Table 2.
Methods of GTV delineation on PET in correlation with surgical specimens.
| Patient no. | Method of GTV delineation on PET | Correlation between CT, PET, PET/CT, and pathological tumor size | |
|---|---|---|---|
| Lin et al. (22) | 37 | Halo for tumor observed in fused PET-CT images | Stronger correlation between GTV and pathological tumor dimensions were observed with PET/CT |
| Mean SUV of the external margin of halo was 2.41 ± 0.73 | |||
| T stage and histology significantly influenced SUV at the edge of the halo | |||
| Yu et al. (23) | 52 | SUV of 2.5 | FDG-PET/CT has significantly better correlation with surgical specimens than CT or PET alone, especially in the presence of atelectasis |
| Yu et al. (24) | 15 | Best correlation between PET GTV and the actual tumor was found at the SUV threshold of 31 ± 11%, and absolute SUV cut-off of 3.0 ± 1.6 | |
| Wu et al. (25) | 31 | Thresholding with 20–55% of SUVmax | Maximal primary tumor dimension was more accurately predicted by CT at the window-level of 1,600 and −300 HU than PET GTVs (best correlation with pathological tumor volume at 50% SUVmax) |
| Schaefer et al. (27) | 15 | Tumor threshold = A*mean SUV70% + B*background | Pathological tumor volume: 39 ± 51 mL |
| PET tumor volume: 48 ± 62 mL | |||
| CT tumor volume: 60.6 ± 86.3 mL | |||
| Both CT and PET volumes are highly correlated with pathological volumes (p < 0.001). | |||
| Increased variation between PET and pathological tumor volumes were observed in lower lobes | |||
| van Baardwijk et al. (28) | 33 | Source-to-background ratio auto-segmentation | Maximal tumor diameter of the PET GTV is highly correlated with that in surgical specimens (CC = 0.90). Auto-segmented GTVs are smaller than manually contoured GTVs on PET/CT |
| Wanet et al. (31) | 10 | Gradient-based method | Comparison of both CT and PET GTV |
| Fixed threshold at 40 and 50% of the SUVmax. | Gradient-based method led to the best estimation of the GTV | ||
| Adaptive thresholding based on the source-to-background ratio | PET GTVs were smaller than CT GTVs in general | ||
| Cheebsumon et al. (32) | 19 | Absolute SUV cut-off (2.5) | Adaptive 50% and gradient-based methods generated the most consistent maximal tumor dimension, which had a fair correlation with the pathological tumor size |
| Fixed threshold at 50% and 70% SUVmax | |||
| Adaptive thresholding 41–70% SUVmax | |||
| Contrast-oriented algorithm | |||
| Source-to-background ratio | |||
| Gradient-based method |