Author: Gisell Ruiz Boiset, Paulo ROBERTO Costa, Luuk J Oostveen, Elsa Bifano Pimenta, Ioannis Sechopoulos, Alessandra Tomal π¨βπ¬
Affiliation: Radboud University Medical Center, University of SΓ£o Paulo (USP), Institute of Physics, Universidade Estadual de Campinas. Instituto de FΓsica Gleb Wataghin π
Purpose: The study evaluated the accuracy and precision of lung nodule volume measurements, specifically solid nodules (SNs) and ground-glass opacities (GGOs) in different imaging settings.
Methods: Handmade GGOs, 3D-printed SNs, and a patient-based anthropomorphic lung phantom were used. Ground-truth nodule volumes (GT) were established using an ultra-high-resolution ΞΌCT scanner. The phantom was scanned using an energy-integrating CT (EICT) and a CdZnTe photon-counting CT (PCCT) at equivalent dose levels (CTDIvol=1.4 mGy). Ten scans per protocol were acquired, including phantom repositioning to mimic clinical variability. PCCT images were reconstructed with normal resolution (NR) and high resolution (HR) using the FC51 (moderate spatial resolution and noise reduction) and FC52 (Similar to FC51 but with slightly higher spatial resolution) kernels, while EICT using the FC86 (High spatial resolution, fine detail, higher noise levels), and FC52 kernels. Images were reconstructed with hybrid iterative reconstruction (HIR). Accuracy and precision were evaluated through relative error (RE) and interquartile range (IQR), respectively.
Results: The PCCT with the FC51/NR and FC52/HR kernels demonstrated superior performance in both accuracy and precision. PCCT/FC52/HR was most accurate, with RE% ranging from -1.3%-8.0%, while PCCT/FC51/NR was the most precise, showing IQRs between 0.4-7.1. The EICT with the FC52/NR exhibited greater bias and variability, with RE% ranging from -1.9%-12.8% and IQRs between 0.6-11.1. The Wilcoxon-Signed-Rank test revealed statistically significant differences for GGO nodules, with notable distinctions between PCCT and EICT equipment (p=0.040), HR and NR resolutions (p=0.012), and FC51 and FC52 kernels in PCCT (p=0.012). For SNs, significant differences were observed between HR and NR resolutions (p=0.006) and FC52 and FC86 kernels in EICT (p=0.037).
Conclusion: PCCT systems, especially with high-resolution settings and kernel FC52, demonstrated enhanced precision and reduced bias for SN and GGO nodules. These findings emphasize the need for selecting specific protocols according to clinical application.