2 Clinical use of 4D CBCT: ITV volume and amplitude study

Abstract

Introduction After the validation of the respiratory correlated CBCT using a dynamic thorax phantom, investigations on the accuracy of the 4D CBCT were made on clinical data (i.e. patients 4D CBCT). A first study consisted in the evaluation of the difference in terms of positioning corrections between 3D CBCT and 4D CBCT images. Then, a second study compared the ITV volume between different imaging modalities (4D CT, 3D CBCT and 4D CBCT) and estimated the interfraction changes in breathing amplitude. Materials and Methods Five patients with a lung tumor treated in ITV strategy were included in the first study. They had both 3D CBCT and 4D CBCT pre-treatment images. A total of 44 CBCT images were retrospectively analyzed. A 4D CBCT image was first acquired and a 3D region-of-interest (ROI) mask registration was computed and applied in order to pre-position the patient. A 3D-CBCT image was then acquired and 3D ROI mask registration was computed but not applied. Only the patients treated with SBRT were included in the second study. A physician delineated ITV on different average images (4D CT, 3D CBCT and 4D CBCT) using a MonacoSim (Elekta) workstation. These contours were compared using a dice coefficient. Tumor motion amplitude from daily 4D CBCT was directly obtained with the on-line reconstruction software Symmetry (Elekta) and the amplitude variation was then computed. Results The average residual positioning corrections between 4D and 3D CBCT images were 0.06, 0.08 and 0.07 cm for LR, SI and AP direction respectively; and were respectively 0.41, 0.62, 0.21° for pitch, roll and yaw. From our preliminary results, the dice coefficients found were 0.77 ± 0.13, 0.70 ± 0.14 and 0.84 ± 0.11 between 4D CT and 4D CBCT, 4D CT and 3D CBCT and 3D CBCT and 4D CBCT images respectively. Interfraction changes in breathing amplitude were negligible in the LR and AP directions. The average breathing amplitude variation was 0.31 cm in SI direction (maximum of 0.44 cm). Conclusion Similar positioning corrections were found using a 3D ROI registration on either a 3D CBCT image or a 4D CBCT image. Since 4D CBCT allows a better visual inspection, it is now systematically used for lung SBRT patient positioning. ITV contours study suggests an overestimation of the ITV from 4D CT images compared to 4D CBCT images. The dice coefficient showed a better correspondence between ITV from 4D CT compared to 4D CBCT than between ITV from 4D CT and 3D CBCT. However, a satisfying correlation between 3D CBCT and 4D CBCT ITV contours was found. Interfractional changes in breathing amplitudes need to be further investigated and compared to treatment margins.