Google Launches AI-Based Dermatology Assist Tool

Google has developed an AI-powered dermatology assist tool designed to detect conditions of the skin, hair and nails based on phone images taken by the user.

The app uses many of the same techniques involved in CT scans that detect diabetic eye disease or lung cancer. The tool allows users to identify dermatologic issues such as a rash on the arm or nail discoloration using a phone camera.

Google is set to launch the pilot app later this year.

Google says it gets ten billion Google Searches related to skin, nail and hair issues every year. In addition, the tech giant says while almost two  billion people worldwide suffer from dermatologic issues, there’s a global shortage of specialists. This leads people to consult with general practitioners that offer lower diagnostic accuracy.

Recognizing the need for better ways of helping identify skin issues people may be having, Google decided to develop a deep learning system (DLS)-, image-based tool that would allow users to gain quick and convenient insight into what type of a skin condition they may be dealing with.

Skin conditions are often difficult to describe in just words and the new AI phone tool offers a way for users to get more accurate assessments based on images of their condition.

Related: Google’s AI Detects Breast Cancer

The tool is installed as an app on a user’s phone and the phone’s camera is used to take three images of the skin, hair or nail concern from various angles. The user is then asked to answer questions about their skin type and symptoms related to the issue they’re experiencing.

The AI model analyzes the provided images and information and links to a database of 288 conditions, offering a list of possible matching conditions that users can research further.

Google says that for each matching condition, the tool will provide dermatologist-reviewed information and common questions and answers, along with matching images from the internet.

The tool is not intended to make a clinical diagnosis or act in lieu of medical advice. Many skin conditions require an in-person examination by a dermatologist and additional testing like a blood test or biopsy. What the tool does is offer access to trusted, verified information related to matching conditions so that users can make informed decisions about the next steps they should take.

Development of the AI Dermatology Health Tool

Google developed its AI-powered, web-based application tool over three years, which included deep machine learning research and product development.

The company has published several peer-reviewed studies that validate the AI model. Of these is a landmark study published in Nature Medicine that shows the AI system can achieve accuracy that is on par with US board-certified dermatologists. Another recent study published in JAMA showed that the tool can also assist primary care practitioners (physicians and nurses) to diagnose skin conditions more accurately in primary care settings.

The deep learning system was developed using data (images and clinical information) over 16,000 de-identified cases from a teledermatology practice covering 17 sites. This culminated in a database consisting of 65,000 images and case data of diagnosed skin conditions, millions of curated skin concern images (such as acne, wrinkles, large pores, pigmentation, redness and sun damage) and thousands of examples of healthy skin across different demographics.

The tool can distinguish between 26 common skin conditions, which represent 80 percent of cases that present in primary care. It also offers a secondary prediction expanded to include 419 skin conditions.

The AI model takes into account factors such as age, sex, race and skin types.

The product has been clinically validated and received CE marking as a Class I medical device in the EU. It is currently not available in the US as the US Food and Drug Administration (FDA) has yet to review its safety and efficacy.

Google has been upping its game in the health app arena by leveraging deep machine learning technology to build AI health tools. Last year, it developed a tool to assist oncologists in the screening of breast cancer. It is predicted that AI tools will become more widespread in healthcare, allowing for improvements in diagnostics, treatment and overall care.