Abstract
546
Objectives To compare the performance of 18F-FDG PET-CT and PET-MR in patients with pulmonary tuberculosis (PTB) in terms of visual detection and quantification accuracy.
Methods Ten subjects with confirmed tuberculosis (age: 55.1 ± 9.6 years [mean ± s.d.]; range: 42.1 - 74.6 years) were recruited. Subjects received an 18F-FDG dose of 129.3 ± 4.0 MBq (mean ± s.d.). Five of the ten subjects underwent a PET-MR scan on a Siemens mMR, followed by PET-CT on a Siemens mCT, while the remaining five were imaged first on the mCT followed by the mMR imaging. The PET-MR studies were performed using two 12-channel surface coils. MR imaging included Dixon imaging for MR-based Attenuation Correction (MRAC), T1 VIBE (pre- and post-contrast), T2 HASTE, T2 PACE, T2 TRUFI and DWI (b50 and 800). PET data were acquired for 15 minutes at 65.9 ± 15.0 minutes (mean ± s.d.) p.i. when performing PET-MR imaging first, and 98.2 ± 15.7 minutes when PET-MR data were acquired second. The PET data were reconstructed using Ordinary-Poisson (OP) OSEM with 3 iterations and 21 subsets. A Gaussian post-smoothing filter of 6 mm FWHM was applied. The matrix size was 172 × 172, with a voxel size of 4.17 × 4.17 mm and slice thickness of 2.03 mm. Subjects underwent CT covering the whole lung in either one or two bed positions. PET data were acquired for 10 minutes per bed position. The PET-CT study was performed at 68.2 ± 17.7 minutes p.i. (114.5 ± 7.5 minutes as the second acquisition). Images were reconstructed with OP-OSEM (3 iterations, 24 subsets) and a 6 mm post-smooth filter. The reconstructed PET image matrix was 200 × 200 with a voxel size of 4.07 × 4.07 mm and slice thickness of 2.03 mm. Both the time-of-flight (ToF) and point-spread function (PSF) modelling capabilities of the mCT were not applied in order to produce images that were comparable with those of the mMR. A visual lesion detection task and SUV analysis were performed. SUV iso-contour (50% maximum) volumes of interest (VOIs) were defined on the PET from the PET-CT study in PMOD (v.3.6 PMOD Technologies, Zürich). These VOIs were then propagated to the PET/MR space via non-rigid registration. Calculation of SUVs was performed in the PET space of both studies. A maximum of 10 lesions per subject were included to avoid any individual subject being over-represented. A Two-Way ANOVA was used (R; v. 3.1.2) to assess both the difference between scanners and the influence of scanning order. A p-value < 0.05 was considered significant.
Results A total of 113 PTB lesions were detected on PET-MR and 118 on PET-CT. SUV analysis was performed on 78 of these lesions. Strong correlations were observed for both SUVmean and SUVmax, with r2=0.64 and r2=0.73 respectively. SUVmean and SUVmax were significantly lower on PET-MR (SUVmean: 2.37 ± 1.31; SUVmax: 4.07 ± 2.21) than PET-CT (SUVmean: 3.49 ± 1.38; SUVmax: 5.18 ± 2.20). The scan order was also found to have an effect (SUVmean: p = 0.015; SUVmax: p = 0.027), suggesting that FDG accumulation may continue to increase in PTB beyond 60 minutes. Visual assessment of the Dixon MRAC mu-maps demonstrated that PTB lesions tend to be assigned the linear attenuation coefficient (LAC) for lung.
Conclusions PET-MR visual performance was shown to be comparable to PET-CT in terms of the number of PTB lesions detected. Measured mean and maximum SUVs were significantly lower on PET/MR, although strongly correlated with PET/CT. Dixon-based attenuation correction under-estimates the LACs of PTB lesions, resulting in lower SUVs compared to PET/CT.