Detection of respiration in central venous pressure using state machine

Reliable information from patient monitors enhances treatment for critically ill patients. Redundant sources for information would aid identification of faulty sensors and leads, and improve presentation of physiological data. Respiratory information can be obtained from several sources, including airway pressure and central venous pressure (CVP). CVP signals have been analyzed using frequency information to isolate the respiration related part of the signal or to obtain statistics about respiration. This study uses a state machine algorithm to detect the timing of each cycle of respiration. A state machine has advantages of enforcing a predictable cycle of expiration and inspiration. The detection of respiratory cycles can be done in real-time, allowing identification of irregular periods between inspirations and prolonged periods with no inspiration, for which an alert may be issued. The algorithm was tested on data obtain from the PhysioNet database of recordings from intensive care patients. The airway pressure signal was used to determine the “true values” of the timing of each respiratory cycle for checking the accuracy of the algorithm analyzing the CVP signal. Parameters of the algorithm were found that would result in a true positive value of above 98% for detection of each cycle of respiration from analysis of the CVP signal, compared to analysis of the RESP signal.