Modeling of hemoglobin response to Erythropoietin therapy through constrained optimization

Hemoglobin response modeling is critical for accurately predicting the hemoglobin level of anemic patients and it is the foundation of optimal control based Erythropoietin (EPO) therapy. In this paper, a novel Hemoglobin response modeling method is proposed. This method was inspired from a study on the physiological model of hemoglobin response to Erythropoietin. Insights on the effect of EPO is first gained from the physiological model and then transformed into an auto-regressive with exogenous input (ARX) model parameter pattern. The ARX model parameter estimation problem is formulated as a constrained optimization problem. Tests on simulated data using the physiological model and the real clinical data both show that the proposed modeling method has superior performance while also being less complex than the physiological model.