Respiratory motion prediction from CBCT image observations using UKF

In this paper, we propose a prediction model for breathing pattern based on observations from CBCT raw projection images. From the raw CBCT projections the diaphragm apex position is measured, which in turn is used for the state estimation. We use a novel state space model followed by an Unscented Kalman Filter (UKF). Our method is compared with one of the successful models called Local Circular Motion (LCM). The initial results show that, our model outperforms the LCM model in terms of prediction accuracy.