Summary 203 adaptations in conjunction with a radiation oncologist. To assess the feasibility, RTTs performed the contour check and adaptations in 30 patients treated with 5 x 7.25 Gy. For the analyses, we focused on the target contour: the Clinical Target Volume (CTV). One-hundred-and-fifty CTV contours, created by RTTs, were included. The contours were independently judged by two physicians (observers) and manually edited if deemed necessary, creating three CTV contours per fraction (1 RTT, 2 observers). The analysis of all contours (RTTs and observers) consisted of three parts: [1] the relative volume differences of the delineated CTV contours were compared between the RTTs and both observers. [2] The interobserver Dice’s Similarity Coefficient (DSC) was calculated between the RTTs and the observers as well as between both observers. [3] A senior physician judged all RTT contours based on acceptability for clinical treatment. The results of this study showed that there was a high agreement between the observer and the RTT contours, as reflected by the high DSC values. This was supported by the fact that in 94.2% of the cases, the RTT contours were directly clinically usable. Furthermore, for only one of the outliers (for one fraction out of five) there was a significant impact on the CTV coverage of the observer-adapted contour. Therefore, it was concluded that contour editing and approval by RTTs is feasible for prostate cancer treatment when RTTs are sufficiently trained. Automation of the online workflow Since manual contour adaptation for the daily adaptive treatment turned out to be quite labourintensive and time-consuming, as shown in chapter 2, our work focussed on optimising the online clinical workflow. By improving the interfraction propagated target and OAR contours, the workload could potentially be reduced. This is mainly important when, during a single fraction, contour propagation will be used multiple times to update the treatment plan. In chapter 3, we described and assessed the quality and clinical usability of propagated contours that were created by an inhouse developed deformable image registration (DIR) algorithm (‘EVolution’). For ten prostate cancer patients, intrafraction contour propagation was performed using the daily MR images. In total, 60 imaging pairs were available for analysis. Two physicians (observers) judged all contours based on two criteria: [1] The need for and extent of manual adaptation of each of the three contours (CTV, rectum, and bladder), ranging from “none” (score of 1) to “multiple major adaptations needed” (score of 4). And [2] the feasibility of adapting all contours within 3 min. Results were stratified by the time between two MR scans (< 10 min versus ³ 10 min, ‘short’ versus ‘long’). The results showed that EVolution produced excellent results for the CTV and bladder contours, which needed none or only minor editing in most cases (³ 97%). For the rectum, the observers judged that more extensive editing was needed in 12-23%. Concerning criterium [2], adaptation times were estimated to be < 3 min for ³ 93% of the cases. Stratification by time interval generally showed better results (i.e. less extensive editing needed) for the short interval cases. The EVolution algorithm is well suited for integration into an online adaptive workflow, as it is fast, accurate, and easy to use for the operator since the algorithm does not need online configuration once it has been configured for a particular MRI sequence. These promising results pave the way for exploring online adaptive MRI-guided workflows that use intrafraction DIR, deformable contour propagation, and re-planning. Clinical feasibility of such workflows still needs to be tested. 10
RkJQdWJsaXNoZXIy MTk4NDMw