AI systems mining big data to make big impact in healthcare

Healthcare is amassing bigger and richer data, and analytics and artificial intelligence (AI) are stepping in to make the most of them with the goal of improving patient care.

Lots of innovation has been happening in healthcare, which is everything from better nutrition to improved sleeping to and less invasive cardiac surgery, and Verily CEO Ashraf Hanna believes that there is still more to come.

“I think the area where [technology] can help really make a big difference is tying those [innovations] together and also being able to tie our behaviors and our interventions to outcomes—to be able to monitor people long enough and in unobtrusive ways, to know that using [one] parameter for treatment of a disease has a better long-term or medium-term impact than using another parameter,” Hanna says.

All these things are very difficult to do today, he acknowledges, but the right sensors with an ability to handle the data could allow us to do those in the very near future. 

Rising to the challenge

Hanna is optimistic about the future of digital health and big data. He cites a couple of places where it could be useful, one of which is improving data availability and avoiding test repetition.

“It’s a lot of waste if you go to hospital A to do something and you go to hospital B to do something else, and they repeat a lot of the tests or programmes,” he says. “Whether you're in France or Singapore or California, your doctor [should be able to] get access to [your data] in a timely way and have a better picture of what's happening in a crisis.”

The second is monitoring data and using it to flag irregularities. Hanna mentions a patch that monitors glucose continuously and sends alerts if blood sugar gets too low so patients can avoid getting rushed to the emergency department or some kind of diabetic coma, but equally important is so they can manage their health better.

Pattern recognition or diagnosis is where data use in healthcare can be brought to a whole new level. This is the third and most advanced, Hanna notes. “You start to diagnose patterns, and you can start to see behaviors and use that potentially to predict [outcomes].”

In reading retinal scans, for example, he describes an experiment they conducted, where a panel of seven leading experts in the US were given a set of images to rate from zero (no disease) to five (severe disease) and had to agree on what the rating was. “It's kind of like a supreme court.”

Interestingly, individual physicians agreed with the panel about 60 percent of the time, while the AI, about 96 percent of the time. So in this case, using an AI translates to “getting the benefit of the seven leading experts in the field. It's not just cheaper and faster, it's also just higher quality,” he explains.

For Hanna, an algorithm’s capabilities can even surpass that of a seasoned doctor. This is because even experts are prone to making errors as a result of “a car accident on the way to work” or “an argument with a spouse.” Things like these do not affect the performance of AI.

“But I see those kinds of complexities, [such as access, consistency of knowledge, and cost] increasing overtime to eventually get to the diagnosis and where we feel more comfortable with the AI making that decision for us,” he admits.

Expectations post–COVID-19

We’ve seen lots of those trends in digital health accelerate recently, Hanna observes.

“In terms of access, people are much more comfortable accessing their healthcare provider and having discussions virtually. So that's been a really important aspect,” he says. “In terms of standardizing care, there's just been a lot more information on what works the best, what aspects to try, and which drugs and which clinical trials.” 

Seeing people come together and support each other is what gives Hanna optimism. Companies and governments are working together to solve the unique problem that the pandemic has created. People go into very dangerous situations, where they could get infected. Others work 30 days in a row without taking a break to build the systems to bolster healthcare response.

“I would hope those areas all stick post-COVID crisis, that we can do a lot more of that together,” he says. “The digital aspects [of AI] are fantastic. I hope they allow us to share our humanity more broadly.”