Dose Accumulation Based on Optimized Motion Field Estimation Using Non-Linear Registration in Thoracic 4D CT Image Data

Respiratory motion is a main issue in current radiation therapy of lung cancer. To adequately account for respiratory motion a detailed knowledge about breathing dynamics and its impact on dose distributions is necessary. Within this context we propose a non-linear registration scheme to estimate motion fields in 4D CT images of lung cancer patients. We focus on handling discontinuities in lung and chest wall motion as motivated by the physiology of breathing. Accuracy and plausibility of the registration scheme are evaluated. Based on 3D conformal radiotherapy (3D CRT) treatment plans the optimized registration method is applied for dose accumulation purposes, i.e., to simulate the impact of breathing motion on the patient-specific dose distribution. Designed as a proof-of-concept study, results are obtained for 5 patients with tumors of different size and location. The motion field estimation is shown to be significantly improved both in terms of registration accuracy and plausibility when accounting for the discontinuities in lung and chest wall motion. The mean target registration error of 1.6±1.1 mm is in the order of voxel size. Based on the optimized registration scheme, dose accumulation yields that motion-induced effects on the dose distribution highly depend on the patient-specific tumor mobility. Qualitatively, the dose distributions are shown to be blurred due to breathing motion. Quantitatively, voxel-wise differences between the planned dose distribution and the distribution obtained by dose accumulation range from <5% (patient with tumor motion of 2 mm) up to 47% (tumor motion: 12 mm) of the prescribed dose.