Abstract Book

ESTRO 37

S518

PO-0949 Effective and efficient rescanning method for PBS proton therapy O. Actis 1 , D. Meer 1 , A. Mayor 1 , D.C. Weber 1,2 1 Paul Scherrer Institute, Center for Proton Therapy, Villigen PSI, Switzerland 2 University Hospital of Zürich, Radiation Oncology, Zürich, Switzerland pencil beam scanning (PBS) delivery due to dynamic beam delivery/motion-interference during irradiation. A common method to mitigate these motion effects is to re-scan the treatment volume multiple times. Rescanning can be performed either on the entire target volume (volumetric rescanning (VR)) or sequentially within each energy layer (layered rescanning (LR)). Several studies demonstrate the higher effectiveness of VR [1,2]. This method leads, however, to an increase of the treatment time due to the high number of energy switches and magnet initializations (ramping) between scans. The latter are necessary to maintain the required precision of the delivered beam. The objective of this work is a technical implementation and validation of the VR strategy with resolved timing issue and no loss on the The starting point of our work was the current VR implementation at PSI Gantry 2. It is performed with decreasing energy sequence and takes about 6s/scan thanks to a fast energy switch of 100 ms while ramping adds 8 s more. The effectiveness of rescanning is typically reached with more than 4 rescans resulting in > 56 s of treatment time. We developed beamline settings for reverse energy sequence and removed the full ramping between scans. This dynamic beam delivery leads to non-negligible beam position errors of larger than 1.5 mm which are compensated by field specific corrections calculated during the patient verification run. The method was validated using our position monitoring system in the gantry nozzle [3] and a similar system at the isocenter. Both monitors are ion strip chambers covering the whole scanning area of Gantry 2 and capable to reconstruct the full range of beams offered by the system. For validation we used a 3-field patient plan of a total dose of 1.8 Gy applied with 4 rescans. Results We developed a strategy for an efficient volumetric rescanning at our therapy system. The new concept was validated using a patient file with 4 rescans. The presented VR concept has demonstrated an increase of the treatment efficiency by more than a factor of 2 without compromising the delivery quality. Dose spots containing 99.5% of the total dose are delivered with a precision of better than 0.5 mm fulfilling clinical requirements. The measured position of the other spots does not deviate by more than 1.5 mm from the nominal value remaining within our tolerance limits. Conclusion In addition to being an effective mitigation method for small tumour motions the presented VR implementation can be applied to treat targets with larger motion amplitudes in combination with other techniques such as gating or breath-hold. Here the time benefit will be even more prominent leading to a more efficient clinical workflow. 1. A. Schätti et. al, 2014, DOI :10.1088/0031- 9155/59/19/5707 2. G. Klimpki et. al, 2017, DOI:10.3929/ethz-a-010883563 Purpose or Objective Intra-fractional motion is limiting the precision of beam delivery quality. Material and Methods

3. O. Actis et. al, 2016, DOI:10.1088/1748- 0221/9/12/C12037

PO-0950 Improved tumor motion monitoring accuracy using piece-wise rigid 2D/3D registration H. Furtado 1 , D. Georg 1 1 Medical University Vienna, Department of Radiooncology & Christian Doppler Laboratory for Medical Radiation Research for Radiation Oncology, Vienna, Austria Purpose or Objective Real-time 2D/3D registration is a promising approach for intra-fraction motion management in radiation oncology. For lung tumors, bone present in x-rays typically has a negative impact on registration accuracy, especially when the ribs occlude the tumor. The registration relies on DRRs generated from a static (sub-)volume of a 4D CT, even if in an intra-fractional x-ray the tumor position relative to the ribs can vary as a function of breathing cycle and breathing pattern, respectively. In this case, a good match between DRR and x-ray is hard to obtain. Dual energy x-rays can be used to eliminate the bone by subtracting the images from the different energies. However, motion between acquisition of the two x-rays introduces artifacts which can also negatively influence accuracy. We propose a method to render bone and soft tissue separately therefore creating a DRR with the correct displacement between ribs and soft tissue. Material and Methods We implemented a piece-wise rigid registration approach in our in-house developed software FIRE, where soft tissue and bone are rendered from different volumes with separate rigid transforms and combined in the final DRR. Sliding away soft-tissue in relation to the ribs enables convergence of the optimization that matches x-rays acquired at any given breathing phase. To evaluate the method we used datasets from 5 patients which consisted of 4D planning CTs (pCT) and fluoroscopy sequences of about 40s at a 5Hz acquisition rate in AP position. We separated each pCT in a boneCT and a soft-tissueCT by simple thresholding. Next we manually annotated the tumor position in each x-ray, which served as ground truth tumor displacement. Registration was then performed using the standard rigid approach (with the pCT) and the proposed piece-wise rigid approach. The results were compared with the ground truth.