Google and its sister organization Verily, the life sciences division of Alphabet, are touting a new artificial intelligence tool that’s enabling wider access to care and screening for diabetic retinopathy and diabetic macular edema for patients at a hospital in India.
WHY IT MATTERS
In a blog post this week, Kasumi Widner, program manager at Google, and Sunny Virmani, product manager at Verily, described the new machine learning algorithm the companies have developed over the past three years to help expand access to – and efficacy of – screening for DR and DME in India, where there’s drastic shortage of more than 100,000 eye doctors and just 6 million of the 72 million diabetics are screened for eye disease.
“Working closely with doctors both in India and the U.S., we created a development dataset of 128,000 images which were each evaluated by three-to-seven ophthalmologists from a panel of 54 ophthalmologists,” wrote Google research engineers when the algorithm was first announced back in 2016.
“This dataset was used to train a deep neural network to detect referable diabetic retinopathy. We then tested the algorithm’s performance on two separate clinical validation sets totalling ~12,000 images, with the majority decision of a panel seven or eight U.S. board-certified ophthalmologists serving as the reference standard. The ophthalmologists selected for the validation sets were the ones that showed high consistency from the original group of 54 doctors.”
That algorithm has since been refined and is now in clinical use at the Aravind Eye Hospital in Madurai, in the southern state of Tamil Nadu – where it’s helping ease the burden on both physicians and improving the patient experience.
Dr. R. Kim, chief medical officer and chief of retina services at Aravind Eye Hospital, which sees thousands of patients daily, tells Google that the AI tool helps physicians “have more time to work closely with patients on treatment and management of their disease, while increasing the volume of screenings we can perform.”
Verily has received EU clearance for medical devices, Widner and Virmani said, and the company plans to build on deployments such as this one to “expand our research and clinical efforts globally, with the goal of screening more people and preventing disease.”
THE LARGER TREND
In India, access to care is often very difficult to come by, and illness, diagnosis and treatment are often mismatched, as Dr. John Halamka, CIO of Beth Israel Deaconess Medical Center, explained earlier this month at the HIMSS Precision Medicine Summit.
Halamka had recently returned from a trip to Bihar, in the northern part of the country, where help helped read images and diagnose diseases such as tuberculosis. He noted that he also sees big promise for AI in helping overworked physicians their make better assessments for rare illnesses that are more common in India, such as abdominal TB. (You can read more about Halamka’s trip on his blog.)
As for diabetic retinopathy, it has emerged as one of the most immediately promising use cases for use of AI and machine learning in healthcare.
In 2018, for instance, of IDx Technologies was the first company to gain clearance from the FDA for an autonomous AI diagnostic device, IDx-DR, that deploys algorithm to help detect the disease. It’s already in use at some U.S. health systems.
And in China, meanwhile, hospitals are rolling out their own versions of AI-enabled imaging technology to assess patients for DR, glaucoma and macular degeneration.
Still, some experts are tempering their enthusiasm. And even IDx Founder and CEO Michael Abramoff has emphasized the huge importance of strenuous clinical validation processes for these new techniques.
“Technology used in a lab does not directly transfer to what we do in healthcare,” he said this past November. “Patient safety is paramount. And if we don’t do it right, (the technology) will be pushed back and we will lose all of the advantages that AI can have in healthcare, for better quality, lower costs and better accessibility.”
ON THE RECORD
“Automated grading of diabetic retinopathy has potential benefits such as increasing efficiency, reproducibility, and coverage of screening programs; reducing barriers to access; and improving patient outcomes by providing early detection and treatment,” according to a 2016 JAMA study describing Google’s work at Aravind Eye Hospital. “To maximize the clinical utility of automated grading, an algorithm to detect referable diabetic retinopathy is needed.”
The report concluded that the “algorithm based on deep machine learning had high sensitivity and specificity for detecting referable diabetic retinopathy,” but noted that more research is needed to “determine the feasibility of applying this algorithm in the clinical setting and to determine whether use of the algorithm could lead to improved care and outcomes compared with current ophthalmologic assessment.”
Healthcare IT News is a HIMSS Media publication.