Part I | Chapter 4 68 Dataset and acquisition Each MRI-guided radiotherapy fraction started with the acquisition of a T2-weighted MR-scan of 2 min. This so-called pre-treatment scan was used to assess the patient's daily anatomy, i.e. interfraction changes. Delineations from the planningMR-scan or froma previous fraction were then adapted to the current scan with the use of deformable registration and adjusted were necessary to match the current anatomy. Next, plan re-optimisation was started. Just prior to the end of the plan re-optimisation period, a position verification (PV) T2-weightedMR-scan was obtained. This scan was used to assess whether the Clinical Target Volume (CTV) was still within the Planning Target Volume (PTV). When this was the case, irradiation was started. During the beam-on period, continuous 3D cine-MR imaging was acquired with a balanced turbo field echo (bTFE) sequence. A single 3D cine-MR frame is referred to as a ‘dynamic', and these socalled cine-MR dynamics have a temporal resolution of 9.4 sec, a field of view of 448 x 448 x 45 voxels and an acquired voxel spacing of 2 x 2 x 2.2 mm.3 Exemplary slices from the cine-MR sequence are provided in Figure 1. Additional technical details of the cine-MR sequence are provided in Table S1 (Supplementary A). Average patient treatment time (beam-on period) was 10.1 min, while the full on-couch period of a patient during a single fraction was on average 40 min. An extensive workflow description including timings was described previously.5 Registration Registration was based on masks created from seminal vesicle delineations. These delineations were created by a clinician for each seminal vesicle separately on the first cine-MR dynamic of every imaging dataset. All masks were thus based on delineations of the daily anatomy. An example of these delineations is provided in Figure S1 (Supplementary B). Seminal vesicle intrafraction motion was determined by applying a mask based, rigid registration method. This method was developed from our previously published method of soft-tissue contrastbased prostate tracking13 which was applied and validated on MR-Linac data.5 Additional improvements were made to the original prostate registration algorithm as described in Supplementary A. These improvements allowed the method tomore accurately track small volumes with relatively large displacements, such as the seminal vesicles. Tracking was performed retrospectively and applied to the left and right seminal vesicle separately. The registrations were performed rigidly, and all transformations are described by a translation and rotation about the centre of mass of the mask created by the clinician. Motion in the right, posterior and cranial translation direction are given as positive values. Motion in the left, anterior and caudal direction are given as negative values. Rotation about the left-right axis is equal to pitch, whereas rotation about the anterior-posterior axis is equal to roll and rotation about the caudal-cranial axis is equal to yaw rotation. Pearson's correlation coefficients were calculated, between the intrafraction motion of both seminal vesicles and for the intrafraction motion of the seminal vesicles with respect to the prostate. Two-
RkJQdWJsaXNoZXIy MTk4NDMw