Illinois Medicine

Volume-24 - Summer 2023

Issue link:

Contents of this Issue


Page 16 of 43

Illinois Medicine | 15 Artificial intelligence (AI) and machine learning (ML) are rapidly transforming healthcare and medical practice. With the establishment of our new Center for Health Equity using Machine Learning and Artificial Intelligence in November 2022, the College of Medicine is harnessing the power of these emerging technologies to reduce health disparities in our communities through research, training, education and innovation. UR BRAINS DO A LOT OF COMPUTING— figuring, sorting, processing—in ways that we don't necessarily fully understand," says Niranjan S. Karnik, MA '95, MD '02, PhD '03, visiting professor of psychiatry and interim director of the new Center for Health Equity using Machine Learning and Artificial Intelligence (CHEMA). "Similarly, the computer also can be set to search data for patterns. "The computer is able to lean into mass data in ways that we can't usually predict beforehand because the amount of information is so vast, often comes from different sources and is, most times, delivered faster than can be synthesized in a formal way," Karnik points out. "With AI and ML, it doesn't matter what you're looking for as long as you tell the computer what is a true case and what is not: The computer will try to find connections and patterns from past data to try to predict these endpoints." To start, researchers define a training-data set to look at what they want to study. Using COVID-19 as an example, if a computer is programmed to look for patients with respiratory illness in the ICU, it will offer a multitude of responses, as COVID-19 looks like many other respiratory illnesses. The training data needs to be as specific as possible so as to distinguish COVID-19 from pneumonia or the flu, for instance, and then set to search for those patterns that create risk. According to Karnik—who is also B Y C A R L A B E E C H E R Harnessing for Health Equity director of the Institute for Juvenile Research and co-director of the Digital Mental Health Program in the department of psychiatry—some results can be anticipated: Older adults and people living in group settings are more likely to be at risk. But he found other unanticipated risks in his own research that looked at COVID-19 patients with opioid use disorder: "We found that there was almost a direct correlation between the degree of opioid use and the severity of the COVID-19 illness among hospitalized patients." But what really surprised him was that, as a computer searches for patterns, it can become just as biased as the people who develop the training data. "Sometimes when we code data, we build in our biases," Karnik explains. AI "O

Articles in this issue

Archives of this issue

view archives of Illinois Medicine - Volume-24 - Summer 2023