A quantitative comparison of the performance of three deformable registration algorithms in radiotherapy

Abstract

We present an evaluation of various non-rigid registration algorithms for the purpose of compensating interfractional motion of the target volume and organs at risk areas when acquiring CBCT image data prior to irradiation. Three different deformable registration (DR) methods were used: the Demons algorithm implemented in the iPlan Software (BrainLAB AG, Feldkirchen, Germany) and two custom-developed piecewise methods using either a Normalized Correlation or a Mutual Information metric (featureletNC and featureletMI). These methods were tested on data acquired using a novel purpose-built phantom for deformable registration and clinical CT/CBCT data of prostate and lung cancer patients. The Dice similarity coefficient (DSC) between manually drawn contours and the contours generated by a derived deformation field of the structures in question was compared to the result obtained with rigid registration (RR). For the phantom, the piecewise methods were slightly superior, the featureletNC for the intramodality and the featureletMI for the intermodality registrations. For the prostate cases in less than 50% of the images studied the DSC was improved over RR. Deformable registration methods improved the outcome over a rigid registration for lung cases and in the phantom study, but not in a significant way for the prostate study. A significantly superior deformation method could not be identified.

Mitigating differential baseline shifts in locally advanced lung cancer patients using an average anatomy model

“…Weistrand et al reported the target registration error of their ANACONDA DIR algorithm implemented in their treatment planning software RayStation (RaySearch laboratories) for CBCT to CT registration in lung, which was 1.17 ± 0.87 mm in vector length when the lungs are used as the focus regions for the DIR, and 3.11 ± 3.23 mm when no focus regions is used, which compares reasonably with our results . In another study, they have compared three DIR methods (the Demons algorithm and two custom‐developed piecewise methods using either a normalized correlation or a mutual information metric) for CBCT to CT registration of lung and reported the Dice index and Hausdorff distance in terms of the number of pixels for a few specific organs . Veiga et al used the diffeomorphic Morphon algorithm for adaptive proton therapy for lung cancer based on the validation of this algorithm for head and neck cases .…”

“…33 In another study, they have compared three DIR methods (the Demons algorithm and two custom-developed piecewise methods using either a normalized correlation or a mutual information metric) for CBCT to CT registration of lung and reported the Dice index and Hausdorff distance in terms of the number of pixels for a few specific organs. 34 Veiga et al used the diffeomorphic Morphon algorithm for adaptive proton therapy for lung cancer based on the validation of this algorithm for head and neck cases. 35 Han et al assessed the dosimetric uncertainty on the DIR algorithm of MIM (MIM Software Inc., Cleveland, OH), and found this to be negligible.…”

A design of a DICOM‐RT‐based tool box for nonrigid 4D dose calculation

“…Studies 25 , 26 , 27 have evaluated the performance of DIR registration and showed that deformable image registration not only improves the accuracy of image localization but also the accumulated target dose distribution as compared to the ridge‐based image registration. Janssens et al (27) verified the accuracy of DIR for 4D dose accumulation at deformation field based on phantom‐based dosimetric measurement.…”

Gradient maintenance: A new algorithm for fast online replanning

“…Particularly, for organs with low soft tissue contrast and/or situations with significant deformations between different imaging sessions, these algorithms almost always fail and generate clinically unacceptable results. [7][8][9] This would necessitate manual contouring or at least editing/verification of autogenerated contours, which is a rather lengthy process. GM largely alleviates this problem, by not requiring the OAR contours on the daily images.…”