ESTRO 38 Abstract book

S533 ESTRO 38

PO-0976 Validation of respiratory motion modeling through repeated 4DMRI in the abdomen: preliminary results G. Meschini 1 , C. Paganelli 1 , G. Fontana 2 , A. Pella 2 , A. Mancin 3 , A. Vai 4 , M. Riboldi 5 , F. Valvo 6 , G. Baroni 1 1 Politecnico di Milano, Department of Electronics Information and Bioengineering, Milan, Italy ; 2 National Centre of Oncological Hadrontherapy, Bioengineering Unit, Pavia, Italy ; 3 National Centre of Oncological Hadrontherapy, Medical Radiology Technicians Unit, Pavia, Italy ; 4 National Centre of Oncological Hadrontherapy, Medical Physics Unit, Pavia, Italy ; 5 Ludwig-Maximilians-Universität, Chair of Experimental Physics - Medical Physics, Munich, Germany ; 6 National Centre of Oncological Hadrontherapy, Radiation Oncology Unit, Pavia, Italy Purpose or Objective Respiratory motion modeling in radiotherapy provides the estimation of anatomical changes induced by breathing, when continuous 3D imaging feedback is not available. However, the evaluation of the model performance in case of intra- and inter-fraction variability of the respiratory pattern is still a challenge. In this study, we exploited repeated 4DMRI acquisitions to create the training and testing datasets for respiratory motion model validation. Material and Methods Repeated 4DMRI of a pancreas (P_pan) and a liver cancer patient (P_liv) were acquired. Patients underwent two scans, one after the other, before irradiation and another scan after irradiation, in order to image intra- and inter- fraction variations, respectively (Table). Specifically, multi-slice acquisition of sagittal images (pixel size: 1.33mm×1.33mm, slice thickness: 5mm) during free breathing and retrospective 4DMRI reconstruction 1 were performed. The motion model 2,3 was trained on the first 4DMRI, thus establishing a correlation between a respiratory surrogate (here image-based) and motion information extracted through deformable image registration (DIR). The model was tested on the subsequent intra- and inter- 4DMRI, estimating the motion by means of the surrogate only. Inter-fraction baseline shifts were compensated 3 (Table). The estimated and the imaged ground truth (GT) motions were finally compared. Results The tumor center of mass (COM) distance between GT and model output had median values below 2.2mm both for P_pan and P_liv (Figure). Median errors below 2mm were observed consistently for the considered organs at risk (kidney and liver, respectively). For P_liv, higher variability of tumor COM distances was quantified in the inter-fraction scenario, with errors up to 6.5mm for respiratory phases close to end-inhale. This is likely due to the baseline shift, the higher range of motion with respect to the training dataset (Table) and to DIR uncertainty.

Conclusion A validation framework based on repeated 4DMRI is proposed for motion modeling in presence of intra- and inter- fraction changes. Preliminary results suggest the feasibility of motion modeling in the abdomen, with median residual errors comparable to the voxel size. The compensation of inter-fraction variations by re-training the model on a newly acquired dataset could improve its performance. Patient acquisitions is currently ongoing, aiming at extending the validation to a larger patient cohort and assessing the impact of DIR error. 1 Meschini et al 2017 A multi-dimensional approach for retrospective 4DMRI sorting, 4D Treatment Planning Workshop 2 Meschini et al 2017 Evaluation of residual abdominal tumour motion in carbon ion gated treatments through respiratory motion modelling, Phys Med 3 Fassi et al 2015 Surrogate-driven deformable motion model for organ motion tracking in particle radiation therapy, PMB PO-0977 Improved 4D proton dosimetry via correlation with beam delivery details using log-files N. Kostiukhina 1 , H. Palmans 2 , S. Waid 3 , M. Stock 2 , D. Georg 1 , B. Knäusl 1 1 Medical University of Vienna / AKH Vienna, Department of Radiotherapy and Christian Doppler Laboratory for Medical Radiation Research for Radiation Oncology, Vienna, Austria ; 2 EBG MedAustron GmbH, Medical Physics, Wiener Neustadt, Austria ; 3 EBG MedAustron GmbH, Therapy Accelerator- Commissioning and Development, Wiener Neustadt, Austria Purpose or Objective For implementing pencil beam scanning treatment in the presence of target motion, detailed insight in the treatment delivery sequence is essential. In this work, a 4D framework was implemented and applied for different scenarios, allowing to study the correlation between time- resolved dosimetry and beam delivery (BD) log files. Material and Methods A pulsed scanned proton beam with varying intensity was employed for the experiments. To represent target motion in a realistic human anatomy, an anthropomorphic breathing phantom, dedicated to proton dosimetry, was used. Two patient-specific parameters were varied: tumor volume (TV) and motion pattern. For TVs of 12cm³ (TV1) and 97cm³ (TV2) static and sinusoidal (0.6cm and 2.0cm amplitude) tumor motions were investigated. For both TVs a CT based treatment plan was created with the phantom in static position and delivered for all motion scenarios. The dose was determined in the target volume and the penumbral region with 0.5s time intervals using four (for TV1) and five (for TV2) pinpoint (PP) ionization chambers. The start of the phantom’s motion was triggered by a pre-