Session Item

When rigidly registering daily Cone Beam CT (CBCT) to the planning CT (pCT), the derived positioning offsets will be dependent on the Region of Interest (ROI) and/or Degree of Freedom (DOF) selected.  We aim to investigate the geometric and dosimetric impact of ROI/DOF selection for proton treatments of skull-based and Head-and-Neck (H&N) tumours. 

For 7 different ROIs (Fig.1(a)) and for both 3D and 6D DOF corrections (3DOF: only translation; 6DOF: translation and rotation), pCT was rigidly registered (in Eclipse) to the CBCTs from fx1 and fx30 for 4 skull-base and 4 H&N cancer patients respectively.  To quantify positioning accuracy, Mutual information (MI) and Normalized Cross Correlation (NCC) between pCT and each transformed CBCT (trCBCT) were calculated. Three bony landmarks were defined by a radio-oncologist, and their Euclidean distances between pCT and trCBCT were computed. To estimate dosimetric consequences of residual positioning errors, the initial plan parameters, including isocenter, field direction and structures were transformed to the CBCT coordinate, based on the inversed transformation of the averaged registration results from all scenarios. The pCT was transformed to the CBCT coordinate as well, but based on individual inversed rigid registration of each ROI/DOF scenarios. Using MatRad, daily dose distributions were then re-calculated on each inversed transformed pCT by considering each derived positioning offset. Finally, dose variations due to the ROI/DOF induced rigid registration differences were quantified as DVH uncertainty bands and voxel-wise max-min dose difference distributions (Fig.1(d-g)).

Positioning errors from 6DOF are generally smaller than 3DOF corrections for all patients, whereas variations between MI and NCC are less than 0.5% for all 16 cases (8 patients x 2 CBCTs) (Fig.1(b)). The image similarities between pCT and CBCT2(fx30) are slightly lower than pCT and CBCT1(fx1) due to the larger anatomy changes during therapy. The mean residual landmark distances (Fig.1(c)) are 1.1±0.4/1.5±0.6 mm for 6/3 DOF scenarios respectively, with variations being more pronounced for 3DOFs among both patients and ROI scenarios. Moreover, among all patients (Fig.2(a)(b)), the dosimetric differences on PTV-D95 are less than 2%, except for one case (3DOF of patient 4 CBCT2). However, noticeable uncertainties in DVHs were observed for small OAR’s (e.g. mean/max dose of 1.4/5% for  brainstem). Highest dose variations were more often observed in the high dose gradient region (Fig.2(e)). 

We have developed a framework to evaluate the impact of ROI/DOF definitions on CBCT based patient positioning from image, geometric and dosimetric aspects. The results indicated that ROI selection only marginally influences the dosimetric result for proton treatment of skull-based and H&N tumours. However, enabling the 6DOF daily positioning offset calculation is important to reduce residual position uncertainties.