Validity of contact skin temperature sensors under different environmental conditions with and without fabric coverage: characterisation and correction.

ABSTRACT

Contact skin temperature (Tsk) sensors are calibrated under uniform thermal conditions but used in the presence of a skin-to-environment temperature gradient. We aimed to characterise the validity of contact Tsk sensors when measuring surface temperature under a range of environmental and fabric coverage conditions, to estimate practical temperature limits for a given measurement bias and to explore correcting for bias. Using two types of contact Tsk sensors (thermistors, n = 5; iButtons, n = 5), we performed experiments in three phases (1) conventional calibration (uniform thermal environment) over 15-40 °C in 5 °C steps (at t = 0, and 24 h, 12 weeks later), (2) surface temperature measurements of a purpose-made aluminium plate (also 15-40 °C) at different environmental temperatures (15, 25, 35 °C) with different sensor attachments and fabric coverings to assess measurement bias and calculate correction factors that account for the next-to-surface microclimate temperature and (3) surface measurements (33.1 °C in 20 °C environment) for assessing generated corrections. The main results were as follows (1) after initial calibration, Tsk sensors were valid under uniform thermal conditions [mean bias < 0.05 °C, typical error of the estimate < 0.1 °C]. (2) For the surface measurements, bias increased with increasing surface-to-microclimate temperature difference for both sensor types. The range of surface temperatures possible to remain within given bias limits could be estimated for the various conditions. (3) For a given measurement, using corrections encompassing the microclimate temperature (mean difference - 0.1 to 0.5 °C) performed better than conventional calibration alone (mean difference - 2.1 to - 0.3 °C). In conclusion, the bias of Tsk sensors is influenced by the microclimate temperature and, therefore, body coverings. Where excessive bias is expected, the validity can be improved through sensor and attachment selection and by applying corrections that account for the local temperature gradient.