ESTRO 2021 Abstract Book

S1414

ESTRO 2021

for accurate interpretation of dose and NTCP differences during ART. The proposed semi auto-segmentation allowed NTCP prediction within 1.5% accuracy for 90% cases.

PO-1687 Evaluation of the CT eFOV image distortion and the effect on VMAT plan dosimetry S. Savva 1 , N. Piracha 1 1 Northampton General Hospital NHS Trust, Radiotherapy Physics, Northampton, United Kingdom Purpose or Objective The aim of the study was to evaluate the extended field-of-view (eFOV) of the Siemens Somatom Confidence CT scanner, by comparing scans extending into the eFOV with the limits of 2mm geometric distortion and ±20HU recommended by IPEM; and evaluating the effects on VMAT plan dosimetry. The scanner has a 50cm scan FOV (sFOV) and up to 80cm eFOV. Materials and Methods A semi-circular phantom was custom-designed and 3D-printed using an Axiom20 (Airwolf 3D) printer and PLA material, with regularly spaced 4mm holes to check the geometric distortion. The Gammex RMI 465 phantom was used for HU checks (and geometry checks of soft tissue). Each phantom waswere scanned at the centre (P0), maximum and intermediate positions of the eFOV in the left, right and anterior directions. Dimensions in the phantom scans were measured and compared to those at P0. The RMI phantom was used to establish the furthest position from the centre where the geometric distortion was less than 2mm. The HU measurements were done using the maximum eFOV positions, comparing the resulting HU values of the involved RMI inserts to their values at P0. For the dosimetric evaluation, clinical cases were used to evaluate the effect of the HU artefacts and image distortion with eFoV. Results Geometrically, the scans at maximum eFOV position had many errors. The cylindrical holes were distorted and caused streak artefacts. The intermediate position gave 5-10 mm difference in the expected width of the phantom. The furthest position where the diameter of the RMI phantom was less than 2mm different than expected was 26cm from the midline, which is 1cm outside the sFOV. The HU difference of most inserts exceeded the ±20HU tolerance; they were however within or close to ±50HU, recommended by ESTRO. Exceptions were the Lung (LN-450) and Polyethylene inserts. See the graph below for the eFOV Vs sFOV values, with the linear fit representing the values at P0. Dosimetrically, the dose differences between most clinical plans recalculated on the RMI were within the ±2% normally accepted. Visual DVH inspection suggested no significant difference. However, when the RMI phantom was at 30cm from the midline, the results were worse; hence it is better if the arcs avoid these areas. The HU artefacts had an insignificant effect.

Conclusion Although the distance and HU deviations exceeded the tolerances used for the sFOV, the eFOV could be used in VMAT plans for the majority of patients. The continuous rotational beam entry of VMAT reduces the effect of image distortion on the plan dosimetry resulting in acceptable results in most cases. Exceptions would be extreme cases, for example patients extending into the eFOV by more than 4.5 cm. Static gantry plans should be considered in a case-by-case basis. PO-1688 Automatic Detection of Circular Contour Errors Using Convolutional Neural Networks N. Futakami 1 , T. Nemoto 2 , E. Kunieda 2 , Y. Matsumoto 3 , A. Sugawara 4 1 Tokai university hachioji hospital, Radiation oncology, Tokyo, Japan; 2 Keio university, Radiation oncology, Tokyo, Japan; 3 Tokai university, Radiation oncology, kanagawa, Japan; 4 Tokai university, Radiation oncology, Kanagawa, Japan

Purpose or Objective