Research ArticlePhysics and Instrumentation

Reproducibility and Comparability of Preclinical PET Imaging Data: A Multicenter Small-Animal PET Study

Julia G. Mannheim, Martin Mamach, Sybille Reder, Alexander Traxl, Natalie Mucha, Jonathan A. Disselhorst, Markus Mittelhäuser, Claudia Kuntner, James T. Thackeray, Sibylle Ziegler, Thomas Wanek, Jens P. Bankstahl and Bernd J. Pichler
Journal of Nuclear Medicine October 2019, 60 (10) 1483-1491; DOI: https://doi.org/10.2967/jnumed.118.221994

Article Figures & Data

Figures

  • FIGURE 1.
    FIGURE 1.

    Image quality phantom data. RCs for different rod sizes (A), SOR for water and air (B), and %SD as a measure of noise (C) for all 4 scanners are depicted. Data were reconstructed using OSEM2D without attenuation correction. Exp = experiment.

  • FIGURE 2.
    FIGURE 2.

    Experiment 1: laboratory-specific 18F-FDG standard imaging protocols. Quantitative analysis (%injected dose/mL [%ID/mL]) of 18F-FDG uptake is depicted for the brain (A), left ventricle (B), liver (C), and muscle (D) for the 4 laboratories (laboratories A, B, and D: n = 10, laboratory C: n = 7). Data were reconstructed using OSEM2D without attenuation correction. Box plots show group means, 25% and 75% confidence intervals, 1 upper and lower SD of the mean, and all individual data points. Test results that were statistically significant using the Tukey–Kramer test (with α = 5%) after Bonferroni–Holm correction (applied separately for each organ) are marked with asterisk.

  • FIGURE 3.
    FIGURE 3.

    Representative images of 18F-FDG biodistribution. (A) Experiment 1: laboratory-specific 18F-FDG standard imaging protocols. (B) Experiment 2: influence of animal handling and personnel on the data comparability. Lab = laboratory.

  • FIGURE 4.
    FIGURE 4.

    Glucose levels. (A) Laboratory-specific 18F-FDG standard imaging protocols. Blood glucose levels for the animals before 18F-FDG injection (laboratories A, B, and D: n = 10, laboratory C: n = 7) are shown. Glucose parameters were measured with standard glucose device of each laboratory. Tukey–Kramer test was used to determine statistical significance. (B) Influence of animal handling and personnel on data comparability. Blood glucose levels for the animals before 18F-FDG injection are shown for the 3 setups (protocol and personnel laboratory D, protocol laboratory A and personnel laboratory D, protocol and personnel laboratory A). The 1-sample t test was used to determine significant differences. Box plots show group means, 25% and 75% confidence intervals, 1 upper and lower SD of the mean and all individual data points. Test results that were statistically significant (with α = 5%) after Bonferroni–Holm correction (applied separately for each test) are marked with asterisk.

  • FIGURE 5.
    FIGURE 5.

    Experiment 2: Influence of animal handling and personnel on data comparability. Quantitative analysis (%injected dose/mL [%ID/mL]) of 18F-FDG uptake is depicted for the brain (A), left ventricle (B), liver (C), and muscle (D) for the 3 studies (n = 9) in laboratory D and compared with data from laboratory A from experiment 1 (n = 10). Data were reconstructed using OSEM2D without attenuation correction. Box plots show group means, 25% and 75% confidence intervals, 1 upper and lower SD of the mean, and all individual data points. Test results that were statistically significant (with α = 5%) after Bonferroni–Holm correction (applied separately for each organ) using the 1-sample t test (comparison of data acquired in laboratory D) and the 2-sample t test (comparison of data acquired in laboratory D to data acquired in laboratory A) are marked with asterisk. Lab = laboratory.

  • FIGURE 6.
    FIGURE 6.

    Experiment 3: Reproducibility of PET image analysis. Comparison of quantitative analysis (%injected dose/mL [%ID/mL]) of 18F-FDG uptake in brain (A), left ventricle (B), liver (C), and muscle (D) are depicted. Same datasets (n = 9) were analyzed individually by trained investigators from laboratory A (analysis 1) and from laboratory D (analysis 2). Data were reconstructed using OSEM2D without attenuation correction. Box plots show group means, 25% and 75% confidence intervals, 1 upper and lower SD of the mean, and all individual data points. Test results that were statistically significant using the 1-sample t test (with α = 5%) are marked with asterisk.

Tables

  • TABLE 1

    Overview of the Used Scanner Types, Software Versions, Energy and Timing Windows, Attenuation Methods, As Well As of the Used Reconstruction Parameters

    Scanner/laboratoryScanner typeSoftwareEnergy window (keV)Timing window (ns)Reconstruction algorithmMatrix/zoomReconstructed voxel size (mm3)Attenuation method, acquisition time, acquired counts
    AInveon dedicated PETInveon Acquisition Workplace 1.5.0.28350–6503.432OSEM2D (4 iterations, 16 subsets)128 × 128 × 159 zoom 10.776 × 0.776 × 0.79657Co source (63.9 MBq), 803 s
    BInveon MultiModality PET/CTInveon Acquisition Workplace 1.5.0.28350–6503.432OSEM2D (4 iterations, 16 subsets)128 × 128 × 159 zoom 10.776 × 0.776 × 0.796CT, 80 keV, 500 µA, 120 projection steps over 220° rotation, 200 ms exposure time and binning 4, 300 s
    CFocus 220microPET Manager 2.5.0.0250–7506OSEM2D (4 iterations, 16 subsets)128 × 128 × 95 zoom 4.7450.400 × 0.400 × 0.79657Co source (50.3 MBq), 604 s
    DInveon dedicated PETInveon Acquisition Workplace 2.0350–6503.4375OSEM2D (4 iterations, 16 subsets)128 × 128 × 159 zoom 10.776 × 0.776 × 0.79657Co source, 935 s (experiment 1, 60.2 MBq), 827 s and 6,588 s (experiment 2, 16.0 MBq)

Additional Files

Back to top

In this issue

Print
Download PDF
Article Alerts
Email Article
Citation Tools
Bookmark this article

Jump to section

Keywords