SinaInnovations 2019: A Close-up Look at Artificial Intelligence
Keynote speaker Michael Snyder, PhD
What is the role of artificial intelligence (AI) in medicine, and how is it changing the practice of health care as we know it? That was the subject of the eighth annual SinaInnovations Conference, held Tuesday, October 15, and Wednesday, October 16, in Stern Auditorium. The event featured leading physicians and scientists from academia and industry who spoke about their work in deploying AI—the most powerful technology under development—to augment discovery and clinical use.
Experts shared their experiences in using AI in a variety of ways, from medical imaging, to predicting disease, to keeping people healthy, and highlighted the massive transformation taking place in health care and medicine, where software is driving innovation.
Michael Snyder, PhD, Chair of the Department of Genetics, and Director of the Center for Genomics and Personalized Medicine at Stanford University, a keynote speaker, discussed the role of AI in advancing personalized medicine. “I see a world where, with genome sequencing and continuous monitoring using wearable devices, we can better manage people’s health and hopefully do this at an individual level, and have personal machine-learning algorithms that follow people and their health state,” Dr. Snyder said. “We’re very capable of measuring more things, and here’s an area where AI can make a big impact.”
Melissa A. Haendel, PhD, Director of Translational Data Science at Oregon State University, spoke about her work in leading the federally funded Monarch Initiative, which is building sophisticated algorithms that integrate a multitude of data about rare diseases in order to improve research and clinical care. “No one group is actually annotating a disease model that has all the same attributes,” Dr. Haendel said. “We can’t even count the number of rare diseases.” Her team’s goal, she said, is to pull all of the data together and use it to build models that help physicians make earlier diagnoses, identify biomarkers of disease, and find better treatments.
Conference participants Heather J. Lynch, PhD, Associate Professor of Ecology and Evolution, Stony Brook University, left, and Georgia D. Tourassi, PhD, Director, Health Data Sciences Institute, Oak Ridge National Laboratory.
David Sontag, PhD, Associate Professor of Electrical Engineering and Computer Science at Massachusetts Institute of Technology, discussed how AI can be used to redesign electronic medical records so they can yield more reliable information on the patient’s risk for various diseases. In one case, he said, his team developed a machine-learning algorithm to help an infectious disease clinician at Massachusetts General Hospital and Brigham and Women’s Hospital reduce the number of unnecessary prescriptions for antibiotics.
AI is already playing a role in augmenting radiology. Keith J. Dreyer, DO, PhD, Vice Chairman, Radiology, at Massachusetts General Hospital and Chief Science Officer of the American College of Radiology, told the audience that “AI has huge value” and will be increasingly useful over time as the field matures.
In his keynote address, Pieter Abbeel, PhD, an entrepreneur and Professor of Electrical Engineering and Computer Science at the University of California, Berkeley, showed how deep machine learning takes place through constant repetition. In one example, he illustrated how a robot learns to run. After 2,000 iterations, it will become proficient. By comparison, a healthy human child would learn to run proficiently after roughly two weeks of practice. In many cases, he said, machines have achieved human-level error rates.
Among the many algorithms Stanford University is working on is one that recognizes the photo of a radiological image taken with a mobile phone, according to Curtis Langlotz, MD, PhD, Director of Stanford’s Center for Artificial Intelligence in Medicine and Imaging. This technology would allow general practitioners and other health care professionals in remote areas to use their mobile phones to access an algorithm that would assist them in making medical decisions when a radiologist is not available. For example, they would be able to determine whether a patient with, say, tuberculosis, should be discharged from the hospital.
Speakers and attendees at the conference agreed that AI is both promising and challenging. Suresh Venkatasubramanian, PhD, Professor of Computing at the University of Utah, cautioned that inherent bias in the data will create bias in the algorithms. “Models are fragile,” he said. “The Achilles heel is that the more sophisticated a system gets, especially with deep learning, the more sensitive it gets to small perturbations, and this could wreak havoc on the system.”
Greg Zaharchuk, MD, PhD, Professor of Radiology (Neuroimaging and Neurointervention) at Stanford University, concluded his talk with a nod to the future. “I think we’re only scratching the surface. This is a moment of extreme creativity, and it’s a very exciting time to be in the field.” Rather than replacing radiologists and other medical specialists, he added, AI “is really going to extend our abilities as physicians.”
New Gift Supports Young Entrepreneurs at Mount Sinai
This year, for the first time, a nonprofit biotech accelerator company founded by five former postdocs at the Icahn School of Medicine at Mount Sinai presented the school with a five-year, $50,000 gift to support young entrepreneurs in the New York City area whose science is being used to create therapies, devices, and diagnostics that support human health. The gift from The Keystone for Incubating Innovation in Life Sciences (KiiLN) went to Raymond A. Alvarez, PhD, Assistant Professor of Medicine (Infectious Diseases) at the Icahn School of Medicine at Mount Sinai, who is working on a platform that identifies and studies the antibodies of individuals who are immune to hantaviruses, which are spread by rodents and have a 38 percent mortality rate. Currently, there are no vaccines or treatments for hantaviruses.
