The rate of premature ventricular contractions (PVC rate) and the dynamics of its response to the intravenous infusion of the antiarrhythmic drug lidocaine was modeled to facilitate the design of a closed-loop drug infusion system. An autoregressive moving average model structure with the lidocaine infusion rate as an exogenous input (ARMAX structure) was used. System identification experiments were performed using a canine model of myocardial infarction. The drug infusion rate was varied using a pseudorandom binary sequence (PRBS) test input to excite the system about an operating point established by an exponential infusion regimen. PVC’s were detected in real time with a correlation technique. Model parameters were estimated in an offline data analysis. First-order models provided an adequate representation of the PVC rate in eight of ten runs. A second-order model was validated for two runs. © 1987 IEEE
