Page 31 - Organ motion in children for high-precision radiotherapy - Sophie Huijskens

Image registration
We collected and stored the imaging data in a centralized database for image analysis. We used Elekta X-ray Volume Imaging (XVI) software (version 4.5; Elekta Oncology Systems) for a two-step rigid registration performed by a single experienced observer. First, a region of interest (ROI) was defined encompassing the vertebral column representative for the chosen area closest to the kidney. The CBCTs were then registered with the refCT using the automatic chamfer match algorithm. This first match of the bony anatomy of each CBCT enabled a consistent quantification of the organs with respect to the bony anatomy, irrespective of daily positioning variations. The kidneys were, prior to registration, delineated as separate ROIs including the whole kidney volume. This enabled to obtain a resolution greater than the slice thickness acquired in the refCT. Second, the kidney was automatically registered using the shaped ROI and a grey value algorithm. Results of the automatic procedure were inspected by the observer and when automatic registration failed, for example due to artifacts, manual registration was conducted (n=81/622 registrations). Registrations resulted in interfractional deviations relative to bony anatomy, expressed as composite vectors in the left-right (LR), cranio- caudal (CC) and anterior-posterior (AP) directions, for the right and left kidney separately. The + and – signs respectively indicate right/caudal/posterior and left/cranio/anterior directions. Registration of the diaphragm was only feasible in the CC direction. The observer registered each CBCT manually. The anterior-posterior plane used for diaphragm registration in each CBCT was chosen at the position of maximum kidney length. Measurements were corrected for rotations by assessing the center of mass (COM) coordinates for each kidney. We equated these coordinates to the refCT to determine the exact distance and direction of the interfractional deviations.
Statistical data analysis
For each patient, we determined organ specific mean, absolute median and standard deviation (SD) of the interfractional motion relative to bony anatomy for each direction. Inter-patient variation was analyzed by considering the spread of all patients’ interfractional kidney motion in the three orthogonal directions. To investigate the equality of variance between patients in each direction and studied organs, we tested the variance of each single patient versus the overall group variance using an F-test (p < 0.01 was considered as statistically significant). For the patient group as a whole, the results of the interfractional motion are expressed as the group mean, the group systematic error (Σ; the SD of the individual means of all patients) and the group random error (σ; root mean square value of the individual SDs) [6], for each direction.
Linear regression analysis
We used linear regression analysis to assess the relationship between renal and diaphragmatic motion and patient-specific factors (including age and height) and to investigate the possible correlation between renal and diaphragmatic motion (p < 0.05 was considered as statistically significant). We also divided the patient cohort in two subgroups based on median values and analyzed the effects of age and height on the systematic and random error. Since tumors had been surgically removed before radiation treatment, we have not investigated a possible relation between organ motion and tumor size. All statistical analyses were done using R version 3.2.1. (R Foundation for Statistical Computing, USA).
28