Dr Google will check your skin, hair, nails

Google has announced an AI-powered tool that alerts you to issues with your skin, hair and nails.

Google has announced an AI-powered tool that helps people worried about skin conditions. It was unveiled at the Google I/O conference underway in the US.

Google has announced an artificial-intelligence powered tool that alerts you to issues with your skin, hair and nails. It could be a rash on your arm or a menacing spot. You take three snaps of the area in question with your phone, upload them, and Google’s AI algorithms do the rest.

You are asked questions about your skin type, how long you’ve had the issue, and about other symptoms.

Google announced the new tool at its Google I/O annual conference underway in the US.

The AI model matches your images against 288 conditions, and offers you a list of possible conditions that you can research further.

Google doesn’t claim to offer definitive diagnosis. It is steering clear of giving online AI-inspired medical advice. Instead it throws up photos of conditions it deems similar to your own, and suggests you see a doctor to take an issue further.

“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,” said Peggy Bui, a Google product manager and lead of Google’s work in dermatology. Her team works applying machine learning to healthcare problems.

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

Dr Bui saw the need for this tool after encountering a patient who didn’t seek help earlier about a small skin lesion on her toe. Several months later it was diagnosed as melanoma, the cancer had spread and she died. “This case is just one example of a missed skin cancer, but skin diseases as a category are an enormous global burden,” Dr Bui said.

She said users had turned to Google already to research their skin conditions. “We see that two billion people are affected, and most cases are actually curable, but half the world‘s population faces a critical shortage of dermatologists.”

She said users already made ten million search queries about skin conditions each year, but in a study of 1100 cases, online users arrived at the correct skin condition only 30 per cent of the time. Three quarters didn’t know what follow-up care they needed.

Checking a skin condition using Google.

The tool was therefore created to turn this flurry of searches into a fully-fledged AI-powered dermatology tool.

Google gives you dermatologist-reviewed information and answers to commonly asked questions, along with similar matching images from the web. Dr Bui said the most relevant skin conditions would be in the first three matches. Users could peruse more photos of similar conditions to theirs and read more about them.

Dr Bui said the tech giant didn’t use personal information from this tool for advertising.

She said three years of machine learning research and product development had gone into creating the tool. The deep learning methodology was written up for Nature Medicine, and Google had published peer-reviewed papers that validated its AI model, which took into account age, sex, race and skin type.

The machine learning phase processed around 65,000 images and case data of diagnosed

skin conditions, millions of curated skin concern images and thousands of healthy skin images.

Google software engineer and technical leader Yuan Liu, who applies machine learning to health problems, said it was “highly challenging” to develop robust machine learning models to identify dermatological conditions.

“This is because it‘s such a complex field, with over 2000 different skin, hair and nail conditions, located in 45 different body parts and each condition can have a different presence across age, sex, ethnicity, and six different skin types.”

Dr Liu said the machine learning model had to factor in different lighting angles and field of views that normally appear in smartphone images. “Because of those challenges, it took us over three years of fundamental research to develop a good ML model.”

Google will seek regulatory approval in Australia for the tool which may be available to testers this year before wider availability.

Google is not the first to do something like this. For example, SkinVision in partnership with Novartis Poland and Melanoma UK has an app for iOS and Android that offers a risk assessment and advice on what to do after it analyses a photo of your skin.

But Google is everywhere, its research on this topic already is long-running and well recognised, and millions globally turn to Google to research skin conditions already.