Armed with reams of data, a patient’s father convinced Dr. Anthony C. Chang, his daughter’s pediatric cardiologist, to proceed with her surgery.
“A data scientist, the father of one of my congenital heart defect patients, really wanted to give me more data than we typically get so he tabulated the pulse oximetry readings on his daughter and plotted it out for me and convinced me that we needed to push ahead with surgery,” Chang said. The chief intelligence and innovation officer and medical director of the Heart Failure Program at Children’s Hospital of Orange County (CHOC), spoke Tuesday during a lecture at the University of California, Irvine, sponsored by the California Institute for Telecommunications and Information Technology (Calit2). “This is 2017. It’s just not acceptable to me that a parent has to go through this kind of data and analytics to help me make a decision.”
Doctors should be able to make those decisions more effectively, using artificial intelligence (AI), Chang said, stressing the need for humans and machines to collaborate so health care can be improved. It was Valentine’s Day in 2011, the night that IBM’s Watson supercomputer beat human contestants on “Jeopardy!” that Chang realized AI was on the scene in a big way. That night, he said, he downloaded the Stanford University data science program application and decided that “artificial intelligence is here to stay.”
The Importance of Collaboration
For health care to benefit from AI, he said, it’s important for clinicians and data scientists to spend time together in order to learn what the needs are and how to best use the tons of data now available but hard to mine efficiently.
“The hottest topic in AI and health care today is reinforcement learning,” Chang said. “Data mining is actually pretty straightforward, particularly in health care. It is the clarity and the organization of the data that’s the main problem. So, how you create an interface between clinical staff and data science is the challenge, but an opportunity also.”
It’s impossible for a human clinician to know everything about his or her specialty, especially as the amount of medical knowledge available now is doubling every 18-24 months, he added. “So, it’s not conceivable for a human clinician to effectively treat patients with the latest data without help…. Until we have a liberated data sharing policy for clinicians, patients and hospitals, we are not going to be able to do analytics the way we all like to do. We need to get more sophisticated about AI in medicine way beyond IBM Watson.”
Chang said the next generation of physicians need to work with data scientists. “I’d like to propose we start thinking about not evidence-based learning, although it has a role in medicine, but more intelligence-based learning, where it’s deep learning, reinforcement learning and analytics focused. … The world is open to you once you have the data science lens to do some incredible things.
“I’m looking forward to the day that AI will help me read the hundreds of echocardiograms I need to read,” he said, adding that in the next decade, patients will start insisting on more informed, precise treatments for their ailments. “As a cardiologist, it really doesn’t make sense for me to prescribe the same heart medication in the same dose to every patient because we’re all so different.”
Chang gave an overview of CHOC’s AIMed (Artificial Intelligence in Medicine) and how it’s addressing biomedical diagnostics, medical imaging, drug discovery, cloud computing and big data, digital medicine and wearable and robotic technology. To learn more, visit the website or attend its second International Multidisciplinary Symposium on Artificial Intelligence in Medicine conference Dec. 11-14 in Laguna Niguel. The following are a few videos from the first conference, held last year. To watch more and download the speakers’ slides, visit the website.
