Figure 5.

(a) Hausdorff distance values for the phantom CT-CT deformable registration. Here two deformations are included. It can be seen that on average the featurelet_NC method reveals best performance on the bladder and the prostate structures. The featurelet_MI method is not improving the registration result substantially and the iPlan method only improve the Hausdorff-distance values for the prostate structure. (b) Hausdorff distances for the phantom CT-CBCT deformable registration, here three deformations are included, it can be observed that for all the structures no method improves the Hausdorff distance in comparison to the rigid registration. RR corresponds to the rigid registration starting point, featurelet_NC to the featurelet deformable registration method using normalized correlation metric, featurelet_MI to the featurelet deformable registration method using mutual information metric and iPlan corresponds to the deformable registration performed using the iPlan -adaptive application.