Comparison of video-based methods for respiration rhythm measurement

The aim of this work is to characterize the di erences in the respiratory rhythm obtained through three video based methods by comparing the obtained respiratory signals with the one obtained with the gold standard method in adult population. The analysed methods are an RGB camera, a depth camera and a thermal camera while the gold standard is an inductive thorax plethysmography system (Respiband system from BioSignals Plux). 21 healthy subjects where measured, performing 4 tests for each subject. The respiratory rhythm and its variability was obtained from the four respiratory signals (3 video methods and gold standard). The signal acquisition was performed with custom and proprietary algorithms. To characterize the respiratory rhythm and its variability obtained with the di erent video sources and gold standard, the instantaneous frequency, Bland-Altman plots and standard deviation of the error between video methods and the gold standard have been computed. The depth and RGB camera present high agreement with no statistical di erences between them, with errors when comparing with the gold standard in the range of mHz. The thermal camera performs poorly if compared with the two other methods, nevertheless it cannot be discarded directly because some errors produced by the subjects head movement could not be corrected. From these results we conclude that the depth and RGB camera, and their respective acquisition algorithms) can be used in controlled conditions to measure respiration rhythm and its variability. The thermal camera on the other hand, although it can not be discarded directly, performed poorly if compared with the other two methods. Further studies are needed to con rm that these methods can be used in real life conditions.