| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | RSS | TABLE OF CONTENTS |
|
|
|
||||||||
|
||||||||||
Clinical Investigation |
1 Department of Molecular and Medical Pharmacology, David Geffen School of Medicine, University of California, Los Angeles, California; and 2 Department of Radiological Sciences, David Geffen School of Medicine, University of California, Los Angeles, California
Correspondence: For correspondence or reprints contact: Christiaan Schiepers, MD, PhD, David Geffen School of Medicine, University of California, CHS AR-259, MC 694215, Los Angeles, California 90095-6942. E-mail: cschiepers{at}mednet.ucla.edu Guest Editor: Tom R. Miller, Mallinckrodt Institute
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
Key Words: PET/CT • 18F-FDG • SPN • lung cancer
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
Noninvasive evaluation of SPNs is usually done by chest radiography, CT, MRI, PET, and SPECT (5–8). Invasive procedures to allow histologic evaluation of SPNs include fiber-optic bronchoscopy, transthoracic needle-aspiration biopsy, video-assisted thoracoscopy, video-assisted thorascopic surgery, or thoracotomy, all of which are associated with considerable costs and morbidity. Surgery has a reported mortality of 0.5% (exploratory) to 4% (curative) (9).
CT provides anatomic and morphologic information. In addition, CT can confirm whether a lesion is truly solitary and can suggest the likelihood of malignancy. Computer-aided diagnosis may further enhance the performance of CT (10). Lesions are usually considered benign if they meet the following criteria: central, concentric calcifications, round, or no nodule growth by CT after 2 y of observation (11,12). Malignant features of a nodule typically include nondemarcated borders, eccentric appearance, spiculated pattern, and a doubling time of <10 mo.
PET with 18F-FDG is used to evaluate glucose metabolism in tumors. Because of increased metabolism, malignant tissues typically demonstrate higher 18F-FDG uptake than benign lesions and normal tissue. The utility of PET for the characterization of SPNs has been reported in a large number of publications. A metaanalysis by Gould et al. (13) has confirmed that PET using 18F-FDG is an accurate, noninvasive method for diagnosing SPNs, with an overall sensitivity of 96.8% and a specificity of 77.8%. The performance of dual-modality imaging with PET/CT for the evaluation of SPNs has not been investigated to date. The goal of this study was to assess the accuracy of PET/CT in the diagnosis and characterization of SPNs compared with PET and CT alone.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
Forty-two patients (12 men, 30 women; age range, 35–84 y; mean age ± SD, 67 ± 11 y) met the criteria and were included for further analysis. The mean nodule diameter ± SD was 15 ± 6 mm (range, 7–30 mm) by CT. Nine lesions were <10 mm in axial diameter. Of the lesions, 57% were found in the upper lobes, with 26% in the right upper lobe and 31% in the left upper lobe, whereas 43% were found in the lower lobes, with 19% in the left lower lobe and 24% in the right lower lobe.
None of the patients had diabetes mellitus. Serum glucose levels were measured by finger stick, but only 35 values were recorded and available in the charts. The serum glucose levels varied from 75 to 116 with a mean of 94 mg/dL.
PET/CT Image Acquisition
Patients were imaged using a whole-body PET/CT scanner (Siemens/CTI Reveal). The CT was a dual-slice system. CT images of diagnostic quality were acquired—that is, 130 kVp, 130 mA·s, 1-s rotation, 4-mm collimation, 8-mm bed feed. No intravenous contrast enhancement was used. The PET system was a lutetium oxyorthosilicate detector, 3-dimensional system. PET images were obtained 60 min after intravenous injection of 7.5 MBq/kg (0.2 mCi/kg) of 18F-FDG. The acquisition time was weight dependent and lasted 2–4 min per bed position as described (14). Multiple bed positions were acquired from approximately the patients' middle thighs to the base of the skull. CT-based attenuation correction was applied. PET and CT images were acquired during shallow breathing.
CT Image Interpretation
The interpretation of CT scans was performed by a thoracic radiologist who was unaware of each patient's history and PET image findings (J.G.). Each nodule was characterized in terms of its attenuation, shape, and margin characteristics using previously well-described criteria developed by the American College of Radiology Imaging Network for its lung cancer screening trials (15,16). For attenuation, the nodule was assessed as being soft tissue, ground glass, mixed ground glass and soft tissue, fluid water, fat, or calcified. If calcified, the amount and position of calcification were described. For shape, the lesion was described as being round, oval, or lobulated. The margin was characterized as being smooth, spiculated, or poorly defined. Finally, the perpendicular longest axial dimensions of the lesion were measured to the nearest millimeter. On the basis of integration of these characteristics (attenuation, shape, margin, and size) the reader then ascribed a 5-point likart scale that was believed to best represent the likelihood of the lesion being malignant. SPNs were scored as follows: 1, not suspicious for malignancy (e.g., soft tissue, round, well-defined margins, calcified, <10 mm); 2, low suspicion; 3, intermediate suspicion (e.g., mixed or soft-tissue attenuation, noncalcified, smooth margins, >10 mm); 4, moderately high suspicion; and 5, high suspicion (e.g., mixed attenuation, lobulated, spiculated, poorly defined, >10 mm). SPNs with a CT score of 1, 2, or 3 were considered benign and 4 or 5 were considered malignant.
PET Image Interpretation
PET images were interpreted by 2 nuclear medicine physicians who were unaware of the patient's history and anatomic imaging results (M.A., C.S.). The location of areas of abnormal tracer uptake were recorded and the degree of uptake was scored using a 5-point scale: 0, no uptake; 1, less than mediastinum; 2, equal to mediastinum; 3, between mediastinum and liver; and 4, higher than liver. Lesions with an uptake of 0 or 1 were considered benign and an uptake of 2, 3, or 4 was considered malignant. In addition, a region of interest was placed around areas of abnormal uptake and the maximum standardized uptake value (SUVmax) was determined and analyzed separately.
PET/CT Image Interpretation
Lesion size, location, and type were verified. Images were interpreted as benign or malignant by consensus. In case of concordance between CT and PET interpretation, the PET/CT classification did not differ from the CT and PET findings. If PET and CT ratings were discordant as to lesion type, the PET, CT, and fusion images were jointly reviewed. The PET/CT interpretation could be described with the following rules: If the CT score was 5 (high suspicion), then the lesion was considered malignant, regardless of the 18F-FDG uptake (Fig. 1). If the PET score was 2, 3, or 4 (moderate uptake to uptake greater than the liver), then the lesion was considered malignant—that is, the PET findings overruled the CT findings (Fig. 1). All but 2 lesions could be classified as either malignant or benign by applying these rules. The remaining 2 lesions, both with a CT score of 4 and PET score of 1, were reexamined and by consensus the PET/CT was read as benign in 1 case and malignant in the other based on the appearance of the fused images.
|
|
|
|
|
2 statistic. |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
PET
Although PET had the same overall accuracy as CT (74%), the sensitivity and specificity were 69% and 85%, respectively (Table 2). For nodules <10 mm in diameter, PET had a sensitivity and specificity of 50% and 100%, whereas the remaining nodules were detected with a sensitivity of 72% and a specificity of 75% (Table 3). The 9 false-negative findings included adenocarcinoma (n = 7), neuroendocrine carcinoma (n = 1), and BAC (n = 1). The average size of these 9 lesions on CT was 13 mm (range, 9–21 mm). There were 2 false-positive findings related to inflammatory conditions: 1 granuloma and 1 anthracosis with inflammation.
By semiquantitative analysis, the SUVmax for all positive PET lesions (n = 31) averaged at 3.0 ± 3.0 (range, 0.5–17.2) (Table 4). When using a cutoff of 2.0 for malignancy, SUV analysis yielded the same results as visual analysis for PET/CT. Correction for lean body mass (according to the formula published by Zasadny and Wahl (17)) did not improve accuracy but, rather, changed 1 true-positive result into a false-negative one. The corrected SUVmax was on average 2.1 ± 2.0 (range, 0.4–12.1). In the true-positive PET subgroup, the lesion size ranged from 8.5 to 29.5 mm, with a mean SUVmax of 3.8 (Table 4).
Dual-Modality Imaging
PET/CT had the highest accuracy, 93% (P < 0.05 for both CT and PET), with a sensitivity of 97% (P = not significant for CT, and P < 0.05 for PET) and a specificity of 85% (P < 0.05 for both CT and PET). There was 1 false-negative PET/CT result, an adenocarcinoma of 12 mm, in an area with emphysematous changes. PET/CT had 8 discordant interpretations compared with PET alone, all of which were true-positive. PET-only interpretations missed 4 lesions because there was no visible uptake. Of the other 4 lesions, 3 had low 18F-FDG uptake (SUVs of 1.2, 1.3, and 1.2), whereas the fourth had an SUV of 2.4 but was visually scored as 1 (benign, Fig. 1).
An ROC analysis was performed for PET and PET/CT, which is shown in Figure 2. The top 2 curves were based on the consensus interpretation. The additional value of PET/CT over PET-alone is clearly seen and appears to hold over the entire range of sensitivity and specificity. In addition, the SUVmax of each SPN was used to classify the lesions—that is, an analysis based only on PET uptake measurements. Cutoff levels for malignancy were set at 1, 2, 3, 4, and 5 SUVs and the ROC calculated. This curve is labeled PET SUV in Figure 2 and shows the optimal tradeoff between sensitivity and specificity for PET to be at cutoff points 3 and 4, corresponding to an SUVmax between 2 and 3. In addition, this ROC illustrates that visual interpretation of PET is similar to quantitative analysis by SUV.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
10 mm in a study from The Netherlands (20). In our study, the 9 nodules with a diameter of <1 cm were detected by PET with a specificity of 100% but with a sensitivity of only 50%. Conversely, CT had a sensitivity of 100% but a much lower specificity of 20% (Table 3). Upon consensus, PET/CT was found to have 100% sensitivity and 100% specificity for small lesions in our sample of 9 patients. There were 4 true-positive findings (2 NSCLC, 1 adenocarcinoma, 1 neuroendocrine cancer) and 5 true-negative findings (1 granuloma, 2 pneumonia, 2 inflammation). In this study, the final interpretation of PET/CT results was dichotomized, and only benign or malignant was used as classification of the SPN. "Equivocal" scores—that is, 3 on CT and 2 on PET—were categorized on the basis of the decision rules (Fig. 1, if discordant). In particular, CT is affected by this type of analysis because the so-called indeterminate nodule with an equivocal interpretation is not considered (5). In addition, ground-glass opacifications are generally interpreted as malignant in this scheme. As is shown in Table 2, dual-modality imaging with PET/CT seems to overcome many of the equivocal findings. The PET/CT results were a consensus interpretation. The scores obtained from the independently read CT and PET studies provided the rules shown in Figure 1 and were congruent with the consensus reading except for 2 cases. It has been suggested that simultaneous interpretation of PET/CT is not only superior to CT alone or PET alone but also superior to a nonsimultaneous interpretation of CT and PET information (21). Our study design does not permit comparison of these reading strategies.
Lesion characterization can be accomplished through visual or quantitative analysis. Although visual assessment by an experienced radiologist or nuclear medicine specialist is sufficient, the lesion's SUV has been reported to be helpful in characterization and prognosis. Lesions with an average SUV higher than 2.5 (3) have been considered malignant. In the current study, the 31 positive PET lesions were evaluated in a semiquantitative manner, using a cutoff criterion of 2.0 for the SUVmax. Similar to Hashimoto et al. (22), who used a PET-only system and an SUVmax cutoff of 2.5, we found that characterization based on quantitative analysis produced the same results as visual interpretation. In addition, accuracy did not improve after lean body mass correction and, in fact, changed 1 true-positive finding into a false-negative one. This indicates that for clinical situations, visual interpretation is at least sufficient if not superior.
The ROC of Figure 2 shows that, indeed, our readers performed as expected. Using the uptake in the liver (around 3 SUVs) and mediastinal blood pool (around 2 SUVs) as reference produces the PET curve in Figure 2, which is comparable with the PET SUV curve (based solely on SUVmax) over a large range (specificities, 5%–85%). Statistical analysis by the AUC (Fig. 2) revealed that the quantitative analysis was not significantly different from the PET-alone interpretation. PET/CT was superior and significantly different from PET alone: AUC = 0.90 versus 0.79 (P < 0.03, Fig. 2).
Although SPNs have been documented to have a 40% malignancy rate, this percentage varies with age, smoking history, location, and patient selection (6,8,11,12). For example, lung cancer is extremely rare in patients under the age of 30 y, whereas patients over 50 y of age have higher malignancy rates. In our study, the mean patient age was 67 y, thereby offering one explanation for the 69% malignancy rate. In addition, our population was selected from a larger pool of patients who were referred for a PET/CT study because they had pathologic verification of their lung lesion. This induces a selection bias, and our malignancy rate was almost twice as high as that in the general population. The prevalence of disease heavily affects the positive predictive value (PPV) and negative predictive value (NPV); therefore, our values should be used with caution as they do not reflect the PPV or NPV in screening of the general population. Confirmation bias may also have a small effect on sensitivity and specificity.
One of the drawbacks of metabolic imaging is that abnormal 18F-FDG uptake is not specific for malignancy. Benign lesions such as bacterial pneumonia, active sarcoidosis, infectious granulomas, acute pyogenic abscesses, cryptogenic fibrosing alveolitis, and so forth have been known to produce false-positive readings on PET (6). In California, the specificity and PPV of PET scans are known to be lower because of the prevalence of fungal infections such as coccidiomycosis. Although the metaanalysis of Gould et al. (13) found a sensitivity of 94% and a specificity of 86% in characterization of SPNs (<4 cm in size) with 18F-FDG PET, our institution found earlier a sensitivity of 80% and specificity of 75% (23). In the current study, 18F-FDG PET alone had a sensitivity and a specificity of 69% and 85%, respectively.
The sensitivity of 18F-FDG PET to detect malignant lesions is compromised for an SPN with low metabolic activity or an SPN that is small in size. For example, pure BACs are typically hypometabolic and, as a result, are usually false-negative. In 2002, Yap et al. (24) found that PET had a sensitivity of 33% in the characterization of 6 pure BAC lesions. These authors found a PET sensitivity of 89% for adenocarcinomas with up to 75% BAC component. A typical false-negative PET example is shown in Figure 3. Small lesions are challenging because of PET's limited spatial resolution, and lesions <7 mm are seen only when the metabolic activity is very high. In our population, PET-only interpretation failed in 9 lesions: 4 had no visible uptake, 3 had low metabolism (1 BAC and 2 neuroendocrine), and 2 were <1 cm.
|
Potential errors due to needle-biopsy sampling were likely small in this study because of the multiple cuts through the nodule. Only half of the SPNs were confirmed by biopsy, which is a limitation of our study.
On a similar note, 1 lesion was initially classified as false-positive by PET/CT, confirmed by biopsy and bronchial lavage (Fig. 4). Eighteen months after the first PET/CT, the patient had another PET/CT examination, which showed that the lesion had grown in size and increased its metabolic activity. A left lobectomy was subsequently performed, which confirmed that the mass was a BAC with an area of poorly differentiated adenocarcinoma. In this case, the initial PET/CT detected and characterized the lesion correctly, whereas the initial biopsy and pathology produced a false-negative result. This is a sampling error related to inadequate tissue for histology (5).
|
This study substantiates PET/CT in the characterization of solitary pulmonary lesions. The combination of anatomic and metabolic imaging preserves the sensitivity from CT and maintains the specificity from PET, resulting in an overall improved accuracy. In other words, there is synergy between anatomic and metabolic imaging when performing PET/CT. There were significant differences between the performance of PET/CT and either PET alone (sensitivity, specificity, accuracy, PPV) or CT alone (specificity, PPV), as demonstrated in Table 2.
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONReferences |
|---|
Related articles in JNM:
This article has been cited by other articles:
|
|
 |
|
  R. E. Fisher PET for the Evaluation of Solitary Pulmonary Nodules J. Nucl. Med., February 1, 2009; 50(2): 326 - 326. [Full Text] [PDF]
|
|
|
|
 |
|
 K. Kodama, M. Higashiyama, K. Takami, K. Oda, J. Okami, J. Maeda, M. Koyama, and T. Nakayama Treatment strategy for patients with small peripheral lung lesion(s): intermediate-term results of prospective study Eur. J. Cardiothorac. Surg., November 1, 2008; 34(5): 1068 - 1074. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. W. Fletcher, S. M. Kymes, M. Gould, N. Alazraki, R. E. Coleman, V. J. Lowe, C. Marn, G. Segall, L. A. Thet, K. Lee, et al. A Comparison of the Diagnostic Accuracy of 18F-FDG PET and CT in the Characterization of Solitary Pulmonary Nodules J. Nucl. Med., February 1, 2008; 49(2): 179 - 185. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Tian, X. Yang, L. Yu, P. Chen, J. Xin, L. Ma, H. Feng, Y. Tan, Z. Zhao, and W. Wu A Multicenter Clinical Trial on the Diagnostic Value of Dual-Tracer PET/CT in Pulmonary Lesions Using 3'-Deoxy-3'-18F-Fluorothymidine and 18F-FDG J. Nucl. Med., February 1, 2008; 49(2): 186 - 194. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | RSS | TABLE OF CONTENTS |
| JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY | THE JOURNAL OF NUCLEAR MEDICINE |
). The 5-point PET/CT scale was created by applying the rules of 
) is based on consensus interpretation. PET SUV curve (
) is based on maximum uptake in the SPN with cutoff levels of 1, 2, 3, 4, and 5 for the SUV. AUC was 0.90 for PET/CT, 0.82 for PET SUV, and 0.79 for PET.
