From: A review of PET attenuation correction methods for PET-MR
Study | Method | Validation data†(Tracer) | Region | Reported error (%) | |
---|---|---|---|---|---|
Proposed method | Vendor method | ||||
Siegel et al. [122]* | Segmentation of body contour from sinogram data |  ~ 300 (2-[18F]FDG, [18F]DOPA, [15O]H2O and [13N]Ammonia) | Whole body |  < 8.9% (Brain 2-[18F]FDG), ± 10% (renal flow ([13N]Ammonia) | N/A |
Chang et al. [145] | Segmentation of body organs from NAC PET images | 14 (2-[18F]FDG) | Thorax | 6 ± 7 (lung lesions), 3 ± 6 (bone lesions) ▪ | N/A |
MLAA with prior MR spatial and CT statistical information using Gaussian mixture model (MLAA-GMM) | 5 (2-[18F]FDG) | Whole body | −11.6 ± 6.0 (cerebrum), −3.5 ± 6.6 (lung), −12.6 ± 8.6 (bone) | Dixon: −18.5 ± 11.3, Dixon: −5.4 ± 12.0, Dixon: −19.8 ± 8.4 | |
Mehranian et al. [137] | MLAA-GMM including lung segmentation from CT images | 17 (2-[18F]FDG and 2 with [18F]Choline) | Whole body | −0.8 ± 6.3 (lung), −5.7 ± 3.2 (myocardium), −2.0 ± 2.1 (bone) | Dixon: −5.2 ± 7.1, Dixon: −9.5 ± 4.3, Dixon: −14.1 ± 2.5 |
Benoit et al. [143] | MLAA using UTE and T1 images as anatomical prior information for updating the attenuation map | 204 (2-[18F]FDG) | Brain | −1.5 ± 3.5 | Dixon: −11.8 ± 4.5, UTE: −6.5 ± 2.0 |
Mehranian et al. [133] | MLAA with spatial constrains from MR images using penalty functions during image reconstruction for updating μ values (P-MLAA+) and both μ-map and activity distribution (P-MLAA++) | 8 (2-[18F]FDG) | Brain | −3.0 ± 3.5 (P-MLAA+), −4.2 ± 3.6 (P-MLAA++)□ | Dixon: −13.5 ± 3.1 |