used to screen for skin cancer and some research results are promising. Research published in Nature, the International Journal of Science, demonstrated artificial intelligence could be trained to achieve an accuracy in classifying skin lesions that was on par with board-certified dermatologists. There are consumer screening apps but some of the more popular ones, like SkinVision, haven’t yet been approved for availability in North America.
In recent research published in the Lancet machine-learning algorithms were more accurate in diagnosing skin lesions than a group of human viewers, which included board-certified dermatologists, dermatology residents and general practitioners.