Part I | Chapter 2 46 just one fraction (patient 4, fraction 4), the adapted CTV contours showed clear under dosage near the seminal vesicles. We have to keep in mind that these numbers are all scaled to the complete five-fraction scheme. Since this only occurred during one fraction, these effects would be smaller in reality. There are several strengths to our study. First, we have included delineation data from 30 patients, covering 150 fractions, created in the online MR-Linac setting with corresponding time pressure. Second, the retrospective nature of our study implicates that RTTs were not aware of the study. Hence, their efforts during contour adaptation were not influenced. Third, contours were not only assessed by numerical parameters, but a blinded judgement of the clinical acceptability and need for adaptations of the contours was also part of our evaluation. Together with the DVHs analyses, this allows insight into potential clinical implications. Conversely, a limitation of the study is the lack of a ground truth CTV contour for each of the fractions. We therefore chose to perform an interobserver comparison. Of course, both the adaptations by Observer 1 and 2, as well as the judgement by Observer 3, all reflect interobserver variability and we think that the clinical implications are limited. Only one of the seven fractions that were judged to need larger adaptations with potential clinical implications, showed a significant impact on the CTV coverage. Adding to this, we only calculated DSC and relative volume differences. Many other mathematical parameters exist that might help understand better how large the variation is between different observers, such as Hausdorff distance and mean absolute surface distance.14,15 However, these parameters, just like DSC, do not always correspond with clinical applicability or relevance, since a high DSC does not exclude the possibility of clinically relevant differences in a small part of the contour.20 We therefore have chosen to focus on two numerical parameters and the blinded clinical judgement by three physicians, with the inherent limitation of subjectivity. Finally, we have only assessed potential dosimetric effects for the CTV coverage and did not assess any effects on OAR doses, as this was not the primary goal of this study. Conclusion Concluding, based on our evaluation of both DSC and clinical judgement, CTV contours that are adapted and approved in the online MR-Linac setting by RTTs are well suited for radiotherapy treatment of prostate cancer. Adaptations were mostly performed in areas that are known for their interobserver variability and clinical implications are thought to be minimal. The few outliers that were observed were relatively small and mostly occurred only for a single fraction. The transition of this task from the treating radiation oncologist to the RTTs is feasible when RTTs are sufficiently trained and confident in their new task.
RkJQdWJsaXNoZXIy MTk4NDMw