Heat illness detection with heart rate variability analysis and anomaly detection algorithm

Incidence of heat illness has been increasing dramatically due to the progression of global warming. Preventing severe heat illness, called heatstroke, is crucial because it can lead to long-term multiple organ damage, including the brain, and results in more than 600 deaths per year in the United States. It has been reported that heat stress affects heart rate variability (HRV), which is the fluctuations of the R-R interval (RRI) on an electrocardiogram (ECG). We propose a method for detecting symptoms of heat illness based on HRV analysis in order to prevent exacerbation of heat illness. In the proposed method, monitoring abnormal changes in HRV caused by heat stress is monitored. Multivariate statistical process control (MSPC), a commonly used anomaly detection method in machine learning, is adopted for training the heat illness detection method. To validate the proposed method, we recruited 103 healthy volunteers with risks of heat illness development: employees working in hot environments, athletes, and amateur marathon runners. Data collection was performed using our wearable heart rate sensor and smartphone app. The result of applying the proposed method showed that a sensitivity of 75% (21 out of 28 cases) and a false-positive rate of 1.02 times per hour were achieved. The proposed heat illness detection method will be used in daily life because RRI data can be easily measured by a wearable sensor. The proposed method will contribute to receiving appropriate treatment for heat illness before exacerbation, which contributes to protecting people’s health. [Display omitted] •An HRV-based heat illness detection AI algorithm was proposed.•An anomaly detection framework was adopted for constricting the AI.•The proposed AI achieved sensitivity of 75% and FP rate of 1.02 times/h.•We developed a custom wearable heart rate sensor and smartphone app.