Respiratory Motion Compensation with Topology Independent Surrogates

We present a method for organ motion compensation based on a statistical motion model. The novelty of our method is, that the surrogates for prediction can be independent of the model’s topology i.e. the surrogate signal does not have to correspond to a point of the motion model. By non-linear regression, we grasp the correlation between a captured signal during free breathing and the motion model parameters. Such a signal could be a spirometer, a breathing belt or an abdominal ultrasound signal. Our method performs on par with stateof-the-art methods, where model points have to be tracked. However, we solely incorporate one-dimensional and even topology independent surrogates. In our experiments on the right liver lobe we have achieved an average motion prediction accuracy of 2-3mm using population models and below 1mm with patient specific models.