Repeatability and reproducibility of spectral results with clinical photon-counting CT: multi-center phantom study

Background: Photon-counting CT utilizes a new detector technology that enhances quantitative imaging but introduces new system sensitivities that may affect quantitative reproducibility. For biomarkers, this quantitative reproducibility across protocol parameters, between institutions, and over time boosts their generalizability and clinical adoption for improved disease stratification. While stability across protocol parameters and over time have previously been evaluated, inter-institution reproducibility with PCCT remains unknown. Purpose: This study aimed to evaluate the repeatability and reproducibility of PCCT spectral results across four institutions using the same scanner model, phantom model, and protocol parameters. Methods: Four institutions scanned the same material inserts from the same multi-energy CT phantom model on the same PCCT model at a tube voltage of 140 kV and a volumetric dose index (CTDIvol) of 10 mGy. For all scans, virtual monoenergetic images (VMI) from 40 to 190 keV, iodine maps, and virtual non-contrast maps (VNC) were analyzed to determine the mean error relative to the expected value for each insert, spectral result, phantom copy, and institution. For all spectral results, the standard deviation across phantom copies and institutions (σip) was calculated for each material insert to highlight variation from both phantom copies and institutions. Additionally, repeatability (RC) and reproducibility coefficients (RDC) were determined. Results: As VMI energy increased, quantitative variation between institutions decreased and stabilized. At VMI 70 keV, the σip was 1.3, 3.8, and 2.7 HU for the blood 70, adipose, and iodine 5 mg/mL, respectively. The σip increased to 7.4 HU at VMI 50 keV and decreased to 1.5 HU at VMI 150 keV. Similarly, the RDC for the iodine 5 mg/mL insert decreased from 20.6 to 4.1 HU as VMI energy increased from 50 to 150 keV. Iodine maps and VNC also exhibited reproducibility across the four institutions with RDC ranging from 0.08 to 0.49 mg/mL and 3.5 to 10.9 HU, respectively, across inserts. Conclusion: This multi-institutional study evaluated the variation of PCCT across a range of inserts and spectral results and demonstrated decreasing variation with increasing VMI energy levels. The characterized repeatability and reproducibility not only further reinforce the benefits of PCCT to support clinical translation and adoption but also provide guidance to ensure generalizable and clinically useful biomarkers for a range of clinical applications.