Biomechanical quality assurance criteria for deformable image registration algorithms used in radiotherapy guidance

Image-guided radiation therapy (IGRT) allows radiation dose deposition with a high degree of geometric accuracy. Previous studies have demonstrated that such therapies may benefit from the employment of deformable image registration (DIR) algorithms, which allow both the automatic tracking of anatomical changes and accumulation of the delivered radiation dose over time. In order to ensure patient care and safety, however, the estimated deformations must be subjected to stringent quality assurance (QA) measures. In the present study we propose to extend the state-of-the-art methodology for QA of DIR algorithms by a set of novel biomechanical criteria. The proposed biomechanical criteria imply the calculation of the normal and shear mechanical stress, which would occur within the observed tissues as a result of the estimated deformations. The calculated stress is then compared to plausible physiological limits, providing thus the anatomical plausibility of the estimated deformations. The criteria were employed for the QA of three DIR algorithms in the context of abdominal conebeam computed tomography and magnetic resonance radiotherapy guidance. An initial evaluation of organ boundary alignment capabilities indicated that all three algorithms perform similarly. However, an analysis of the deformations within the organ boundaries with respect to the proposed biomechanical QA criteria revealed different degrees of anatomical plausibility. Additionally, it was demonstrated that violations of these criteria are also indicative of errors within the dose accumulation process. The proposed QA criteria, therefore, provide a tissue-dependent assessment of the anatomical plausibility of the deformations estimated by DIR algorithms, showcasing potential in ensuring patient safety for future adaptive IGRT treatments.