Estimation of core body temperature by near-infrared imaging of vein diameter change in the dorsal hand

Abstract

Core body temperature ( T core ) is a key indicator of personal thermal comfort and serves as a monitor of thermal strain. Multi-parametric sensors are not practical for estimating core temperature because they require long data collection times and a wide variety of settings. This study introduces dorsal hand vein dynamics as a novel indicator along with heart rate (HR) and dorsal hand skin temperature ( T hand ) for predicting T core during rest following T core elevation. Twelve healthy males aged 27 ± 9 years old participated in the experiment. The experimental procedure consisted of a 10-min rest followed by 60 min of passive heat stress induced by leg immersion in hot water at 42°C and a 40-min thermal relaxation period after the legs were removed from the water. A near-infrared (NIR) imaging system was configured to monitor the dorsal hand veins during the entire experimental session. The values of HR , T hand , and T core were continuously monitored while the ambient temperature and relative humidity (RH) were maintained in a climate chamber at 20°C and 50%, respectively. Our selected predictor parameters demonstrated similar patterns in the T core such that the value increased as a result of passive heat stress and decreased in the thermal relaxation phase. The experimental data were divided into two phases: thermal stress and relaxation. At the resting condition, inclusion of the hand vein diameter ( VD ) improved the multiple linear regression value ( R 2 ) about 26%. At the relaxation phase, however, training regressions R 2 = 0.68 and R 2 = 0.94 were observed in the regression model with and without considering VD , respectively. The test regression value of R 2 = 0.88 and the root mean square error (RMSE) of 0.18°C showed good agreement with the predicted values. These findings demonstrate acceptable validity of the non-invasive T core estimation at the resting condition. In particular, the inclusion of VD as a predictor in the regression analysis increases the prediction accuracy with a lower RMSE value.