When will doctors and well-trained algorithms work together to get a patient the fastest and most appropriate care?
The question struck Maithra Raghu with renewed intensity when, after severely injuring herself skiing, she found the medical system unable to quickly pinpoint the exact source of her pain.
As a senior researcher at Google Brain, Raghu has been working for years to build an interface between clinical artificial intelligence systems and human medical experts. Her injury — which turned out to be a torn ACL — spurred her to work toward a future where tools driven by machine learning and AI can help diagnose or triage patients so they can get the care they need faster.
“My accident hammered home that there are shortages of medical professionals who can get inundated with so much information that triage is really important,” Raghu said.
To Raghu, that doesn’t mean a dystopian world where superintelligent AI tools wipe out human doctors. Rather, she envisions a more practical future where AI systems serve to enhance clinicians’ inherent humanity, both by taking over menial tasks like data entry and by flagging potential issues on an X-ray or an MRI — such as the spot in Raghu’s knee where she’d been hurt skiing.
One of Raghu’s recent projects involved presenting a set of clinical cases separately to a group of doctors and a trained AI system and soliciting their opinions. Then, they showed the cases to the doctors and the AI tool together so that the clinicians could take the tool’s take into account.
“My accident hammered home that there are shortages of medical professionals who can get inundated with so much information that triage is really important.”
“The combined approach performed way better than either one alone,” Raghu said. “I’m super excited about the ramifications of that.”
— Erin Brodwin
