Development of Algorithm to Extract Features from Capnograph Signal

Cardiopulmonary diseases are generally detected using spirometry and peak expiratory flow meter which is tedious for a patient for the fastest blow required. Hence doctors recommend using Capnography which detects the CO2 saturation level from exhaled breath. This paper gives an algorithm to detect vital features from the Capnographic datasets (available in capnobase) which might be suitable for classifying cardiopulmonary diseases like Asthma, COPD, hypoventilation, hyperventilation, etc. Some of the signals are composed of artifacts while some are of spontaneous breath which makes it difficult to identify and extract the features properly. We propose a method to extract a total of 25 features from the Capnograph automatically using digital signal processing. We also identified the significant differences for apnea and cardiogenic oscillation. Among the features, some of them are slopes that properly fit with the original signal with 99.95% sensitivity for regular Capnogram and about 97-99% for irregular Capnogram.