Computer-assisted Doppler monitoring to enhance detection of air emboli.
Currently, two of the most sensitive clinical approaches commonly used to monitor for venous air embolism, i.e., precordial Doppler audio and capnography, require the attention of the anesthesiologist's eye or ear, which is a distraction from other aspects of care. To assess the feasibility of allowing the computer to relieve the necessity for continuous human monitoring, we developed a computer algorithm for monitoring the precordial Doppler audio. This algorithm extracted (1) the amplitude of certain higher-frequency components of the Doppler audio, (2) a measure of the average value of the envelope of Doppler audio, and (3) the ratio between the average value of the Doppler envelope and the amount of envelope signal variation at heart rate frequency and its multiples. These three features were monitored by an adaptive pattern recognition algorithm that compared each new value for each feature with the previously developed mean and standard deviation for that feature. If the changes in the three features exceeded a detection threshold, an alarm (indicating suspected air embolism) was activated. Implemented as a prototype system, the algorithm was given preliminary testing in 2 dogs and activated alarms at levels of air well below those reported to cause clinically significant hemodynamic changes in dogs. While decreasing the distraction for the anesthesiologist, this early prototype alarm system alerts its user to the need for analysis of the Doppler signals when it senses an air embolus.