At the most recent I / O event, Google has announced it has a new AI deep learning tool that can help detect, identify and assess health problems such as skin conditions or tuberculosis (TB) more efficiently. And you can use it on your smartphone.
The tools use your device̵
The Dermatology Assist Tool
Google’s AI powered tool for dermatology makes it easier for you to better understand common problems with your skin, nails and hair. It uses many of the same techniques used to detect diabetic eye disease or lung cancer in CT scans. And Google uses it to help you get answers about things like a rash or weird spots on your skin.
Google provides answers for more than 10 million skin, nail and hair-related problems every year, proving that most people search online for answers before seeing a doctor. This tool will then take the form of a web-based application to be launched later this year.
Once started, use your device’s camera to take three photos of the affected area from three different angles. From there, answer a short series of questions about your skin type and how long you’ve had the problem or symptoms. Google’s AI model analyzes that information and runs it against its database of 288 conditions, allowing it to query a list of possible matching conditions.
For each matching condition that Google returns, the tool will show you dermatologist-reviewed information, along with FAQs and similar matching images. From there, you can continue your research or make the decision to see a doctor yourself. The tool is not intended to be a substitute for your doctor, an in-person examination, or testing; Google says earlier, “we hope this gives you access to authoritative information so you can make an informed decision about your next step.”
Using the tool to improve tuberculosis screenings
In addition to the dermatology support tool, Google has also shared research on how it uses its AI-based screening tool to “ identify potential tuberculosis patients for follow-up testing. ” Google also contributes to the “The End TB StrategyTo help reduce the incidence of the disease.
Currently, tuberculosis affects about 10 million people every year, and disproportionately many people in low- to middle-income countries. Early detection is key, but it’s still quite difficult because the symptoms are much the same as those of other common respiratory illnesses. And while cost-effective screening (like chest X-rays) helps, experts aren’t always around to interpret the results. Google’s AI tool can change that, saving time and money.
The company’s deep learning system can successfully and accurately identify patients most likely to have active pulmonary tuberculosis based on an X-ray. The screening tool is implemented within the process as a step before ordering a more expensive diagnostic test. This could potentially save patients 80% of the cost per positive TV case.
The tool has a false negative and false positive rate comparable to 14 radiologists, even in patients with HIV (making it more difficult to detect). Google has also tested the tool on anonymized data from patients in five countries to make it work more accurately for a wider variety of races and ethnicities.
To apply these findings in the real world, Google calibrated the AI system’s thresholds, which returns a number between 0 and 1 as a TB risk indicator. The research “suggests that any clinic could start with this default threshold and be confident that the model will perform in the same way as radiologists, making it easier to deploy this technology. From there, clinics can adjust the threshold based on local needs and resources. “
With global efforts underway, the World Health Organization hopes that, along with previous screenings, this will help reduce the number of future cases over the next decade.