Every year, Google sees almost ten billion searches related to skin, nail and hair problems. There is no shortage of people suffering from dermatological issues, and simply typing in what you see on your skin cannot be helpful enough. At the recently concluded I/O summit, Google announced its latest venture into health care: an AI-powered tool that would help people identify conditions concerning skin, hair and nails. Google also unveiled another AI-backed healthcare initiative: research that could turn into an AI-powered screening tool for tuberculosis. 

Artificial intelligence (AI) in healthcare has found a great significance in recent times  and Google has been pioneering in this field of work for a while now.  For instance, earlier this year, Google collaborated with Northwestern Medicine to explore the possibility of using AI models to reduce the time between breast cancer development and diagnosis. 

Now Google’s new AI tool will allow users to identify dermatological issues, like an arm rash, using their phone’s camera. Upon launching the device, user’s would have to take three pictures of their concern from different angles. The AI tool will then ask questions about skin type, how long the issue in question has bothered the user and other symptoms that the user may be experiencing—to filter out the algorithm. The AI model is reported to know 288 conditions and considers age, skin type, sex, and race in its skin condition analyses. It will also provide users with further information on the conditions the AI matches to their concern. This information would consist of dermatologist-reviewed details, frequently asked questions and their answers, matching images on the internet, and will allow the user to look up the ailment further. 

The system builds upon Google Health’s past work on using AI tools to identify skin conditions. Google first published its study on the topic last year in Nature Medicine, which showed how the tool could identify 26 common skin conditions just as accurately as dermatologists and perhaps more accurately than primary care physicians.

Google’s team spent three years developing the ML model, which was trained on millions of images on skin problems, thousands of pictures displaying healthy skin and 65,00 photos from clinical settings. Google Health then tested the tool on around 1000 images of skin problems from a diverse range of patients. As per Google, the system identified the correct condition among its top-three options 84 percent of the time and included the correct condition as a possible issue 97 percent of the time.

Google is currently working with a Stanford University research team to see how well the platform works in a health care setting. The tech giant even received a Class I medical device mark for the AI tool in the European Union—allowing it to operate as a low-risk medical device. The tool is yet to be evaluated by the American Food and Drug Administration (FDA) and is not available in the US yet. 

However, Google has also admitted that this tool cannot offer the additional testing or screening some conditions may require. Still, it hopes to achieve a platform that allows people access to information without spending hours researching potentially unreliable information all over the internet.

AI For Tuberculosis

Google’s second health initiative from the I/O is an AI-powered tuberculosis (TB) screening aid. As per Google, cost-effective screening—especially chest X-rays—are an excellent way to improve the TB screening process, but these methods require experts to interpret them. Considering that TB disproportionately affects low-to-middle income countries, this expert advice might not be readily available to everyone. Hence, Google’s work revolves around catching TB early and working towards eradicating it. Google Health researchers have developed an AI-based tool that builds on their existing work in medical imaging to identify TB patients for further testing. The system produces a number between 0 and 1 which indicates the risk of TB.

The AI-based tool was reportedly able to detect active pulmonary tuberculosis cases accurately and had false positives and negatives at rates similar to 14 radiologists. Google also tested the system on HIV-positive patients—who are at a higher risk of developing TB and are more difficult to test on due to having chest X-rays different to other TB patients—and maintained its accuracy.

Google stated that the screening tool could prove an excellent preliminary test before applying more expensive tests on patients. A study conducted by the tech company displayed that the tool could save up to 80 percent of the cost per positive TB case detected. The initiatives mentioned above are only a few among Google Health’s numerous fantastic initiatives to prevent and effectively solve multitudes of troubling health problems.