Biotechnology HPC Software Applications Institute (BHSAI)

BHSAI’s New Algorithm Could Help Reduce the Risk of Heat Injuries

December 31, 2018   |  Download PDF

We often overestimate our ability to stay active in scorching weather—hence, our surprise when we hear of people dying of heat stroke during a summer heat wave. Warfighters who take part in military training or combat missions under such conditions face a similar threat, as their risk of incurring heat injury increases (~2000 Service members/year for the past several years).

A marked increase in core body temperature heightens the risk of heat injury. In principle, we could mitigate the risks posed by intense physical activity in hot and humid weather by developing a wearable system that accurately measures core body temperature and alerts the wearer before it increases to dangerous levels. However, such a system has been difficult to implement in practice.

One reason is that core body temperature is difficult to measure accurately. The rectal temperature probe has long been the gold standard. Measurements obtained at other sites (mouth, armpit, etc.) do not accurately reflect the time course of changes in core body temperature; hence, probes placed at such sites are inadequate for alerting the individual in a timely fashion. A major problem of rectal probes, however, is that they are invasive, uncomfortable, and impractical for long-term monitoring in the field.

Schematic of BHSAI’s AI algorithm to provide individualized estimates of core body temperature in real time, using non-invasive physiological and environmental data.Schematic of BHSAI’s AI algorithm to provide individualized estimates of core body temperature in real time, using non-invasive physiological and environmental data.


Fortunately, the increasing power of commercial devices, such as smartwatches, for monitoring physiological measurements, has raised the prospects of developing a non-invasive system that uses multiple physiological measurements to estimate core body temperature in real time. What has remained lacking, however, is a mathematical model—simple enough to be computationally efficient, but also powerful enough to account for the large individual differences in sensitivity to heat stress.

In the June issue of the Journal of Applied Physiology, a team of scientists from TATRC’s Biotechnology High Performance Computing Software Applications Institute (BHSAI), led by Dr. Jaques Reifman and supported by the Military Operational Medicine Research Area Directorate of the U.S. Army Medical Research and Materiel Command, described a model that meets these requirements. The model incorporates a newly developed artificial intelligence (AI) algorithm that uses vital signs from a smartwatch and adjusts the model parameters to estimate an individual’s core body temperature. In doing so, the algorithm automatically “learns” how each individual responds to heat stress.

With multiple collaborators at the Uniformed Services University of the Health Sciences, Sheba Medical Center, Israel Defense Forces Institute of Military Physiology, Tel Aviv University, University of Otago, and University of Pimorska, the BHSAI team used the model to retrospectively analyze core body temperature data for 166 subjects trained under different environmental and exertional heat-stress conditions in three separate studies. They showed that the algorithm could automatically learn how each of the subjects responded under the different conditions. Dr. Srinivas Laxminarayan, a BHSAI staff scientist and lead author of the study, added that, “the new AI system clears the path to develop a practical non-invasive system that will help reduce the risk of heat injury among Service Members.”


This article was published in the July 2019 issue of the TATRC Times.


Links