Advancing Medical Diagnosis with Machine Learning and Digital Cameras October 25, 2022 In the News Artificial IntelligenceHealth Sciences Share this page: Twitter Facebook LinkedIn Email From left, Richard Sowers, Rachneet Kaur and Manuel Hernandez. Photo by Fred Zwicky. By Megan Meave Johnson Imagine you live in a rural area. The nearest hospital is 50 miles away, and the nearest neurosurgeon is 150 miles away. It takes months, maybe even years, to get an appointment. You know if you had a diagnosis, you could get treatment, but relief will have to wait until you can see that specialist. This may seem extreme, but for people suffering from undiagnosed multiple sclerosis or Parkinson’s disease, especially those in areas where neurosurgeons are in short supply, it’s the reality they have to live with. This was the problem that a team of researchers led by Manuel Hernandez, a professor of kinesiology and community health from the University of Illinois Urbana-Champaign wanted to solve. Rounding out the team was graduate student Rachneet Kaur and industrial and enterprise systems engineering (ISE) and mathematics professor and NCSA and CDA affiliate Richard Sowers. The team used digital cameras to capture the gait of adults, both with these diseases and without. Using treadmills and digital cameras, they gathered hours of video data to be analyzed. By watching certain locations on the moving bodies, Kaur was able to develop a method for tracking a pattern of body movement over time. She then fed her collected data into more than a dozen machine-learning and deep-learning algorithms to test how accurate her method was. When the team applied their approach to new test subjects, they found several of these algorithms resulted in an impressive 75% accuracy at detecting MS and Parkinson’s from the observed gait of a patient. The NCSA is uniquely positioned to carry out research like this. Machine learning computations, like the ones in this work, depend on an incredible number of calculations to train various mathematical models. I’m always grateful to have the NCSA resources available at the University of Illinois. Richard Sowers, NCSA and CDA affiliate, professor of ISE & Mathematics This breakthrough study will help to improve access to diagnostic tools for these diseases in a quick and relatively inexpensive way, making access to treatment more expedient for patients waiting for those hard-to-find specialists. Shorter wait times for diagnoses are good news for everyone, but especially for patients who have access to so few doctors in their area.You can read more about this study in the full article published by the Illinois News Bureau.