Part I | Chapter 7 134 Patients were eligible for SBRT treatment in case of low- or intermediate-risk prostate cancer (NCCN classification), IPSS £ 20, and good clinical condition (WHO performance status 0-2). No restrictions with respect to age and/or prostate size were applied. Seven patients with high-risk prostate cancer, including two patients with cT3a prostate cancer, were treated off-protocol at the physician’s discretion (Table 1). Treatment was delivered in five fractions over the course of 2.5 weeks (two fractions per week). A dose of 36.25 Gy was prescribed to the Planning Target Volume (PTV) (Supplementary A). The Clinical Target Volume (CTV) included the prostate body, the Gross Tumour Volume (GTV) with a 4 mm margin excluding OARs, and up to 1/3rd of the seminal vesicles. An isotropic CTV to PTV margin of 5 mm was applied. All patients were treated with a so-called ‘Adaptto-Shape’ workflow.9 In short, during each fraction, an initial daily T2-weighted 3D MR scan (MRinitial) was acquired. Contours were propagated non-rigidly from the pre-treatment MR to the current daily MR scan. Next, the operator visually checked the propagated contours and – if needed – manually adapted them.18 After contour approval, treatment plan optimisation was initiated. Before the end of plan optimisation, an additional MR scan was acquired for position verification (PV) purposes (MRPV). In all patients, an additional virtual couch shift (VSC), also known as ‘Adapt-to-Position’ (ATP), was applied before treatment delivery in case of prostate shifts of > 1 mm between MRinitial and MRPV (for details see 19). Dose accumulation and dosimetry parameters The accumulated dose (Doseacc) was reconstructed using the daily dose distributions (Dosefx1-5) corresponding to the daily PV scans (MRPV1-5), so that the latest anatomy prior to beam-on time was considered (Figure 1). As previously reported, the time between MRPV and beam-on in our cohort was on average 5 min compared to approximately 27 min between MRinitial and beam-on. 19 Therefore, dose calculation on MRPV will ensure a better estimation of the actual delivered dose as compared to usingMRinitial or the pre-treatment plan only. For image registration purposes and dosevolume analyses, the inner and outer bladder wall structures were delineated by two physicians on MRPV1-5 (Supplementary B). The bladder was defined as the entire volume circumscribed by the outer bladder wall, including the bladder content. The bladder wall was defined as the hollow structure bordered by the outer and inner bladder wall contours. An in-house developed deformable image registration (DIR) algorithm and pipeline was used (‘EVolution’20), which was extended with dose accumulation possibilities.21 The algorithm was chosen based on its previous successful employment for registering longitudinally-acquired MR images in prostate cancer patients.21–23 The delineated outer and inner bladder wall contours were used to guide the registration process. Dosefx1-5 for each scan (MRPV1-5) was calculated based on the corresponding treatment plan (Figure 1). Next, MRPV2-5 were registered to MRPV1 (reference), resulting in four deformation vector fields (DVF) for each patient. The DVF were used to map the dose distribution to the reference image using the energy-per-mass transfer technique.21 Dice similarity coefficient (DSC) and Hausdorff distance (HD) were calculated for the bladder and bladder wall structures (delineated versus propagated) to evaluate the DVF of MRPV2-5 to MRPV1. For each case, the mean DSC and HD (DSCmean and HDmean) over the four registrations were calculated. Outliers were identified based on DSC and HD and visually checked to assess the location and extent of mis-
RkJQdWJsaXNoZXIy MTk4NDMw