Thermal Imaging for Respiration Monitoring in Sleeping Positions: A Single Camera is Enough

Polysomnography, the current gold standard for sleep monitoring, uses multiple obtrusive contact sensors for the assessment of respiration and flow. This could partially be overcome by using thermal cameras. For accurate estimation of breathing parameters, it is important to understand how many cameras should be used and where they should be placed during overnight recordings. This study used low-cost/low-resolution thermal cameras to collect videos from 25 healthy subjects to evaluate the optimal camera placement for the monitoring of respiration. The accuracy of the camera setup was measured for a different number of cameras and positions. For the different setups, the results show that the mean absolute error ranges between 17% and 22%, the precision between 83% and 85%, the sensitivity is 96%, and the F1-scores lie between 88% and 90%. Results for all setup variations are not statistically significantly different. These findings indicate that there is no necessity to compromise between experimental feasibility and signal quality, as the various setups examined did not exhibit a significant disparity in the tested parameters.