Non-invasive estimation of left ventricle elastance using a multi-compartment lumped parameter model and gradient-based optimization with forward-mode automatic differentiation (2205.12330v1)

Abstract: Accurate estimates of left ventricle elastances based on non-invasive measurements are required for clinical decision-making during treatment of valvular diseases. The present study proposes a parameter discovery approach based on a lumped parameter model of the cardiovascular system in conjunction with optimization and non-invasive, clinical input measurements to approximate important cardiac parameters, including left ventricle elastances. A subset of parameters of a multi-compartment lumped parameter model was estimated using 1st order Adam gradient descent and hybrid Adam and 2nd order quasi-Newton limited-memory Broyden-Fletcher-Goldfarb-Shanno optimization routines. Forward-mode automatic differentiation was used to estimate the Jacobian matrices and compared to the common finite differences approach. Synthetic data of healthy and diseased hearts were generated as proxies for non-invasive clinical measurements and used to evaluate the algorithm. Twelve parameters including left ventricle elastances were selected for optimization based on 99% explained variation in mean left ventricle pressure and volume. The hybrid optimization strategy yielded the best overall results compared to 1st order optimization with automatic differentiation and finite difference approaches, with mean absolute percentage errors ranging from 6.67% to 14,14%. Errors in left ventricle elastance estimates for simulated aortic stenosis and mitral regurgitation were smallest when including synthetic measurements for arterial pressure and valvular flow rate at approximately 2% and degraded to roughly 5% when including volume trends as well. However, the latter resulted in better tracking of the left ventricle pressure waveforms and may be considered when the necessary equipment is available.