ESTRO 37 Abstract book

S1187

ESTRO 37

Purpose or Objective The MR-Linac is a new technology that uses onboard MRI to directly guide radiotherapy. Geometric fidelity is very important for both the MR-Linac and the MR-Sim used in the treatment planning phase. Here, we evaluate whether there are significant differences between the 1.5T MR-Sim and MR-Linac systems in our radiotherapy department. Material and Methods Geometric fidelity in MRI is primarily determined by two factors: System-specific gradient non-linearities and patient-induced B0 inhomogenity. We expect patient- induced distortions to be similar in the MR-Sim and MR- Linac since both have 1.5T B0 fields. Distortions caused by gradient non-linearities are system specific, and may differ substantially due to the split gradient design of the MR-Linac. A Philips Geometric-QA phantom was imaged on a Philips Ingenia 1.5T MR-Sim and an Elekta Unity 7MV, 1.5T MR- Linac. The main body coil was used for RF transmission and signal detection. Two T1-weighted 3D gradient-echo sequences were acquired with readout along the AP and PA directions respectively. The Philips QA software was then used with the acquired images to calculate distortion in each orthogonal direction. In order to remove distortion due to B0 inhomogenity induced by the phantom, the arithmetic mean of the marker locations in the AP and PA images was calculated. The 1.5T Philips Ingenia system was used to acquire B0 maps in the Brain and Pelvis during clinical patient scanning using a duel echo spoilt gradient echo, with an echo spacing of 4.6 ms. The B0 maps were unwrapped using the FSL Prelude algorithm to eliminate phase wraps. As distortion due to B0 inhomogenities are dependent on acquisition parameters, displacement maps were calculated using the readout gradient value of a 3D T1-weighted TFE scan. Patient-induced distortions, that occur along the readout direction, were added to the corresponding dimension of the system-distortion maps, and the total distortion vector was calculated the RSS of the displacements along all three orthogonal maps. Results Figure 1 shows the system distortions, and the total distortions in brain and pelvis scans. The system distortions appear better for the MR-Linac, with only the edges of the FOV displaying greater distortion than the MR-Sim. These findings are confirmed by Table 1, where the mean and maximum values in the brain are lower for the MR-Linac. While in the larger pelvis scan, the maximum value is slighter larger for the MR-Linac, the mean and P90 values are also lower.

were found when appliying the calibration curves to the same phantom for both SE and DE (Fig 2). Considerable deviations were observed when applying the SE CIRS M or S phantom HU-RED calibration to the Catphan and vice versa due to the differences in Z eff between the phantoms. These differences were greatly reduced using the DE scan (Fig. 2). On the other hand, applying the calibration of the CIRS M/S to the CIRS S/M yielded larger deviations for DE compared to SE.

Conclusion The ΔHU-RED calibration method has been successfully implemented on a linac integrated CBCT scanner. It allows the discrimination of materials that have the same HU but different electron densities. The robustness with respect to the object size due to changes in scatter conditions is reduced compared to the SE method. Further evaluation of the off-axis accuracy and robustness is needed and the trade-off between dose, image quality and HU-RED calibration has to be explored.

[1] Med.Phys ; 2012 ;Apr; 39 ( 4 ):2021-30 [2] Med.Phys;2016;Mar;43(3):1057-64

EP-2147 Comparison of Spatial-Distortion Maps for MR- Sim Versus MR-Linac in the Brain and Pelvis at 1.5T R.J. Goodburn 1,2 , R.H.N. Tijssen 2 , M.E.P. Philippens 2 1 Cambridge University Hospital NHS Foundation Trust,

Medical Physics, Cambridge, United Kingdom 2 University Medical Center Utrecht, Dept. of Radiotherapy, Utrecht, The Netherlands