Researchers explore how AI tracks breathing, predicts air quality

In two recent studies, University of Texas at Dallas researchers demonstrated how artificial intelligence (AI) and machine learning can be used to address a variety of issues from a social science policy perspective.

Dr. Dohyeong Kim, a researcher in the School of Economic, Political and Policy Sciences (EPPS), and collaborators in South Korea have developed a wearable stethoscope that uses AI to monitor a patient’s breathing sounds for wheezing. Kim is also part of a team using machine learning to predict levels of airborne bacteria and fungi in indoor environments.

“In EPPS, we have multiple scholars working on AI issues,” said Kim, a professor of public policy, geospatial information sciences (GIS), and social data analytics and research, and senior associate dean of graduate education for EPPS. “AI applications have been primarily the domain of computer scientists or engineers, but it is getting more important to understand how AI can be applied in social science, health care, education, the environment and other areas.”

Kim and his colleagues previously developed a novel AI-based method for counting wheezing events in patients that can indicate breathing trouble that needs medical attention. The wearable stethoscope, described in a new article in the journal Engineering, is a wireless, skin-attachable, low-power device that includes a lung-sound monitoring patch (LSMP).

The LSMP monitors respiratory function through a mobile app and classifies normal and problematic breathing by comparing their unique acoustic characteristics. In the study, which included corresponding authors from South Korea, the LSMP sensor was tested in pediatric patients with asthma and elderly patients with chronic obstructive pulmonary disease (COPD).

The AI-based breathing-event counter was able to distinguish more than 80% of abnormal events, especially wheezing, in the COPD patients.

“In the previous study, we developed a method of training the algorithm with the wheezing sounds, but at the time we had not fully developed the wearable devices,” Kim said. “With the stethoscope fully developed, we can use this AI algorithm to automatically detect in real time whether the breathing sounds are normal. We can monitor and see the intensity and frequency of those wheezing sounds.”

In a related study, published in the Feb. 15 issue of the journal Building and Environment, researchers used machine learning to examine the combined effect of temperature and humidity on indoor bioaerosol concentrations.

“We found that we can use the temperature and humidity as a good predictor of the potential presence of bacteria and mold,” Kim said.

Exposure to airborne bioaerosols, such as bacteria and fungi, presents significant health risks, especially for vulnerable populations like children, the elderly and those with compromised immune systems. Bioaerosol exposure can aggravate respiratory and allergic conditions, underscoring the need for real-time monitoring in indoor environments.

The researchers analyzed data collected from 4,048 samples across 10 types of multiuse facilities, including day care centers and libraries in South Korea, and showed that temperature and humidity jointly and significantly affected concentrations of bacteria and mold.

Kim said the findings provide guidelines for controlling indoor bioaerosol levels and creating safer and healthier indoor environments by adjusting temperature and humidity.

In addition to Kim, Gloria Geevarghese BS’24 is an author of the Building and Environment study, along with researchers from Yonsei University, Seokyeong University and Korea University.

Additional authors of the Engineering study included researchers from the Korea Institute of Science and Technology, Ajou University, Kosin University College of Medicine and Seokyeong University.