Calculating Cardiovascular Lumped-Parameter Model Values by Injecting Small Volume Perturbations in an Isovolumic Heart
Abstract
Diagnosing cardiac patient problems contains many uncertainties, and to fully diagnose the patient's condition usually requires a lengthy drug regimen to see what works and what does not. Compounding this problem is that even after the correct drug regimen has been discovered, the underlying cause for the problem may remain a mystery. Thus, the uncertainty and the length of time required to provide an accurate and adequate solution makes it very difficult to provide quality care to the patient. Templeton and others have shown that lumped cardiac muscle parameters can be extracted from an isolated heart by injecting small volumes at high frequencies relative to the heart rate and measuring the pressure response to this volume change. Using the Hill muscle model of two springs and a dash pot to portray the different elements of the cardiac muscle, the pressure and volume relationship makes it possible to calculate these muscle parameters using frequency response analysis techniques. The hypothesis to be tested is "Is it possible to develop a method to test cardiac muscle for stiffness, resistance, and contractile force from measuring ventricular pressure and injected flow?" To test this hypothesis, an isovolumic heart model is developed and allowed to develop pressure, along with a small volume injected to create a pressure response. Analysis of the pressure and flow waveforms produces a measured value of the cardiac model parameter values to compare to the model values. Results from injecting small volume changes into a mathematical heart model show that it is possible to extract the muscle model parameters of non-linear resistance, inertia of the fluid and muscle, and stiffness of the muscle while filling and contracting. The injected frequency and volume were varied to find usable conditions, both with regard to the calculations and the practical limits. Analyzing the error between the measured and model values for a large number of different combinations of model parameters shows an average error of less than 1%.