Google’s new skin check app ‘could lead to misdiagnosis’, dermatology experts say

• Google has developed an app to help users find answers to common skin conditions
• It uses algorithms and artificial intelligence (AI) to match symptoms to diagnoses
• Local experts caution that the app can't be flawless, and that it's best to consult your doctor about skin concerns

It’s no secret that many of us turn to Google to find a diagnosis for our health concerns. But once we go down that rabbit hole, it can become overwhelming and the habit may leave us stressed, uncertain, and even fearful.

Aware of the millions of searches each month, Google has developed an artificial intelligence (AI) “dermatology assist tool” for people with skin concerns. According to a statement by the tech giant, there are nearly 10 billion Google searches for skin, nail, and hair issues each year.

The company explains: “[The app] helps you understand what’s going on with issues related to your body’s largest organ: your skin, hair and nails. Using many of the same techniques that detect diabetic eye disease or lung cancer in CT scans, this tool gets you closer to identifying dermatologic issues – like a rash on your arm that’s bugging you – using your phone’s camera.”

User-friendly 

The process of using the app, named Derm Assist, is simple: users download and launch the app, and use their phone’s camera to take three images of the skin condition of concern from different angles. The app will then ask users questions about their skin type, how long they have had the issue, and any other symptoms that may help to accurately identify the condition.  

The app will analyse this information and search its database of 288 skin conditions to provide them with a list of possible matching conditions that they can then research further. However, Google asserted that their app is not meant to provide its users with a final diagnosis.  

“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,” it said. “Rather we hope it gives you access to authoritative information so you can make a more informed decision about your next step.”

Shortage of dermatologists

With a global shortage of specialists and two billion people worldwide suffering from dermatologic issues, according to the statement, the AI-powered dermatology assist tool’s guidance would be welcomed by many people.  

The statistics are even more concerning within a local context. “We do understand the need of the population to get easy, accurate and fast diagnosis of skin conditions,” Dr Willie Visser, Head of Dermatology at the University of Stellenbosch, Tygerberg Academic Hospital, told Health24. 

“We currently have a crisis in South African dermatology care with only 177 dermatologists (not all are working full-time) to serve a population of 60 million people. The Royal College of Physicians recommends one full-time equivalent consultant dermatologist per 62 500 population. This adds up to 1.6 consultants for every 100 000 of the population.” 

The ratio in South Africa is a shocking 0.3 dermatologists per 100 000, he said, and the current crisis is even worse in the public sector – a very small number of full-time dermatology consultants are attending to the majority of the South African population who are usually underprivileged. 

He added: “This is proven by the fact that the average waiting time at a private dermatologist is between two and three months and several established dermatologists are not even seeing any new patients. Therefore, many patients are deprived of any dermatology care and are suffering from the consequences of poor future planning.” 

Given the above, it is understandable that people are turning to the internet to try and diagnose their skin, hair, or nail problems, Visser said.

Studies showing diagnostic potential 

Previous studies by the company have shown the diagnostic potential of AI-powered apps. Research published in the journal Nature in May in 2020 found  that deep learning systems (DLS) could identify common skin conditions that were comparable with the skill of US board certified dermatologists.

Another study, published in JAMA, showed how non-specialist doctors (primary care physicians and nurse practitioners) used AI-based technology to increase their ability to differentiate between various skin conditions.  

But while these studies are promising, there are concerns and limitations to be aware of, said Visser. 

“Photos are not always taken in the same way and distance from the skin. The lighting and focus could be a problem. Skin tones and the extent of the condition is not always taken into account,” which could all lead to a misdiagnosis, and unnecessary anxiety and visits to a dermatologist, he added. 

Visser highlighted an even bigger fear: If a serious condition is missed by the app, it can delay a necessary visit to a dermatologist.

More than just a clinical picture 

Dr Sian Hartshorne, president of the South African Society For Dermatologic Surgery (SASDS), also cautioned that although AI-tools have been shown to have some success in assessing tumours of the skin, dermatological conditions have many subtle differences which help dermatologists to make diagnoses.  

“Most of these medical conditions need more than just a clinical picture to make a diagnosis. Careful questioning of each patient is vital in confirming a diagnosis,” she said. 

Hartshorne mentioned another worry: “Dermatological conditions in patients with darker skin tones also remain a major issue in creating AI platforms to aid diagnosis of dermatological conditions.” 

While Google has attempted to be inclusive of all skin types in the app’s database, which has used 65 000 images, including “pale skin that does not tan to brown skin that rarely burns”, it has faced criticism because of its “biased sampling”.  

A Vice report notes that out of the thousands of skin conditions pictured, only 3.5% came from patients with Fitzpatrick skin types V and VI – in others words, those representing brown skin and dark brown or black skin. In this sense, the app could end up either over-or under-diagnosing people who are non-white, and leaves experts feeling uncomfortable with the app being put in the hands of the general public. 

“AI will, without doubt, form a part of our future in medicine but we need to take care and put our patients first without misleading data or missing important dermatological conditions,” said Hartshorne.

'Can be useful in future'

“I do, however, think these dermatological AI-systems can be useful in the future for dermatologists and other healthcare providers as a tool to improve diagnostic accuracy,” said Visser.

AI technology to evaluate worrisome moles are already being used in everyday practice, he said. But Visser asserted these systems must be seen as auxiliary diagnostic tools and will have to be interpreted on an individual basis. 

Currently, there are South African apps, such as Dr Derma, available where a user can securely send their symptoms and pictures of their skin condition to a real dermatologist, Visser said, and explained that these systems are more accurate, and if your condition is serious, you’ll know that your dermatologist will have to prioritise your in-office visit.   

“There is no doubt that this technology will improve dramatically in the near future and will form part of the management of dermatological conditions,” he said.   

The Derm Assist app has been in the works for three years and is not yet available for use. For a sneak peek into how the app works, you can check out the company's blog post.