Session Item

We have recently introduced ultra-fractionated SBRT for prostate cancer (5 x 7.25 Gy) on a 0.35T ViewRay MRIdian MR-Linac system. As part of the MR-guided radiotherapy (MRgRT) workflow, a simulation scan is performed on the MR-Linac using a 3D balanced steady-state free-precession (aka TrueFISP) imaging sequence. This scan differs substantially from a standard 2D T2w Turbo spin-echo (2D T2w-TSE) 1.5T MRI simulation scan that is typically used in our conventional, non-MRgRT, workflow. The aim of this study was therefore to investigate whether this transition has an impact on the delineations performed by the radiation oncologist. To our knowledge this is the first study investigating the influence on delineation between 1.5T MRI-sim and 0.35T MRgRT imaging.

Both 1.5T T2w-TSE (res. = 0.8 x 0.8 x 3 mm) and 0.35T TrueFISP (res. = 1.5 mm3 isotropic) scans were acquired in 10 patients with localized prostate cancer. For this interim analysis, 4 experienced radiation oncologists (RTOs) delineated the prostate CTV and base of the seminal vesicles in five patients on both scans. The 0.35T scan was delineated first during which the RTO was allowed to have access to the unmatched diagnostic MRI scan (not MRI-sim) as per clinical workflow. The 1.5T scans were delineated in a second round, once all the 0.35T scans had been delineated.
With the lack of a histopathological ground truth, the intra- and inter-modality consistency between multiple observers and the potential discrepancy in delineated volumes were assessed. Intra-modality differences were calculated by comparing the specialists’ contours within one patient. Inter-modality differences were calculated by comparing the differences between a specialists’ 0.35T and 1.5T contours within one patient. Differences in contours were quantified by three metrics: CTV volume differences, Dice similarity/mismatch coefficients and Hausdorff distances.

Fig. 1 shows example slices on both systems and the delineations in one patient. On both modalities the main differences in contours were found at the apex and the distal part of the seminal vesicles. Contours on 0.35T showed higher consistency than on 1.5T, as depicted by the Dice similarity coefficients and Hausdorff distances (Fig. 2). Inter-modality differences were fairly consistent over all specialists (data not shown). No significant volume difference (2.3% median with 15% inter-quartile range) was found between delineated CTVs on 1.5T and 0.35T MR images.

Intra-modality contour differences where smaller with the 0.35T images. Therefore, there is no indication that a 1.5T simulation scan is necessary for the standard workflow of the 0.35T MRIdian.

Figure 1: Two examples of a 1.5T (left-top) and 0.35T (left-bottom) MRI scans and an impression of four contours on a 1.5T image in the sagittal plane (right). Each color represents the delineation of one RTO.

Figure 2: Intra-modality results showing a higher consistency between specialists' contours for the 0.35T images