Introduction

Lung cancer is the leading cause of cancer-related death in both men and women worldwide. In China, the mortality rate is 30%, which ranks first [1]. In recent years, morbidity and mortality have been increasing in China [1]. For patients with lung cancer, accurate N-staging is critical for developing individualized treatment plans and predicting prognosis [2]. In addition to the loss of surgical options for stage N3 patients, the 5-year survival rate is reduced to 6% [3]. Therefore, improving the accuracy of lung cancer N-staging to reduce the false-positive rate is one of the issues that has attracted clinical attention.

18F-fluorodeoxyglucose (FDG)–positron emission tomography/computed tomography (PET/CT) plays an important role in the differential diagnosis, staging, response assessment, and prognosis of lung cancer. The sensitivity and specificity of FDG PET/CT in the differential diagnosis and accurate staging of lung cancer are limited. Because FDG is not a tumor-specific imaging agent, the standard uptake value (SUV) is affected by a variety of factors [4, 5]. In China, as an endemic region for endemic infectious diseases, some benign lung diseases such as tuberculosis, infections, and inflammatory and granulomatous diseases cause FDG-avid, leading to a decrease in the specificity of FDG PET/CT in N-staging of lung cancer [6,7,8,9]. In recent years, the use of dynamic FDG PET/CT (dPET/CT) imaging in oncology has received much attention. Dynamic metabolic parameters, such as net influx rate (Ki) and tumor blood flow (K1), obtained based on the two-tissue irreversible compartment model approach, better describe the different metabolic stages of FDG and thus reflect the pathophysiological mechanisms of the disease [10, 11].

Previously, we conducted a series of studies related to the clinical application of dPET/CT in lung cancer [12,13,14,15]. In our preliminary study [12,13,14,15], we investigated the clinical value of dPET/CT in the differential diagnosis, N-staging, and epidermal growth factor receptor (EGFR) status prediction of lung cancer. The results of our preliminary study showed that the dynamic metabolic parameter Ki has a better differential diagnostic value in lung cancer differential diagnosis (cutoff value of 0.0250 ml/g/min) and EGFR status prediction (cutoff value of 0.0350 ml/g/min), especially improved specificity [13]. Of particular note, in a previous study [12], we compared the value of static metabolic parameters [SUVmax, lymph node (LN)–SUVmax/primary tumor (PT)–SUVmax] and dynamically visualized metabolic parameters (Ki and Ki/K1) in metastatic and non-metastatic FDG-avid LNs of lung cancer. We tentatively concluded that the dynamic metabolic parameters Ki and Ki/K1 with high specificity at cutoff values of 0.022 ml/g/min and 0.093, respectively (specificity of 0.918 and 0.776, and AUCs of 0.672 and 0.673, respectively) were able to better discriminate the metastatic and non-metastatic LNs from the FDG-avid LNs. For static metabolic parameters, SUVmax and LN-SUVmax/PT-SUVmax showed high sensitivity (0.826 and 0.999) at cutoff values of 4.050 and 0.236, respectively, but the specificity was suboptimal (specificity 0.388 and 0.204, and AUC 0.596 and 0.566, respectively).

To our knowledge, there are fewer previous relevant studies on the cutoff value of i. Based on our previous studies [12,13,14,15], we found that although Ki has good specificity in differential diagnosis, the sensitivity is not very reasonable. Therefore, in this study, we further investigated the clinical valueK of (SUVmax and LN-SUVmax/PT-SUVmax) and dynamic metabolic parameters (Ki and Ki/K1) in the differential diagnosis of FDG-avid LNs in lung cancer based on the cutoff value in our previous findings. The clinical value of each metabolic parameter in single or combined detection was investigated, and the results of the preliminary study were validated.

Materials and methods

Patients

The study was approved by the Ethics Committee of Cancer Hospital and Shenzhen Hospital, Chinese Academy of Medical Sciences (KYLH2022-1), and all patients signed a written informed consent before FDG PET/CT imaging.

A total of 121 patients underwent dPET/CT (chest, 65 min) + static FDG PET/CT imaging (sPET/CT, whole body, 10–20 min) from April 2022 to August 2023. All patients were found to have pulmonary nodules/masses on chest CT scans and were not receiving anti-inflammatory or anti-tumor therapy.

From the 121 patients, we retrospectively collected 126 FDG-avid LNs from 37 lung cancer patients to perform the present study. All 37 included patients enrolled had pathologically confirmed lung cancer. Postoperative pathology and/or puncture biopsy pathology confirmed 126 FDG-avid LNs as metastatic or non-metastatic LNs. The interval between the FDG PET/CT scan and the pathology results was less than 2 weeks.

Patients' FDG PET/CT scan characteristics of patients were collected, including PT-SUVmax, FDG-avid LN long diameter, short diameter, LN-SUVmax, and dynamic metabolic parameter values (including Ki and K1). Clinical information about the patients was also collected, including age, sex, type of primary lesion pathology, LN pathology results, and TNM staging [16]. LN locations were categorized according to the International Association for the Study of Lung Cancer LN map into the mediastinal region (zones 1–9) and the pulmonary hilar region (zones 10–12) [17]. The LNs enrolled in both dPET/CT and sPET/C corresponded to postoperative pathologic findings and or puncture biopsy sites and results. Finally, the results of the N-staging in dPET/CT and sPET/CT were based on the eighth edition of the TNM lung cancer classification [16].

PET/CT data acquisition, reconstruction, and analysis

Figure 1 shows the dPET/CT + sPET/CT examination process, data acquisition, image reconstruction, and metabolic parameter acquisition for the patient. All scans were performed in a Discovery MI PET/CT (GE Healthcare, Milwaukee, USA).

Fig. 1
figure 1

dPET/CT + sPET/CT scans examination procedure and data processing

Full size image

Dynamic parameters of Ki and K1 were then calculated and obtained based on the two-tissue irreversible compartment model. Ki/K1 was computed subsequently as a separate marker. The image-derived input function (IDIF) was extracted from the ascending aorta by drawing a 10-mm-diameter ROI on six consecutive slices in an image obtained by combining early time frames (0–60 s), where the effects of motion and partial volume are less pronounced than in the left ventricle. In addition, this study did not account for differences in blood and plasma uptake. In this model, we assumed unidirectional uptake of 18F-FDG (i.e., k4 = 0) with irreversible trapping in the tissue as 18F-FDG-6-PO4 [18]. Parametric images of each dynamic scan were generated using voxel-based analysis. Due to the large number of voxels in a whole-body image, the Lawson–Hanson non-negative least-squares algorithm was used to solve a linearized problem instead of the conventional non-linear one [19].

Pathological diagnosis

In this study, postoperative pathology and/or puncture biopsy pathology were the gold standard for follow-up.

All puncture and/or postoperative specimens were fixed in formalin, dehydrated, and embedded in paraffin. Four-micron sections of each tissue were stained with hematoxylin and eosin (H&E) and immunohistochemistry. The diagnosis was based on microscopic appearance and immunohistochemical results. The diagnosis was made independently by two experienced pathologists. In case of disagreement, the diagnosis was clarified after a full departmental discussion.

Statistical analysis

Continuous variables were reported as the median and interquartile range. Categorical variables were described as number and frequency. The χ2 test was used to evaluate the agreement of the individual and combined diagnosis of each metabolic parameter with the gold standard. The receiver-operating characteristic (ROC) analysis was performed for each parameter to determine the diagnostic efficacy in differentiating non-metastatic from metastatic LNs with high FDG-avid. The difference in the area under the curve (AUC) was determined by Delong’s test. A P value of less than 0.05 was considered statistically significant. All statistical analyses were performed with R statistical software (version 4.1.1).

Results

Patients and LN characteristics

Patient and LN characteristics are shown in Table 1. Among the 37 patients who underwent dPET/CT + sPET/CT scans, the mean age was 59.95 (59.95 ± 11.44) years, and the number of male and female patients was 24 (64.87%) and 13 (35.13%), respectively. Among the 126 LNs that were pathologically confirmed, 56 (44.44%) LNs were non-metastatic and 70 (55.56%) LNs were metastatic.

Table 1 Characteristics of the patients and LNs
Full size table

Primary tumors and FDG-avid LNs 18F-FDG PET/CT characteristics summary

Among the 37 patients, the average SUVmax of the primary tumor was 10.95 (10.95 ± 4.48). In the metastatic LNS group (N = 70), the ranges of long diameter, short diameter, SUVmax, LN-SUVmax/PT-SUVmax, Ki, and Ki/K1 were 1.95 cm [1.43;2.55], 1.30 cm [1.10;1.80], 7.75 [5.85;12.38], 0.69 [0.53;1.08], 0.023 [0.024;0.050] ml/g/min, and 0.295 [0.138;0.478], respectively. In the non-metastatic LNS group (N = 56), the ranges of long diameter and short diameter, SUVmax, LN-SUVmax/PT-SUVmax, Ki, and Ki/K1, were 1.30 cm [1.18;1.50], 1.00 cm [0.90;1.10], 3.50 [2.78;4.63], 0.45 [0.28;4.63], 0.012 [0.009;00016] ml/g/min, and 0.056 [0.0248;0.146], respectively.

Diagnostic value of individual and combined tests for each metabolic parameter

In previous studies, we have concluded that the cutoff values of 4.05, 0.236, 0.022 ml/g/min, and 0.093 for the static metabolic parameters SUVmax, LN-SUVmax/PT-SUVmax, and the dynamic metabolic parameters Ki, Ki/K1, respectively. Therefore, based on the above thresholds, we tested static metabolic parameters, and dynamic metabolic parameters individually and in combination. Table 2 and Fig. 2 show the results of ROC analysis of static and dynamic metabolic parameters when tested individually and in combination, respectively.

Table 2 ROC analysis results for each metabolic parameter when tested individually and in combination
Full size table
Fig. 2
figure 2

The results of the ROC analysis of SUVmax and Ki when tested individually and in combination

Full size image

In separate assays, the dynamic metabolic parameter Ki [sensitivity (SEN) of 84.30%, specificity (SPE) of 94.60%, positive predictive value (PPV) of 84.29%, negative predictive value (NPV) of 94.64%, accuracy of 88.89%, and AUC of 0.895] had a better diagnostic value than the static metabolic parameter SUVmax (SEN of 82.90%, SPE of 62.50%, PPV of 82.86%, NPV of 65.50%, accuracy of 74.60%, and AUC of 0.727) in differentiating between metastatic and non-metastatic LNs groups, respectively.

In the combined diagnosis group, the combined SUVmax + Ki diagnosis had a better diagnostic value in the differential diagnosis of metastatic from non-metastatic LNs, with SEN, SPE, PPV, NPV, accuracy, and AUC of 84.3%, 94.6%, 84.29%, 94.64%, 88.89%, and 0.907, respectively.

Accuracy of metabolic parameters SUVmax and K i in N staging

Of the 126 LNs in 37 patients with lung cancer, 70 were metastatic, and 56 were non-metastatic. At the time of sPET/CT diagnosis, N-staging was consistent with pathology in 20 patients, but not in 17 patients, of which 15 patients were over-staged and 2 patients were under-staged. For the combined SUVmax + Ki diagnosis, N-staging was accurate in 30 patients, but N-staging was inconsistent with pathology findings in 7 patients, of which 3 patients were over-staged and 4 patients were under-staged. Table 3 shows the comparison of N-staging based on SUVmax and SUVmax + Ki diagnosis with pathological findings (Fig. 3).

Table 3 Accuracy of sPET/CT and dPET/CT N-staging in 37 patients with lung cancer
Full size table
Fig. 3
figure 3

dPET/CT and sPET/CT images of non-metastatic FDG-avid LNs

Full size image

A 66-year-old female patient. Surgical pathology confirmed squamous cell carcinoma in the upper lobe of the right lung (A, E, I, white arrow, size of 3.5 × 3.1 cm, SUVmax of 14.8). FDG PET/CT scan showed multiple FDG-avid LNs in the zone 11R, 7, and 4R. Among them, the FDG-avid LN in zone 11R (B, F, white arrow) with a size of 1.4 × 1.3 cm, SUVmax of 4.0, and Ki of 0.010 ml/g/min. The FDG-avid LN in zone 7 (C, G, white arrow) with a size of 1.0 × 0.8 cm, SUVmax of 4.4, and Ki of 0.016 ml/g/min. The FDG-avid LN in zone 4R (D, H, white arrow) with a size of 2.1 × 1.3 cm, SUVmax of 4.1, and Ki of 0.013 ml/g/min. Finally, all FDG-avid LNs in zones 11R, 7, and 4R were pathologically confirmed to be cancer-free (Fig. 4).

Fig. 4
figure 4

dPET/CT and sPET/CT images of non-metastatic and metastatic FDG-avid LNs

Full size image

A 66-year-old male patient. FDG PET/CT showed a nodule in the upper lobe of the left lung (A, B), with a size of 1.0 × 0.6 cm, and SUVmax of 4.8. FDG PET/CT scan showed FDG-avid LNs in zones 4L and 4R. Among them, the FDG-avid LN in zone 4L (C, F, white arrow) was pathologically confirmed adenocarcinoma metastasis (D), with a size of 4.2 × 3.8 cm, SUVmax of 34.6, and Ki of 0.135 ml/g/min. The FDG-avid LN in zone 4R (C, F, yellow arrow) was pathologically confirmed non-metastasis (E), with a size of 1.3 × 0.9 cm, SUVmax of 4.0, and Ki of 0.015 ml/g/min.

Discussion

In lung cancer, accurate N-staging is essential for developing personalized treatment plans and determining prognosis and is one of the clinical concerns. In this study, based on the previous study, we further confirmed the high specificity of the dynamic metabolic parameter Ki in the differential diagnosis between metastatic from non-metastatic FDG-avid LNs in lung cancer. The combination of SUVmax and Ki is expected to be a reliable imaging basis for accurate N-staging of lung cancer.

Radiopharmaceutical distribution is a dynamic process that varies widely in diseases and individuals [20]. Previous studies have confirmed that dPET/CT extracts physiological and biochemical parameters that better reflect the pathophysiological mechanisms of disease compared to sPET/CT. These parameters (e.g., Ki) have been shown to discriminate between benign and malignant diseases [10,11,12,13,14,15, 19, 21,22,23]. Previous studies have confirmed the value of dPET/CT in the differential diagnosis of lung cancer and inflammatory lesions, but there are fewer studies in the differential diagnosis of LNs. Previously, in our studies related to dynamic metabolic parameters in the differential diagnosis of lung cancer LNs [12], we discussed the value of dynamic metabolic parameters (including K1, K2, Ki, and Ki/K1) in the N-staging of lung cancer. We concluded that when the cutoff values of Ki and Ki/K1 were 0.022 ml/g/min and 0.093, respectively, there was a good diagnostic value in the differential diagnosis of metastasis and non-metastasis in FDG-avid LNs of lung cancer, especially improving the specificity (AUC of 0.672 and 0.673, and specificity of 0.918 and 0.776, respectively). Therefore, to further investigate the diagnostic value and feasibility of dynamic metabolic parameters in N-staging of lung cancer based on the critical values of our previous study, we conducted this validation study.

In this study, we found that the static parameters SUVmax (cutoff value of 4.05) and LN-SUVmax/PT-SUVmax (cutoff value of 0.236) had high sensitivity (82.5% and 92.7%) and accuracy (74.60% and 61.11%), but the specificity was not satisfactory (62.50% and 17.90%). The addition of the dynamic metabolic parameter Ki (cutoff value of 0.022) resulted in a significant improvement in the differential diagnosis when SUVmax and Ki were combined (AUC of 0.907), with high sensitivity (84.30%), specificity (94.60%), and accuracy (88.89%).

Among the 37 lung cancer patients in our study, at the time of sPET/CT diagnosis, N-staging was consistent with pathology in 20 (54.05%) patients, but not in 17 (45.94%) patients, of which 15 (40.54%) patients were over-staged and 2 (5.41%) patients were down-staged. For the combined SUVmax + Ki diagnosis, N-staging was accurate in 30 (81.08%) patients, but N-staging was inconsistent with pathology findings in 7 (18.92%) patients, of which 3 (8.10%) patients were over-staged and 4 (10.81%) patients were down-staged. After adding the dynamic metabolic parameter Ki for co-diagnosis, eight patients (21.62%) diagnosed as stage N3 by SUVmax were downgraded to stage N0–2, and four (30.77%) patients underwent successful surgical treatment. Therefore, SUVmax + Ki combined diagnosis has high diagnostic value in the differential diagnosis of FDG-avid LN metastasis from non-metastasis in lung cancer, especially to improve the specificity. In particular, it improves the diagnostic accuracy of patients with clinical suspicion of stage N3, seeks surgical opportunities as much as possible, and reduces some unnecessary puncture biopsies. To assist clinicians in accurate N-staging to improve the prognosis and quality of life of lung cancer patients.

Previous studies have reported that [24,25,26], the false-positive rate of squamous cell carcinoma on sPET/CT is significantly higher than that of adenocarcinoma and small cell carcinoma. This may be because squamous cell carcinomas are mostly central bronchogenic carcinomas, often associated with obstructive pneumonia and atelectasis, which can activate macrophages and inflammatory cells and cause reactive hyperplasia of LNs, leading to false-positive sPET/CT results [27]. In our sample size of 21 LNs (21/126) from 5 patients (5/37) with squamous cell carcinoma, the PT-SUVmax ranged from 8.5 to 16.4 and LN-SUVmax ranged from 2.5 to 11.5. N-staging sPET/CT failed to accurately stage all five patients, misstating two and over-staging three patients. Three of these patients were accurately staged with the SVUmax + Ki combination. Three (3/13) of these patients had pathologically confirmed N0, but SUVmax was assessed as N3 (1/3) and N2 (2/3), and with the addition of Ki, the N3 (1/3)/N2 (1/3) stage was reduced to N0 (2/3). Among the patients in the adenocarcinoma group (N = 26), 17 patients (17/26) were accurately staged by sPET/CT, but 4 patients (4/26) were misstated and 5 patients (5/26) were over-staged. Among the over-staged patients (N = 5), PT-SUVmax ranged from 1.3 to 14.8, LN-SUVmax (N = 20) ranged from 2.6 to 17.2, and all were staged as stage N2 by sPET/CT. When SUVmax + Ki was combined with the test, the five patients who were over-staged as described above were accurately staged, with three down-staged to N0 and two down-staged to N1, all consistent with pathologic findings. Among patients with small cell carcinoma (N = 5), 3 patients (3/5) were accurately staged by sPET/CT, but 2 patients (2/5) were over-staged. Among the over-staged patients (N = 2), PT-SUVmax was 7.6 and 9.0, respectively, LN-SUVmax (N = 3) ranged from 2.6 to 4.2, and sPET/CT staging was N3 and N2, respectively. When SVUmax + Ki was detected in combination, the patients originally staged as N3 by sPET/CT were downgraded to N2, and those staged as N2 were downgraded to N0, which was consistent with the pathological results. The inclusion of the dynamic metabolic parameter Ki improved the specificity of the differential diagnosis. The combined SUVmax + Ki is expected to be a reliable metabolic parameter for lung cancer N-staging. In addition, whether the dynamic metabolic parameter Ki differs between pathology types and degree of differentiation is also of interest and one of our ongoing research topics.

There are several limitations to our study. First, we used a retrospective design due to our use of pathologic findings as the gold standard and the fact that the dPET/CT scan takes 65 min, resulting in our small sample size. In addition, prolonged scanning may lead to patient discomfort and feasibility issues. Second, motion correction was not considered in this study. It is known that motion in the chest region can affect not only the SUV but also the kinetic parameters quantification [28,29,30]. Third, SUVmax rather than SUVmean was used in this study, because we believed SUVmax to be more stable and less affected by partial volume effects.

Conclusions

When the cutoff value of the Ki was 0.022 ml/g/min, it had a high diagnostic value in the differential diagnosis between metastasis and non-metastasis in FDG-avid LNs of lung cancer, especially in improving the specificity. The combination of SUVmax and Ki is expected to be a reliable metabolic parameter for N-staging of lung cancer.