Google project managers called Camille Nebeker out of the blue, intrigued by her paper on using artificial intelligence for healthy aging. The tech giant sees a new market for its products, which could generate reams of new data for researchers studying AI in health—provided they can close the gap on consent.
The conversation with Nebeker, an associate professor at the University of California, San Diego’s medical school, has Google considering testing products beyond the 15-to-35-year-old consumer market it had previously targeted.
“And now they just want to go out to the retirement communities and start collecting data from residents to figure out how they can pitch their product to that demographic,” Nebeker said.
From a Google Health algorithm that can spot breast cancer more accurately than doctors to an Amazon.com Inc. deal with Novartis AG to detect potential bottlenecks in manufacturing and delivering new medicines, technology giants have made inroads in health care. Likewise, biomedical researchers have increasingly turning to machine learning to answer pressing biomedical questions.
But there’s a disconnect between the data researchers need for their studies and how tech companies collect their information. Artificial intelligence can churn through reams of data generated by smartwatches and apps like Kardia that can detect abnormal heartbeats, which Nebeker studied for improving the health of aging patients. However, unless the way that data is collected meets the requirements for studying human subjects—known as the Common Rule (45 C.F.R. 46)—it can present ethical conundrums for researchers.
Machine learning is a type of AI that requires large data sets to train networks. That information is often pulled from publicly available data like social media networks. But getting clearance to use that data in a research setting is more complicated than how computer scientists are used to operating.
“The concern is that the data are being used without the originators of the content agreeing to the use,” Susan Gregurick, associate director for data science and director of the Office of Data Science Strategy at the National Institutes of Health, told Bloomberg Law.
It’s a conundrum increasingly important to resolve as more medical researchers are incorporating machine learning into their work.
“It’s a very exciting time that we live in now to take advantage of the technology, to enhance our understanding of human biology and to provide better treatment for our patients,” Aziz Nazha, an oncologist who leads the Cleveland Clinic’s Center of Clinical Artificial Intelligence, said. “The question is: How do we actually work with the technology and how do we put it together to benefit the research?”
Click to Agree
Closing the gap in consent standards would involve carefully defining standards and processes for how a new field of biomedical machine learning would approach data collection and use, Gregurick said.
“This means making participants aware they are participants, explaining how the data may be used and reused, and getting their agreement to collect and use data,” Gregurick added. “It also likely involves creating some standards around the consent process for biomedical machine learning—best practices to follow when building projects and recruiting research participants and to follow throughout the research process.”
A high-level NIH advisory panel recommended last month the agency establish a working group to address the consent issues as research moves forward into biomedical machine learning. Gregurick said the agency is currently working on an implementation plan for all the recommendations, including the one on consent.
Bridging the gap also means helping computer scientists and physicians speak the same language, said Nazha, who has held educational seminars teaching medical students and doctors about machine learning. While computer scientists can build the machine learning models, they don’t often grasp how they are relevant to researchers who may not know the ins and outs of programming and modeling, he said.
Nazha’s confident the current gaps between computer scientists and medicine will be overcome with time.
“There’s a huge interest now in AI in health care. There’s a huge interest in AI in general,” he said. “The interest will continue to grow, and I think that’s a good thing.”