Biomedical research needs data. Google wants to help

This photo shows Left to right: Eric Schmidt, former chairman and CEO of Google, Eric Lander, founding director of the Broad Institute of MIT and Harvard, and Jeff Dean, Google senior fellow and leader of Google AI, onstage at Google Next on July 24, 2018. (Image credit: Bérénice Magistretti)

Left to right: Eric Schmidt, former chairman and CEO of Google, Eric Lander, founding director of the Broad Institute of MIT and Harvard, and Jeff Dean, Google senior fellow and leader of Google AI, onstage at Google Next on July 24, 2018. (Image credit: Bérénice Magistretti)

Data is overflowing in the healthcare sector — think electronic medical records, patient apps, medical imaging, and more. But what’s the point of collecting all this data if it can’t be easily accessed and shared to further biomedical research?

This was a question raised by two panels earlier this week during the Google Cloud Next conference in San Francisco. The answer, drum roll, please, can be found in AI and the cloud.

Healing power

“We’re experimenting with rich data sets across many different data domains and scientific disciplines,” said Andrea Norris, Chief Information Officer at the National Institutes of Health (NIH), in a conversation with Toby Cosgrove, cardiac surgeon and former CEO of the Cleveland Clinic (who just announced that he is joining the Google Cloud healthcare team as an executive advisor), and Greg Moore, Google Cloud’s Vice President for Healthcare. The session was titled Healthcare and Life Sciences in the Cloud.

“But there are challenges,” Norris cautioned, summarizing the circumstances faced by many healthcare providers. “Most of our research data is siloed, and signal computers or servers not integrated, nor connected. We need to make research data more fair, findable, accessible, and reusable.”

The NIH will partner with Google Cloud to provide its biomedical researchers access to advanced computational infrastructures, tools, and services. The initiative, called STRIDES (Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability) aims to reduce economic and technological barriers to accelerate biomedical advances, by making data easy to discover and share.

“Today, it takes about a billion dollars and up to 10 years to get a successful new drug to market,” explained Norris. Centralizing the NIH’s data in the cloud will hopefully accelerate the drug discovery process, allow doctors to recommend better patient treatments, and possibly predict the outcome of those treatments with machine learning and artificial intelligence.   

Another initiative that the NIH is launching is the All of Us research program — an effort to gather lifestyle, environment, and biology-related data from one million or more people living in the U.S. By taking into account individual differences, the NIH hopes to deliver better precision medicine tailored to a person’s needs.

According to Norris, 80% of NIH funds (roughly $35 billion a year) support the scientific research of 300,000 researchers across the U.S., and in some cases, across the globe.

Google’s support of the NIH is just its latest foray into healthcare, where Apple and Amazon are also providing services. Co-founder Sergey Brin was an early backer of genetic testing startup 23andMe, which was created by his ex-wife, Anne Wojcicki (Brin carries the gene that increases the risk for Parkinson’s disease). Alphabet (Google’s parent company) is pursuing many life sciences projects and reportedly filed 186 healthcare-related patents between 2013 and 2017. It also founded Verily, a research organization that focuses on the intersection between tech and healthcare.  


Eric Schmidt, former chairman and CEO of Google, kicked off the AI and the Future of Healthcare panel by asking Jeff Dean, Google’s head of AI, about eyesight: “You are obsessed at the moment about the retina, and in particular with something called the fundus. Why?”

Dean replied that there have been a lot of fundamental advances in general purpose computer vision, which has broad implications across a wide range of industries, especially in relation to medical imaging.

“Now we can actually train machine learning models to take, for example, a fundus image, which is a picture of the back of your retina, and make an assessment of whether that retina shows symptoms of a disease called diabetic retinopathy, which is the fastest cause of blindness in the world,” said Dean.  

“There are about 400 million people at risk in the world, many in countries with very few ophthalmologists. We now have machine learning models that not only are as good as board-certified ophthalmologists, but we’ve gotten them to be as good as retinal specialist at diagnosing this kind of disease,” he added.  

Dean went on to explain that this is a great example of using machine learning in healthcare. What’s more, his research team found that retinal imaging can be a good indicator of many other health-related issues, such as cardiovascular health. Snapping an image of your retina to assess your risk of a heart disease is much faster and more efficient than drawing blood and waiting for test results. A paper about this research had been published in Nature a few months ago.

Yet according to Eric Lander, the founding director of the Broad Institute of MIT and Harvard, who was onstage with Schmidt and Dean, these AI-driven scientific discoveries don’t mean much if they can’t be properly integrated and implemented within the medical system.

“It’s frozen history that has set us back,” Landers said. “Hospitals got into these data systems early and didn’t think about how all these pieces would fit together. … In most environments, people might just migrate to a better solution, but in a highly regulated environment (like healthcare), it’s hard to do that.”

Landers, like Norris of the NIH, revealed the chasm between Google’s optimism and the current challenges faced by healthcare providers. But they also communicated a note of hope. While neither AI nor cloud services provide an instant panacea, they do show a path forward, one that just may lead to the decades-old quest for an easy way to store, retrieve, and share medical data.