“ISI has been working in the health area for many years, and we’ve done very significant work,” said Yolanda Gil, reflecting on USC Viterbi Information Sciences Institute’s (ISI) leadership in managing large-scale data through National Institute of Health (NIH) centers totaling $85M in funding. Now, as director of ISI’s Center on Artificial Intelligence Research for Health (AI4Health), Gil is set to accelerate AI research advances to solve pressing health challenges by leveraging ISI’s depth and breadth of expertise in AI.

A recent $1 million award from USC’s Frontiers of Computing initiative will help drive this mission. Gil, who has been an active member of the center since it was created in 2021, said: “USC is investing in the potential for AI to catapult new AI innovations that will improve patient outcomes.”

“Now is the right time to make a major push in the area of AI and health,” said Craig Knoblock, the Keston Executive Director of ISI. “There are important health problems to be solved, and AI has matured to a point where there are many opportunities where it can be naturally applied.” The challenges, he said, “are getting the right people together from both the AI side and the health side, and finding the right problems to focus on.” ISI has a world-class AI research group, widely-known for seminal contributions in language processing, machine learning, computer vision, causal models, and knowledge technologies.  The center will continue to build on these strengths while leveraging ISI’s strong partnerships with USC health sciences.

As ISI’s Senior Director for Artificial Intelligence and Data Science Initiatives, and an ISI Fellow (the institute’s highest honor), Gil sees AI4Health as a collider of talent, connecting AI researchers with clinicians and health practitioners who are interested in pushing the envelope of AI advances for their field.

Promoting Healthier Lives with AI

“We’re doing research on a broad range of areas that affect our health beyond doctor visits—how to improve nutrition, exercise, and recovery; prevent stress; and help people follow treatments and take medication properly,” Gil explained. This means using AI to study human behavior, focusing on both physical and mental well-being. “We study people’s behaviors and actions in all these aspects of health, but also their attitudes and degree of awareness,” she noted, emphasizing AI’s untapped potential in helping people understand what small changes in their daily routines can meaningfully improve their long-term health.

The center’s data efforts include improving fairness, equity, and accessibility of care. “We have to be very careful about characterizing the scope and nature of the data, and remedying any biases that could affect AI algorithms,” Gil said. AI4Health researchers are also tackling the challenges of multimodal information—such as text, images, and medical records—which require AI systems to connect and integrate varied formats; and federated learning, which allows AI algorithms to access significantly more data by being trained on health records from many locations without compromising individuals’ privacy.

AI-Ready Data to Discover Critical Knowledge

Another key aspect of AI4Health’s work is making health data accessible to AI researchers. “We need what we call AI-ready data: data that is ready for AI to learn and draw insights from,” said Gil. “Data isn’t just tables,” she emphasized. “It’s dynamic images, it’s timeseries, it’s physical objects, it’s medical concepts, all connected together.  And these connections enable AI to get the full picture and discover critical causal knowledge.”

What’s Next for AI4Health?

Gil is passionate about the transformative potential that AI technologies hold. “AI is ready to have a positive, profoundly transformative impact in the world and we can’t run away from that,” she said, pointing to AI algorithms that significantly improve early detection and diagnostic accuracy. “If an AI algorithm can diagnose cancer years before a human radiologist can, the irresponsible thing is to not deploy the AI.” And in a field where mistakes carry life-altering and even life-threatening consequences, Gil underscores that the center’s research in ensuring transparency, fairness, and safety in AI applications is not just important—it’s essential.

Published on November 12th, 2024

Last updated on November 12th, 2024