Mayo Clinic physicians Chieh-Ju Chao and Chia-Chun Chiang will work at Stanford on innovations in AI and medicine.
The development of artificial intelligence (AI) systems for health care often finds data scientists and doctors on opposite sides of a frustrating information divide. A new collaboration with the Mayo Clinic brings physician team — and married couple — Chieh-Ju Chao and Chia-Chun Chiang to the Stanford Institute for Human-Centered AI (HAI) to learn how to bridge the gap.
Improving Treatment for Migraine Patients
At the Mayo Clinic in Rochester, Minnesota, Chia-Chun Chiang splits her medical practice between treating patients who’ve suffered acute stroke and migraine sufferers who come to her to seek help with the intense headaches that can sometimes incapacitate them for weeks, months, or longer. Chiang, a specialist in headache medicine and vascular neurology, not only hopes to improve these daily realities for her patients but also wants to explore how AI may someday lead to better overall treatment through exciting new breakthroughs.
For the next year, Chiang will be a visiting scholar at Stanford HAI, where she’s exploring the intersection of AI and neurology. The appointment is part of a new collaborative program between HAI and the Mayo Clinic and will give Chiang a chance to do what she’s always loved — find innovative solutions to challenging medical questions.
Born in Taiwan, Chiang found her medical horizons were expanded early in her career, both by her neurosurgeon father and a series of influential mentors on two continents.
“My father always encouraged me to think not just about day-to-day realities in medical practice but about making research breakthroughs in medicine — about discovery,” she says. “Arthur Chiou, a professor of biophotonics at National Yang Ming University, opened a window for me to see different types of medical research. Later, I came to the University of California, San Diego, and met Professor Shu Chien in the department of bioengineering. He got me into the world of using math and physics to solve medical research problems. It was inspiring.”
Chiang’s time at Stanford HAI will be spent on several research projects focusing on her special interest in migraine and stroke. She hopes to use AI to help doctors determine which of many preventive migraine medications might help chronic migraine sufferers — those experiencing migraine attacks more than 15 days per month — stave off debilitating attacks.
“Right now there’s no way to predict if a patient will respond to medication A, B, C, D, or E,” she says. “It’s a trial-and-error process, where we start with one medication and wait three months to see if it helps. This can be a long time and a long process when you’re in pain. I’m hoping to be able to apply advanced machine learning technologies to the Mayo Clinic’s established headache database to help predict good matches or to develop personalized treatment plans for patients with migraine.”
Chiang is also collaborating with Nigam Shah, a Stanford professor of medicine and chief data scientist for Stanford Health Care, to be able to more effectively utilize all of a patient’s electronic health record data to understand which migraine patients might be at risk of stroke and cardiovascular disease.
“Things like clinician notes, consultations, and admission notes are written in free text, which traditionally have been hard for data scientists to access,” she says. “Dr. Shah’s lab has developed several advanced machine learning tools tailored to health care notes that can extract and use this sort of information. We hope to be able to apply these tools to databases at the Mayo Clinic and at Stanford to give us some solid research results from two different health care systems.”
Chiang says AI is changing the field of neurology through innovations such as telemedicine and AI-assisted CT analysis, both of which benefit stroke patients, particularly those in isolated regions of the country where small hospitals may not have neurologists on staff. Continuing the introduction of relevant and user-friendly technology into health care, however, will require improved communication between those who develop AI and those who treat patients.
“I’m very excited about the use of AI in my field, which is why I wanted to come to HAI and contribute to this process by stimulating more discussion and work between data scientists and clinicians,” she says. “I see patients who are suffering every day in clinic. It’s crucial for us to continue imagining how AI tools can be developed to improve their quality of life.”
Developing Better Tools for Echocardiography
Mayo Clinic cardiologist Chieh-Ju Chao was a teenager in Taiwan when the college entrance exam set him on a trajectory to medical school. Deep down, though, the 18-year-old harbored a secret ambition.
“I was happy to go into medicine, but I always loved physics,” he says. “It was always in the back of my mind. I love the beauty of the laws and formulas in physics and, as a kid, I dreamed of being the next Einstein or Feynman.” After medical school, Chao took postgraduate physics courses in Taiwan before becoming a postdoctoral fellow at UC San Diego, where he led a research project to build a 3D dynamic heart model. He began to see intersections between physics and medicine, particularly in the field of cardiology. “It was one of the most important periods of my life and one that really changed my career path,” he says.
For the next year, Chao will have a chance to deeply explore his interdisciplinary interests, as a visiting scholar at the Stanford Institute for Human-Centered Artificial Intelligence. As part of a collaborative program between Stanford HAI and the Mayo Clinic, facilitated and guided by HAI Co-Director Fei-Fei Li, Chao will expand his current research on the potential use of AI technology to augment echocardiography, which harnesses ultrasound waves to examine the anatomy and dynamics of the heart. Chao hopes his dual background in medicine and technology will allow him to bridge the knowledge gap that often exists between health care and technology professionals and to accelerate the expansion of AI systems into clinical medicine at the Mayo Clinic and elsewhere.
“The usual pathway for this has always been for clinicians to come up with an idea, then talk to AI engineers,” he says. “But that process can take a long time because communication between the two disciplines is not easy. For the first time, this program between Mayo and HAI sends clinicians out into the AI field to learn some of these skills and take them back. I knew this was something I wanted to do. I see myself as a person who will be able to speak both languages.”
At the Mayo Clinic, Chao studies how AI models might be used to solve clinical questions, such as how to classify different grades of cardiac diastolic dysfunction, which occur when the heart’s damaged ventricles are unable to completely fill with blood. (That work, in fact, resulted in Chao being named a “Young Investigator Award” finalist at the 2022 American College of Cardiology Conference.)
At Stanford HAI, he’ll focus on two primary research projects. The first will incorporate the use of AI to give cardiologists a more realistic view of the heart during procedures such as aortic and mitral valve replacements, which are often done with the use of catheters rather than open-heart surgery. Positioning these catheters currently relies on the 2D images provided by echocardiogram and X-ray. Chao envisions AI models that might be able to compile that information with CT images of the patient’s heart, giving doctors a real-time 3D and even hologram-based picture that would make the procedure smoother and more efficient.
“Our goal is to improve success rates, minimize complications, and lessen radiation exposure through shorter procedures,” Chao says. “It could potentially shorten procedure times from five or six hours to two or three and allow a more efficient use of our procedure room resources.”
Secondly, Chao will explore the potential use of AI visual linguistic models to generate time-consuming echocardiography reports more efficiently. Ideally, the model would have the capacity to do an initial interpretation of the study and generate a preliminary report that could then be reviewed by clinicians.
“This would be an advancement in current AI application in the medical field,” Chao says. “People have been trying to create systems that can report on chest X-rays, which are a single image. In an echo study, we’re dealing with video clips that record the cardiac dynamics from different views. It’s an even more exciting challenge.”
Chao says he’s motivated both by colleagues at the Mayo Clinic who are enthusiastic about this work as well as the range of scholars and technical resources he’s finding at Stanford HAI.
“It’s amazing to work with cutting-edge AI scientists and the latest technology,” he says. “Talking about my clinical ideas with colleagues here in this setting is a very different and immersive learning experience for me, and these conversations are generating new ideas as I learn about what people are doing here and about potential applications I hadn’t thought about. It’s all really exciting.”
Stanford HAI’s mission is to advance AI research, education, policy and practice to improve the human condition. Learn more.