S2097
Physics - Image acquisition and processing
ESTRO 2026
References: [1] Mehta D, Thompson R, Morton T, Diamantaire A, Shefer E. Iterative model reconstruction: simultaneously lowered computed tomography radiation dose and improved image quality. Med Phys Int J. 2013;2(1):147-55.[2] The Phantom Laboratory. Catphan® 504 Manual. Salem (NY): The Phantom Laboratory; 2017. Keywords: Philips Spectral CT7500, extended FoV, IMR/iDose Quantifying the impact of growth and pathology on spatial normalisation for paediatric late effect research Denis Page 1 , Eliana Vasquez Osorio 1 , Angela Davey 1 , Chelsea Sargeant 1 , Peter Sitch 2 , Marcel van Herk 1 , Ed Smith 2,1 , Love Goyal 2,1 , Shermaine Pan 1 , Martin McCabe 1,3 , Marianne Aznar 1 1 Division of Cancer Sciences, The University of Manchester, Manchester, United Kingdom. Digital Poster Highlight 4855 2 Department of Proton Beam Therapy, The Christie NHS Foundation Trust, Manchester, United Kingdom. 3 Department of Clinical Oncology, The Christie NHS Foundation Trust, Manchester, United Kingdom Purpose/Objective: Image-based data mining enables voxel-wise analysis of dose-response relationships by spatially normalising (SN) patient anatomies and dose distributions into a common reference space[1]. However, reference anatomy selection can influence registration accuracy and downstream dose-response analyses. This issue is critical in paediatric neuro- oncology, where age-related brain morphology differences are substantial, particularly between ages 0-12y[2]. Tumour presence and surgical interventions can further affect SN.This study quantifies the impact of age and pathology on SN for paediatric brain data mining. Material/Methods: T1-weighted MR images, acquired at planning, from 103 paediatric patients (n=46 medulloblastoma [age:2- 24, 21F/25M], n=57 ependymoma [age:1-19, 32F/25M]), and 50 healthy volunteers (HV) [age:3-21, 24F/26M] from the PING dataset[3] were analysed. Volunteer scans were age- and sex-matched to the patient cohort. FastSurfer[4] was used to delineate seven key brain structures: brainstem, left/right hippocampi, cerebellum, left/right ventricles, and hypothalamus.All scans were registered to three population-averaged paediatric reference templates (1.75-2.25y, 4.5-8.5y, 7-11y)[5] using the ANTs SyN algorithm[6] (Figure 1). Registrations were visually assessed, and Jacobian determinant analysis used to
mm and 800 mm.Phantom near edgeThe CatPhan504 was positioned with its border consistently 50 mm away from the eFoV perimeter. eFoV images were acquired with diameters of 650 mm, 700 mm and 800 mm.eFoV centre and edge scans were benchmarked against IMR image quality metrics using Pylinac CatPhan504 module. Results: Material HU difference at the edge reached ~20 HU for ≤ 700 mm eFoV and ≥ 50HU at 800 mm eFoV. The HU values at the centre are more consistent with IMR images for all eFoV images. The differences at the larger diameter reconstructions would require mitigation with overrides or restrictions on use.The geometric distortion metric of the CTP404 module did not accurately represent visible distortions. Compared to IMR images, the phantom width shows a consistent 5 mm increase at the 800 mm eFoV across all protocols, while the height remains consistent. This suggests geometric distortion at the edge of eFoV.Uniformity index, low contrast visibility, and spatial resolution deteriorate at the highest reconstruction diameter. Figures 1 and 2 present the abdomen protocol data, which was representative across protocols.
Conclusion: The fact that ≥ 50HU differences are observed at the 800mm reconstruction suggests mitigation using overrides may be required for clinical use. Geometric distortion is noted when the phantom is close to the 800 eFoV, so proton beams should avoid traversing tissues further than 700mm from the isocentre.
Made with FlippingBook - Share PDF online