Dynamic computed tomography to determine cardiac output in patients with left ventricular assist devices

Abstract

The purpose of this study was to develop and test a novel, noninvasive means of estimating cardiac output in patients with continuous-flow left ventricular assist devices. Based on the indicator dilution principle, we estimated cardiac output from signal intensity versus time recordings during first-pass imaging with contrast-enhanced computed tomography. To account for recirculation, a gamma variate function was modified to model the indicator concentration curve. A proportionality constant for 64-slice computed tomography data acquisition was derived relating computed tomography Hounsfield units to indicator concentration. Validity of the equation was tested in ambulatory patients with continuous-flow left ventricular assist devices who underwent simultaneous computed tomography and Swan-Ganz catheter thermodilution-based measurement of cardiac output at various left ventricular assist device rpm settings. Agreement between measurements was testing using kappa statistic. Pairwise comparison of calculated output using the dynamic computed tomography test bolus technique versus thermodilution output measurements yielded good agreement (P = .03). The output calculated using dynamic computed tomography underestimated the thermodilution output measurement by 0.54 +/- 0.37 L/min (95% confidence interval, 0.66-0.94). Noninvasive estimation of cardiac output from left ventricular assist devices is feasible using first-pass dynamic computed tomography. By affording simultaneous good visualization of the device in situ, computed tomography may be useful for noninvasive assessment of location and function of ambulatory patients with left ventricular assist devices.