ESTRO 37 Abstract book

ESTRO 37

S529

Material and Methods Three patients with locally advanced rectal cancer underwent a planning MRI (pMRI) at a diagnostic scanner and weekly scans at the MR-Linac (Unity ATL1, Elekta AB, Stockholm, Sweden) during their chemoradiotherapy. The CTV, comprised of the mesorectal and presacral regions, was delineated on the pMRI as well as the GTV. Subsequently, a population statistics derived LoP consisting of the original (0), two smaller (-2&-1) and two larger CTVs (1&2) was generated. Repeat MR-Linac MR scans (MRL-MRIs) were registered to the pMRI on the pelvic bone. GTV and CTV were delineated on all MRL- MRIs as a surrogate for automatic contour propagation and were used as input for automatic plan selection minimizing the absolute mean distance to the LoP and ‘propagated’ CTV. This choice was then compared to choices by three observers for each individual MRL-MRI. Subsequently, the inter-observer and observer-automatic agreement was determined. Finally, the impact of the locational variation of the GTV was assessed by evaluating the vector length of the center of mass difference between planning and repeat GTV and comparing this to the plan choices. Results Ten MRL-MRIs could be evaluated for plan selection. Plan selection consistency between observers was 93%. Selected plans deviated at most by CTV increment. Former observer variation on cone-beam CT showed that 13% of all selections deviated by 1 plan from the majority choice. On MRI, the original plan was chosen by majority only twice. For seven fractions plan 1 was chosen and plan 2 was chosen once. The automatic plan selection, on the other hand, selected the original plan for each of the ten fractions. With increasing vector length of the GTV between pMRI and MRL-MRI, the observers tended to choose larger plans to ensure adequate coverage of the GTV (Figure 2).

the voxel values are close to the ones of the CT. The accuracy of the CBCT dose calculation was assessed by comparing the dose distribution calculated on the CBCT generated by each of the method to the reference CT dose calculation with Pinnacle v. 9.10 (Philips). The endpoints were dose differences: mean dose (D mean ) for the parotid glands (PG) and maximum dose (D 2% ) for the spinal cord. A spatial dose distribution comparison was performed by a 3D local gamma analysis (γ 3D , criteria: 2%/2 mm, 30% threshold). In total, 20 dose calculations per method were analysed. Results The mean dose differences and γ 3D (pixels percentage with gamma < 1) corresponding to each of the method are given in table 1. Figure 1 illustrates spatial dose discrepancy areas, located near body contour, heterogeneities (bones or air cavities) and regarding shoulders.

Figure 1. LoP plotted on an MRL-MRI. GTV depicted in yellow.

Conclusion In this study, the most accurate method for CBCT dose calculation is the CBCT deformation from the planning CT, presenting a mean dose uncertainty less than 2% for the D mean of the parotid glands and less than 0.5% for the D 2% of the spinal cord. The mean γ 3D of this method was higher than 95% with the lowest dispersion (standard deviation = 3%). This preliminary study suggests the feasibility of using CBCT to monitor the dose in the parotid glands during the treatment course in order to trigger replanning if necessary. PO-0963 MR-Linac: first experience with library of plan technique for rectal cancer patients M.E. Nowee 1 , C. Beekman 1 , A. Betgen 1 , S. Van Beek 1 , J.J. Sonke 1 , U.A. Van der Heide 1 , B. Van Triest 1 , P. Remeijer 1 1 The Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective A potential dosimetric benefit of a library of plan (LoP) selection strategy has been described in rectal cancer patients. In our institute, an online LoP technique is used to account for anatomical changes due to rectal filling. In this study we describe the first experiences with MR based LoP for future use at the MR-Linac and explore the potential and pitfalls of automated plan selection.