Abstract Book

S155

ESTRO 37

delivers according to the daily adapted plan. To this purpose, the use of Electronic Portal Imaging Devices (EPID) for independent in vivo dose verification in the MR-Linac is being developed and a proof of concept at gantry angle 0 is presented. Material and Methods The MR-Linac is a combination of a 1.5T magnetic resonance imaging scanner with a 7 MV linear accelerator and is equipped with an on-board EPID. The existing back projection algorithm used with conventional linacs in our institute has been adapted for the MR-Linac at gantry 0, and has been commissioned with measurements of square fields made under a variety of different phantom setups (varying SSD, phantom thickness and field size) acquired with EPID and ionization chamber. Performance of the back-projection algorithm was verified for square fields and for 25 IMRT fields irradiated to a 23 cm slab phantom. The verification was performed by comparing the dose reconstructed with the EPID to the dose measured with the OCTAVIUS 1500 detector array (PTW, Freiburg, Germany), at 10 cm depth at the isocenter plane. The 2- D reconstructed dose distributions were compared to the measured 2-D dose distributions by means of a global 2-D γ-analysis (3%, 3mm, 20% isodose). In both cases, the dose distributions were only compared in the central region of the EPID images, where there is no strong extra attenuation of the MRI structures in the EPID acquired signal ( Figure 1 ), of 11.5 cm at the isocenter plane in the cranial-caudal direction.

Results An increased foam thickness decreases the surface dose (Figure 1c), from a mean dose of 89.1±6.1 cGy with visible loop imprints using 4 mm thick foam to 54.9±3.1 cGy without loop imprints using 15 mm. The latter amounts to an increase from 0.20 D_max ( D_open ) to 0.34 D_max ( D_coil,15mm ). The depth-dose curves with and without the coil (Figure 2c) show maximal dose differences of 17.9 cGy (<0.04 D_max ) in the build-up region. Accurate surface doses could not be obtained in this setup. From a depth of 1 cm the curves overlap. Conclusion The use of 15 mm foam resulted in the lowest surface dose, i.e. 0.34 D_max . To put this into perspective, the current treatment table causes surface doses of up to 1.35 D_max (results not shown). Dose increases in the depth-dose curves of <5% before D_max and the lack of dose difference deeper than 1 cm suggest that on-body placement of a receive array during a treatment in the MR-linac is feasible. This has distinct advantages for our 64-channel array in terms of imaging performance and acceleration, which are needed for real-time imaging during treatment. OC-0298 EPID dosimetry at the MR-Linac using a back- projection algorithm: proof of concept. I. Torres Xirau 1 , I. Olaciregui-Ruiz 1 , B. J. Mijnheer 1 , B. Vivas-Maiques 1 , U. A. van der Heide 1 , A. Mans 1 1 Netherlands Cancer Institute, Department of Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective The clinical implementation of MRI-guided radiotherapy (Unity ATL1, Elekta AB, Stockholm, Sweden) aims for online and real-time adaptation of the treatment plan. Given the complexity of the machine in combination with novel techniques for daily re-planning, QA is highly desired in these systems to verify that the treatment

Results For the commissioning square fields, X-Y profiles of reconstructed EPID dose distributions were compared to the Octavius 1500 array measured profiles. For the high dose region (above 80% of maximum dose), more than 93% of the points showed local deviations smaller than 4%. The γ-parameters of the commissioning square fields averaged over 20 fields are shown in Table 1 , together with the averaged gamma parameters of 25 IMRT fields. Reconstructions of the dose distributions were calculated in 129.3 ms per segment on average.

Conclusion The EPID dosimetry back projection algorithm was successfully adapted for the MR-Linac at gantry 0. Experiments showed good agreement between measured and EPID reconstructed dose distributions. This result is an essential step towards an accurate, independent, and potentially fast field by field patient specific QA tool for