Page 44 - Organ motion in children for high-precision radiotherapy - Sophie Huijskens

Reference CT and CBCT imaging
According to tumor-based treatment protocols, pretreatment CT scans for planning purposes were acquired for all patients. These scans were considered the reference for organ position (i.e., refCT), and included 3D-CTs (for 60 patients) and 4D-CTs (for one child and nine adults) (LightSpeed RT16; General Electric Company, Waukesha, WI, USA). Slice thickness varied from 2.5 to 3.0 mm; 10 children had 3D-CTs with 5.0 mm slice thickness. The 10 breathing phases of the 4D-CT were averaged to simulate a 3D-CT. For all patients, CBCT images were routinely acquired for setup verification before radiation delivery (Synergy, Elekta Oncology Systems, Crawley, UK). In general for children, an accustomed extended no-action level (eNAL) protocol was used [19]. This yields CBCT imaging and online correction at the first three radiation treatment fractions after which the a-priori set-up correction is adjusted, to be checked at the fourth fraction. From the fifth fraction on the eNAL protocol is followed, acquiring weekly imaging, unless eNAL results exceed tolerance limits [19]. In adults treated for esophageal, gastric, and pancreatic tumors, daily CBCTs were acquired for online position verification. Pediatric CBCT acquisition time varied between 35 and 60 s, with a 200° or 360° rotation, respectively. The acquisition time for adult CBCT was 120 s with a 360° rotation.
Image registration
Imaging data were collected and stored in our database for image analysis. Elekta X-ray volume imaging (XVI) version 4.5 software (Elekta Oncology Systems, Stockholm, Sweden) was used for two- step rigid organ registrations as described previously [17]. First, after defining a region of interest (ROI) including six to seven vertebrae at the level of the diaphragm and the kidneys, the CBCT was registered to the refCT using the automatic chamfer match algorithm for bony anatomy to account for daily setup variations. Bony anatomy registration was then followed by registration of the left and right kidney separately, using distinct ROIs based on delineations including the whole kidney volume, enabling the assessment of smaller values of organ position variation (greater resolution) than the slice thickness acquired in the refCT. For each patient, renal position variation was assessed in the cranial-caudal (CC), left–right (LR) and anterior–posterior (AP) directions. Registration outcomes were visually evaluated and (manually) corrected if necessary. To correct interfractional position variation for rotations, we assessed the center of mass coordinates for the left and right kidney separately, and compared them to the refCT in order to determine the distance and direction of interfractional position variation. Diaphragm position variation was considered as a surrogate for abdominal organ position variation, and also assessed using the two-step rigid registration method. After the automatic registration of the vertebral column as described for renal registration, the diaphragm as one complete structure was manually registered in the CC direction only.
Statistical analysis
Per patient, the mean and SD of interfractional position variation were calculated in three directions for the kidneys, and in the CC direction only for the diaphragm. We calculated median 3D vector lengths as a summarized measure of renal position variation, and used the non-parametric test for independent samples to test for differences between children and adults (significance level p<.05). Subsequently, the group means, group systematic errors (Σ; the SD of patients’ individual means), and group random errors (σ; the root mean square of the individual SDs) for children and adults were
41