Time-Varying Parameter Identification of a Tumor Growth Model Using Moving Horizon Estimation

Abstract

A nonlinear Moving Horizon Estimator (MHE) was developed which can estimate the time-varying parameters of a tumor growth model under chemotherapeutic treatment. We introduce a sequential estimation strategy using the Full Information Estimator (FIE) that is able to approximate an estimate to the average initial model parameters. The algorithm penalizes the estimation error and the deviation of parameters between each consecutive iteration of the FIE. We also describe the tuning process in detail, where we utilized a grid search process to find the best choice for the parameters of the MHE. The algorithm was tuned and validated using time-series data, originating from in vivo mice experiments.