Detection of differences of cardiorespiratory metrics between non-invasive respiratory support modes using machine learning methods

•A machine learning based method was proposed to characterize difference in cardiorespiratory features differed among the non-invasive respiratory support modes for preterm infants.•Eleven groups of cardiorespiratory metrics calculated from ribcage and abdominal movements, electrocardiogram, photoplethysmography and blood oxygen saturation were automatically analyzed.•The used of statistical value calculation and principal component analysis methods made it possible to examine the potential contributions of many variables without the risk of over parameterization.•The use of the random forest approach made it possible to detect nonlinear interactions among features that cannot be detected easily using classical statistical techniques.•This paper made a new finding that the amplitude of respiratory movements and correlation between the ribcage and abdomen movements contributed to distinguishing CPAP vs NIV-NAVA/NIPPV.