Margin Reduction by Virtual Couch Shift in Prostate SBRT on a 1.5T MR-Linac

Abstract

We treat localized prostate cancer on a 1.5T MR-Linac using plan re-optimization to the anatomy of the day in each fraction. We investigated the efficacy of a 2-stage, clinically available workflow to reduce intrafraction motion during the 40-minute patient on-couch time and potential margin reduction. Patients with intermediate risk prostate cancer (NCNN classification) were treated with 5 x 7.25 Gy on an MRL. Each fraction, an initial T2 weighed MR scan (‘INI’, 2 min scan time) was acquired, and pre-treatment contours were propagated and manually adapted to this scan, thus correcting for prostate rotation and organ deformation. Next, dose re-optimization (7 beam IMRT) was applied (Monaco treatment planning system), with a 5 mm CTV-PTV margin. Because the average time between the initial scan and beam on is approximately 30 minutes, prostate intrafraction motion can be considerable. Therefore, during the last minutes of this process, a position verification (‘PV’) MR scan was obtained to determine this motion. In our initial treatments (group 1, 26 patients) a virtual couch shift (VCS) was applied if part of the prostate was outside the PTV, (i.e., a prostate shift > 5 mm). The VCS involved a translation of the planned dose according to the PV to INI scan registration of the prostate, using a fast re-optimization method. In the subsequent 25 patients (group 2), the VCS was always applied, also for shifts below 5 mm, to further reduce the impact of pre-delivery intrafraction motion. Next, the plan was delivered in typically 7 minutes, while 3D cine-MR dynamics were acquired simultaneously (balanced turbo field echo). Each 3D dynamic spanned 9.4 seconds, and consisted of 448x448x45 voxels of 1.0x1.0x2.2mm. The prostate in all dynamics was automatically registered to INI and PV scans using a previously published method (D M de Muinck Keizer et al 2019 Phys. Med. Biol. 64 235008). In group 1, 24% of the patients required a VCS. In this group, intrafraction motion shifts relative to the treatment plan anatomy during dose delivery were 0.9±0.9 mm, 1.8±1.5 mm and 1.6±1.4 mm for respectively LR, AP and CC directions (mean ± SD of absolute values). In group 2, the VCS reduced the corresponding values to 0.4±0.4 mm, 1.4±1.2 mm and 1.3±1.3 mm. The average time between PV scan and beam on was 5 min, of which the VCS calculation time was below 1 min. An analysis of the required PTV margins in the corresponding directions showed that in group 2 margins could be safely reduced to 2, 4 and 4 mm in the LR, CC and AP directions. Daily VCS shortly before dose delivery allowed for a PTV margin reduction below conventionally applied 5 mm margins. However, due to the continuous and stochastical nature of intrafraction motion, margin reduction below 4 mm would require intrafraction plan adaptation. Presently, we develop such methods based on the high quality of the 1.5T 3D cineMR.