Application of Dynamic 18F-FDG PET/CT for Distinguishing Intrapulmonary Metastases from Synchronous Multiple Primary Lung Cancer

It has been a big challenge to distinguish synchronous multiple primary lung cancer (sMPLC) from primary lung cancer with intrapulmonary metastases (IPM). We aimed to assess the clinical application of dynamic 18F-FDG PET/CT in patients with multiple lung cancer nodules. We enrolled patients with multiple pulmonary nodules who had undergone dynamic 18F-FDG PET/CT and divided them into sMPLC and IPM groups based on comprehensive features. The SUVmax, fitted Ki value based on dynamic scanning, and corresponding maximum diameter (Dmax) from the two largest tumors were determined in each patient. We determined the absolute between-tumor difference of SUVmax/Dmax and Ki/Dmax (ΔSUVmax/Dmax; ΔKi/Dmax) and assessed the between-group differences. Further, the diagnostic accuracy was evaluated by ROC analysis and the correlation between ΔSUVmax/Dmax and ΔKi/Dmax from all groups was determined. There was no significant difference for ΔSUVmax/Dmax between the IPM and sMPLC groups, while the IPM group had a significantly higher ΔKi/Dmax than the sMPLC group. The AUC of ΔKi/Dmax for differentiating sMPLC from IPM was 0.80 (cut-off value of Ki = 0.0059, sensitivity 79%, specificity 75%, p < 0.001). There was a good correlation (Pearson r = 0.91, 95% CI: 0.79-0.96, p < 0.0001) between ΔSUVmax/Dmax and ΔKi/Dmax in the IPM group but not in the sMPLC group (Pearson r = 0.45, p > 0.05). Dynamic 18F-FDG PET/CT could be a useful tool for distinguishing sMPLC from IPM. Ki calculation based on Patlak graphic analysis could be more sensitive than SUVmax in discriminating IPM from sMPLC in patients with multiple lung cancer nodules.


Introduction
Lung cancer is among the most common cancers and the leading cause of cancer-related death worldwide [1]. Given the advanced diagnostic and surveillance methods, as well as the increasing aging population, there has been a recent increase in the incidence of synchronous multiple primary lung cancer (sMPLC). It is crucial to discriminate sMPLC from primary lung cancer with intrapulmonary metastases (IPM) due to their varying clinical staging, treatment strategies, management, and prognosis. Patients with sMPLC are staged separately and generally treated with curative surgical resection because of early stage and favorable prognosis while those with IPM in advanced stages receive chemotherapy or radiotherapy with palliative intent [2].
Differential diagnosis between sMPLC and advanced lung cancer is traditionally based on conventional histopathologic features. These include location, morphology, histologic type, time interval, lymphatic invasion, metastases, and clinical manifestation, as described by the Martini and Melamed criteria published in 1975 [3], which was subsequently modified by the American College of Chest Physicians (ACCP) [2]. However, this traditional diagnostic process often involves an overlap in a significant proportion of cases, especially for tumors with a similar morphology or histotype, which impedes the determination of whether they share the same clone origin. Moreover, some patients cannot undergo preoperative histological examination or relative surgery. Further, some tumors cannot be sampled and tested pathologically if the patient has a poor physical condition, including limited cardiopulmonary reserves.
Novel molecular and genomic analyses, including biomarker assessment (driver gene mutations) [4], array comparative genomic hybridization [5], shallow whole-genome sequencing [6], and TP53 mutation analysis [7], can define the between-lesion correlation. These molecular biological techniques allow the detection and analysis of specific molecular markers or mutation sites to determine the heterogenicity of two cancer foci. However, given the substantial misclassification rate, which results from limited sensitivity, stabilization, repeatability, and economic benefits, techniques that could be able to precisely define sMPLC or IPM have not been widely applied in clinic [8].
Preoperative differentiation of primary tumors from metastases is more clinically pivotal than postoperative differentiation. Moreover, preoperative imaging examination plays a significant role. However, there is limited information regarding the imaging and metabolic characteristics of multiple lung cancer nodules [8]. Pretreatment chest highresolution computed tomography (HRCT) is a conventional and cost-effective means of preliminary diagnosis. Since distinguishing IPM from MPLC is empirically based on the morphological pulmonary nodule features, it is difficult to obtain a definitive diagnosis of a primary tumor or metastasis solely based on CT characteristics. Moreover, according to the ACCP clinical practice guidelines, 18 F-fluorodeoxyglucose positron emission tomography/computed tomography ( 18 F-FDG PET/CT) plays a significant role in guiding clinical decisions and is a recommended standard workflow for patients with potentially curable lung cancer [2]. 18 F-FDG PET/CT has been extensively used in patients with lung cancer for diagnosis [9], staging [10], prediction (i.e., monitoring therapy response) [11], and prognosis [12]. These mentioned applications indicate that 18 F-FDG PET could be used to identify early-stage sMPLC, which involves multiple pulmonary sites. Several heterogeneous methods could be used to determine 18 F-FDG uptake. Dynamic PET acquisition, immediately starting from radiotracer injection, measures drug activity change over time. Dynamic PET including multiple frames usually requires long image acquisition time (for 18 F-FDG, generally lasting for 60 minutes) and only allows the assessment of one FOV. Generally speaking, the radioactivity concentration of radiotracer in blood and tissue will be changing most rapidly early after injection; therefore, most dynamic PET acquisition has finer time frames at early time points and wider time frames at later time points [13]. On the other hand, static PET acquisition is a common clinical PET acquisition mode. After radiotracer injection for a period of time, when the physiological metabolism or binding in vivo is stable, the multiple-FOV and single frame static emission scan is acquired. For 18 F-FDG, static emission scan lasting for 10-15 minutes is usually started at 45-60 minutes postinjection. In routine clinical practice, the standardized uptake value (SUV) based on static PET scanning, which is a simple semiquantitative index that reflects the metabolic activity of tumor lesions, is widely applied in PET imaging evaluation. SUV is associated with several tumor characteristics, including histopathological subtypes [14]; tumor proliferation [14], differentiation [15], and aggressiveness [16]; and tumor stage, recurrence, and survival [17,18]. However, the SUV outcome is affected by several factors, including the image acquisition time [19], acquisition mode [20], reconstruction mode [21], serum glucose and insulin levels [22,23], and positive contrast agent [24]. Theoretically, Patlak graphic analysis based on dynamic PET scanning is the most classical fully quantitative measure of glucose metabolism that involves irreversible trapping. It calculates the 18 F-FDG net influx rate constant (i.e., uptake rate constant, K i ) through linear graphical data fitting [25,26]. Being a linear kinetic model, Patlak analysis lacks noise amplification and is independent of uptake time and changes in plasma FDG clearance [27,28].
Since sMPLC with a separate clonal origin often indicates early-stage and less than two-year interval tumorigenesis, we hypothesized that there would be the same SUV or K i for the initial and second primary lung cancers. Since IPM with an identical clonal origin often indicates latestage tumorigenesis, we assumed that the SUV or K i of the primary lung cancer was larger than pulmonary metastases. Moreover, it remains to be assessed whether K i calculation through Patlak graphic analysis has values for differentiating sMPLC from IPM in patients with multiple lung cancer nodules since it provides more robust dynamic information and accurate 18 F-FDG metabolism quantification.
This study is aimed at assessing the differential diagnostic ability of dynamic 18 F-FDG PET scan in patients with lung cancer involving multiple pulmonary sites. This could be able to provide a novel method for sMPLC diagnosis in clinic.

Materials and Methods
2.1. Patients. In our center, firstly, all first-visit patients with related symptoms underwent preliminary screening using chest HRCT. Then, other examinations mainly including serologic examination, tumor-associated antigen, brain magnetic resonance imaging, and cervical and abdominal CT scan further clarified the general condition and clinical TNM classification. Fifty-three patients with multiple 2 Molecular Imaging pulmonary nodules regardless of extrapulmonary metastases received dynamic 18 F-FDG PET scans and had not received any medical treatment before this. After careful preoperative evaluation based on above examination, twenty-seven suitable patients (stage I, stage II, and partial stage IIIA) without medical contraindications received radical surgical resection. Partial inoperable patients (stage IV and partial stage IIIA) were followed by CT-guided percutaneous lung puncture biopsy, bronchoscopic biopsy, or superficial lymph node biopsy mainly based on tumor location. What calls for special attention was that two cases whose △K i /D max lower than 0.0001 were excluded considering quantitative inaccuracies due to minima. Finally, this study enrolled forty-three patients with multiple lung cancer nodules (either suspected or proven by biopsy/resection). Based on histopathologic, clinical, radiologic, and genetic features, two experienced oncologists divided those patients into the sMPLC (19 cases) and IPM groups (24 cases) reference to the ACCP evidencebased clinical practice guidelines [2]. That is, the division criteria of sMPLC are (1) same histology, tumor in different lobe as primary, and no N2 and N3 involvement, and no systemic metastases; (2) or different histology, molecular genetic characteristics or arising from a separate focus of carcinoma in situ [2]. The division criteria of IPM are (1) same histology and multiple systemic metastases; (2) or same histology in different lobe and presence of N2 and N3 involvement; (3) or <2-year interval [2]. The inclusion criteria were (1) patients whose preliminary chest HRCT showing multiple pulmonary nodules; (2) patients who performed dynamic 18 F-FDG PET/CT for pulmonary nodules; (3) at least two lesions were identified as lung cancers based on histopathologic, clinical, radiologic, and genetic features. The exclusion criteria were (1) no pathological validation; (2) having other diseases that present multiple pulmonary nodules; (3) known history of other malignant diseases; (4) diabetes and/or severe cardiovascular disease. Figure 1 presents the flowchart for case inclusion. Specifically, we only included one lesion pair (two lesions: the largest and second-largest lesion) per patient for simplifying comparisons and calculations, especially for patients with more than two lesions in the lung. Based on the anatomic sites of both lesions, two tumor nodules were subgrouped as follows: (1) tumors confined to the same or different unilateral lung lobes and (2) tumors confined to different bilateral lung lobes. Moreover, the IPM group was classified into 3 subgroups based on the primary tumor size as follows: ≤3 cm subgroup; 3-5 cm subgroup; >5 cm subgroup. The study was approved by the institutional review board of the Fifth Affiliated Hospital of Sun Yat-sen University (IRB protocol number ZDWY.FZYX.002). All the included patients provided signed informed consent. The clinical trial registration number is NCT03679936. Baseline clinical characteristics, including sex, age, height, weight, smoking history, and tumor characteristics, were obtained from electronic medical records with permission.

Dynamic PET Data Acquisition and Reconstruction.
Dynamic PET/CT scan was performed using 112-ring digital light guide PET/CT (uMI780, United Imaging, China). The patients were fasted for at least 6 h before scanning. The patient was restricted to movement in the scanner to avoid motion artifacts and conducive to subsequent accurate fusion of PET and CT. The scan covered the region between the thoracic inlet and the lower liver margin. Each PET/CT scan began with a transmission CT scan for 5 seconds that was used Excluded cases (n = 2) Cases whose K i /D max lower than 0.0001 were excluded considering quantitative inaccuracies due to minima Excluded cases (n = 8) (i) None of the lesions was pathologically validated; (ii) Other diseases that present multiple pulmonary nodules except for lung cancer sMPLC group (n = 19) Conclusion mainly depend on: (i) Different histopathologic subtype (n = 12) (ii) Same histopathologic subtype but different gene mutation (n = 1) (iii) Combine histopathology with radiologic features (n = 6) 3 Molecular Imaging for attenuation correction. Next, an 18 F-FDG bolus (range 143-327 MBq) was intravenously injected and a dynamic PET scan was acquired immediately as follows. Dynamic data were collected for 60 min comprising 48 frames with the following dimensions: 18 × 5 s, 6 × 10 s, 5 × 30 s, 5 × 60 s, 8 × 150 s, and 6 × 300 s (Figure 2(a)). The acquired data were corrected for decay, scatter, random, and attenuation; moreover, they were reconstructed using ordered subset expectation maximization.
2.3. PET Data Analysis. Two experienced nuclear medicine physicians analyzed the dynamic 18 F-FDG PET/CT images using Carimas 2.10 software (Turku PET center, Finland). Reconstructed PET images were analyzed as follows: (1) definition of volumes of interest (VOIs); (2) obtaining the timeactivity curve (TAC) of the left ventricle (i.e., plasma input function) and lung cancer nodules; (3) K i calculation; (4) SUV max calculation.
First, three-dimensional VOIs were manually drawn over the left ventricle (arterial blood pool) and tumors with a landmark using the Carimas 2.10 software. VOIs were visually localized using CT images. Moreover, the maximum diameters (D max ) of the two tumors (the largest and second-largest lesion) were measured on multiplanar reconstructed CT images in the lung window via unidimensional measurements based on Response Evaluation Criteria in Solid Tumors 1.0 (RECIST criteria 1.0) [29].
Second, by projecting the VOIs onto the complete dynamic dataset, the Carimas analysis software automatically output relative TAC data of the VOIs. The left ventricle TAC was used for plasma input function, also known as image-derived input function (IDIF).
Third, the TAC of the left ventricle and tumors were used to fit standard Patlak modeling to assess FDG tissue kinetics via the least square regression method using Matlab 2018b (Math-Works Inc., Natick, MA, USA) [25,26]. Patlak analysis was performed using data obtained between 20 and 60 postinjection minutes (i.e., between frames 38 and 48). The Patlak model for 18 F-FDG metabolism in lung cancer ( Figure 2(b)) has been previously described in detail [30].
Fourth, the last frame (at 55-60 min postinjection) of the dynamic scans was used for static analysis to obtain the SUV. The SUV max of each pulmonary malignant lesion on PET/CT was extracted from the dynamic data.

Calculation of Indicators
Based on SUV max measured from 18 F-FDG PET, the fitted K i value obtained through Patlak graphic analysis, and the corresponding maximum diameter (D max ) of two tumors, the absolute between-tumor difference of SUV max /D max and K i /D max (ΔSUV max /D max ; ΔK i /D max ) was calculated in both the sMPLC and IPM groups.

Histopathologic Analysis and Genetic Mutation Analysis.
All the punctured tissue and resected specimens were fixed, dehydrated, embedded, sectioned, and stained for microscopic examination. Two experienced pathologists recorded the histopathologic diagnoses and features, including the histotype; lymphatic, nerve, and pleural invasion; extranodal extension; and regional lymph node metastasis.
Gene sequencing of some patients (18 out of 19 in sMPLC group; 20 out of 26 in IPM group) was carried out by the Beijing Genomics Institute. Gene sequencing projects include (1) Oseq TM-T tumor individualized diagnosis and treatment gene detection: 508 gene coding regions and partial intron regions closely related to solid tumors were detected; (2) Oseq TM-ctDNA noninvasive tumor individualized diagnosis and treatment gene detection: 508 gene coding regions and partial intron regions closely related to solid tumors were detected;   Molecular Imaging 2.6. Statistical Analysis. Between-group differences in baseline characteristics were assessed using an unpaired two-tailed Student's t-test or chi-square test. Quantitative data were analyzed through descriptive statistics (median (minimum-maximum)), scatter plots, Mann-Whitney test, and Pearson correlation (GraphPad Prism 8 software; two-tailed; 95% confidence intervals). Statistical significance was set at p < 0:05.

Patient and Tumor Characteristics and Gene Detection
Results. There were no differences in the baseline characteristics between the sMPLC and IPM groups. Table 1 summarizes the gender distribution, age, height, body weight, FDG injection dose, and smoking history. IPM diagnoses were          (Tables 2-4, Figures 3(a) and 3(b)). Supplementary Table S3 shows the comparison of ΔSUV max /D max and ΔK i /D max between different age groups and no significant difference was found within their respective groups. Considering a small clinical sample size, internal validation by Bootstrap method by SPSS 25.0 further validated the reproducibility. When performing bootstrapping based on 1000 bootstrap samples, the difference was still more significant in ΔK i /D max (p = 0:003) than in ΔSUV max /D max (p = 0:031) between two groups (Supplementary Table S4). However, we did not conduct external validation because of limited included cases at present. There was a good correlation (Pearson r = 0:91, 95% CI: 0.79-0.96, p < 0:0001, Figure 3 Tables 3 and 4 independently present individual results of each case in the sMPLC group and the IPM group including location of tumor, pathological subtype/type, 8th TNM stage, D max , SU V max , K i , ΔSUV max /D max , and ΔK i /D max . Specially, in the sMPLC group, each lesion was independent and staged separately.

Receiver Operating Characteristic (ROC) Curve Analysis.
ROC curve analysis was used to determine the diagnostic accuracy of ΔSUV max /D max and ΔK i /D max . For ΔSUV max / D max , ROC curve (AUC = 0:66, Figure 3(e)) indicated low diagnostic value. For ΔK i /D max , the area under the curve (AUC) was 0.80 (95% CI: 0.67-0.93, p < 0:001) for distinguishing sMPLC from IPM (Figure 3(f)), which suggested that ΔK i /D max had a moderately high diagnostic value. The left upper corner (ΔK i /D max = 0:0059) of the ROC curve with the maximal Youden index was chosen as the optimal cut-off point, which had a sensitivity and specificity of 79% and 75%, respectively. The corresponding falsepositive and false-negative rate was 25% and 21%.

12
Molecular Imaging (OS) of MPLC [31,32]; therefore, we compared the results between unilateral and bilateral sMPLC. The bilateral sMPLC group had a slightly higher ΔK i /D max . However, there was no significant difference between unilateral and bilateral sMPLC with respect to ΔSUV max /D max and ΔK i /D max ( Supplementary  Figures S1(a) and S1(b)). "Primary" represents that tumors were considered as primary tumors. "Metastasis" represents that tumors were considered as metastatic tumors. "Unsure" represents that no definite diagnosis could be made based on CT characteristics. 13 Molecular Imaging 3.5. Unilateral IPM vs. Bilateral IPM. According to the 8 th Edition Lung Cancer Stage Classification, unilateral and bilateral IPM could have different T and M categories, which indicates different stages and OS [33]. Therefore, we compared differences in the indicators between unilateral and bilateral IPM. The bilateral IPM group had a slightly higher ΔSU V max /D max . However, there was no significant difference between unilateral and bilateral sMPLC in terms of ΔSU V max /D max and ΔK i /D max (Supplementary Figures S1(c) and S1(d)).

IPM Subgroup Comparisons Based on Primary Tumor
Size (≤3 cm vs. 3-5 cm vs. >5 cm). Size plays a critical role in defining the T category as indicated by the proposed size cutpoints of the 8 th edition Lung Cancer Stage Classification [34]. We compared differences in the indicators among three IPM subgroups (≤3 cm vs. 3-5 cm vs. >5 cm). Nevertheless, there was no significant among-group difference in ΔSUV max /D max and ΔK i /D max (Supplementary Figure S2).

Diagnostic Results
Based on CT Characteristics. Table 5 summarizes the diagnostic results of at least two experienced radiologists based on CT characteristics in the sMPLC group and the IPM group. Based on CT characteristics, 26 out of 43 cases were diagnosed correctly. Based on the optimal cut-off value of △K i /D max = 0:0059, 35 out of 43 cases were diagnosed correctly. Therefore, △K i /D max performed better than experienced CT diagnosis in differentiating sMPLC from IPM.

Discussion
To our knowledge, this was the first study on the application of dynamic 18 F-FDG PET to discriminate between tumors with common (i.e., IPM) and separate (i.e., sMPLC) lineages. Previous PET studies have predominantly focused on static imaging. Our findings indicated that dynamic 18 F-FDG PET could be able to provide more detailed parameters for sMPLC identification.
The Martini and Melamed criteria are currently the most widely accepted; however, there remains no standard and uniform clinical guidelines for MPLC. There is no consensus among major lung cancer research institutes regarding MPLC classification. It is difficult to discriminate sMPLC from IPM without a lung, lymph node biopsy, or surgical procedure, which is important since the two conditions have significantly different therapeutic regimens and prognosis. Previous studies have reported that the OS of patients with MPLC was significantly better than those with metastatic tumors [31,35]. Patients with MPLC are generally treated with curative surgical treatment while those in advanced stages receive chemotherapy or radiotherapy with palliative intent [2].
In routine clinical settings, preoperative differentiation of primary tumors from metastases is more critical than postoperative differentiation. Moreover, preoperative imaging examination plays a crucial role. It is recommended that patients with multiple pulmonary nodules undergo PET scans for careful systemic assessment based on the ACCP guidelines [2]. There is limited information regarding the imaging and metabolic characteristics of multiple lung cancer nodules [8]. Moreover, there have been few studies on the value of preoperative imaging for distinguishing MPLC from IPM in patients with multiple lung cancers [36].
Several case reports have reported incidental detection of sMPLC using 18 F-FDG PET [37][38][39][40]. Contrastingly, there have been few studies indicating that 18 F-FDG PET could be able to locate the clonal origin of synchronous multiple tumors [41][42][43]. Dijkman et al. reported that the relative betweentumor difference in SUV -ΔSUV%ð½bigger SUV − smaller SUV/bigger SUVÞ could distinguish advanced disease from second primary tumors in patients with synchronous pulmonary lesions [41]. In this previous study, the second primary tumor group included initial primary cancers originating from the lung or other organs with second primary lung cancer. The metastatic disease group (control) included primary lung cancers with intrapulmonary metastases and extrapulmonary cancers with pulmonary metastases. Therefore, their findings reflect the differential diagnostic ability of ΔSUV% from a wider scope that includes cancer originating from the lungs or other organs. Furthermore, Pang et [39]. Kosaka et al. assessed 75 cases of lung cancers with 296 metastases and reported that the SUV ratio (i.e., metastatic SUV to primary SUV) could differentiate primary lung cancer from metastasis [42]. However, most of the metastatic lesions were not pathologically confirmed and they only assessed lung cancers with metastases in the following sites: lymph nodes, bones, liver, pleura/lung, adrenal, kidney, and small intestine. Since this study did not involve a control group, its conclusions were empirical and probabilistic.
Similar to our study, Liu et al. only included patients with both cancers located in the lung and reported significant differences in the SUV max ratio (bigger SUV max /smaller SUV max ) between sMPLC and IPM. This indicated that the betweentumor SUV max ratio could differentiate sMPLC from IPM [43]. In our study, we did not find a significant betweengroup difference in the SUV max ratio and K i ratio. These inconsistent findings could be attributed to differences in the sample size and study population, as well as inclusion and exclusion criteria.
Both Dijkman et al. and Liu et al. suggested that the SUV of tumors with a common lineage were more consistent than those with separate lineage [41,43]. Moreover, they reported that related indicators (e.g., ΔSUV% and SUV max ratio) were significantly higher in the sMPLC group than in the IPM group. Inconsistent with these previous findings, we found that ΔSU V max /D max was numerically, but not significantly, larger in the IPM group than in the sMPLC group. Nevertheless, ΔK i /D max was significantly larger in the IPM group than in the sMPLC group.
14 Molecular Imaging In our study, there was a good correlation between Δ SUV max /D max and ΔK i /D max in the IPM group, but not in the sMPLC group, which could be attributed to the identical and separate clone origin, respectively. The moderately high accuracy of ΔK i /D max indicated by the AUC provided further evidence for the application of dynamic 18 F-FDG PET as an auxiliary modality for discriminating sMPLC from IPM. We considered the size and SUV max of the tumor as indexes since related studies have reported that these indexes can predict postoperative outcomes in patients with sMPLC [44]. Previous studies have reported adenocarcinoma as the major histopathological type in patients with sMPLC [45]. Moreover, the histologic type of initial and second primary lung cancers has been reported to be mostly similar [2,35], which is consistent with our findings. Jiang et al. reported no significant OS differences between unilateral and bilateral MPLC [31]. Contrastingly, Trousse et al. reported that patients with bilateral MPLC had a better outcome than those with unilateral MPLC. Further, in patients with unilateral MPLC, lesions within the same lobe have been associated with better survival compared to those in different lobes [32]. Consequently, we assessed differences between unilateral and bilateral sMPLC. However, there was no significant between-group difference in ΔSU V max /D max and ΔK i /D max . But the bilateral sMPLC group had a slightly higher ΔK i /D max which may indicate a later disease stage.
>According to the 8 th Edition Lung Cancer Stage Classification [46], a separate T, N, M category should be designated to each tumor for second primary lung cancer. However, for patients with IPM, tumor nodules located in the same and different lobes of the unilateral lung and in the bilateral lung are staged as T3, T4, and M1a, respectively. As a result, unilateral IPM and bilateral IPM have different T and M categories, which is indicative of different stages. Regarding the site of the separate tumor nodule relative to the primary tumor in clinically staged patients, tumors in the same lobe are associated with a superior OS than those in different ipsilateral lobes. Moreover, the tumors with different contralateral lobes were associated with the worst OS [33]. Consequently, we determined differences in the indicators between unilateral IPM and bilateral IPM. Although there was no significant difference, the bilateral IPM group had a slightly higher ΔSUV max /D max than the unilateral IPM group, which was indicative of a later disease stage. Size plays a significant role in defining the T category as shown by the proposed size cut-points of the 8 th Edition Lung Cancer Stage Classification. Survival analysis with 1 cm increments in tumor size showed that survival progressively decreased for each 1 cm cut-point [34]. The 3 cm, 5 cm, and 7 cm cutpoints significantly separate T1, T2, T3, and T4, respectively, when only the primary tumor size is taken into consideration [46]. We did not observe a significant difference in ΔSUV max /D max and ΔK i /D max among the three IPM subgroups (≤3 cm vs. 3-5 cm vs. >5 cm) based on the primary tumor size.
We observed that ΔK i /D max was a sensitive indicator, which could be explained by three main differences between SUV and Patlak K i . First, SUV measures phosphorylated and unphosphorylated FDG in the tumor while Patlak K i only considers phosphorylated FDG [47]. Second, SUV is strongly dependent on uptake time and changes in plasma FDG clearance while Patlak K i is independent of both [27,47]. Third, Patlak K i employs the integral under the plasma input function for normalization while SUV approximates this integral using the injected dose divided by the body weight [47]. Consequently, SUV and Patlak K i measure different quantities with the later providing more robust dynamic information and more accurate 18 F-FDG metabolism quantification. FDG metabolism quantification via Patlak analysis has several advantages over using SUV since it involves linear modeling without noise amplification and is independent of the uptake time and changes in plasma FDG clearance [27,28]. However, for calculating fitted K i by dynamic mathematical model, dynamic scan (tissue TAC) and arterial blood sampling (arterial blood TAC, but substitutable with IDIF) are required, which is unsuitable in clinical settings. The reason why it is difficult to perform dynamic PET/CT acquisition in the current clinical practice mainly include the following. (1) Long acquisition time and tedious acquisition process may not be practical for clinical centers with a large number of patients to be examined because of low working efficiency [13]. (2) Many patients, particularly those with advanced cancers, could not be able to tolerate too long acquisition time because the patient is restricted to movement in the scanner and will feel uncomfortable [13]. (3) The conventional and common PET scanner only has the limited axial field of view of about 15-25 cm, confining dynamic PET to a single bed position or one FOV [48]. On the other hand, static analysis only requires a 10-15 min static scan at a fixed time, which is usually 45-60 min after injection. In daily clinic, a single frame static image may provide enough desired information within acceptable error limits in most instances [13]. The selection of ideal analytical method should consider an optimal trade-off between quantitative accuracy and clinical convenience [13,28]. Therefore, SUV is recommended when prioritizing clinical applicability and simplicity. Nevertheless, with the gradual development of the total-body PET scanner equipped with a long axial field of view of 194 cm, short-time dynamic total-body PET imaging with favorable spatial and temporal resolution and signal-to-noise ratio will make dynamic PET imaging more acceptable and popular in clinical application [48,49].
This study has several limitations. Firstly, we employed a small sample size (43 cases); therefore, the results are preliminary and exploratory. Second, most of the cases with metastatic disease were diagnosed without histopathological confirmation based on typical clinical and radiological features, although the primary tumor was pathologically confirmed. Third, partial tumors were <1 cm with low SUV, which could have resulted in SUV biases due to the partial volume effect and statistical noise [50]. Fourth, for the optimal cut-off value of △K i /D max = 0:0059 with a sensitivity of 79% and specificity of 75%, the corresponding false-positive and false-negative rate was 25% and 21% in this study. For this study, the diagnostic ability of a single indicator (△SUV max /D max or △K i /D max ) was not particularly high, but △K i /D max performed better. Meanwhile, further external validation is needed.

Conclusions
Our findings indicated that dynamic 18 F-FDG PET/CT could be a useful tool for distinguishing sMPLC from IPM complemented by histopathologic, clinical, and genetic evaluation, especially during preoperative assessment, which is mainly dependent on clinical and imaging characteristics. K i calculation based on Patlak graphic analysis could be more sensitive for distinguishing metastatic disease from sMPLC in patients with multiple lung cancer nodules. There is a need for further studies to confirm the consistency of our findings.

Data Availability
The data presented in this study are available on reasonable request from the corresponding author. The data are not publicly available due to privacy or ethical issues.

Ethical Approval
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of the Fifth Affiliated Hospital of Sun Yatsen University (IRB protocol number ZDWY.FZYX.002).

Consent
Informed consent was obtained from all subjects involved in the study. Figure S1: there was no significant difference in △SUV max /D max and △K i /D max between the unilateral and bilateral sMPLC group (a, b), as well as the IPM group (c, d). Figure S2: IPM subgroup comparisons based on primary tumor size (≤3 cm vs. 3-5 cm vs. >5 cm). Table S1: individualized diagnosis and treatment gene detection of the sMPLC group. Table S2: individualized diagnosis and treatment gene detection of the IPM group. Table S3: comparison of △SUV max /D max and △K i /D max between different age groups.