Study finds AI algorithm can diagnose blindness-causing disease more accurately than doctors

Dr. Michael Chiang (left) examines an eye scan. He is the co-author of a new study that found a specially trained AI algorithm diagnosed an eye disease more accurately than trained physicians. (OHSU News Photo)

Artificial intelligence is only getting more powerful. Today, we’re able to build computer programs that can master complex strategy games and even hold natural-feeling conversations.

But what about diagnosing disease?

A new study on a specialized AI algorithm found that it was able to automatically diagnose a disease that causes childhood blindness more accurately than trained physicians can, a step towards automating medical tasks that are often bottle-necked by a shortage of doctors.

Jayashree Kalpathy-Cramer. (Photo courtesy of OHSU)

The algorithm was developed and studied by scientists at Oregon Health and Science University and Massachusetts General Hospital. It was trained to diagnose a disease called retinopathy of prematurity (ROP) that, if untreated, will lead to total blindness. It’s the same disease that took the sight of musician Stevie Wonder and is the most common cause of childhood blindness.

The algorithm was shown sample images of eye scans and correctly diagnosed patients with ROP 91 percent of the time. Physicians trained to diagnose the disease had an average accuracy of 82 percent using the same images.

“This algorithm distills the knowledge of ophthalmologists who are skilled at identifying ROP and puts it into a mathematical model so clinicians who may not have that same wealth of experience can still help babies receive a timely, accurate diagnosis,” Jayashree Kalpathy-Cramer, one of the study’s lead authors, said in a press release. Kalpathy-Cramer is a professor at Harvard Medical School and also does work at Massachusetts General Hospital.

The results are particularly notable because there is a shortage of ophthalmologists that are trained and willing to diagnose the disease, co-author Dr. Michael Chiang said in the release.

“This creates enormous gaps in care, even in the United States, and sadly leads too many children around the world to go undiagnosed,” he said. Chiang is a pediatric ophthalmologist and also studies ophthalmology and medical informatics at OHSU.

The study is just a first step towards the tech being used to actually diagnose patients, but it’s an important step nonetheless.

As the healthcare system works to solve challenges including a staggering shortage of primary care doctors and skyrocketing healthcare prices, AI-powered solutions are increasingly coming to the forefront.

98point6, a Seattle-based startup, launched a “virtual clinic” app that relies on AI chatbots to let patients have primary care appointments over text from their smartphone.