Part I | Chapter 4 70 Labelling behaviour types All cine-MR imaging sets were visually inspected and labelled to identify the major behaviour type of the seminal vesicle intrafraction motion. Datasets in which gas pockets were present that influenced the seminal vesicles were labelled as ‘gas pockets’. Datasets in which a drift occurred, due to gradual bladder filling and/or relaxation of the patient were labelled as ‘drift’. For cases were peristalsis of the rectum or shifting rectum filling caused intrafraction motion, the label ‘peristalsis’ was used. Remaining datasets with no visually identifiable motion behaviour type and overall small amounts of seminal vesicle motion (less than 2 mm translation) were labelled as ‘stable’. Exemplary motion paths from these groups are provided in the results. Coverage probability analysis The effect of intrafraction motion of the seminal vesicles and prostate during a fraction was determined using a coverage probability analysis. Such analysis describes the frequency of each target voxel being covered by a certain volume and this principle was previously described in literature.14,15 This concept is useful because it allows taking into account rotations and deformations, whereas conventional margin recipes focus on translations only. Analysis was performed using isometric volume expansion in steps of 1 mm (range 1-10 mm), surrounding the original delineations created by the clinician. As described previously, the obtained intrafraction motion for each cine-MR dynamic was applied with a transformation to the original mask. This results in the position of the targets corresponding with each cine-MR dynamic. A check was then performed to verify if any part of the transformed masks reached outside the expanded volume at any time point. All voxels reaching outside the expanded volume were labelled, allowing to calculate the coverage probability for each point in the original masks. An example of this principle is provided in Figure 4. The coverage probability analysis was performed for each target individually in all cine-MR frames and all fractions. The resulting numbers describe the percentage of volume of each target that was covered at least 95% of the time. Population systematic (å) and population random (s) graphs for the seminal vesicles and prostate centre of gravity were calculated and are provided in the Supplementary C (Figure S2 and S3). The method to determine these errors was described in a previous study.16 The population systematic error describes the spread of the mean displacements over the group of patients at any time point. The population randomerror expresses the corresponding fluctuation with respect to these patientspecific means.17 Moreover, based on these population systematic and random errors for the translation directions, calculations with the van Herk formula (2.5å + 0.7s)18 were performed and are provided and are provided in Supplementary D (Table S2).
RkJQdWJsaXNoZXIy MTk4NDMw