Dual Time Point ¹⁸F-FDG PET for the Evaluation of Pulmonary Nodules

MATERIALS AND METHODS

Thirty-six patients (21 women, 15 men; mean age, 67 y; range, 36–88 y) with known or suspected malignant lung densities, who were referred for routine ¹⁸F-FDG PET scanning, were examined twice: initial whole-body imaging followed by a second scan of the chest. All scans were obtained on a dedicated whole-body C-PET scanner (ADAC Laboratories, Milpitas, CA). Informed consent was obtained from all patients. At the time of ¹⁸F-FDG injection all patients had fasted for at least 6 h and had blood sugar levels of <140 mg/dL.

¹⁸F-FDG was synthesized using the method described by Hamacher et al. (14). Image acquisition for the whole-body scan started at a mean time point of 69 min (range, 55–110 min) after injection of 2.52 MBq/kg of body weight. This first scan (scan 1) started on all patients at the shoulders and included thorax, abdomen, and pelvis. It consisted of 4 or 5 emission frames of 25.6-cm length with an overlap of 12.8 cm covering an axial length of 64–76.8 cm. After an interval of 56 min (range, 49–64 min), a second emission scan of the thorax only (scan 2) was acquired on all patients at a mean time of 122 min after tracer injection (range, 100–163 min). Starting at the same position as scan 1, it covered an axial length of 25.6–38.4 cm (1 or 2 frames). A transmission scan was obtained with both sets of images for attenuation correction. Image reconstruction was performed with an iterative ordered-subsets expectation maximization algorithm with 4 iterations and 8 subsets (15). Attenuation-corrected images were obtained by applying transmission maps, which were acquired after ¹⁸F-FDG injection with a ¹³⁷Cs source interleaved with the emissions scans (16,17).

Regions of interest (ROIs) were overlaid onto the fully corrected PET images of scans 1 and 2 in the area of the known radiographic lung density. This was achieved by direct visual assessment of the lesion position on the CT scan and subsequent identification of the corresponding area on PET scans 1 and 2.

The margins of these ROIs were placed at approximately 50% of the maximal counts of the highest lesion counting density. The ROI was then transferred to the other scan of the same patient. In tumor lesions that extended over several slices in the craniocaudal direction, the ROI was placed in the midportion of the lesion where the maximal counts were measured. If no discernible uptake was present on either PET scan, ROIs were drawn in the presumed location that corresponded best with that of the radiographic density. Because the patients did not have to leave the scanning table between the end of the first and the beginning of the second emission scan, only minimal correction was required in a few cases because of patient motion.

The SUV was calculated according to the following standard formula (3): .

The presence of malignancy was proven by obtaining a biopsy or by resecting the lesion in 19 patients. In 1 patient with clinical and radiographic findings highly suggestive of a malignancy, no tissue diagnosis was established. In addition, this patient responded favorably to radiation therapy, which further enhanced the presumed diagnosis. The lack of a malignant process was established in 2 of 18 patients by resecting the lesion. In the remaining 16 patients, stability (or resolution in 1 patient) of the lesion on radiographic examination over an extended period of time (18–26 mo) was considered as evidence for a benign process.

RESULTS

The patient characteristics and scan results for malignant lesions are shown in Table 1, whereas the data for patients with biopsy-proven benign or highly likely benign lesions are summarized in Table 2. Thirty-eight lesions were evaluated because 2 patients had 2 lesions each with focal ¹⁸F-FDG uptake (lesions 5 and 6 and lesions 37 and 38).

The average SUV in malignant tumors (mean ± SD) was 3.66 ± 1.95 on scan 1 and 4.43 ± 2.43 on scan 2. The change in SUV for this group of patients was 20.5% ± 8.1% (P < 0.01), indicating a significant increase in SUV between the 2 time points for the entire study population. In contrast, the uptake values for the benign lesions were 1.14 ± 0.64 on scan 1 and 1.11 ± 0.70 on scan 2 (P = not significant), suggesting that SUVs for benign lesions between the first and second emission scans predominantly remained constant or declined. Four of 20 malignant lesions had uptake values of <2.5 (range, 1.12–1.69), whereas 2 of 18 nodules categorized as benign revealed SUVs of ≥2.5.

Figure 1 shows a 1.8 × 1.0 cm lesion in the right lower lobe (lesion 9). The SUV was 1.45 on scan 1 and 1.72 on scan 2. The uptake increased by 18.6% between the scans. Biopsy of this lesion revealed a moderately differentiated adenocarcinoma. Figure 2 shows the images of lesion 29. This patient had a 0.8-cm density in the left upper lobe. SUV measurements were 1.54 (scan 1) and 1.50 (scan 2), a minor decrease of 2.6%. The thoracic surgeon elected to resect the nodule, which revealed a benign granuloma.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

A 72-y-old woman presented with right lower lobe nodule measuring 1.8 × 1.0 cm. SUV increased from 1.45 on scan 1 at 65 min after injection (A) to 1.72 on scan 2 at 123 min after injection (B). Biopsy of lesion revealed moderately differentiated adenocarcinoma.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

A 69-y-old woman presented with 0.8-cm density in left upper lobe. SUV was not significantly different between scan 1 at 54 min after injection (A), where SUV was 1.54, and scan 2 at 100 min after injection (B), where SUV was 1.50. Excised lesion was granuloma.

When assessing the diagnostic value of the first emission scan by applying an SUV threshold of 2.5 for separating benign from malignant lesions, the sensitivity was 80% and the specificity was 94%. When the same SUV threshold was applied to the data from the second set of scans, the sensitivity remained unchanged, whereas the specificity decreased to 89%.

The SUV increase methods yielded a most accurate result when an SUV threshold of 10% was used. By adopting an SUV increase of >10% between the first and second scans as a criterion for malignancy, all 20 neoplastic lesions were identified correctly and 16 of 18 lesions were diagnosed correctly as being benign. When an SUV increase of >5% was used as the threshold, the sensitivity, specificity, and accuracy were 100%, 61%, and 82%, respectively. When a threshold of 15% was used, the sensitivity, specificity, and accuracy were 60%, 94%, and 76%, respectively. Therefore, the sensitivity for this method was 100% while the specificity and accuracy remained quite high, equaling 89% and 95%, respectively, when a threshold of 10% was used to diagnose a malignant lung nodule.

DISCUSSION

In the last decade ¹⁸F-FDG PET scanning has proven to be a valuable tool for the assessment of undetermined pulmonary nodules. In addition to visual assessment of the metabolic activity of the nodules, measurement of the SUV for the semiquantitative assessment of ¹⁸F-FDG uptake in pulmonary lesions has proven to assist in differentiating between malignant and benign nodules. Several reports consider it to be a simple and useful tool for this purpose, and most publications conclude that a threshold value of 2.5 is optimal for obtaining a high sensitivity while maintaining a good specificity (1–4,7,8,9).

However, other investigators have found that determination of the SUV suffers from significant limitations. Hamberg et al. (18) showed that the usual scan start times of 45–60 min lead to significant underestimation of the true SUV because, in most tumors, ¹⁸F-FDG uptake continues to rise beyond this period and typically does not reach a plateau for several hours. In untreated tumors, 95% of the plateau value was reached at 298 ± 42 min, with a range of 130–500 min. Although the authors found a positive correlation between the glucose metabolic rate and the SUV, the value was only R² = 0.65, indicating a considerable margin of error.

Lodge et al. (19) came to a similar conclusion in a study of 29 patients with various benign and malignant soft-tissue masses. High-grade sarcomas reached the maximal ¹⁸F-FDG uptake at 4 h, whereas uptake in benign lesions reached its maximal value within 30 min. In this study, the diagnostic value of Patlak or nonlinear regression analysis was not superior to SUV measurements at 4 h. The authors believed that 1 possible explanation might be the poor counting statistics and increased noise several hours after ¹⁸F-FDG injection.

These results are in contrast to a study by Lowe et al. (20), who assessed the change in SUV over time in a cohort of 14 patients with pulmonary abnormalities (10 malignant, 4 benign). On the basis of measurement of the signal-to-noise ratio, the best separation between benign and malignant lesions occurred at 50 min after injection and no improvement was seen at later time points.

In view of the encouraging results by Hustinx et al. (13), who acquired scans at 2 time points for head and neck tumors, we adopted this study approach for the evaluation of pulmonary nodules. The percentage SUV change between the first and second scans with a threshold of 10% increase in measured values provided a higher sensitivity (100% vs. 80%), while maintaining an excellent specificity (89% vs. 94%), than that obtained from a single image acquisition using the usual SUV threshold method.

The results of this study for the sensitivity and specificity measures by adopting the SUV threshold of 2.5 are similar to those published in the literature (1–8). However, there was a clear benefit of calculating the percentage SUV change between the 2 scans compared with the SUV threshold alone. Four neoplastic lesions of 1.5- to 2.0-cm maximal diameter showed relatively low uptake values on scans 1 and 2 but showed a considerable increase in the SUV between these 2 scans. None of these lesions was of a histologic type that is known to frequently exhibit low-uptake values, such as bronchioalveolar carcinoma. The SUV measurements of these 4 relatively small lesions could be underestimated because of volume averaging (10). There was a marked percentage increase in SUV in these 4 lesions (mean, 19.5%), which was not significantly different from that of the entire group of malignant lesions. One explanation for this phenomenon is that, at the time of the second emission scan (approximately 2 h), the peak ¹⁸F-FDG concentration is not reached in most malignant lesions. On the basis of the results of Hamberg et al. (18) and Lodge et al. (19), the likely mechanism appears to be that most neoplasms continue to accumulate ¹⁸F-FDG for several hours after injection.

The results of this study and those of other reports (18,19) provide an argument for adopting an extended scanning protocol, which involves imaging the pulmonary nodules at 2 time points. In particular, small- and medium-sized lesions, which may not be detected because of the limited resolution of the technique or relatively low glucose utilization (secondary to a low rate of cell division), could be identified as malignant on the basis of an increased SUV over time.

Although there was no difference between the standard SUV threshold and the percentage SUV change method for identifying benign lesions in this study, the latter method may prove advantageous when uptake values are close to the threshold of 2.5. Acquiring a second scan that reveals stable or declining uptake values suggests a benign lesion and reduces the need for surgical exploration in such indeterminate cases.

Two lesions (lesions 28 and 35) with a high likelihood of benign etiology showed increases of >10% of their SUVs (12.7% and 29.3%) over time. The patient with significant increase continues to show no evidence of malignant features in her lung lesion. Because of the patient’s general medical condition and the stability of the lesion, no biopsy has been obtained so far to clarify the etiology of the lesion. Given the generally very low uptake values of either of these 2 lesions, we conclude that the application of the threshold technique probably has no value in lesions with SUV of <1.0.

The limitation of our study is the small number of patients. This study did not include any bronchioalveolar carcinoma, which sometimes does not show any increased ¹⁸F-FDG activity. This study included only 18 benign lung nodules with 2 false-positive results. Some benign granulomatous lesions, such as sarcoidosis, aspergillosis, and coccidiomycosis, have been reported to be ¹⁸F-FDG avid and to show increasing uptake over time (20,21), resulting in false-positive studies. Thus, patients with granulomatous disease or bronchioalveolar carcinoma are the main causes of false-positive and false-negative studies.

Finally, if one should make a final interpretation of the study solely on the basis of the change in the SUV, rigorous attention to technique is paramount for the evaluation of pulmonary nodules. A change of 10% between the early and late images is very small when the SUV is small, and any patient’s motion between the 2 scans can significantly affect and falsify the result. The change in the SUV is also dependent of the reproducibility of the ROI between both scans.

On the basis of the data reported in this article, the dual time point scan protocol is relatively simple and is practical in the setting of a clinical PET center.

CONCLUSION

Our results indicate that dual time point ¹⁸F-FDG PET scanning can improve the sensitivity and possibly also the specificity in the evaluation of pulmonary nodules, especially in small- to medium-sized lesions. Further studies are needed to confirm these preliminary results and improve the statistical accuracy.