From: A review of PET attenuation correction methods for PET-MR
Study | Method | Validation data†(tracer) | Region | Reported error (%) | |
---|---|---|---|---|---|
Proposed method | Vendor method | ||||
Kops et al. [166]* | Generation of pseudo-transmission images by bringing the averaged database images to the patients’ space | 4 (2-[18F]FDG) | Brain | 9 | N/A |
Hofmann et al. [188] | Generation of pseudo-CTs using pattern recognition to find similar patches between a brain MR-CT pair database and target MRI | 3 (N/A) | Brain | 3.2 ± 2.5 (VOIs in brain) | N/A |
Malone et al. [167]* | Generation of pseudo-transmission images by registering a brain MR-transmission maps database to the patients’ T1 | 10 (2-[18F]FDG) | Brain | −1.2 ± 1.2 | N/A |
Wollenweber et al. [170] | Pseudo-CT generation by registration of patient’s T1 image to a CT database | 13 ([18F]FDG) | Brain | 0.2 ± 2.6 (VOIs in brain) | N/A |
Arabi et al. [190] | Adaptation of [188] using voxel-wise local normalised cross-correlation | 14 (2-[18F]FDG) | Whole Body |  ~ 5 ± 4 (cerebellum),  ~ 4 ± 10 (lung) | N/A |
Burgos et al. [60] | Generation of pseudo-CTs by registering a MR-CT pair database to patients’ T1 image a using local image similarity measure | 41 (2-[18F]FDG) | Brain | 0.2 ± 2.1 | UTE: −11.8 ± 2.1 |
Izquierdo-Garcia et al. [172] | Generation of 6-class probability maps from patients’ T1 image and coregistration to a similarly created probability map database with paired CT images. Averaged CT is brought back to patients’ space | 15 (2-[18F]FDG or [18F]FET) | Brain | −1.0 ± 2.5 | Dixon: 13.0 ± 10.2 |
Poynton et al. [173] | Registration of 3-class probability maps from CT data-dUTE-T1 database to patients’ corresponding MR images to generate a tissue probability map | 13 (N.A.) | Brain | 1.8 ± 2.4^ | N/A |
Chen et al. [181] | Generation of pseudo-CTs by registering a MR-CT pair database to patients’ T1 image. Air segmentation is performed using an air probabilistic map and sparse regression to identify similar patches between atlas and patients’ MR | 20 ([18F]Florbetapir) | Brain | 2.4 ± 1^ | Dixon: 12.7 ± 2.2^ |
Burgos et al. [177] | Use of convolution-based local normalised correlation coefficient to irregular regions to account for mismatches in the FOV between CT and MRI and extension of the database to include both T1 and T2 images | 15 (2-[18F]FDG or [18F]Florbetapir) | Brain | 1.6 ± 0.6 (2-[18F]FDG), 1.9 ± 0.6 ([18F]Florbetapir) □ | N/A |
Arabi et al. [186] | Pseudo-CT generation by registering an averaged atlas image to the patients’ T1 and use the precomputed transformation matrices to bring all atlas images to the patients’ space | 41 (2-[18F]FDG) | Whole body | 3.5 ± 6.1 (cerebellum), 3.2 ± 9.8 (lung), −4.9 ± 5.9 (bone) | N/A |
Teuho et al. [174] | Assignment of μ values on a tissue probability map generated from T1 images in SPM | 7 (2-[18F]FDG) | Brain | −0.3 ± 2.7 □ | UTE: −7 ± 2.3□ |
Wallstén et al. [187] | Statistical decomposition algorithm to generate pseudo-CTs from the patients’ T2 images | 15 ([11C]acetate) | Pelvis | −2.3(prostate lesion) −4.2 ± 5.7 (bone) | Dixon: −5.6 Dixon: −17.7 ± 8.4 |
Sousa et al. [165]* | A database of paired of 68Ge-transmission AC maps-T1 images registered to the patients’ T1 image and averaging all transmission AC maps | 9 ([11C]PE2I) | Brain | −0.1 ± 3.2 | N/A |