ESTRO 38 Abstract book

S149 ESTRO 38

London, United Kingdom; 2 University College London, Centre for Medical Image Computing- Department of Medical Physics and Biomedical Engineering, London, United Kingdom Purpose or Objective Motion-including dose reconstruction (MIDR) aims at reconstructing the actually delivered dose to the moving anatomy during radiotherapy. However, the time-resolved patient anatomy during treatment is generally unknown and commonly estimated using 3D or 4D pre-treatment images. In this study, we reconstructed the delivered dose on a ground-truth, fully time-resolved anatomy (GT-MIDR) and used this to evaluate the accuracy of MIDR based on The digital XCAT phantom was used to generate three regularly breathing thorax phantoms, each with a lung tumour moving according to its location (Fig 1). A 4DCT was generated and treatment plans were created for either a mid-ventilation approach (midVent) or treatment delivery with dynamic MLC tracking (tracking) (9-beam step-and-shoot IMRT, RTOG 1021). Treatment delivery under regular motion or regular motion and continuous drift was simulated in our in-house software. For MIDR, the treatment fluence is discretized into sub- beams; each sub-beam is associated with the anatomy instance that it was delivered to and shifted to account for residual tumour position difference between the estimated anatomy and the actual target position if any. I.e. motion is modelled by a sub-beam isocenters shift, to emulate the actual relative target/beam position. The dose for each instance is then calculated in a research treatment planning system (TPS) and accumulated on the reference anatomy via deformable registration. For GT-MIDR, ground-truth anatomy instances were generated from the XCAT. For 3D-MIDR, only one anatomy instance, the midVent 4DCT phase, was used and motion was accounted for by sub-beam isocenter shifts only. For 4D-MIDR, anatomy instances were chosen as the 4DCT phase where the tumour is closest to the actual tumour position and sub-beam isocenter shifts accounted for residual position differences. Results Differences between GT-MIDR and planning (Fig 2) show the effect of motion (midVent) or motion and mitigation (tracking) on dose delivery. Target underdosage was highest for tumour A. Tracking resulted in higher dose to the spinal cord and heart (tumour A, B) or aorta (tumour C). 3D and 4D CT images. Material and Methods

Differences between 3D or 4D and GT-MIDR (Fig 3) show the accuracy of the respective methods. Reconstructed target dose errors above 1Gy were observed for 3D-MIDR. For tumour A, reconstructed dose to the organs-at-risks (OAR) were underestimated by 3D-MIDR. For tumour B, dose to the oesophagus and aorta was overestimated by 3D-MIDR. For tumour C, dose to the spinal cord and the oesophagus were overestimated by 4D-MIDR and dose to the aorta was underestimated by 3D and 4D-MIDR.

Conclusion In the first demonstration of GT-MIDR, we calculated the delivered dose to target and OAR for a range of lung