ESTRO 2022 - Abstract Book

S1376

Abstract book

ESTRO 2022

Conclusion State of the art 3D CNNs produce clinical acceptable automatic segmentations of i-LNs and LN levels, bringing the irradiation of i-LNs closer to clinical implementation. The UNet outperformed DM on both the segmentation of i-LNs and LN levels with better matching contours. However, the overestimation of predicted i-LNs was smaller while using DM compared to UNet. Still a high sensitivity is the most important factor with respect to i-LNs irradiation, which were high for both networks.

PO-1594 The use of LEGO to assess CT-MRI fusion accuracy for stereotactic radiosurgery

G. Heyes 1 , R. Flintham 2 , H. Augustus 1 , R. Stange 1

1 Queen Elizabeth Hospital Birmingham, Radiotherapy Physics, Birmingham, United Kingdom; 2 Queen Elizabeth Hospital Birmingham, Medical Physics, Birmingham, United Kingdom Purpose or Objective Magnetic Resonance (MR) at 3T is used to define targets for stereotactic radiosurgery (SRS) treatments. Images are fused to Computed Tomography (CT) for treatment planning. The delivery accuracy of the CyberKnife system is <1mm. MR images can suffer spatial distortions >1mm, and can be highly variable according to the local MR equipment, protocols and patient setups. For small targets (<0.5cc), or in multiple brain metastases where accurate fusion across the whole brain is required, it is critical to assess these distortions locally and their impact on dosimetric accuracy for SRS. Materials and Methods A phantom was made using 520 LEGO (Billund, Denmark) bricks with each brick manufactured to a tolerance of 20 µ m. The resulting cuboid (125mm x 126mm x 160mm) of similar dimensions to a human brain was encased in a custom-made Perspex phantom filled with food-grade mineral oil to give a uniform MR signal at 3T. The cuboid was scanned on a Siemens Skyra 3T MR (Erlangen, Germany) using our standard-of-care SRS planning sequence, and on a 1mm Philips CT. Further MR scans were performed with the centre of the phantom offset in 16mm steps (equivalent to 1 Lego brick width) from the centre of the MR bore in the Sup-Inf plane to maximum offset of 64mm. Each MR dataset was fused to CT using the Accuray Precision radiotherapy planning system. This software was used to assess the displacement of 30 bricks from the true position as defined by CT.