Correlation of solid proportion and lymph node metastasis in lung cancers ≤30 mm in diameter

Introduction

Lung cancer is the leading cause of cancer-related death worldwide (1). Lymph node status is an important factor in lung cancer staging and prognosis, exerting a substantial influence in treatment decisions (2,3). Patients with N0 status may not require lymph node dissection or only need sentinel lymph node exploration to improve their quality of life (4,5). However, excessive lymph node dissection leads to prolonged air leakage time, excessive thoracic tube drainage, and weakened immune function (6). For patients with resectable N1 or N2 lung cancer, if surgical treatment is required, systematic lymph node dissection should be performed (7). Other comprehensive treatments (e.g., chemotherapy, radiotherapy, targeted therapy, and immunization) should be performed if resection is not feasible (8,9). Therefore, accurate preoperative assessment of lymph node metastasis (LNM) is essential for the appropriate treatment of patients with lung cancer.

Previous studies using logistic regression analysis have identified factors predictive of LNM in lung cancer, including solid components [i.e., solid proportion (SP)], serum carcinoembryonic antigen level, tumor size, histological subtype (micropapillary or solid), and maximum standardized threshold (10-13). However, some of these factors are determined via postoperative pathology assessment. In recent years, artificial intelligence (AI) technologies have been used to build predictive models for LNM in lung cancer. Some studies have shown that the prediction efficacy of AI model is better than that of conventional linear regression models (14,15), whereas other AI models demonstrate low predictive efficacy (<0.80) (15,16). Overall, research in this area has the following three limitations: (I) insufficient refinement of SP in pulmonary nodules; (II) reliance on postoperative results for determining indicators, which hinders preoperative prediction; and (III) small sample sizes in AI models, highlighting the need for increased prediction efficacy.

To address these limitations, our study categorized pulmonary nodules into six subgroups according to SP according to a step increment of 25%, and the frequency and distribution of LNM were analyzed in each subgroup. Additionally, computed tomography (CT) images in combination with clinical data were used to develop predictive models for LNM and predict model efficacy, with the goal of improving preoperative lymph node status assessment. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1077/rc).

Methods

Study population

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the ethics committees of Taizhou Municipal Hospital (approval No. LWYJ2024194; center 1), Tongde Hospital of Zhejiang Province (approval No. MR-33-24-041194; center 2), Anqing Medical Center of Anhui Medical University (approval No. 83230471; center 3), and The First Affiliated Hospital of Bengbu Medical College (approval No. 2023-440; center 4). The requirement for written informed consent was waived due to the retrospective observational nature of the research.

We initially analyzed the data of 4,369 patients with surgically confirmed lung cancer and lymph node resection who underwent treatment from January 2014 to September 2023 at one of the four participating centers. A total of 2,074 patients were enrolled according to the following inclusion criteria: (I) radical surgery and lymph node dissection, with lung cancer confirmed via postoperative pathology; (II) plain chest CT scan completed within 1 month before the operation; (III) quality of CT image sufficient for analysis; and (IV) CT manifestations of ground-glass nodules (GGNs) and solid nodules (SNs) with a diameter of less than 3 cm. Meanwhile, the exclusion criteria were as follows: (I) lung cancer confirmed by needle biopsy, (II) preoperative radiotherapy or chemotherapy, (III) CT scan thickness greater than 2 mm, and (IV) no or incomplete clinical and pathological data. Figure 1 illustrates the inclusion and exclusion criteria, pathological types of lung cancer, metastatic lymph node distribution, and number of lymph nodes in this study.

Figure 1 Flowchart of patient selection, pathological types, and distribution of metastatic lymph node zones and numbers. AAH, atypical adenomatous hyperplasia; AIS, adenocarcinoma in situ; CT, computed tomography; IAC, invasive adenocarcinoma; LN+, lymph node positive; LN−, lymph node negative; MAC, mucinous adenocarcinoma; MIA, minimally invasive adenocarcinoma; SCC, squamous cell carcinoma.

CT inspection methods

All patients underwent routine CT examination with one of the following CT scanners: SOMSYOM Emotion 16 (Siemens Healthineers, Erlangen, Germany), Definition Flash (Siemens Healthineers), Force CT (Siemens), Optima CT540 or 680 (GE HealthCare, Chicago, IL, USA), or Light Speed 16 or 64 (GE HealthCare). The CT scanning parameters were as follows: tube voltage, 100–120 kV; tube current, 200–280 mA; pitch, 0.8–1.0; collimator, 0.6–0.65; and matrix, 512×512. The scanning stratification and reconstruction thicknesses were 5–10 and 1–2 mm, respectively; the reconstruction was conducted with standard or bone algorithms.

Clinical and radiological analysis

The following clinical data were extracted from medical records: sex, age, smoking history, surgical history, and disease statuses (high blood pressure, diabetes mellitus, cardiovascular diseases, tumor indicators, emphysema/bulla, heterogeneous ventilation or perfusion, interstitial lung disease, bronchiectasis, and multiple lung comorbidities). Hypertension was considered present if the systolic blood pressure was ≥140 mmHg or the diastolic blood pressure was ≥90 mmHg as measured at rest for 3 days on different days without the use of antihypertensive drugs (17). The CT images were independently evaluated by two senior radiologist with 12 and 18 years of experience in chest imaging, respectively. A consistency analysis was conducted between the two observers, with an intraclass correlation coefficient >80% indicating good consistency. For inconsistent results, the two radiologists reached an agreement after discussion. The following CT features were analyzed: SP, distribution, morphology, lobulation, spiculation, airspace, air bronchogram, pleural tag, calcification, rim sign, satellite nodules, halo sign, mixed signs, long diameter (LD), short diameter (SD), CT maximum density value (CTmax), CT minimum density value (CTmin), CT mean density value (CTmean), and CT standard deviation of density value (CTsd). Nodules were categorized into six subgroups according to their internal SP: SP₀ (SP =0), SP₁ (0< SP ≤25%), SP₂ (25%< SP ≤50%), SP₃ (50%< SP ≤75%), SP₄ (75%< SP <100%), and SP₅ (SP =100%). Additionally, the LD and SD of each lesion were measured on axial images. We used RadiAnt DICOM Viewer (64-bit) software (Medixant, Poznan, Poland) on plain scan CT images to delineate the region of interest (ROI) and measure CT values, avoiding the blood vessels, calcification, cystic, and tumor margin areas within the lesion. We measured the area with the highest density on the maximum cross-section of the lesion and recorded the CTmax and CTmin. CTmean, and CTsd. Pleural tags were categorized into four types: 0, no contact; I, direct contact; II, pleural indentation; and III, a mixture of types I + II. SP was defined as the SP diameter divided by the maximum tumor diameter. Heterogeneous ventilation or perfusion was defined as uneven lung density presenting as a mosaic-like pattern (Figure S1A). The rim sign was defined as a lesion edge density ≥3-fold greater than its internal density (Figure S1B). Satellite nodules were defined as those within 3 cm of the main nodules but not crossing lobar boundaries.

Model construction

In this study, all patients’ nodules were classified as GGNs and SNs based on different SPs. All patients were further divided into three groups (GGN group, SN group, and GGN + SN group). Univariate and multivariate logistic regression analyses of the clinical data and radiological signs were conducted. The independent risk factors for LNM of lung cancer in each group were screened out, and prediction models (model 1, model 2, and model 3) were subsequently constructed. The use of predictive models is an accessible, effective, and noninvasive preoperative approach for predicting LNM and can inform the selection of appropriate treatment methods based on the assessed risk level.

Statistical analysis

Data were analyzed with SPSS 25.0 statistical software (IBM Corp., Armonk, NY, USA). Continuous variables are expressed as the mean ± standard deviation. Independent samples t-tests were used to compare normally distributed continuous variables, whereas the Mann-Whitney test was used to compare nonnormally distributed continuous variables. Categorical variables are expressed as frequencies and were compared with the Chi-squared test. Binary logistic regression and receiver operating characteristic curve analyses were performed on statistically significant variables to determine the area under the curve (AUC), accuracy, sensitivity, and specificity of each model. The Delong test was used to compare modal efficacies, with a significance threshold of α<0.05. Histograms were used to represent lung cancer subtypes, differences in SP, and lymph node interval numbers.

Results

General clinical data and CT features

The general clinical data and CT features of the study population are presented in Tables 1,2, respectively. The cohort comprised 2,074 patients (807 men and 1,267 women) with a mean age of 59.22 years (range, 17–86 years). In this study, there were 264 cases (12.7%) of minimally invasive adenocarcinoma (MIA), 1,688 cases (81.4%) of invasive adenocarcinoma (IAC), 22 cases (1.06%) of squamous cell carcinoma (SCC), 55 cases (2.65%) of mucinous adenocarcinoma (MAC), and 45 cases (2.17%) of other types. There were 1,656 patients (78.85%) in the GGN group and 418 patients (21.15%) in the SN group. LNM occurred in 6.65% (138/2,074) of the patients, accounting for 1.78% (37/2,074) of patients in the GGN group and 4.87% (101/2,074) in the SN group. Across the six SP subgroups (SP_0-5), the incidence of LNM increased from 0% to 24.2%, as shown in Figure 2. In the GGN group, the probability of LNM gradually increased from group SP₀ to group SP₄, and there were statistically significant pairwise differences between group SP₃ and each of the SP_0-2 groups. These findings suggested that SP >50% can serve as a threshold for predicting lymph node positivity, as depicted in Figure 3. The numbers and distributions of metastatic lymph nodes in the different SP groups are shown in Figure 4. Interestingly, in the SP₃ group, the number of lymph nodes that metastasized to the fifth and sixth station of mediastinal lymph nodes was the smallest, and the regional differences compared with other stations were statistically significant. Lung cancers with similar SPs can present without LNM, and such cancers are difficult to clinically distinguish solely on the basis of CT imaging findings, as indicated in Figure 5.

Figure 2 Distribution of lymph node status in patients with lung cancer in groups SP_0-5. LN+, lymph node positive; LN−, lymph node negative; SP, solid proportion.

Figure 3 Comparison of lymph node status differences between groups SP_0-4. The differences between group SP₃ and each group from SP₀ to SP₂ supported the selection of SP₃ as a critical indicator. **, P value is less than 0.01; ***, P value is less than 0.001. LN+, lymph node positive; LN−, lymph node negative; SP, solid proportion.

Figure 4 Comparisons of LNM distribution and lymph node number across and within groups. *, P value is less than 0.05; **, P value is less than 0.01; ***, P value is less than 0.001. LN+, lymph node positive; LNM, lymph node metastasis; SP, solid proportion.

Figure 5 Illustration of difficulty in distinguishing LNM in lung cancer patients with similar SPs through use of CT images alone. (A,C) Wrong predictions. (B,D) Correct predictions. The reason for the prediction error in (A) may be due to LNM (SP ≤50%) being the only case, which was relatively rare. The reason for the prediction error in (C) might be due to the real component being less than that in (D). The arrows indicate lung cancer with similar SPs. CT, computed tomography; LNM, lymph node metastasis; SP, solid proportion.

Predictive factors for LNM

In model 1, multivariate logistic regression analysis showed that SP was the sole independent risk factor [odds ratio (OR) 4.396; 95% confidence interval (CI): 2.964–6.521; P<0.001]. In model 2, multivariate logistic regression analysis revealed the following independent risk factors: SD (OR 1.128; 95% CI: 1.073–1.185; P=0.023), CTmean (OR 0.988; 95% CI: 0.981–0.996; P=0.023), heterogeneous ventilation or perfusion (OR 0.437; 95% CI: 0.202–0.945; P=0.040), and high blood pressure (OR 0.216; 95% CI: 0.063–0.736; P=0.019). In model 3, multivariate logistic regression analysis identified the following independent risk factors: SP (OR 3.012; 95% CI: 2.444–3.712; P<0.001), spiculation (OR 1.464; 95% CI: 1.118–1.918; P=0.006), pleural tag (OR 1.436; 95% CI: 1.167–1.768; P=0.001), age (OR 0.973; 95% CI: 0.952–0.995; P=0.015), and rim sign (OR 0.479; 95% CI: 0.289–0.794; P=0.004).

Comparison of model performance

The AUC, accuracy, sensitivity, and specificity in terms of distinguishing lung cancer-negative and -positive lymph nodes were, respectively, 0.929, 0.850, 97.3%, and 84.7% for model 1; 0.733, 0.735, 61.4%, and 77.3% for model 2; and 0.904, 0.751, 97.8%, and 73.5% for model 3. The receiver operating characteristic curves of the models are presented in Figure 6. Pairwise comparisons of diagnostic efficacy (Table 3) indicated no statistically significant difference between models 1 and 3 (P>0.05); however, model 2 significantly differed from both models 1 and 3 (P<0.05). Meanwhile, the diagnostic performance of model 1 was comparable to that of model 3. Confusion matrices for the models are provided in Figure 7. These matrices show the distribution of the two categories predicted by each model. Among the three prediction models, models 1 (84.7%) and 3 (97.8%) exhibited the highest accuracy in terms of predicting lymph node negativity and positivity, respectively.

Figure 6 AUCs of distinguishing LNM in lung cancer for models 1 to 3 were 0.929, 0.733, and 0.904, respectively. Model 1, GGN group; Model 2, SN group; Model 3, GGN + SN group. AUC, area under the curve; LNM, lymph node metastasis.

Figure 7 Confusion matrices for the three models. LN+, lymph node positive; LN−, lymph node negative.

Discussion

To our knowledge, this is the first study to investigate the relationship between clinical CT imaging features and LNM in lung cancers of ≤30 mm diameter according to SP (six categories). We found that the probability of LNM gradually increased from SP₀ to SP₅. Statistical analysis revealed that SP >50% could serve as the threshold for lymph node positivity. Therefore, we recommend lymph node dissection for patients in groups SP_3-5. Moreover, there was a clear relationship between LNM probability and distribution with a step increment of 25% across the six SP subgroups. The diagnostic efficacies of models 1 and 3 in this study were particularly high (AUC =0.929 and AUC =0.904, respectively). Model 1, which included only SP as an independent risk factor, offers sufficient simplicity for clinical application and exhibited relatively high accuracy in predicting negative LNM (84.7%). Meanwhile, model 3 demonstrated relatively high accuracy in predicting positive LNM (97.8%).

Previous studies have demonstrated a strong relationship between SP in pulmonary nodules and LNM. Research (18-20) indicates that pure ground-glass pulmonary nodules often do not exhibit LNM. The probability of LNM occurrence for nodules with SP >0% but ≤50% is 0.837–6.9%; for nodules with SP >50% but <100%, this probability is 8.9–9.1%. In the studies by Hashizume et al. (21) and Ohde et al. (22), no lymph node involvement was evident for nodules with SP <50%. However, these previous studies are inconsistent in terms of the relationship between SP in lung cancer and LNM probability. In our large-scale study, the overall LNM rate was 6.65% (138/2,074). For nodules with SP ≤50% and those with SP >50%, the LNM rates were 0.149% (1/672) and 12.67% (36/284), respectively. Pure SNs exhibited an LNM rate of 24.16% (101/418). Notably, in nodules with SP >50%, the probability of LNM significantly increased (from 0% and 0.758% in groups SP_1-2 to 10.0% and 15.28% in groups SP_3-4, respectively), which is consistent with previous findings (23). The large sample size and refined grouping strategy are the strengths of our study and may allow for a more robust explanation of the relationships between SP values and both the probability and distribution of LNM, which thereby supports previous reports. The findings suggest that the amount of SP in nodules is closely related to the invasiveness of lung cancer: the greater the SP, the higher the possibility of lymphatic metastasis. Therefore, the probability of LNM is very low in patients with nodules with SP ≤50%, and for such patients, only sentinel lymph node exploration may be needed, not lymph node dissection. The probability of LNM increases for patients with nodules with SP >50%, and lymph node sampling or dissection is recommended for such patients.

Spiculation is widely regarded as a characteristic of malignant pulmonary nodules (24,25), and its presence is associated with vascular invasion, pathologic aggressiveness (26), and histological subtypes associated with poor prognosis (27). Although spiculation was not an independent risk factor for LNM in models 1 and 2 in our study, its predictive potential with respect to LNM in lung cancer should not be ignored. We further found that pleural tags constituted an independent risk factor in model 3, with types II and III pleural tags being associated with a higher probability of LNM. Pleural indentation (type II) has been reported to be a strong predictor of visceral pleural invasion in patients with lung cancer (28,29). When visceral pleural invasion occurs, tumor cells can easily metastasize to lymph nodes via lymphatic drainage because numerous lymphatic vessels within the visceral pleura drain into the mediastinum (30). This phenomenon may explain why pleural indentation is predictive of hilar and mediastinal LNM in patients with lung cancer. Moreover, the rim sign (OR =0.518) was identified as a predictor of lymph node negativity in our study. The formation of a rim sign is analogous to the development of a “shell” around the nodules, which limits outward tumor infiltration and growth, reducing the likelihood of LNM. To our knowledge, rim sign in pulmonary nodules has not been reported elsewhere. Previous studies have shown that LNM probability in lung cancer increases according to tumor size (11,19). Our results corroborate these findings, as the LD and SD of pulmonary nodules were associated with LNM. Multivariate analysis confirmed that nodule SD in model 2 was significantly associated with LNM in lung cancer. Univariate logistic regression revealed significant differences in factors such as lobulation, LD, CTmin, and CTsd, between lymph node-negative and -positive patients; however, these differences did not remain statistically significant in the multivariate logistic regression analysis, possibly because the multivariate analysis approach eliminated the effects of confounding factors, allowing identification of more important predictors of LNM.

In this study, we developed predictive models for LNM in lung cancer through refined analyses of preoperative clinical and CT image indices. Specifically models 1 and 3 demonstrated high diagnostic efficacy (AUC =0.929 and AUC =0.904, respectively), superior to that of conventional models (AUC =0.701–0.804) (11,12) and AI models (AUC =0.754–0.899) (14-16). In this study, the efficacy of model 2 was lower than that of models 1 and 3. Although four independent risk factors were screened out in model 2, the overall OR value was not particularly high, which explains the relatively low model efficacy.

This study involved several limitations that should be acknowledged. First, this study did not examine pulmonary glandular prodromal lesions (AIS and atypical adenomatous hyperplasia), which would narrow the scope of the findings’ application. Second, we only used plain scan data and did not analyze the enhanced data. In future work, the model will be constructed through the enhanced data, with the expectation of higher model performance. Third, the multicenter nature of this study, involving different CT scanners and different scanner models, made it challenging to standardize scanning parameters, which might have impacted the CT evaluation results. Finally, the participants included were individuals with lung cancer generally, and there was no clear classification of lung cancer types. We look forward to building models for different types of lung cancer in the future.

Conclusions

Our proposed predictive models (models 1 and 3) offer a noninvasive preoperative approach for efficiently predicting LNM in lung cancers ≤30 mm in diameter with varying SPs (six categories). Models 1 and 3 demonstrated robust efficacy in LNM prediction. Model 1 had a single index and exhibited the highest accuracy in predicting LNM negativity, while model 3 showed the highest accuracy in predicting LNM positivity. These complementary models can effectively assess LNM risk in patients with lung cancer, providing valuable reference information for clinical diagnosis and treatment planning.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1077/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1077/dss

Funding: This study was supported by the Medical Health Science and Technology Project of Zhejiang Province (grant Nos. 2022KY702 to J.W. and 2022KY439 to C.H.), and Taizhou Science and Technology Plan Project (grant Nos. 21ywa35 to C.H. and 24ywb81 to D.Y.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1077/coif). D.Y. reports that this study was supported by Taizhou Science and Technology Plan Project (grant No. 24ywb81). C.H. reports that this study was supported by the Medical Health Science and Technology Project of Zhejiang Province (grant No. 2022KY439) and Taizhou Science and Technology Plan Project (grant No. 21ywa35). J.W. reports that this study was supported by the Medical Health Science and Technology Project of Zhejiang Province (grant No. 2022KY702). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committees of Taizhou Municipal Hospital (No. LWYJ2024194), Tongde Hospital of Zhejiang Province (No. MR-33-24-041194), Anqing Medical Center of Anhui Medical University (No. 83230471), and The First Affiliated Hospital of Bengbu Medical College (No. 2023-440). The requirement for individual consent was waived due to the retrospective nature of the analysis.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer statistics, 2023. CA Cancer J Clin 2023;73:17-48.
2. Huang J, Osarogiagbon RU, Giroux DJ, Nishimura KK, Bille A, Cardillo G, et al. The International Association for the Study of Lung Cancer Staging Project for Lung Cancer: Proposals for the Revision of the N Descriptors in the Forthcoming Ninth Edition of the TNM Classification for Lung Cancer. J Thorac Oncol 2024;19:766-85.
3. Krantz SB, Lutfi W, Kuchta K, Wang CH, Kim KW, Howington JA. Improved Lymph Node Staging in Early-Stage Lung Cancer in the National Cancer Database. Ann Thorac Surg 2017;104:1805-14.
4. Nakamura K, Saji H, Nakajima R, Okada M, Asamura H, Shibata T, Nakamura S, Tada H, Tsuboi M. A phase III randomized trial of lobectomy versus limited resection for small-sized peripheral non-small cell lung cancer (JCOG0802/WJOG4607L). Jpn J Clin Oncol 2010;40:271-4.
5. Vansteenkiste J, De Ruysscher D, Eberhardt WE, Lim E, Senan S, Felip E, Peters SESMO Guidelines Working Group. Early and locally advanced non-small-cell lung cancer (NSCLC): ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol 2017;24:iv89-98.
6. Allen MS, Darling GE, Pechet TT, Mitchell JD, Herndon JE, Landreneau RJ, Inculet RI, Jones DR, Meyers BF, Harpole DH, Putnam JB Jr, Rusch VW. ACOSOG Z0030 Study Group. Morbidity and mortality of major pulmonary resections in patients with early-stage lung cancer: initial results of the randomized, prospective ACOSOG Z0030 trial. Ann Thorac Surg 2006;81:1013-20.
7. Kalata S, Lam GT, Alnajjar RM, Clark MJ, He C, Welsh R J, Chang AC, Lagisetty KH. Adequate lung cancer surgery lymphadenectomy within a statewide quality collaborative: Quality improvement in action. J Thorac Cardiovasc Surg 2023;166:241-50.e3.
8. Pisters K, Kris MG, Gaspar LE, Ismaila N. Adjuvant Systemic Therapy and Adjuvant Radiation Therapy for Stage I to IIIA NSCLC Guideline Expert Panel. Adjuvant Systemic Therapy and Adjuvant Radiation Therapy for Stage I-IIIA Completely Resected Non-Small-Cell Lung Cancer: ASCO Guideline Rapid Recommendation Update. J Clin Oncol 2022;40:1127-9.
9. Kris MG, Gaspar LE, Chaft JE, Kennedy EB, Azzoli CG, Ellis PM, Lin SH, Pass HI, Seth R, Shepherd FA, Spigel DR, Strawn JR, Ung YC, Weyant M. Adjuvant Systemic Therapy and Adjuvant Radiation Therapy for Stage I to IIIA Completely Resected Non-Small-Cell Lung Cancers: American Society of Clinical Oncology/Cancer Care Ontario Clinical Practice Guideline Update. J Clin Oncol 2017;35:2960-74.
10. Ye B, Cheng M, Li W, Ge XX, Geng JF, Feng J, Yang Y, Hu DZ. Predictive factors for lymph node metastasis in clinical stage IA lung adenocarcinoma. Ann Thorac Surg 2014;98:217-23.
11. Seok Y, Yang HC, Kim TJ, Lee KW, Kim K, Jheon S, Cho S. Frequency of lymph node metastasis according to the size of tumors in resected pulmonary adenocarcinoma with a size of 30 mm or smaller. J Thorac Oncol 2014;9:818-24.
12. Aokage K, Suzuki K, Wakabayashi M, Mizutani T, Hattori A, Fukuda H, Watanabe SI. Predicting pathological lymph node status in clinical stage IA peripheral lung adenocarcinoma. Eur J Cardiothorac Surg 2021;60:64-71.
13. Hung JJ, Yeh YC, Jeng WJ, Wu YC, Chou TY, Hsu WH. Factors predicting occult lymph node metastasis in completely resected lung adenocarcinoma of 3 cm or smaller. Eur J Cardiothorac Surg 2016;50:329-36.
14. Wu Y, Liu J, Han C, Liu X, Chong Y, Wang Z, Gong L, Zhang J, Gao X, Guo C, Liang N, Li S. Preoperative Prediction of Lymph Node Metastasis in Patients With Early-T-Stage Non-small Cell Lung Cancer by Machine Learning Algorithms. Front Oncol 2020;10:743.
15. Shimada Y, Kudo YJ, Maehara S, Shimada Y, Kudo Y, Maehara S, Fukuta K, Masuno R, Park J, Ikeda N. Artificial intelligence-based radiomics for the prediction of nodal metastasis in early-stage lung cancer. Sci Rep 2023;13:1028.
16. Tian WW, Yan QQ, Huang XY, Feng R, Shan F, Geng DY, Zhang ZY. Predicting occult lymph node metastasis in solid-predominantly invasive lung adenocarcinoma across multiple centers using radiomics-deep learning fusion model. Cancer Imaging 2024;24:8.
17. Liu LSWriting Group of 2010 Chinese Guidelines for the Management of Hypertension. Zhonghua Xin Xue Guan Bing Za Zhi 2011;39:579-615.
18. Zha JY, Xie D, Xie HK, Zhan LP, Zhou FY, Ying PQ, Zhao LL, Jiang GN, Fei K, Chen C. Recognition of “aggressive” behavior in “indolent” ground glass opacity and mixed density lesions. J Thorac Dis 2016;8:1460-8.
19. Ding N, Mao Y, Gao S, Xue Q, Wang D, Zhao J, Gao Y, Huang J, Shao K, Feng F, Zhao Y, Yuan L. Predictors of lymph node metastasis and possible selective lymph node dissection in clinical stage IA non-small cell lung cancer. J Thorac Dis 2018;10:4061-8.
20. Tsutani Y, Miyata Y, Nakayama H, Okumura S, Adachi S, Yoshimura M, Okada M. Appropriate sublobar resection choice for ground glass opacity-dominant clinical stage IA lung adenocarcinoma: wedge resection or segmentectomy. Chest 2014;145:66-71.
21. Hashizume T, Yamada K, Okamoto N, Saito H, Oshita F, Kato Y, Ito H, Nakayama H, Kameda Y, Noda K. Prognostic significance of thin-section CT scan findings in small-sized lung adenocarcinoma. Chest 2008;133:441-7.
22. Ohde Y, Nagai K, Yoshida J, Nishimura M, Takahashi K, Suzuki K, Takamochi K, Yokose T, Nishiwaki Y. The proportion of consolidation to ground-glass opacity on high resolution CT is a good predictor for distinguishing the population of non-invasive peripheral adenocarcinoma. Lung Cancer 2003;42:303-10.
23. Zhang Y, Fu F, Wen Z, Deng L, Wang S, Li Y, Chen H. Segment Location and Ground Glass Opacity Ratio Reliably Predict Node-Negative Status in Lung Cancer. Ann Thorac Surg 2020;109:1061-8.
24. McWilliams A, Tammemagi MC, Mayo JR, Roberts H, Liu G, Soghrati K, et al. Probability of cancer in pulmonary nodules detected on first screening CT. N Engl J Med 2013;369:910-9.
25. Shi LL, Sheng MH, Wei ZC, Liu L, Zhao JL. CT-Based radiomics predicts the malignancy of pulmonary nodules: a systematic review and meta-analysis. Acad Radiol 2023;30:3064-75.
26. Shimada Y, Yoshida J, Hishida T, Nishimura M, Ishii G, Nagai K. Predictive factors of pathologically proven noninvasive tumor characteristics in T1aN0M0 peripheral non-small cell lung cancer. Chest 2012;141:1003-9.
27. Liang M, Tang W, Tan FW, Zeng H, Guo CY, Feng FY, Wu N. Preoperative prognostic prediction for stage I lung adenocarcinomas: Impact of the computed tomography features associated with the new histological grading system. Front Oncol 2023;13:1103269.
28. Wang Y, Lyu D, Fan L, Liu SY. Research progress in predicting visceral pleural invasion of lung cancer: a narrative review. Transl Cancer Res 2024;13:462-70.
29. Onoda H, Higashi M, Murakami T, Tao H, Yokoyama S, Kunihiro Y, Kawano R, Tanabe M, Tanaka N, Matsumoto T. Correlation between pleural tags on CT and visceral pleural invasion of peripheral lung cancer that does not appear touching the pleural surface. Eur Radiol 2021;31:9022-9.
30. Imai K, Minamiya Y, Saito H, Nakagawa T, Ito M, Ono T, Motoyama S, Sato Y, Konno H, Ogawa J. Detection of pleural lymph flow using indocyanine green fluorescence imaging in non-small cell lung cancer surgery: a preliminary study. Surg Today 2013;43:249-54.

(English Language Editor: J. Gray)