The data of this study is collected from two PET centers, a phantom study is used to examine the SUV measurement on both scanners. The experiment indicates that SUV from different scanners under the same image protocols and same scintillation detector type (BGO for both scanners) can be quite different in value. However, they follow very similar trends as size increases, the SUV value increased despite all spheres having the same T/B activity ratios, which is consistent with our clinical result. Accordingly, we recommend that the follow up scans to evaluate treatment response or re-stage the disease be performed on the same scanner to be comparable. The difference in SUV on different scanners despite the same T/B activity ratios might be attributed to the difference in calibration and machine-identity-features. Although, there was a difference in the SUVmax value between our two scanners of a factor of ~1.3× in the phantom study, we chose not to apply an adjustment of SUVmax for our clinical result because the average SUVmax of each nodule group from both centers were close to each other, particularly for group 1 and group 2. The averages of the SUVmax of group 1 and group were 3.03 and 5.28 for MC-1, respectively, and 3.3 and 5.43 for MC-2, respectively. In addition, overall accuracy using an SUVmax cutoff of 2.5 were similar. The accuracies were 77% and 75% for MC-1 and MC-2, respectively. The trendline, linear regression equation and R2 of malignant and benign nodules for MC-1 and for MC-2 demonstrate the same relation between nodule size and SUVmax. The relation is stronger for malignant than benign lesions. Consequently, we selected to keep the clinical data as it is without adjustment of SUVmax between the two scanners.
The results of the present study indicate that there is a relation between the size of pulmonary nodules and the SUV value. The linear regression equation and R2 for malignant nodules and for benign nodules, as well as the trendlines for malignant and benign nodules demonstrated that the slope of the regression line was greater for malignant than for benign nodules. In Figure 2, it can be seen that on the left side of the graph, where the small nodules (≤ 1 cm) are plotted, the nodules mixed randomly with no predominant areas for benign or malignant nodules. No SUVmax cutoff can separate them. However on the middle and right side, where larger size nodules (> 2.0 cm) are plotted, the nodules become more polarized, and the malignant nodules predominate in the upper portion of the plot area where the SUV is high, while the benign nodules predominate in the lower portion of the plot area where SUV is lower. Determination of an SUV cutoff for larger nodules is more feasible but not definite in the diagnosis of pulmonary nodules.
When the SUVmax cutoff of 2.5 was used to differentiate between malignant and benign pulmonary nodules. The sensitivity, specificity and accuracy of nodules for group 2 was 91%, 47%, and 79%, respectively. For group 3 it was 94%, 23%, and 76%, respectively. For group 4 it was 100%, 17%, and 82%, respectively. Although, the sensitivity and accuracy of the test increased with the increase in the size, reaching 100% and 82% respectively for nodules greater than 3.0 cm, the specificity declined from 47% for group 2 to 17% for group 4. The accuracy of differentiating large pulmonary nodules (> 1.0 cm) using SUVmax cutoff of 2.5 seems reasonable. However, no predetermined fixed SUVmax cutoff is able to differentiate pulmonary nodules as definitely benign or definitely malignant, regardless of the nodule's size.
One of the main findings of the present study was that the small nodules (≤ 1 cm) tend to have lower SUVs than larger nodules. The small benign pulmonary nodules have average SUV as equal as to malignant nodules. Thus, maximum or mean SUV is not accurate tool in the evaluation of small pulmonary nodules. Only 54% of the time was the test able to differentiate between malignant and benign nodules. Attempting to lower SUVmax to less that 2.5, such as 1.8 might increase the sensitivity of the test, however, the specificity is decreased resulting in no clinically significant improvement in the accuracy of the test to differentiate between the malignant and benign nodules. The sensitivity, specificity, and accuracy of a cutoff of 1.8 were 100%, 0.0%, and 46%, respectively. This result reflects the fact that FDG is not a specific tracer for malignancy. In our study, a variety of small benign nodules (≤ 1 cm) presented with mean and maximum SUV more than 2.5 and resulted in a false positive PET scan. (e.g., the SUVmax was 5.3 for squamous metaplasia, 4.6 for rheumatoid nodules, 4.2 for lymphoid tissue and 3.9 for TB). Other benign nodules such as granuloma, chronic inflammation, cryptococcus infection, reactive nodules and atypical hyperplasia also presented with high SUVmax leading to reading a false positive PET scan. On the other hand, some of well-differentiated and slow growing malignant nodules presented with SUVmax less than 2.5 (1.34 for squamous cell carcinoma, 1.77 for adenocarcinoma and 2.15 for small cell lung cancer).
The data above support that although, the SUVmax cutoff of 2.5 is a useful tool in the evaluation of large pulmonary nodules (> 1.0 cm), it has no or minimal value in the evaluation of small pulmonary nodules (≤ 1.0 cm). However, the combination of flexible value of SUVmax cutoff according to the size of the nodule, visual assessment, and CT characteristics of the nodules, in addition to pretest probability of malignancy, is the most appropriate approach to characterize small pulmonary nodules. To increase the sensitivity of the test of SUVmax cutoff for characterizing small nodules (≤ 1 cm), we recommend reducing the cutoff of less than 2.5
The limitation of this study is the exclusion of the negative PET scans. We exclude negative PET scan because the SUV of a non-FDG-avid nodule cannot be measured. Thus, the specificity of PET scan using an SUVmax cutoff of 2.5 calculated on this study is not reflecting the actual specificity of PET in the characterizing of pulmonary nodules
The introduction of dedicated PET/CT scanners to the clinical arena in early 2001 , has resulted in improved accuracy in the characterization of pulmonary nodules , by maintaining the synergism between the anatomic sensitivity of CT, and metabolic specificity of PET.
Although, FDG-PET/CT is a valuable diagnostic tool, it has multiple pitfalls that limit its accuracy in the evaluation of pulmonary nodules, particularly small nodules. There are three potential directions for future research to improve PET/CT accuracy in the evaluation of pulmonary nodules. One direction involves improvement of PET/CT scanner to provide better sensitivity, resolution and co-registration which potentially enhance its sensitivity to detect small pulmonary nodules, in addition to provide better quantitative and qualitative evaluation of pulmonary nodules. The second direction of future research involves imaging processing and display formats that might enhance the reader delectability. A PET/CT with virtual bronchoscopy provides virtual 3-dimensional images which enhances the intraluminal lesions . The third direction involves development and investigation of new PET radiotracers that might have better sensitivity and specificity to differentiate pulmonary nodules. Both 18F-fluorothymidine (18F-FLT) and 18F-fluorocholine (18F-FCH) have been developed and investigated for use in lung cancer [16–18], however neither tracer has shown clear improvement over 18F-FDG. Eventually, these three directions of future research will improve the delectability and categorization of the pulmonary nodules.