S2050
Physics - Image acquisition and processing
ESTRO 2026
References [1] Spadea, M. F., Maspero, M., Zaffino, P., & Seco, J. (2021). Deep learning based synthetic-CT generation in radiotherapy and PET: A review. Medical physics, 48(11), 6537– 6566. https://doi.org/10.1002/mp.15150 [2] Zaffino, P., Raggio, C. B., Thummerer, A., Marmitt, G. G., Langendijk, J. A., Procopio, A., Cosentino, C., Seco, J., Knopf, A. C., Both, S., & Spadea, M. F. (2024). Toward Closing the Loop in Image-to-Image Conversion in Radiotherapy: A Quality Control Tool to Predict Synthetic Computed Tomography Hounsfield Unit Accuracy. Journal of Imaging, 10(12), 316. https://doi.org/10.3390/jimaging10120316 Metal artefact reduction for spinal implants: A quantitative comparison of Photon-Counting and Energy-Integrating CT Tsz Yan Lee, Kar Wei Natalie Yip, Tsz Ching Fok, Tin Lok Chiu, Siu Ki Yu Medical Physics Department, Hong Kong Sanatorium & Hospital, Hong Kong SAR, China Purpose/Objective: The presence of high density implants often poses challenges in radiotherapy treatment planning by degrading the accuracy of CT number near the implants. This study quantitatively compares the metal artefact reduction (iMAR) performance on a Photon- Counting CT (PCCT) and an Energy-Integrating Detector CT (EID CT) in the presence of various high density implant materials at the spine region. Material/Methods: Titanium, steel, and alumina (Al ₂ O ₃ ) rods were inserted bilaterally along the spine of an anthropomorphic phantom. Scans were acquired on Siemens PCCT (Naeotom Alpha) and EID CT (Somatom Drive) systems using a clinical spine protocol (120kVp, Br40 kernel). PCCT images were reconstructed to polychromatic images (referred to T3D by the manufacturer), 70keV, and 190keV virtual monoenergetic image (VMI) settings with iterative metal artefact reduction (iMAR) algorithm. Images without implant of the corresponding reconstruction setting were served as baseline for comparison.The artefact strength of images of different high density implants was quantified using the artefact Index (AI)and robust Digital Poster 1069 artefact index ( ρ -index) 1, a metric that is less sensitive to extreme outliers and a more reliable measure of artefact severity that emphasizes severe dark streaking artefacts.
Results Fig. 2 shows that the error-prediction framework systematically improved the quality of sCTs across all tested cases. The magnitude of the improvement was proportional to the initial error level, with the largest MAE improvements observed in cases having an initial worst conversion accuracy (maximum MAE improvement was equal to 42 HU). Bias correction was most effective in cases with positive initial bias, achieving a maximum bias improvement of 28 HU, while negligible changes were observed when the original bias approached zero. For cases with low initial signed errors, the correction remained stable or showed mild overcorrections.
Conclusion Preliminary investigations of voxel-wise signed error prediction proved its capability in sCT correction, improving the synthetic image accuracy. This method could also serve as a pre-treatment verification step in radiotherapy. Future work will extend validation to larger and multi-center datasets, as well as different anatomical sites, to further assess its generalizability.
CT number difference between the implant scenario and
Made with FlippingBook - Share PDF online