Intrascanner Reproducibility of an SPM-Based Head MR-Based Attenuation Correction Method

Abstract

Recently, an exhaustive examination of 11 state-of-the-art MR-based attenuation correction (AC) concluded that there are currently a few methods showing similar results compared to the gold-standard, CT-based AC. While the study presented a thorough portfolio of metrics to quantify accuracy (bias) and quality, it lacked one of the most important metrics to quantify robustness that is critical for its clinical applicability: intrascanner reproducibility (repeatability). In this paper, we provide for the first time a study of the repeatability of one of the outperforming brain MR-based AC methods: the statistical parametric mapping (SPM)-based pseudo-CT approach. Twenty two subjects undergoing 3 18F-FDG PET/MRI visits within two months were retrospectively analyzed in this paper. Pseudo-CT mu-maps were obtained from the coregistered MR images for all three visits and the PET data from visit 1 was reconstructed using all three mu-maps. Relative changes (RC), intraclass correlation coefficient, reproducibility coefficient (RDC95%) and Bland-Altman limits of agreement were used to measure repeatability. Voxel-based and regions of interest-based results showed that absolute RC for the reconstructed PET images are within ~2%. The brain cortex and the cerebellum were the regions with the largest variability (~3%). The differences across visits were not statistically significant ( ${p=0.90}$ ). In conclusion, this paper shows for the first time the repeatability of the SPM-based pseudo-CT approach for brain MR-based AC. These results, in addition to the ease of implementation and the quality and robustness previously demonstrated, confer this SPM-based method an ideal candidate for routine brain PET/MRI research and clinical studies.