The Enterprise Innovation Institute’s Brandy Nagel is part of a team that received $100K to study racial bias in infrared medical devices
Brandy Nagel, program manager for the Georgia Minority Business Development Agency Business Center, attended what she thought was an information session on artificial intelligence for social justice. Three hours later she was part of a team that brought home $100,000 to study and correct racial bias in infrared medical devices.
Temporal (forehead) thermometers and pulse oximeters became everyday pieces of medical equipment during the COVID-19 pandemic. But studies show these devices and others that use infrared technology are not as accurate when used on people with darker skin as they are on people with lighter skin. Pulse oximeters miss low oxygen levels in 11% of Black patients and temporal thermometers miss 23% of fevers in Black patients. These inaccuracies have immense public health implications, not just in the U.S., but around the world.
“I showed up because it was described as artificial intelligence for social justice,” Nagel said about the AI.Humanity with a Social Justice Lens Pitch Competition. “I know almost nothing about AI, but I understand social justice as a problem to be solved. I thought the event would help me understand how AI can be applied.”
Nagel was expecting a lecture or a seminar at the event, which was sponsored by the Emory-Georgia Tech collaborative research seed grant program, AI.Humanity, and held at Coda at Tech Square.
“When I got there, there was a conversation already going, so I joined a group,” she said. “We were talking about a problem and how AI could be applied to solve that problem. And then the organizers said, go to your breakout rooms. You’ve got 45 minutes to write your pitch.”
Despite knowing little about the medical technology under discussion, Nagel knew she could contribute to the team. “I know about pitches, and I know how to prepare for a pitch,” she said. She was able to help hone the idea for the pitch committee.
Funding for the grant comes equally from Emory and Georgia Tech and teams include members from both institutions. Nagel’s team includes from Emory University: Dr. Sivasubramanium Bhavani, assistant professor of medicine in the Emory University School of Medicine; Dr. Michele Sumler, associate professor of anesthesiology in the med school; and Cassidy Puckett, assistant professor in the Department of Sociology. The members from Georgia Tech are: Molei Tao, associate professor in the School of Mathematics; Sheila Isbell, research scientist in the Georgia Tech Research Institute Information and Communications Laboratory (GTRI-ICL); and Nagel.
Isbell was drawn to the event because of the opportunity to explore using AI for the good of humanity.
“In much of the news, people are concerned about how AI is taking jobs or about bias in AI,” Isbell said. “But I was interested in working with other people to see how we can use AI to help. I was also interested because it’s a collaboration between Georgia Tech and Emory, and we’ll be doing something in the health and wellness sphere, which I really love to do.”
With Nagel at the white board, the team mapped out its approach. “We talked about where would we get the data? Because this is something where the data is a significant part of figuring out how to solve the problem,” Nagel said. “Is the solution in how the device is designed? Or is the solution in a different policy? We don’t know.”
Isbell’s forte is dealing with data, making her an integral part of a team that has so many different perspectives, a strength that made them successful in their pitch, she said.
“We had sociologists. We had an anesthesiologist. We had someone who does AI. I do data,” she said. “It was important that we had the diversity of thought and experience in our team for this project. If we could help doctors know about what their devices are doing, then they can make better decisions. We don’t want people to mis-interpret the signals about their patients, therefore causing more health disparities.”
Following the 45-minute discussion period, the team was ready. The pitch covered three objectives with the goal of creating more equitable medical devices:
-
- Perform a systematic review of infrared technologies and devices that are currently used in healthcare systems to understand the scope of the problem.
- Conduct a pilot observational study to measure the discrepancy between the information provided by infrared medical devices and that of gold standard devices.
- Develop an AI algorithm to calibrate medical device measurements with skin pigmentation using results from the pilot study.
With the pitch competition won, what happens next?
Nagel, who isn’t a medical device expert, does understand design thinking. “The first step of design thinking is empathy,” she said, “deeply understand the customer or the beneficiary’s experience, not just their point of view, but what they experience. In this case, the beneficiary might be a patient, but it could also be a medical professional who’s trying to quickly triage in an emergency room and figure out who needs attention first.
“One thing that I think is exciting about this is that our team can be informed by going into an emergency room and seeing how this works,” she said. “What’s the real-world experience that may help inform how we solve the problem? I think it’s AI. But I think there’s also going to be a personal touch.”
A second team also won $100,000 for its pitch AI-Assisted Social Justice in Tissue and Organ Biomanufacturing. The team, comprised of both Georgia Tech and Emory researchers, will use domain-specific AI-based approaches to develop tissue bioprinting processes that are optimized for patients of various racial and ethnic backgrounds.
“I think what really intrigued me about this is that I saw it as a solvable problem and something that could have a significant impact,” Nagel said. “The pulse oximeter and the forehead thermometer, these two devices were used to help diagnose millions of people with COVID, particularly in the early days of the pandemic. If we look back on the deaths, and we see that a disproportionate number of darker skinned people died, then we might say it was because of a bad diagnostic tool. That sounds like an important problem to solve.”