ESTRO 37 Abstract book

ESTRO 37

S523

Results A total of 191 CBCTs were analysed. Results are illustrated in table 1. There was a significant difference between mean patient motion using the different registration strategies (p<0.05). In both rigid and non-rigid approximation, anterior and lateral motion was largest for the GTVt in the upper and middle rectum and anterior motion was largest for the upper mesorectum. DVF displacement of up to 29 mm occurred anteriorly for GTVt in the upper rectum and the upper mesorectum. These are not accounted for in PTV margins using rigid margin recipes.

Hospital NHS Trust, Academic Radiotherapy, Sutton, United Kingdom 2 St George's Hospital NHS Trust, Physics, London, United Kingdom Purpose or Objective To quantify internal target motion of the gross tumour volume (GTVt) and mesorectum using both rigid and non- rigid approximation in cone beam CT (CBCT) of patients treated with radiotherapy for rectal cancer. Material and Methods 1. CBCTs were acquired in 18 patients receiving neoadjuvant chemoradiation for rectal cancer on days 1-3 and then weekly thereafter. Rigid registration between the planning CT and each CBCT was performed based on bony anatomy. GTVt, mesorectum, rectum and bladder were manually contoured on each image set. Each CBCT was deformed to the planning CT by computed deformable image registration (DIR) on a commercial treatment planning system (RayStation v5, Raysearch Laboratories). GTVt and mesorectum were divided into target structures in the upper, middle and lower rectum and upper and lower mesorectum (Figure 1a and 1b). Target motion using non-rigid approximation was determined by deformation vector field (DVF) displacement of voxels within a 1mm internal wall of the target sub-divisions. This was divided into quarters to create structures representing the right, left, anterior and posterior walls (figure 1c). Rigid approximation of target motion was determined by translation of the most extreme right, left, anterior and posterior positions of the target structures, relative to the mid femoral head. These measurements were taken at each 2.5mm slice on which these structures were defined (figure 1d). Mean target motion using both methods was calculated for each patient and compared using paired t-test. Patient and population systematic (Σ) and random (σ) error were calculated for internal motion. 10. For the rigid approximation data, the margins required to account for the observed errors were calculated using van Herk’s rigid margin recipe and the PTV to encompass the population target motion and set-up error was determined. 3. 4. 5. 6. 7. 8. 9. 2.

Conclusion Anisotropic margins are required for the GTVt and mesorectum. Margins should be largest for GTVt and mesorectum located superiorly in the pelvis. Complex geometric changes occur that are not account ted for in margin recipes using rigid approximation and new margin recipes that account for deformation are indicated. The variability in target motion means that population margins will be large and these patients would benefit from adaptive strategies, particularly in the context of dose escalation to the GTVt to increase the rate of complete pathological response. PO-0956 A self-gated coronal 4D-MRI method for daily imaging of liver lesions on an MR-Linac T. Van de Lindt 1 , M. Fast 1 , M. Nowee 1 , E. Jansen 1 , V. Pelt 1 , U. Van der Heide 1 , J.J. Sonke 1 1 Netherlands Cancer Institute, Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective To develop an automated self-gated 4D-MRI method for daily patient set-up and tumor motion quantification for liver radiotherapy on an MR-Linac, providing a respiratory surrogate for amplitude binning, tumor visibility without the use of a contrast agent and a fast acquisition- reconstruction time (<5min). Material and Methods One patient with a liver metastasis underwent four MRI examinations in our ongoing IRB approved MR-Linac (Unity ATL1, Elekta AB, Sweden) imaging trial. Two 4D-MRI scans per examination were acquired: a repeated coronal 2D multi-slice TFE (TR/TE=4.3/2.2ms, SENSE=2, shot- length=365ms) and single-shot TSE (TE=60ms , SENSE=2.5, shot-length=220ms) scan, both with interleaved slice acquisition order, 2x2x5mm 3 voxels, 25 slices, 30 repetitions and total acquisition of ~4min. For reconstruction of the respiratory signal, the median intensities in AP and LR direction over all acquired slices were calculated to obtain a median coronal and sagittal image. Then, the gradient in CC direction was c omputed and a parabola was fitted on the liver dome of both