International Business Machines Corp.’s decision to explore a sale of its Watson Health business highlights the challenges of building data sets that unlock the full value of artificial intelligence in solving healthcare problems.
From the start, IBM said massive data sets were core to its mission to help doctors tackle big problems like diagnosing and curing cancer. A company spokeswoman Tuesday said that IBM Watson Health is committed to leading the market in healthcare data sets, and that Watson Health offers databases that include “real-world, longitudinal clinical, operation and financial data and analytic tools.”
Even so, some experts found that it can be difficult to apply AI to treating complex medical conditions. Having access to data that represents patient populations broadly has been a challenge, experts told the Journal, and gaps in knowledge about complex diseases may not be fully captured in clinical databases.
“I believe that we’re many years away from AI products that really positively impact clinical care for many patients,” said Bob Kocher, a partner at venture-capital firm Venrock who focuses on healthcare IT and services investments and who was a White House health adviser under President Barack Obama.
Software that makes recommendations on personal medical treatments needs data on what actions have worked in the past. But data on medical histories and treatment outcomes aren’t always complete, may be recorded in different formats, and may be sitting in various systems owned by insurance carriers, health providers and other organizations.
It is hard to build an AI technology that tells somebody what cancer treatment they should get, “because if you’re wrong, they’re going to die,” Dr. Kocher said. “So it needs to be perfect.” Since the data used to train the AI model is imperfect, the tool will fall short of expectations, he said.
Some problems just aren’t computable, according to Dr. Andrew Rosenberg, chief information officer of Michigan Medicine, a health system affiliated with the University of Michigan.
Dr. Rosenberg said he met with Watson eight years ago. “One of the issues that came out of those discussions was computation,” he said. “How do you compute the 25-year-old saying they don’t want the Covid vaccine? How do you compute that and predict that?”
“If you look at where there’s been a lot of success in medical, it’s been around using AI to analyze an MRI, or an X-ray,” said Stephen Messer, co-founder and vice chairman of Collective[i], an AI company focused on optimizing the sales process in a range of markets.
Once the AI is trained, he said, “it’s very good at whatever it needs to find. It’s repetitive. These are the places where you’re having huge wins.”
Despite the broader challenges, the application of artificial intelligence to healthcare could find more success in certain areas, some experts say.
“What’s getting people excited is neural network technology,” Mr. Messer said.
Neural networks are composed of layers of interconnected artificial “neurons” that automatically learn about the features of a specific object based on large amounts of data.
Almost all the successful AI players are using networks of data to leverage neural nets, Mr. Messer said.
Wearable devices are collecting huge amounts of data, Mr. Messer said. Devices are going to start to identify early on that someone is at risk of getting atrial fibrillation, he said. A new owner of Watson Health should look at different ways it could leverage an increased healthcare data pool, and then apply new technologies such as neural nets, he added.
Health and fitness applications can collect vast amounts of data from people with smartphone and smartwatches and represent one avenue to apply neural networks. Researchers are looking to see if AI can analyze heartbeats, sleep patterns and speech to identify everything from heart problems to early-stage dementia.
As more companies figure out how to pool sets of health data to create even more scale, the application of deep learning to healthcare may continue to develop, according to Mr. Messer.
—David Uberti contributed to this article.
Write to John McCormick at firstname.lastname@example.org
Copyright ©2020 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8