Evaluation of contactless respiratory rate measurement: Thermography vs. rPPG

Respiratory rate (RR) is one of the most pivotal vital signs in the foundational assessment of patients. Its applicability spans a diverse range of healthcare sectors, from the sleep monitoring to the critical care medicine. Non-contact methods using cameras for the RR measurement can avoid discomforts of patients, match hygiene requirements better, and is more suitable for telemedicine and remote monitoring. In this work, based on a multimodal camera system, we conduct comparative study on RR measurement based on thermal signal and remote photoplethysmography (rPPG) signals obtained from multichannel (RGB, 780 nm, 940 nm, thermal) videos. We discuss the quality of RR-related information carried by these two different sources and provide respective methods for RR estimation. Experimental results demonstrate that the rPPG-based method achieved a mean absolute error (MAE) of 3.50 breath per minute (bpm) in our experimental dataset with ground truth RR ranging from 6 bpm to 42 bpm, whereas the method based on thermal signals reached a MAE of 1.21 bpm. Additionally, our method based on thermal signal demonstrated good agreement with polysomnography in a clinical validation in sleep laboratory.