Combining Computed Tomography Radiomics and Clinical Features to Predict Lymph Node Metastasis in Patients with Lung Cancer

Combining Computed Tomography Radiomics and Clinical Features to Predict Lymph Node Metastasis in Patients with Lung Cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Combining Computed Tomography Radiomics and Clinical Features to Predict Lymph Node Metastasis in Patients with Lung Cancer Peiqi Wang, Hao Hu, Bin Yang, Yubo Wang, Yadan Yin, Yang Fu, Bosen Xie, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7815825/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 16 Mar, 2026 Read the published version in BMC Medical Imaging → Version 1 posted 10 You are reading this latest preprint version Abstract Background This study aimed to develop a predictive model for lymph node metastasis in patients with lung cancer using non-contrast computed tomography (CT). Methods A total of 403 patients with lung cancer who met the inclusion criteria were randomly divided into training (n = 282) and test (n = 121) sets. Clinical information was collected, and radiomic features were extracted from non-contrast chest CT images using the “Radiomics” toolkit in 3D Slicer software. Subsequently, least absolute shrinkage and selection operator regression analysis was employed to reduce the number of variables and establish a prediction model for lymph node metastasis in patients with lung cancer based on non-contrast CT scans. The predictive performance and clinical utility of the model were evaluated using the area under the curve (AUC) and decision curve analysis, and Shapley additive explanations analysis was applied to enhance interpretability. Results Lymph node metastasis was present in 35.5% (143/403) of patients. Two clinical features and 16 radiomic features most strongly associated with lymph node metastasis were identified, and nine models were constructed. The receiver operating characteristic curves of the combined clinical–radiomic model demonstrated favorable predictive performance. The clinical–radiomic SVM model demonstrated the best performance in predicting lymph node status (AUC = 0.927 in the training set, 0.852 in the internal test set, and 0.812 in the external test set).Decision curve analysis indicated that the prediction model provided substantial clinical benefits. Conclusion The radiomics model based on non-contrast CT demonstrated good diagnostic performance in predicting lymph node metastasis in patients with lung cancer and may provide guidance for individualized targeted therapy. lung cancer lymph node metastasis computed tomography radiomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Background Lung cancer is a malignant tumor with the highest incidence and mortality rates worldwide. In 2022, there were 2.48 million new cases and 1.817 million deaths worldwide, accounting for 12.4% and 18.7% of all malignant tumor cases and deaths, respectively [ 1 ] . Lung cancer is characterized by uncontrolled cell growth in lung tissue and has a high fatality rate. Without treatment, cancer cells continue to proliferate and spread beyond the lungs [ 2 ] . Currently, lung cancer staging is performed using the tumor-node-metastasis system [ 3 ] , which evaluates the size and location of the primary tumor, presence and location of affected lymph nodes, and presence of distant metastasis. For patients with suspected lung cancer, lymph node metastasis (LNM) is critically important and plays a profound role in guiding postoperative treatment strategies and surgical decision-making [ 4 , 5 ] . Patients with early-stage lung cancer without LNM may undergo wedge resection or lobectomy to preserve as much healthy lung tissue and pulmonary function as possible. This approach can achieve a 5-year survival rate of > 70% [ 6 , 7 ] . Some guidelines recommend multiple treatment modalities for patients with lung cancer and LNM, including preoperative induction chemotherapy with or without radiotherapy, postoperative adjuvant chemotherapy, and postoperative radiotherapy [ 8 , 9 ] . Nevertheless, the 5-year survival rate in these patients is only 26–53% [ 10 ] . The current gold standard for the preoperative diagnosis of LNM is mediastinoscopy and endobronchial ultrasound-guided transbronchial needle aspiration. However, because these procedures are invasive, they are not routinely recommended [ 11 ] . Computed tomography (CT) is the most commonly used noninvasive diagnostic tool for lung cancer, providing information on tumor location, size, and metastatic spread [ 12 ] . Typically, radiologists use a short-axis diameter > 10 mm in mediastinal lymph nodes as the criterion for metastasis [ 13 ] . However, some lymph nodes may enlarge owing to inflammatory changes, limiting diagnostic accuracy [ 14 ] . In recent years, ¹⁸F-FDG positron emission tomography (¹⁸F-FDG PET) has been widely used in the preoperative evaluation of patients with lung cancer [ 15 ] . It provides both metabolic and anatomical information, with a sensitivity of 77% and specificity of 86% for mediastinal lymph node staging [ 16 ] . Nonetheless, some researchers have reported that ¹⁸F-FDG PET still has a high false-positive rate (5%–21%) for detecting malignancy in normal-sized lymph nodes and for excluding malignancy in patients with concurrent inflammatory conditions [ 17 – 19 ] . Moreover, its high cost, limited availability, and additional radiation exposure hinder its widespread use in patients with lung cancer. Therefore, a key scientific question addressed in this study was how to utilize noninvasive methods to identify and screen more comprehensive and effective features of tumor heterogeneity and the microenvironment, thereby enabling accurate prediction of LNM in patients with lung cancer. In recent years, it has become clear that the information obtained from medical images through visual inspection is limited, and significant amounts of valuable data remain hidden [ 20 ] . Artificial intelligence technology can extract rich information from imaging data, autonomously detect relevant lesion characteristics, and make effective automated judgments using extensive training data and algorithmic models [ 21 ] . Various artificial intelligence models have demonstrated promising performance in diagnosing and treating LNM in patients with lung cancer [ 22 – 24 ] . This study aimed to develop a machine learning model that integrates common clinical characteristics and CT-based radiomic features of tumors to predict LNM in patients with lung cancer. The goal was to provide a simpler and more economical method to guide clinicians in delivering individualized and precise treatments, thereby improving patients' quality of life. Methods Study design This study aimed to develop a machine learning model that integrates common clinical characteristics and CT-based radiomic features of tumors to predict LNM in patients with lung cancer. Clinical information was collected, and radiomic features were extracted from non-contrast chest CT images. least absolute shrinkage and selection operator regression analysis was employed to reduce the number of variables and establish a prediction model for lymph node metastasis in patients with lung cancer based on non-contrast CT scans. The predictive performance and clinical utility of the model were evaluated using the area under the curve (AUC) and decision curve analysis, and Shapley additive explanations analysis was applied to enhance interpretability(Fig. 1 ). Participants We retrospectively collected data from 403 patients with lung cancer treated at Panzhihua Traditional Chinese Medicine and Western Medicine Hospital between March 2020 and March 2025. Among them, 143 had mediastinal lymph node metastases, whereas 260 had no lymph node metastases. Among the 403 patients with malignant tumors, the pathological subtypes were as follows: 369 cases of adenocarcinoma, 23 cases of squamous cell carcinoma, seven cases of small cell lung cancer, and three cases not otherwise specified. The inclusion criteria were as follows: ① age > 18 years; ② presence of a single pulmonary nodular lesion; ③ confirmation of lung cancer via paraffin-embedded pathological section; ④ complete case data. The exclusion criteria were as follows: ① concomitant malignant tumors; ② concomitant diffuse lung diseases, such as pulmonary tuberculosis or silicosis; ③ history of preoperative neoadjuvant radiotherapy/chemotherapy or molecular targeted therapy; ④ history of previous malignant tumors. The patients were randomly divided into development (n = 282) and test (n = 121) sets at a ratio of 7:3 (Fig. 2 ). Basic information, laboratory data, and pathological data were collected from the patients’ electronic medical records. Basic information included age, sex, smoking history, and family history. Laboratory parameters included alpha-fetoprotein, carcinoembryonic antigen (CEA), carbohydrate antigen CA125, carbohydrate antigen CA199, neuron-specific enolase, and squamous cell carcinoma-related antigen. Pathological examinations included tumor type and immunohistochemical indicators. In addition, data from 30 patients at Panzhihua Central Hospital were retrospectively collected as an external test set. Analysis of Clinicopathological Factors and Prediction Model Construction Statistical analyses of demographic and medical record data were performed using R software version 4.3.2. Continuity correction chi-squared tests (for categorical variables) and one-way analysis of variance (for normally distributed continuous variables) were used for statistical testing. Factors with a stronger association with LNM status (p < 0.05) were selected for multivariate analysis(Table 1 ). Table 1 Group statistics of clinical data in the test set and internal test set Variables level Train Test p n 282 121 Gender (%) male 157 ( 55.7) 54 ( 44.6) 0.054 female 125 ( 44.3) 67 ( 55.4) Age (median [IQR]) 66.00 [55.00, 74.00] 68.00 [54.00, 74.00] 0.699 Smoke (%) no 180 ( 63.8) 88 ( 72.7) 0.105 yes 102 ( 36.2) 33 ( 27.3) Family_history (%) no 257 ( 91.1) 110 ( 90.9) 1 yes 25 ( 8.9) 11 ( 9.1) T_Stage (%) T1 76 ( 27.0) 34 ( 28.1) 0.314 T2 107 ( 37.9) 48 ( 39.7) T3 42 ( 14.9) 10 ( 8.3) T4 57 ( 20.2) 29 ( 24.0) M_stage (%) M0 169 ( 59.9) 73 ( 60.3) 0.307 M1 113 ( 40.1) 47 ( 38.8) M2 0 ( 0.0) 1 ( 0.8) Clinical_stage (%) IA 59 ( 20.9) 27 ( 22.3) 0.489 IB 112 ( 39.7) 39 ( 32.2) IIA 14 ( 5.0) 8 ( 6.6) IIB 7 ( 2.5) 5 ( 4.1) IIIA 1 ( 0.4) 0 ( 0.0) IIIB 0 ( 0.0) 1 ( 0.8) IV 89 ( 31.6) 41 ( 33.9) CEA (median [IQR]) 2.90 [2.10, 4.10] 2.90 [2.30, 4.60] 0.765 Pathological_type (%) adenocarcinoma 258 ( 91.5) 112 ( 92.6) 0.238 squamous cell carcinom 19 ( 6.7) 4 ( 3.3) small cell lung cancer 3 ( 1.1) 4 ( 3.3) other 2 ( 0.7) 1 ( 0.8) CT Image Acquisition and Feature Extraction All patients were placed in the supine position and scanned from the thoracic inlet to the diaphragmatic base. Scanning parameters were as follows: slice thickness: 5.0 mm, tube voltage: 120 kVp, and tube current: 80–300 mAs. All images were displayed using the standard lung (window width: 1200 HU; window level: −600 HU) and mediastinal (window width: 350 HU; window level: 40 HU) window settings. All images were saved on a hard disk in DICOM format and imported into 3D Slicer software (version 5.8.1). Regions of interest were manually delineated by a radiologist with over 5 years of experience. The radiologists were blinded to clinical and pathological information. In total, 851 radiomic features were extracted using the SlicerRadiomics module, including: 1. 14 shape features describing tumor morphology and geometric structure, such as edge detection and contour description; 2. 93 texture features describing image texture information, such as the gray-level co-occurrence matrix, gray-level difference matrix, and gray-level run length matrix; 3. 744 frequency-domain features describing tumor characteristics in the frequency domain, such as Fourier transform coefficients and wavelet transform coefficients. Selection of Radiomic Features Prior to feature selection, all radiomic features were standardized using Z-score normalization. Owing to the large number of radiomic features, three methods were employed for dimensionality reduction to avoid issues, such as model overfitting and multicollinearity. First, variance thresholding was used to remove features with a variance 0.05). Finally, the least absolute shrinkage and selection operator regression algorithm was used to sparsify the feature set, select indicators most relevant to the research objective, and obtain their weights. Key radiomic features were selected (Fig. 3 ). Ultimately, 16 radiomic features were retained (Table 2 ), including: wavelet-LLH_glcm_Imc2, wavelet-HHH_glcm_ClusterProminence, original_shape_Sphericity, wavelet-LLL_glcm_MCC, wavelet-HLH_glcm_Imc2, original_glcm_ClusterShade, original_gldm_SmallDependenceHighGrayLevelEmphasis, original_glszm_ZoneEntropy, wavelet-LLL_gldm_SmallDependenceHighGrayLevelEmphasis, original_firstorder_90Percentile, original_glszm_SizeZoneNonUniformity, wavelet-LHL_firstorder_Maximum, wavelet-LLH_firstorder_Mean, wavelet-LLL_glszm_GrayLevelVariance, and original_firstorder_Minimum. After dimensionality reduction, the radiomics score (RadScore) was calculated for each patient using the following formula. The RadScore, which comprehensively represents radiomic features, was incorporated into model construction. The formula for calculating the RadScore is: RadScore = Intercept + feature₁ × coefficient₁ + … + featureₙ × coefficientₙ where the intercept is the fitted intercept term and coefficient[i] is the feature weight coefficient generated by least absolute shrinkage and selection operator regression. (A) shows the mean squared error of the radiomic features displayed by the Lasso regression analysis in the model. The two vertical lines indicate the lambda values when the number of variables is reduced to the two lowest levels. (B) shows the trajectory of each radiomic feature, with the horizontal axis representing the log lambda for each radiomic feature and the vertical axis representing the coefficient of each radiomic feature. Table 2 LASSO feature selection table Features Coeffieints wavelet-LLH_glcm_Imc2 -0.025579894 wavelet-HHH_glcm_ClusterProminence -0.019328046 original_shape_Sphericity -0.013027585 wavelet-LLL_glcm_MCC -0.010390812 wavelet-HLH_glcm_Imc2 -0.007564503 original_glcm_ClusterShade -0.007167992 original_gldm_SmallDependenceHighGrayLevelEmphasis -0.00049788 original_glszm_ZoneEntropy 0.0011592808893695907 wavelet-LLL_gldm_SmallDependenceHighGrayLevelEmphasis 0.0027629811348519767 original_firstorder_90Percentile 0.009111013539256278 original_glszm_SizeZoneNonUniformity 0.03503650252922063 wavelet-LHL_firstorder_Maximum 0.03734605667166642 wavelet-LLH_firstorder_Mean 0.03901754470939138 wavelet-LLL_glszm_GrayLevelVariance 0.05842791980640918 original_firstorder_Minimum 0.0622995314014254 Model Construction To predict LNM, filtered clinical and radiomic features were used as inputs to construct a radiomics model for the differential diagnosis of LNM. All data were randomly divided into development and internal test sets (split ratio: 7:3). The training set was used for model training, whereas the internal and external test sets were used to assess the generalization ability of the model. Logistic regression, support vector machine (SVM), and random forest (RF) algorithms were employed to construct models for the differential diagnosis of LNM. Receiver operating characteristic curves were plotted, and the area under the curve (AUC) was calculated to evaluate accuracy, precision, sensitivity, and specificity. Construction of the Radiomics Nomogram A radiomics nomogram was constructed by combining RadScores and significant clinicopathological factors in a multivariate logistic regression model. Receiver operating characteristic curves were plotted and AUC values were calculated to evaluate accuracy, precision, sensitivity, and specificity. Additionally, a calibration curve was plotted to assess calibration and discrimination. Decision curve analysis was performed to quantify the net benefit of the nomogram and evaluate its clinical utility. Results Clinical Characteristics The demographic and clinicopathological characteristics of patients in the development and internal test sets are detailed in Table 1 . The training set comprised 282 patients (median age: 66 years; interquartile range: 55–74 years; 157 men). Univariate and multivariate analyses revealed that LNM status was significantly associated with T stage and smoking history. However, no statistically significant differences were observed between patients with and without metastasis in the training set regarding sex, age, family history, metastasis stage, clinical stage, or CEA levels (all p > 0.05, Table 3 ). Table 3 Univariate and Multivariate Analyses Variables Univariate analysis Multivariate analysis OR CI.lower CI.upper P value OR CI.lower CI.upper P value Gender 0.614 0.404 0.926 0.021 0.982 0.568 1.712 0.948 Age 1.027 1.01 1.045 0.002 1.018 0.999 1.037 0.065 Smoke 1.961 1.279 3.01 0.002 1.959 1.115 3.481 0.02 Family_history 0.901 0.423 1.827 0.778 T_stage 1.883 1.545 2.311 < 0.001 1.726 1.365 2.197 < 0.001 M_stage 2.255 1.494 3.42 < 0.001 1.543 0.811 2.897 0.18 Clinical_stage 1.182 1.103 1.268 < 0.001 1.025 0.911 1.154 0.679 CEA 0.994 0.985 0.999 0.092 Pathological_type 0.614 0.404 0.926 0.021 Model Construction and Predictive Performance Three machine learning algorithms—logistic regression, SVM, and RF—were selected to construct prediction models. Models were built using clinical features, radiomic features, and a combination of both (clinical–radiomic features). In the training set, the SVM model with combined clinical–radiomic features demonstrated the best performance (AUC = 0.927). The model based solely on radiomic features also performed well (AUC = 0.921), approaching the performance of the combined model. The model based solely on clinical features exhibited low performance (AUC = 0.7). In the internal test set, the SVM model with combined clinical–radiomic features achieved the best performance (AUC = 0.852), whereas the RF model based solely on radiomic features achieved an AUC of 0.844. In the external test set, the RF model with combined clinical–radiomic features performed best (AUC = 0.833) (Fig. 4 ). The clinical utility of the models was assessed using decision curve analysis curves. Within a reasonable threshold probability range, the net benefit of the combined model was higher than that of the clinical-only and radiomic-only models (Fig. 5 ). A. Development set: The LR model with clinical features alone (AUC=0.693), the LR model with radiomic features alone (AUC=0.806), and the LR model with combined features (AUC=0.833).B. Internal test set: The LR model with clinical features alone (AUC=0.650), the LR model with radiomic features alone (AUC=0.812), and the LR model with combined features (AUC=0.817).C. External test set: The LR model with clinical features alone (AUC=0.644), the LR model with radiomic features alone (AUC=0.767), and the LR model with combined features (AUC=0.794).D. Development set: The SVM model with clinical features alone (AUC=0.725), the SVM model with radiomic features alone (AUC=0.921), and the SVM model with combined features (AUC=0.927).E. Internal test set: The SVM model with clinical features alone (AUC=0.707), the SVM model with radiomic features alone (AUC=0.824), and the SVM model with combined features (AUC=0.852).F. External test set: The SVM model with clinical features alone (AUC=0.687), the SVM model with radiomic features alone (AUC=0.840), and the SVM model with combined features (AUC=0.812).G. Development set: The RF model with clinical features alone (AUC=0.711), the RF model with radiomic features alone (AUC=0.879), and the RF model with combined features (AUC=0.884).H. Internal test set: The RF model with clinical features alone (AUC=0.733), the RF model with radiomic features alone (AUC=0.844), and the RF model with combined features (AUC=0.838).I. External test set: The RF model with clinical features alone (AUC=0.663), the RF model with radiomic features alone (AUC=0.799), and the RF model with combined features (AUC=0.833). DCA curves for the clinical feature model, radiomic feature model, and clinical-radiomic combined feature model. Within the threshold probability range of 0.2–0.8, the combined model yielded a higher net benefit. Model Interpretability Analysis SHapley Additive exPlanations (SHAP) plots were generated to visualize the impact of different features on individual predictions. Tumor stage was identified as the most critical variable, followed by radiomic features reflecting tumor gray-level distribution and shape and smoking history. All features exhibited positive SHAP values, indicating an overall tendency to increase the predicted probability of LNM. This demonstrates the synergistic effect of clinical and radiomic features, providing a basis for understanding model decisions, investigating metastatic mechanisms, and optimizing diagnostic and treatment strategies (Fig. 6 ). A: SHAP feature importance heatmap. The x-axis represents individual instances (samples), and the y-axis lists key features (including T stage, radiomic features, smoking history, etc.). Color represents the SHAP value (red: positive, blue: negative, indicating a promoting or inhibitory effect on LNM prediction, respectively). The curve at the top represents the model output \(f(x)\).B: SHAP summary plot. The y-axis lists key features (including T stage, radiomic texture/shape features, smoking history, etc.). The x-axis represents the impact of the feature value on the model output (SHAP value). Color indicates the feature value (red: high, blue: low). Nomograms A nomogram was constructed to present the model visually (Fig. 7 ). From top to bottom, the left side of the nomogram displays the risk factor variables, total points, and predicted probability of LNM. Points were assigned for each risk factor, and the sum corresponded to the total points, which were then used to estimate the probability of LNM based on the multivariate logistic regression model. Discussion In this study, we retrospectively collected the clinicopathological characteristics of patients and extracted radiomic features from non-contrast chest CT scans to develop nine models for predicting LNM in lung cancer, utilizing clinical, radiomic, and combined features. The clinical–radiomic SVM model demonstrated the best performance in predicting lymph node status (AUC = 0.927 in the training set, 0.852 in the internal test set, and 0.812 in the external test set), whereas the model based solely on clinical features showed low predictive efficacy (AUC = 0.700). This may be attributed to the superior generalization capability of the SVM algorithm in classification tasks, its faster computational speed compared with that of other machine learning algorithms, and its low sample size requirement [ 25 ] . Previous studies have identified strong associations between LNM in lung cancer and factors, such as age, smoking status, tumor size, histology, degree of differentiation, CEA level, vascular invasion, and pleural involvement [ 26 , 27 ] . Ryuichi et al. developed a clinical prediction model based on serum CEA and CYFRA levels, maximum solid tumor diameter, age, and consolidation/tumor ratio, which achieved an AUC of 0.75 [ 28 ] . Hu constructed a RF model using pathology, degree of differentiation, maximum short-axis diameter of lymph nodes, and tumor diameter, yielding an AUC of 0.83 [ 29 ] . Tang developed a nomogram incorporating Cyfra21-1, D-dimer, tumor size, percentage of solid components, and lesion location, which achieved an AUC of 0.904 [ 30 ] . By contrast, the clinical model in our study exhibited low performance (AUC of the SVM model: 0.725 in the training set, 0.707 in the internal test set, and 0.687 in the external test set). Furthermore, no statistically significant differences were observed between patients with and without metastasis in terms of age, family history, histology, degree of differentiation, or CEA levels, which is inconsistent with previous findings. This discrepancy may be because of the relatively small sample size and potential selection bias in our study. Radiomics converts digital images containing tumor pathophysiological information into measurable and quantifiable data, thereby improving the accuracy of imaging diagnosis [ 31 ] . Zhang developed a nomogram based on CT radiomic features to predict occult LNM, achieving an AUC of 0.782 for the training set and 0.813 for the test set [ 32 ] . Geng further utilized deep learning radiomic features extracted via neural network-based automatic segmentation to construct a model for predicting LNM, achieving an AUC of 0.859 [ 33 ] . Compared with the information provided by non-contrast CT, PET-CT offers additional hemodynamic and metabolic parameters, such as maximum standardized uptake value, mean standardized uptake value, metabolic tumor volume, and total lesion glycolysis, which better reflect tumor aggressiveness. Gabriela combined clinicopathological characteristics with radiomic features extracted from ¹⁸F-FDG PET-CT images to develop a gradient boosting machine model, which achieved an AUC of 0.94 [ 34 ] . In comparison with the aforementioned studies, the present study developed a nomogram for predicting LNM in lung cancer based on readily accessible clinical features (including age, smoking history, and tumor markers) combined with radiomic features extracted from preoperative non-contrast CT scans. This model demonstrated superior performance in predicting lymph node status (AUC = 0.927 in the training set, 0.852 in the internal test set, and 0.812 in the external test set), with a specificity of 0.93 and sensitivity of 0.81. Its specificity was notably higher than that of conventional CT (specificity = 0.55, sensitivity = 0.81) [ 16 ] . The predictive performance of our model exceeded that of the model developed by Boris based on contrast-enhanced CT imaging features [ 35 ] . These results indicate that combined clinical–radiomic features can provide more comprehensive information, improve the accuracy of classification models, reduce reliance on single-modality data, and enhance applicability across diverse clinical scenarios. Our study identified T stage as a key factor influencing LNM in lung cancer. Large tumors exhibit a high propensity for metastasis [ 36 ] , and our results similarly showed that tumor stage was a high-impact predictor in the nomogram. Decision curve analysis demonstrated that the nomogram could effectively predict LNM when the threshold probability ranged from 0.2 to 0.8, indicating promising potential for clinical application. The strengths of this study are as follows: Current research on predicting LNM status in lung cancer primarily focuses on clinicopathological data and PET/CT, with limited studies based on non-contrast CT. Compared with contrast-enhanced CT and PET/CT, non-contrast CT involves lower radiation exposure and is noninvasive. Moreover, this study used easily accessible clinical data combined with non-contrast CT features to develop a model for predicting LNM in lung cancer. By leveraging multimodal fusion, the accuracy of the model was improved, offering a more economical and widely applicable method for predicting LNM status in patients with lung cancer. Additionally, external validation was performed using data from another hospital to verify the applicability and generalization capability of the model. This study had several limitations. First, because of its retrospective nature, selection bias may exist. Second, the sample size was relatively small, precluding the use of deep learning algorithms to enhance model performance. Therefore, future multicenter, large-sample prospective studies are warranted to further validate and optimize the model, thereby improving its applicability and generalizability. Conclusion This study integrated CT-based radiomic features and clinical variables and applied multiple machine learning algorithms to construct a predictive model for LNM in lung cancer. Among these models, the SVM model incorporating combined features demonstrated the best performance with high predictive accuracy. These findings provide a novel approach for individualized risk assessment and intervention in patients with lung cancer. Abbreviations AUC Area under the curve CEA Carcinoembryonic antigen CT Computed tomography LNM Lymph node metastasis PET ¹⁸F-FDG positron emission tomography RadScore Radiomics score RF Random forest SHAP SHapley Additive exPlanations SVM Support vector machine. Declarations Ethics approval and consent to participate The research protocol was approved by the Ethics Committee of Panzhihua Traditional Chinese Medicine and Western Medicine Hospital (PCWEC-C-003). Given the retrospective nature of this study, the requirement for informed consent was waived. All methods were carried out in accordance with relevant guidelines and regulations. Our study adhered to the principles of the Declaration of Helsinki. Consent for publication Not Applicable. Competing interests The authors declare no competing interests. Funding This work was supported by the Program for 2023 Kunming Health Science and Technology Talent Training Project (Medical Technology Center) (No.2023-SW-13). National Natural Science Foundation of China (NSFC)(No.82160348), 2024 Senior Health Technology and Medical Discipline Leader of Yunnan Provincial Health Commission(No. D-2024056). Author Contribution PQW and HH contributed equally to the study concept and design, acquisition of subjects and data, analysis and interpretation of data, and preparation of the manuscript and need therefore be granted both first authorship. YBW was involved in data collection, data analysis and interpretation and reviewed the manuscript. LY assisted in data acquisition and reviewed the manuscript. YDY and YF contributed to the study concept and design and reviewed the manuscript. BSX and JGL were involved in the study design, data analysis and interpretation, and manuscript review. MXK contributed to the study design and manuscript review. CYW participated in the study design, data analysis and interpretation, and manuscript preparation. BY was involved in the study design, data collection, data analysis and interpretation, and manuscript preparation. All the authors read and approved the final manuscript. Acknowledgements We thank Panzhihua Traditional Chinese Medicine and Western Medicine Hospital and Panzhihua Central Hospital for providing data records. Data Availability The data that supports the findings of this study are available in the supplementary material of this article. References Freddie B, Mathieu L, Hyuna S, Jacques. F,Rebecca LSl.Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.CA Cancer. J Clin 2024 May-Jun;74(3):229–63. 10.3322/caac.21834 Zhang Y, Li ZJ, Chen YX, Xiao H, Zhou YK, Du SS, et al. Outcomes of Image-Guided Moderately Hypofractionated Radiotherapy for Stage III Non-Small-Cell Lung Cancer. J Oncol. 2021;2021(0):2721261. 10.1155/2021/2721261 . Wanyin L, Carole AR,Andrew GN,Saeed M. The 8(th) lung cancer TNM classification and clinical staging system: review of the changes and clinical implications. Quant Imaging Med Surg. 2018;8(7):709–18. 10.21037/qims.2018.08.02 . Ziv GM. Staging in Non-Small Cell Lung Cancer.Surg Oncol Clin. N Am. 2016;25(3):493–502. 10.1016/j.soc.2016.02.004 . Tian Y, He Y, Li X. Liu XW.Novel nomograms to predict lymph node metastasis and distant metastasis in resected patients with early-stage non-small cell lung cancer. Ann Palliat Med. 2021;10(3):2548–66. 10.21037/apm-20-1756 . Zhang ZR, Feng HX, Zhao H, Hu J, Liu LX, Liu Y, et al. Sublobar resection is associated with better perioperative outcomes in elderly patients with clinical stage I non-small cell lung cancer: a multicenter retrospective cohort study. J Thorac Dis. 2019;11(5):1838–48. 10.21037/jtd.2019.05.20 . Babatunde AYK, Chi-Fu JY, Brian CG, Li XC, Michael SM,Gu L, et al. A national analysis of wedge resection versus stereotactic body radiation therapy for stage IA non-small cell lung cancer. J Thorac Cardiovasc Surg. 2017;154(2):675–e6864. 10.1016/j.jtcvs.2017.02.065 . Megan ED, Navneet S, Nofisat I,Mara BA, Douglas. AA,JeffreyB,et al. Management of Stage III Non-Small-Cell Lung Cancer: ASCO Guideline. J Clin Oncol. 2022;40(12):1356–84. 10.1200/JCO.21.02528 . Postmus PE, Kerr KM, Oudkerk M, Waller SSDA, Vansteenkiste J, et al. Early and locally advanced non-small-cell lung cancer (NSCLC): ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28(0):iv1–21. 10.1093/annonc/mdx222 . Hwayoung S, Sung H Y,Junghoon K, Jihang K, Kyoung WL, Woojoo L, et al. Application of N Descriptors Proposed by the International Association for the Study of Lung Cancer in Clinical Staging. Radiology. 2021;300(2):450–7. 10.1148/radiol.2021204461 . Noeul K, Sun HS, HongseokY,Byung WJ, Kyungjong L, Sang-Won U, et al. Infectious complications of EBUS-TBNA: A nested case-control study using 10-year registry data. Lung Cancer. 2021;161(0):1–8. 10.1016/j.lungcan.2021.08.016 . Annalori P, Paolo S. Lung Cancer and Radiological Imaging. Curr Radiopharm. 2020;13(3):238–42. 10.2174/1874471013666200523161849 . Cong MD, Feng H, Ren JL, Xu Q, Cong LL,Hou ZZ, Wang YY, et al. Development of a predictive radiomics model for lymph node metastases in pre-surgical CT-based stage IA non-small cell lung cancer. Lung Cancer. 2020;139(0):73–9. 10.1016/j.lungcan.2019.11.003 . Fieke H, David JH, Johannes MD, Naomi B, Rob A E M T, Michel WJM, W, et al. Variation Between Multidisciplinary Tumor Boards in Clinical Staging and Treatment Recommendations for Patients With Locally Advanced Non-small Cell Lung Cancer. Chest. 2020;158(6):2675–87. 10.1016/j.chest.2020.07.054 . Michael KG, Ware GK, Chara ER, Courtney CM, Anita ND, Hidenobu S, et al. Test performance of positron emission tomography and computed tomography for mediastinal staging in patients with non-small-cell lung cancer: a meta-analysis. Ann Intern Med. 2003;139(11):879–92. 10.7326/0003-4819-139-11-200311180-00013 . Sobic-Saranovic D, Petrusic I, Artiko V, Pavlovic S, Subotic D, Saranovic D, Nagorni-Obradovic L, Petrovic N, et al. Comparison of 18F-FDG PET/CT and MDCT for staging/restaging of non-small cell lung cancer. Neoplasma. 2015;62(2):295–301. 10.4149/neo_2015_035 . Sang Woo Lee,Seong-Jang Kim. Is Delayed Image of 18F-FDG PET/CT Necessary for Mediastinal Lymph Node Staging in Non-Small Cell Lung Cancer Patients? Clin Nucl Med. 2022;47(5):414–21. 10.1097/RLU.0000000000004110 . Huang M, Li SL, Li NLYM, Zhang SY,Lu FL,Yang Y, et al. Occult mediastinal lymph node metastasis in FDG-PET/CT node-negative lung adenocarcinoma patients: Risk factors and histopathological study. Thorac Cancer. 2019;10(6):1453–60. 10.1111/1759-7714.13093 . Sarah JG, Anthony W K,Jonathan TP, Kyle B, Frank CD, Daniel JB, et al. Indications for invasive mediastinal staging in patients with early non-small cell lung cancer staged with PET-CT. Lung Cancer. 2017;109(0):36–41. 10.1016/j.lungcan.2017.04.018 . Yi Y, Jiang YH,BaiBB,Zhang CF. Emerging role of deep learning-based artificial intelligence in tumor pathology. Cancer Commun (Lond). 2020;40(4):154–66. 10.1002/cac2.12012 . Philippe L, Ralph T H L,Timo MD, Jurgen P, Evelyn EC, van de Jong T et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol 2017;14(12):749–762. 10.1038/nrclinonc.2017.141 Allan FF, A,Sérgio AS, Raul L Jr, Tarcísio A R,Agláia M G X,Erica R. Combining machine learning and texture analysis to differentiate mediastinal lymph nodes in lung cancer patients. Phys Eng Sci Med. 2021;44(2):387–94. 10.1007/s13246-021-00988-2 . Fabian CL, Piotr W, Thorsten AB, Mirjam S, Miriam R, Mathilda W, et al. Transfer-Learning Deep Radiomics and Hand-Crafted Radiomics for Classifying Lymph Nodes from Contrast-Enhanced Computed Tomography in Lung Cancer. Cancers (Basel). 2023;15(10). 10.3390/cancers15102850 . Kensuke K, Hironobu S, Kyoichi O, Toshiteru T, Hyungeun Y, Shinji A. Impact of the number of dissected lymph nodes on machine learning-based prediction of postoperative lung cancer recurrence: a single-hospital retrospective cohort study. BMJ Open Respir Res. 2024;11(1). 10.1136/bmjresp-2023-001926 . Evan R, Brian C, Mousumi B. Classif J Appl Stat. 2019;46(16):2987–3007. 10.1080/02664763.2019.1625876 . Chen B, Xia WJ, Wang ZQ, Zhao H, Li XF, Liu LX, Liu Y, Hu J, et al. Risk analyses of N2 lymph-node metastases in patients with T1 non-small cell lung cancer: a multi-center real-world observational study in China. J Cancer Res Clin Oncol. 2019;145(11):2771–7. 10.1007/s00432-019-03006-x . Shan LY, Zhang L, Zhu XL,Wang ZL,Fang SH,Lin JF, et al. Chinese never smokers with adenocarcinoma of the lung are younger and have fewer lymph node metastases than smokers. Respir Res. 2022;23(1):293. 10.1186/s12931-022-02199-z . Ryuichi Y, Yoshitaka E, Takuya A, Hiroyuki D,Makoto T, Wataru S, et al. Diagnostic artificial intelligence model predicts lymph node status in non-small cell lung cancer using simplified examination. J Thorac Dis. 2024;16(11):7320–8. 10.21037/jtd-24-1067 . Hu SL, Luo M, Li YL. Machine Learning for the Prediction of Lymph Nodes Micrometastasis in Patients with Non-Small Cell Lung Cancer: A Comparative Analysis of Two Practical Prediction Models for Gross Target Volume Delineatio. Cancer Manag Res. 2021;13(0):4811–20. 10.2147/CMAR.S313941 . Tang YBA, Garu,Chen X, Guan ZY,Cai XD,Huang HX, et al. Risk factors for mediastinal lymph node metastases in early-stage non-small-cell lung cancer and prediction model establishment. Am J Cancer Res. 2024;14(12):5952–64. 10.62347/DIZG4944 . Robert JG, Paul EK, Hedvig H. Radiomics: Images Are More than Pictures. They Are Data[J] Radiol. 2016;278(2):563–77. 10.1148/radiol.2015151169 . Epub 2015 Nov 18. Zhang R, Zhang RR,Luan T,Liu BW,Zhang YM, Xu YP, et al. A Radiomics Nomogram for Preoperative Prediction of Clinical Occult Lymph Node Metastasis in cT1-2N0M0 Solid Lung Adenocarcinoma. Cancer Manag Res. 2021;13(0):8157–67. 10.2147/CMAR.S330824 . Geng MF, Geng. MS,WeiR,Chen MW.Artificial intelligence neural network analysis and application of CT imaging features to predict lymph node metastasis in non-small cell lung cancer. J Thorac Dis. 2022;14(11):4384–94. 10.21037/jtd-22-1511 . Kyongmin SB, Bomi G, Sae J, Na,Ji HH, Sang HC, Ho JA, et al. DeepCUBIT: Predicting Lymphovascular Invasion or Pathological Lymph Node Involvement of Clinical T1 Stage Non-Small Cell Lung Cancer on Chest CT Scan Using Deep Cubical Nodule Transfer Learning Algorithm. Front Oncol. 2021;11(0):661244. 10.3389/fonc.2021.661244 . Boris G, Philipp H B,Alexander DJ, B,Alexander H,Julian MM, R,Christian F, et al. Inferring FDG-PET-positivity of lymph node metastases in proven lung cancer from contrast-enhanced CT using radiomics and machine learning. Eur Radiol Exp. 2022;6(1):44. 10.1186/s41747-022-00296-8 . Hu SL, Luo M, Li YLM. Learning for the Prediction of Lymph Nodes Micrometastasis in Patients with Non-Small Cell Lung Cancer: A Comparative Analysis of Two Practical Prediction Models for Gross Target Volume Delineation. Cancer Manag Res. 2021;13(0):4811–20. 10.2147/CMAR.S313941 . Additional Declarations No competing interests reported. Supplementary Files Attachment.xlsx THEEXTERNALTESTSET.xlsx Cite Share Download PDF Status: Published Journal Publication published 16 Mar, 2026 Read the published version in BMC Medical Imaging → Version 1 posted Editorial decision: Revision requested 12 Nov, 2025 Reviews received at journal 11 Nov, 2025 Reviewers agreed at journal 06 Nov, 2025 Reviews received at journal 03 Nov, 2025 Reviewers agreed at journal 22 Oct, 2025 Reviewers invited by journal 20 Oct, 2025 Editor assigned by journal 20 Oct, 2025 Editor invited by journal 17 Oct, 2025 Submission checks completed at journal 17 Oct, 2025 First submitted to journal 17 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7815825","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":537519118,"identity":"c4d4d47d-3f14-45ca-b835-1940a533c5fc","order_by":0,"name":"Peiqi Wang","email":"","orcid":"","institution":"Department of Clinical Medical of Dali University","correspondingAuthor":false,"prefix":"","firstName":"Peiqi","middleName":"","lastName":"Wang","suffix":""},{"id":537519120,"identity":"a06669d5-9acb-424b-bd42-fd7f9cbf1a55","order_by":1,"name":"Hao Hu","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Hu","suffix":""},{"id":537519121,"identity":"55fe81ea-c046-4705-998e-90607664942a","order_by":2,"name":"Bin Yang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIiWNgGAWjYJACZgYDMNXA8MHAxo4ULYwNjDMK0pKJ1MIA0cLM8+EQYwMh5QbHzx5+XVBwR95curHxsY3BAWYG9sNHN+DVciYvzXqGwTPDnXMONhvnGNzhY+BJS7uBV8uBHDNjHoPDjBtuJLZJ5xg8Y2aQ4DHDr+X8G7AWe6CW9t8WQL0NBLXcyDF+DNSSCLIFGHREaJG88caMeYbB4WSQXyR7DNKS2Qj5he98jvHngj+HbbdLNx/88OOPjR0/++FjeLUoHGBgkwC7UAIqwoZPOQjINzAwf0DRMgpGwSgYBaMAHQAAGjVRO/3g9UQAAAAASUVORK5CYII=","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":true,"prefix":"","firstName":"Bin","middleName":"","lastName":"Yang","suffix":""},{"id":537519124,"identity":"8ff55d3c-ec32-4075-9803-522dbc5559e1","order_by":3,"name":"Yubo Wang","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yubo","middleName":"","lastName":"Wang","suffix":""},{"id":537519127,"identity":"b1b058b1-54ea-429f-871b-b2acb5668bd4","order_by":4,"name":"Yadan Yin","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yadan","middleName":"","lastName":"Yin","suffix":""},{"id":537519131,"identity":"2bc69f6e-3826-4222-90aa-352755905f19","order_by":5,"name":"Yang Fu","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Fu","suffix":""},{"id":537519135,"identity":"2c8dad8d-58c4-4ade-9b03-022500491659","order_by":6,"name":"Bosen Xie","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bosen","middleName":"","lastName":"Xie","suffix":""},{"id":537519136,"identity":"9bdfe0b2-11a4-4531-a804-a0785bd41b84","order_by":7,"name":"Jiageng Li","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jiageng","middleName":"","lastName":"Li","suffix":""},{"id":537519137,"identity":"35138012-00cb-46a4-87c2-aa6e12273c47","order_by":8,"name":"Mengxue Kong","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Mengxue","middleName":"","lastName":"Kong","suffix":""},{"id":537519138,"identity":"c6dce0e1-7622-4275-89c6-f2dfd7683eee","order_by":9,"name":"Chunyuan Wei","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Chunyuan","middleName":"","lastName":"Wei","suffix":""},{"id":537519139,"identity":"89e5ae06-b620-40a0-b29a-403f1eb64e4c","order_by":10,"name":"Lei Yue","email":"","orcid":"","institution":"Panzhihua Central Hospital","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Yue","suffix":""}],"badges":[],"createdAt":"2025-10-09 09:38:31","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7815825/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7815825/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12880-026-02262-x","type":"published","date":"2026-03-16T15:59:22+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":94857983,"identity":"74f923c4-a669-4d20-b43c-2d6bb999d90a","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2850083,"visible":true,"origin":"","legend":"","description":"","filename":"CombiningComputedTomographyRadiomicsandClinicalFeaturestoPredictLymphNodeMetastasisinPatientswithLungCancer.docx","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/0eab85c97d6390ad5df8e165.docx"},{"id":94858081,"identity":"09353e29-0a4f-48e7-9e23-98fad9ea1345","added_by":"auto","created_at":"2025-10-31 12:35:11","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11867,"visible":true,"origin":"","legend":"","description":"","filename":"351611635ec54897a8deac4e3ef9e369.json","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/43ac1739a94ec5006f32ee80.json"},{"id":94857985,"identity":"9952ba06-5cf1-4c65-9413-af5088bd756b","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4620379,"visible":true,"origin":"","legend":"","description":"","filename":"Attachment.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/6deeaf2d98ae20ba5fa9a90c.xlsx"},{"id":94857979,"identity":"9f3de7a6-224b-4e71-abd7-a6917a4a49f6","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":379998,"visible":true,"origin":"","legend":"","description":"","filename":"THEEXTERNALTESTSET.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/5f6de750bb64b7d2a891641f.xlsx"},{"id":94857982,"identity":"741e5697-908c-4b5a-b539-2e9afaf7f4b7","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"xml","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":125415,"visible":true,"origin":"","legend":"","description":"","filename":"351611635ec54897a8deac4e3ef9e3691enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/3f703596520352d3d001b962.xml"},{"id":94857992,"identity":"2cb84a88-4034-4921-afa4-41f59b7bb972","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":154628,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/4fb57098050a6a6e6877253f.png"},{"id":94985152,"identity":"9ba7c113-a60d-4858-807b-9f6d90fc3acc","added_by":"auto","created_at":"2025-11-03 06:57:36","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":165338,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/2ee09b2d995b6eb96b04663e.png"},{"id":94857987,"identity":"1a8a9c62-8881-482c-b17d-172f8c688fb2","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":155107,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/d8fbc0d400a60231293d6443.jpeg"},{"id":94858034,"identity":"561f230f-d653-4ec0-ae3b-ecd1fc8e48f5","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170445,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/5313b737f9164df6f461859e.jpeg"},{"id":94985892,"identity":"2eda935f-ee54-453f-a3ae-38713cf9c7dd","added_by":"auto","created_at":"2025-11-03 06:59:12","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":168748,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/aaccfc524c98a13ec703a13c.jpeg"},{"id":94857991,"identity":"a478cd7c-8934-4863-9c05-e3fbaf331454","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"jpeg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":168298,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/0179a31f8aada715f4b9d53d.jpeg"},{"id":94985260,"identity":"7376f87c-846d-47d9-a0ca-3f6bd4e40081","added_by":"auto","created_at":"2025-11-03 06:57:48","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1100685,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/84c2bb2d02777524ade2739f.png"},{"id":94858031,"identity":"250f46fa-35f1-4e04-a854-1a187e5c358d","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"jpeg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":299758,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/46c22254947a79be8348518c.jpeg"},{"id":94858080,"identity":"11a73827-6941-4ca6-8373-2774c6789ea4","added_by":"auto","created_at":"2025-10-31 12:35:00","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":821699,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/5d9567744a4099f8ff1edc38.png"},{"id":94985764,"identity":"3e750fa2-0d11-4494-b43b-2cb5a837da43","added_by":"auto","created_at":"2025-11-03 06:58:51","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30511,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/3e101e680786c4e4d5641bde.png"},{"id":94858036,"identity":"b0f3207e-9f54-4e87-a16c-f966b59f6bda","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35935,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/575f084c036eb2a8d07fd4ea.png"},{"id":94858018,"identity":"92949add-088d-4344-bafd-02286886c09b","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43011,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/49303039e581cd45f2a48501.png"},{"id":94858032,"identity":"c0ebb3eb-ee93-4a84-ad5b-ee5e41791a3e","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51799,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/8ff09bdb2675ea33a5bf876d.png"},{"id":94858037,"identity":"b94fcb78-1870-4d9c-a005-25d2fa84c0d7","added_by":"auto","created_at":"2025-10-31 12:34:50","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52505,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/768b3c51e803bf1a93ba65cb.png"},{"id":94858038,"identity":"e9d29fa1-811d-4e4f-811a-897e5791c842","added_by":"auto","created_at":"2025-10-31 12:34:50","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50664,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/97a5e14d7b415d36c9a31e0b.png"},{"id":94858002,"identity":"5fb5efbe-fecf-4410-a4b6-8ea5b27984ff","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11844,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/6d572063020b2f7ca7733c2d.png"},{"id":94858040,"identity":"da8b14a6-0ef3-42a3-a11d-067323ec9df7","added_by":"auto","created_at":"2025-10-31 12:34:50","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":66934,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/368077db81c9c88ac81b8b64.png"},{"id":94986543,"identity":"97e48ef0-65fe-4d10-9a57-9e972d52c132","added_by":"auto","created_at":"2025-11-03 07:00:24","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8980,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/9aa0656923324e69fd75ba0b.png"},{"id":94858030,"identity":"2c08a527-c07e-4eb9-a826-f76bc8f6a63b","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"xml","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":123199,"visible":true,"origin":"","legend":"","description":"","filename":"351611635ec54897a8deac4e3ef9e3691structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/fb1e4017ec74355b94c24b68.xml"},{"id":94858042,"identity":"01dc9cd3-f498-41f6-9727-041881400130","added_by":"auto","created_at":"2025-10-31 12:34:50","extension":"html","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":132955,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/7da0db4092ebd6fe33b51964.html"},{"id":94857978,"identity":"455d8f2a-3807-47d4-9336-3e53d600ac39","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":187106,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eOverall flowchart\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/979f2e1e4685fc69cf920fa7.png"},{"id":94858043,"identity":"d1b6f804-afb9-40f4-b9ff-55b4312d00f7","added_by":"auto","created_at":"2025-10-31 12:34:50","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":142655,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003ePatient inclusion and exclusion criteria and enrollment flowchart\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/4976a78bfdf7aa2fdcd5a7bb.png"},{"id":94858045,"identity":"77e3b1e1-ce2f-4f59-bc7f-13928f309fd7","added_by":"auto","created_at":"2025-10-31 12:34:52","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":278580,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eRadiomic features selected using the LASSO regression model\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/98b0ddd54dd86e5cd39175c2.png"},{"id":94857989,"identity":"f50b71e7-7fe0-4c0f-83f0-be968510644e","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":201118,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eReceiver operating characteristic (ROC) curves of the models\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/8189fa3c34832ce60cb6f632.png"},{"id":94985804,"identity":"9272d4e0-4428-4be4-af42-163c881a3b9b","added_by":"auto","created_at":"2025-11-03 06:59:02","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":43213,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eDecision curve analysis (DCA)\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/e537159fdb9a70152bee2de5.png"},{"id":94985360,"identity":"9210194a-4988-450d-8f76-a070a3a0b8f6","added_by":"auto","created_at":"2025-11-03 06:58:01","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":365199,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eSHAP plots for predicting lymph node metastasis\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/f3c18e06096e5f2c3365c31d.png"},{"id":94986007,"identity":"f71622f7-dd54-4599-be48-fb6847f73890","added_by":"auto","created_at":"2025-11-03 06:59:33","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":24963,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eNomogram for predicting lymph node metastasis\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/8ae3faa423f3f432e89f2f20.png"},{"id":105223434,"identity":"b0efa08b-b6d6-4fb2-af25-54e224a6b8c1","added_by":"auto","created_at":"2026-03-23 16:06:28","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2308909,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/ddb1dc5d-7ba4-414e-acd7-84b2da0ce923.pdf"},{"id":94857981,"identity":"9ca22213-b896-4089-832e-442a731fa3ce","added_by":"auto","created_at":"2025-10-31 12:34:49","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":4620379,"visible":true,"origin":"","legend":"","description":"","filename":"Attachment.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/93986fde066dcb858f9a1665.xlsx"},{"id":94858044,"identity":"370269a8-602f-44ae-ab55-fbd0751a33d3","added_by":"auto","created_at":"2025-10-31 12:34:52","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":379998,"visible":true,"origin":"","legend":"","description":"","filename":"THEEXTERNALTESTSET.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7815825/v1/01857a7fa524510cc8a0936d.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Combining Computed Tomography Radiomics and Clinical Features to Predict Lymph Node Metastasis in Patients with Lung Cancer","fulltext":[{"header":"Background","content":"\u003cp\u003eLung cancer is a malignant tumor with the highest incidence and mortality rates worldwide. In 2022, there were 2.48\u0026nbsp;million new cases and 1.817\u0026nbsp;million deaths worldwide, accounting for 12.4% and 18.7% of all malignant tumor cases and deaths, respectively \u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Lung cancer is characterized by uncontrolled cell growth in lung tissue and has a high fatality rate. Without treatment, cancer cells continue to proliferate and spread beyond the lungs \u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. Currently, lung cancer staging is performed using the tumor-node-metastasis system \u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e, which evaluates the size and location of the primary tumor, presence and location of affected lymph nodes, and presence of distant metastasis. For patients with suspected lung cancer, lymph node metastasis (LNM) is critically important and plays a profound role in guiding postoperative treatment strategies and surgical decision-making \u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Patients with early-stage lung cancer without LNM may undergo wedge resection or lobectomy to preserve as much healthy lung tissue and pulmonary function as possible. This approach can achieve a 5-year survival rate of \u0026gt;\u0026thinsp;70% \u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. Some guidelines recommend multiple treatment modalities for patients with lung cancer and LNM, including preoperative induction chemotherapy with or without radiotherapy, postoperative adjuvant chemotherapy, and postoperative radiotherapy \u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. Nevertheless, the 5-year survival rate in these patients is only 26\u0026ndash;53% \u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe current gold standard for the preoperative diagnosis of LNM is mediastinoscopy and endobronchial ultrasound-guided transbronchial needle aspiration. However, because these procedures are invasive, they are not routinely recommended \u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. Computed tomography (CT) is the most commonly used noninvasive diagnostic tool for lung cancer, providing information on tumor location, size, and metastatic spread \u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. Typically, radiologists use a short-axis diameter\u0026thinsp;\u0026gt;\u0026thinsp;10 mm in mediastinal lymph nodes as the criterion for metastasis \u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. However, some lymph nodes may enlarge owing to inflammatory changes, limiting diagnostic accuracy \u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. In recent years, \u0026sup1;⁸F-FDG positron emission tomography (\u0026sup1;⁸F-FDG PET) has been widely used in the preoperative evaluation of patients with lung cancer \u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. It provides both metabolic and anatomical information, with a sensitivity of 77% and specificity of 86% for mediastinal lymph node staging \u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. Nonetheless, some researchers have reported that \u0026sup1;⁸F-FDG PET still has a high false-positive rate (5%\u0026ndash;21%) for detecting malignancy in normal-sized lymph nodes and for excluding malignancy in patients with concurrent inflammatory conditions \u003csup\u003e[\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. Moreover, its high cost, limited availability, and additional radiation exposure hinder its widespread use in patients with lung cancer. Therefore, a key scientific question addressed in this study was how to utilize noninvasive methods to identify and screen more comprehensive and effective features of tumor heterogeneity and the microenvironment, thereby enabling accurate prediction of LNM in patients with lung cancer.\u003c/p\u003e\u003cp\u003eIn recent years, it has become clear that the information obtained from medical images through visual inspection is limited, and significant amounts of valuable data remain hidden \u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. Artificial intelligence technology can extract rich information from imaging data, autonomously detect relevant lesion characteristics, and make effective automated judgments using extensive training data and algorithmic models \u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. Various artificial intelligence models have demonstrated promising performance in diagnosing and treating LNM in patients with lung cancer \u003csup\u003e[\u003cspan additionalcitationids=\"CR23\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThis study aimed to develop a machine learning model that integrates common clinical characteristics and CT-based radiomic features of tumors to predict LNM in patients with lung cancer. The goal was to provide a simpler and more economical method to guide clinicians in delivering individualized and precise treatments, thereby improving patients' quality of life.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eStudy design\u003c/h2\u003e\n \u003cp\u003eThis study aimed to develop a machine learning model that integrates common clinical characteristics and CT-based radiomic features of tumors to predict LNM in patients with lung cancer. Clinical information was collected, and radiomic features were extracted from non-contrast chest CT images. least absolute shrinkage and selection operator regression analysis was employed to reduce the number of variables and establish a prediction model for lymph node metastasis in patients with lung cancer based on non-contrast CT scans. The predictive performance and clinical utility of the model were evaluated using the area under the curve (AUC) and decision curve analysis, and Shapley additive explanations analysis was applied to enhance interpretability(Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u0026nbsp;\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eParticipants\u003c/h3\u003e\n\u003cp\u003eWe retrospectively collected data from 403 patients with lung cancer treated at Panzhihua Traditional Chinese Medicine and Western Medicine Hospital between March 2020 and March 2025. Among them, 143 had mediastinal lymph node metastases, whereas 260 had no lymph node metastases. Among the 403 patients with malignant tumors, the pathological subtypes were as follows: 369 cases of adenocarcinoma, 23 cases of squamous cell carcinoma, seven cases of small cell lung cancer, and three cases not otherwise specified. The inclusion criteria were as follows: ① age\u0026thinsp;\u0026gt;\u0026thinsp;18 years; ② presence of a single pulmonary nodular lesion; ③ confirmation of lung cancer via paraffin-embedded pathological section; ④ complete case data. The exclusion criteria were as follows: ① concomitant malignant tumors; ② concomitant diffuse lung diseases, such as pulmonary tuberculosis or silicosis; ③ history of preoperative neoadjuvant radiotherapy/chemotherapy or molecular targeted therapy; ④ history of previous malignant tumors. The patients were randomly divided into development (n\u0026thinsp;=\u0026thinsp;282) and test (n\u0026thinsp;=\u0026thinsp;121) sets at a ratio of 7:3 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). Basic information, laboratory data, and pathological data were collected from the patients\u0026rsquo; electronic medical records. Basic information included age, sex, smoking history, and family history. Laboratory parameters included alpha-fetoprotein, carcinoembryonic antigen (CEA), carbohydrate antigen CA125, carbohydrate antigen CA199, neuron-specific enolase, and squamous cell carcinoma-related antigen. Pathological examinations included tumor type and immunohistochemical indicators. In addition, data from 30 patients at Panzhihua Central Hospital were retrospectively collected as an external test set.\u003c/p\u003e\n\u003ch3\u003eAnalysis of Clinicopathological Factors and Prediction Model Construction\u003c/h3\u003e\n\u003cp\u003eStatistical analyses of demographic and medical record data were performed using R software version 4.3.2. Continuity correction chi-squared tests (for categorical variables) and one-way analysis of variance (for normally distributed continuous variables) were used for statistical testing. Factors with a stronger association with LNM status (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) were selected for multivariate analysis(Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eGroup statistics of clinical data in the test set and internal test set\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003elevel\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTrain\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTest\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003en\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e282\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e121\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e157 ( 55.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e54 ( 44.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003efemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e125 ( 44.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 ( 55.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge (median [IQR])\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e66.00 [55.00, 74.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e68.00 [54.00, 74.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.699\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoke (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eno\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e180 ( 63.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88 ( 72.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.105\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eyes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e102 ( 36.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33 ( 27.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFamily_history (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eno\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e257 ( 91.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e110 ( 90.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eyes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25 ( 8.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11 ( 9.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT_Stage (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e76 ( 27.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e34 ( 28.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.314\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e107 ( 37.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e48 ( 39.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e42 ( 14.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 ( 8.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e57 ( 20.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e29 ( 24.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eM_stage (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eM0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e169 ( 59.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e73 ( 60.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.307\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eM1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e113 ( 40.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e47 ( 38.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eM2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0 ( 0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 ( 0.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eClinical_stage (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e59 ( 20.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e27 ( 22.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.489\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e112 ( 39.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e39 ( 32.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIIA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14 ( 5.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8 ( 6.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIIB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7 ( 2.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 ( 4.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIIIA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 ( 0.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0 ( 0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIIIB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0 ( 0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 ( 0.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e89 ( 31.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e41 ( 33.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCEA (median [IQR])\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.90 [2.10, 4.10]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.90 [2.30, 4.60]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.765\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePathological_type (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eadenocarcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e258 ( 91.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e112 ( 92.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.238\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003esquamous cell carcinom\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e19 ( 6.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 ( 3.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003esmall cell lung cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 ( 1.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 ( 3.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eother\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 ( 0.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 ( 0.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003ch3\u003eCT Image Acquisition and Feature Extraction\u003c/h3\u003e\n\u003cp\u003eAll patients were placed in the supine position and scanned from the thoracic inlet to the diaphragmatic base. Scanning parameters were as follows: slice thickness: 5.0 mm, tube voltage: 120 kVp, and tube current: 80\u0026ndash;300 mAs. All images were displayed using the standard lung (window width: 1200 HU; window level: \u0026minus;600 HU) and mediastinal (window width: 350 HU; window level: 40 HU) window settings.\u003c/p\u003e\n\u003cp\u003eAll images were saved on a hard disk in DICOM format and imported into 3D Slicer software (version 5.8.1). Regions of interest were manually delineated by a radiologist with over 5 years of experience. The radiologists were blinded to clinical and pathological information. In total, 851 radiomic features were extracted using the SlicerRadiomics module, including: 1. 14 shape features describing tumor morphology and geometric structure, such as edge detection and contour description; 2. 93 texture features describing image texture information, such as the gray-level co-occurrence matrix, gray-level difference matrix, and gray-level run length matrix; 3. 744 frequency-domain features describing tumor characteristics in the frequency domain, such as Fourier transform coefficients and wavelet transform coefficients.\u003c/p\u003e\n\u003ch3\u003eSelection of Radiomic Features\u003c/h3\u003e\n\u003cp\u003ePrior to feature selection, all radiomic features were standardized using Z-score normalization. Owing to the large number of radiomic features, three methods were employed for dimensionality reduction to avoid issues, such as model overfitting and multicollinearity. First, variance thresholding was used to remove features with a variance\u0026thinsp;\u0026lt;\u0026thinsp;0.8. Second, univariate selection was applied to filter out non-significant features (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). Finally, the least absolute shrinkage and selection operator regression algorithm was used to sparsify the feature set, select indicators most relevant to the research objective, and obtain their weights. Key radiomic features were selected (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). Ultimately, 16 radiomic features were retained (Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e), including: wavelet-LLH_glcm_Imc2, wavelet-HHH_glcm_ClusterProminence, original_shape_Sphericity, wavelet-LLL_glcm_MCC, wavelet-HLH_glcm_Imc2, original_glcm_ClusterShade, original_gldm_SmallDependenceHighGrayLevelEmphasis, original_glszm_ZoneEntropy, wavelet-LLL_gldm_SmallDependenceHighGrayLevelEmphasis, original_firstorder_90Percentile, original_glszm_SizeZoneNonUniformity, wavelet-LHL_firstorder_Maximum, wavelet-LLH_firstorder_Mean, wavelet-LLL_glszm_GrayLevelVariance, and original_firstorder_Minimum. After dimensionality reduction, the radiomics score (RadScore) was calculated for each patient using the following formula. The RadScore, which comprehensively represents radiomic features, was incorporated into model construction.\u003c/p\u003e\n\u003cp\u003eThe formula for calculating the RadScore is:\u003c/p\u003e\n\u003cp\u003eRadScore\u0026thinsp;=\u0026thinsp;Intercept\u0026thinsp;+\u0026thinsp;feature₁ \u0026times; coefficient₁ + \u0026hellip; + featureₙ \u0026times; coefficientₙ\u003c/p\u003e\n\u003cp\u003ewhere the intercept is the fitted intercept term and coefficient[i] is the feature weight coefficient generated by least absolute shrinkage and selection operator regression.\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003cp\u003e(A) shows the mean squared error of the radiomic features displayed by the Lasso regression analysis in the model. The two vertical lines indicate the lambda values when the number of variables is reduced to the two lowest levels. (B) shows the trajectory of each radiomic feature, with the horizontal axis representing the log lambda for each radiomic feature and the vertical axis representing the coefficient of each radiomic feature.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eLASSO feature selection table\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eFeatures\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eCoeffieints\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-LLH_glcm_Imc2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.025579894\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-HHH_glcm_ClusterProminence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.019328046\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_shape_Sphericity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.013027585\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-LLL_glcm_MCC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.010390812\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-HLH_glcm_Imc2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.007564503\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_glcm_ClusterShade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.007167992\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_gldm_SmallDependenceHighGrayLevelEmphasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.00049788\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_glszm_ZoneEntropy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0011592808893695907\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-LLL_gldm_SmallDependenceHighGrayLevelEmphasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0027629811348519767\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_firstorder_90Percentile\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.009111013539256278\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_glszm_SizeZoneNonUniformity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.03503650252922063\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-LHL_firstorder_Maximum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.03734605667166642\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-LLH_firstorder_Mean\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.03901754470939138\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ewavelet-LLL_glszm_GrayLevelVariance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.05842791980640918\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eoriginal_firstorder_Minimum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0622995314014254\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003ch3\u003eModel Construction\u003c/h3\u003e\n\u003cp\u003eTo predict LNM, filtered clinical and radiomic features were used as inputs to construct a radiomics model for the differential diagnosis of LNM. All data were randomly divided into development and internal test sets (split ratio: 7:3). The training set was used for model training, whereas the internal and external test sets were used to assess the generalization ability of the model. Logistic regression, support vector machine (SVM), and random forest (RF) algorithms were employed to construct models for the differential diagnosis of LNM. Receiver operating characteristic curves were plotted, and the area under the curve (AUC) was calculated to evaluate accuracy, precision, sensitivity, and specificity.\u003c/p\u003e\n\u003ch3\u003eConstruction of the Radiomics Nomogram\u003c/h3\u003e\n\u003cp\u003eA radiomics nomogram was constructed by combining RadScores and significant clinicopathological factors in a multivariate logistic regression model. Receiver operating characteristic curves were plotted and AUC values were calculated to evaluate accuracy, precision, sensitivity, and specificity. Additionally, a calibration curve was plotted to assess calibration and discrimination. Decision curve analysis was performed to quantify the net benefit of the nomogram and evaluate its clinical utility.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n\u003ch2\u003eClinical Characteristics\u003c/h2\u003e\n\u003cp\u003eThe demographic and clinicopathological characteristics of patients in the development and internal test sets are detailed in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. The training set comprised 282 patients (median age: 66 years; interquartile range: 55\u0026ndash;74 years; 157 men). Univariate and multivariate analyses revealed that LNM status was significantly associated with T stage and smoking history. However, no statistically significant differences were observed between patients with and without metastasis in the training set regarding sex, age, family history, metastasis stage, clinical stage, or CEA levels (all p\u0026thinsp;\u0026gt;\u0026thinsp;0.05, Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab3\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eUnivariate and Multivariate Analyses\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eVariables\u003c/p\u003e\n\u003c/th\u003e\n\u003cth colspan=\"4\" align=\"left\"\u003e\n\u003cp\u003eUnivariate analysis\u003c/p\u003e\n\u003c/th\u003e\n\u003cth colspan=\"4\" align=\"left\"\u003e\n\u003cp\u003eMultivariate analysis\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eOR\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCI.lower\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCI.upper\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eP value\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eOR\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCI.lower\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCI.upper\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eP value\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eGender\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.614\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.404\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.926\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.021\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.982\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.568\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.712\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.948\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAge\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.027\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.01\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.045\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.002\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.018\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.999\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.037\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.065\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSmoke\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.961\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.279\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e3.01\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.002\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.959\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.115\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e3.481\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u003cstrong\u003e0.02\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eFamily_history\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.901\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.423\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.827\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.778\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eT_stage\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.883\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.545\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e2.311\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.726\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.365\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e2.197\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.001\u003c/strong\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eM_stage\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e2.255\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.494\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e3.42\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.543\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.811\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e2.897\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.18\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eClinical_stage\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.182\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.103\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.268\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.025\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.911\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.154\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.679\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eCEA\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.994\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.985\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.999\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.092\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003ePathological_type\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.614\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.404\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.926\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.021\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n\u003ch2\u003eModel Construction and Predictive Performance\u003c/h2\u003e\n\u003cp\u003eThree machine learning algorithms\u0026mdash;logistic regression, SVM, and RF\u0026mdash;were selected to construct prediction models. Models were built using clinical features, radiomic features, and a combination of both (clinical\u0026ndash;radiomic features). In the training set, the SVM model with combined clinical\u0026ndash;radiomic features demonstrated the best performance (AUC\u0026thinsp;=\u0026thinsp;0.927). The model based solely on radiomic features also performed well (AUC\u0026thinsp;=\u0026thinsp;0.921), approaching the performance of the combined model. The model based solely on clinical features exhibited low performance (AUC\u0026thinsp;=\u0026thinsp;0.7).\u003c/p\u003e\n\u003cp\u003eIn the internal test set, the SVM model with combined clinical\u0026ndash;radiomic features achieved the best performance (AUC\u0026thinsp;=\u0026thinsp;0.852), whereas the RF model based solely on radiomic features achieved an AUC of 0.844. In the external test set, the RF model with combined clinical\u0026ndash;radiomic features performed best (AUC\u0026thinsp;=\u0026thinsp;0.833) (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eThe clinical utility of the models was assessed using decision curve analysis curves. Within a reasonable threshold probability range, the net benefit of the combined model was higher than that of the clinical-only and radiomic-only models (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eA. Development set: The LR model with clinical features alone (AUC=0.693), the LR model with radiomic features alone (AUC=0.806), and the LR model with combined features (AUC=0.833).B. Internal test set: The LR model with clinical features alone (AUC=0.650), the LR model with radiomic features alone (AUC=0.812), and the LR model with combined features (AUC=0.817).C. External test set: The LR model with clinical features alone (AUC=0.644), the LR model with radiomic features alone (AUC=0.767), and the LR model with combined features (AUC=0.794).D. Development set: The SVM model with clinical features alone (AUC=0.725), the SVM model with radiomic features alone (AUC=0.921), and the SVM model with combined features (AUC=0.927).E. Internal test set: The SVM model with clinical features alone (AUC=0.707), the SVM model with radiomic features alone (AUC=0.824), and the SVM model with combined features (AUC=0.852).F. External test set: The SVM model with clinical features alone (AUC=0.687), the SVM model with radiomic features alone (AUC=0.840), and the SVM model with combined features (AUC=0.812).G. Development set: The RF model with clinical features alone (AUC=0.711), the RF model with radiomic features alone (AUC=0.879), and the RF model with combined features (AUC=0.884).H. Internal test set: The RF model with clinical features alone (AUC=0.733), the RF model with radiomic features alone (AUC=0.844), and the RF model with combined features (AUC=0.838).I. External test set: The RF model with clinical features alone (AUC=0.663), the RF model with radiomic features alone (AUC=0.799), and the RF model with combined features (AUC=0.833).\u003c/p\u003e\n\n\u003cp\u003eDCA curves for the clinical feature model, radiomic feature model, and clinical-radiomic combined feature model. Within the threshold probability range of 0.2\u0026ndash;0.8, the combined model yielded a higher net benefit.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n\u003ch2\u003eModel Interpretability Analysis\u003c/h2\u003e\n\u003cp\u003eSHapley Additive exPlanations (SHAP) plots were generated to visualize the impact of different features on individual predictions. Tumor stage was identified as the most critical variable, followed by radiomic features reflecting tumor gray-level distribution and shape and smoking history. All features exhibited positive SHAP values, indicating an overall tendency to increase the predicted probability of LNM. This demonstrates the synergistic effect of clinical and radiomic features, providing a basis for understanding model decisions, investigating metastatic mechanisms, and optimizing diagnostic and treatment strategies (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eA: SHAP feature importance heatmap. The x-axis represents individual instances (samples), and the y-axis lists key features (including T stage, radiomic features, smoking history, etc.). Color represents the SHAP value (red: positive, blue: negative, indicating a promoting or inhibitory effect on LNM prediction, respectively). The curve at the top represents the model output \\(f(x)\\).B: SHAP summary plot. The y-axis lists key features (including T stage, radiomic texture/shape features, smoking history, etc.). The x-axis represents the impact of the feature value on the model output (SHAP value). Color indicates the feature value (red: high, blue: low).\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n\u003ch2\u003eNomograms\u003c/h2\u003e\n\u003cp\u003eA nomogram was constructed to present the model visually (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e). From top to bottom, the left side of the nomogram displays the risk factor variables, total points, and predicted probability of LNM. Points were assigned for each risk factor, and the sum corresponded to the total points, which were then used to estimate the probability of LNM based on the multivariate logistic regression model.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we retrospectively collected the clinicopathological characteristics of patients and extracted radiomic features from non-contrast chest CT scans to develop nine models for predicting LNM in lung cancer, utilizing clinical, radiomic, and combined features. The clinical\u0026ndash;radiomic SVM model demonstrated the best performance in predicting lymph node status (AUC\u0026thinsp;=\u0026thinsp;0.927 in the training set, 0.852 in the internal test set, and 0.812 in the external test set), whereas the model based solely on clinical features showed low predictive efficacy (AUC\u0026thinsp;=\u0026thinsp;0.700). This may be attributed to the superior generalization capability of the SVM algorithm in classification tasks, its faster computational speed compared with that of other machine learning algorithms, and its low sample size requirement \u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003ePrevious studies have identified strong associations between LNM in lung cancer and factors, such as age, smoking status, tumor size, histology, degree of differentiation, CEA level, vascular invasion, and pleural involvement \u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e. Ryuichi et al. developed a clinical prediction model based on serum CEA and CYFRA levels, maximum solid tumor diameter, age, and consolidation/tumor ratio, which achieved an AUC of 0.75 \u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e. Hu constructed a RF model using pathology, degree of differentiation, maximum short-axis diameter of lymph nodes, and tumor diameter, yielding an AUC of 0.83 \u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e. Tang developed a nomogram incorporating Cyfra21-1, D-dimer, tumor size, percentage of solid components, and lesion location, which achieved an AUC of 0.904 \u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. By contrast, the clinical model in our study exhibited low performance (AUC of the SVM model: 0.725 in the training set, 0.707 in the internal test set, and 0.687 in the external test set). Furthermore, no statistically significant differences were observed between patients with and without metastasis in terms of age, family history, histology, degree of differentiation, or CEA levels, which is inconsistent with previous findings. This discrepancy may be because of the relatively small sample size and potential selection bias in our study.\u003c/p\u003e\u003cp\u003eRadiomics converts digital images containing tumor pathophysiological information into measurable and quantifiable data, thereby improving the accuracy of imaging diagnosis \u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e. Zhang developed a nomogram based on CT radiomic features to predict occult LNM, achieving an AUC of 0.782 for the training set and 0.813 for the test set \u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e. Geng further utilized deep learning radiomic features extracted via neural network-based automatic segmentation to construct a model for predicting LNM, achieving an AUC of 0.859 \u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e. Compared with the information provided by non-contrast CT, PET-CT offers additional hemodynamic and metabolic parameters, such as maximum standardized uptake value, mean standardized uptake value, metabolic tumor volume, and total lesion glycolysis, which better reflect tumor aggressiveness. Gabriela combined clinicopathological characteristics with radiomic features extracted from \u0026sup1;⁸F-FDG PET-CT images to develop a gradient boosting machine model, which achieved an AUC of 0.94 \u003csup\u003e[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn comparison with the aforementioned studies, the present study developed a nomogram for predicting LNM in lung cancer based on readily accessible clinical features (including age, smoking history, and tumor markers) combined with radiomic features extracted from preoperative non-contrast CT scans. This model demonstrated superior performance in predicting lymph node status (AUC\u0026thinsp;=\u0026thinsp;0.927 in the training set, 0.852 in the internal test set, and 0.812 in the external test set), with a specificity of 0.93 and sensitivity of 0.81. Its specificity was notably higher than that of conventional CT (specificity\u0026thinsp;=\u0026thinsp;0.55, sensitivity\u0026thinsp;=\u0026thinsp;0.81) \u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. The predictive performance of our model exceeded that of the model developed by Boris based on contrast-enhanced CT imaging features \u003csup\u003e[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e. These results indicate that combined clinical\u0026ndash;radiomic features can provide more comprehensive information, improve the accuracy of classification models, reduce reliance on single-modality data, and enhance applicability across diverse clinical scenarios.\u003c/p\u003e\u003cp\u003eOur study identified T stage as a key factor influencing LNM in lung cancer. Large tumors exhibit a high propensity for metastasis \u003csup\u003e[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e, and our results similarly showed that tumor stage was a high-impact predictor in the nomogram. Decision curve analysis demonstrated that the nomogram could effectively predict LNM when the threshold probability ranged from 0.2 to 0.8, indicating promising potential for clinical application.\u003c/p\u003e\u003cp\u003eThe strengths of this study are as follows: Current research on predicting LNM status in lung cancer primarily focuses on clinicopathological data and PET/CT, with limited studies based on non-contrast CT. Compared with contrast-enhanced CT and PET/CT, non-contrast CT involves lower radiation exposure and is noninvasive. Moreover, this study used easily accessible clinical data combined with non-contrast CT features to develop a model for predicting LNM in lung cancer. By leveraging multimodal fusion, the accuracy of the model was improved, offering a more economical and widely applicable method for predicting LNM status in patients with lung cancer. Additionally, external validation was performed using data from another hospital to verify the applicability and generalization capability of the model.\u003c/p\u003e\u003cp\u003eThis study had several limitations. First, because of its retrospective nature, selection bias may exist. Second, the sample size was relatively small, precluding the use of deep learning algorithms to enhance model performance. Therefore, future multicenter, large-sample prospective studies are warranted to further validate and optimize the model, thereby improving its applicability and generalizability.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study integrated CT-based radiomic features and clinical variables and applied multiple machine learning algorithms to construct a predictive model for LNM in lung cancer. Among these models, the SVM model incorporating combined features demonstrated the best performance with high predictive accuracy. These findings provide a novel approach for individualized risk assessment and intervention in patients with lung cancer.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eArea under the curve\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eCEA\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eCarcinoembryonic antigen\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eCT\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eComputed tomography\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eLNM\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLymph node metastasis\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003ePET\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003e\u0026sup1;⁸F-FDG positron emission tomography\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eRadScore\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eRadiomics score\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eRF\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eRandom forest\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSHAP\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSHapley Additive exPlanations\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eSVM\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSupport vector machine.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eEthics approval and consent to participate\u003c/h2\u003e\u003cp\u003e The research protocol was approved by the Ethics Committee of Panzhihua Traditional Chinese Medicine and Western Medicine Hospital (PCWEC-C-003). Given the retrospective nature of this study, the requirement for informed consent was waived. All methods were carried out in accordance with relevant guidelines and regulations. Our study adhered to the principles of the Declaration of Helsinki.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003cp\u003eNot Applicable.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003ch2\u003eCompeting interests\u003c/h2\u003e\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThis work was supported by the Program for 2023 Kunming Health Science and Technology Talent Training Project (Medical Technology Center) (No.2023-SW-13). National Natural Science Foundation of China (NSFC)(No.82160348), 2024 Senior Health Technology and Medical Discipline Leader of Yunnan Provincial Health Commission(No. D-2024056).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003ePQW and HH contributed equally to the study concept and design, acquisition of subjects and data, analysis and interpretation of data, and preparation of the manuscript and need therefore be granted both first authorship. YBW was involved in data collection, data analysis and interpretation and reviewed the manuscript. LY assisted in data acquisition and reviewed the manuscript. YDY and YF contributed to the study concept and design and reviewed the manuscript. BSX and JGL were involved in the study design, data analysis and interpretation, and manuscript review. MXK contributed to the study design and manuscript review. CYW participated in the study design, data analysis and interpretation, and manuscript preparation. BY was involved in the study design, data collection, data analysis and interpretation, and manuscript preparation. All the authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e\u003cp\u003eWe thank Panzhihua Traditional Chinese Medicine and Western Medicine Hospital and Panzhihua Central Hospital for providing data records.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data that supports the findings of this study are available in the supplementary material of this article.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eFreddie B, Mathieu L, Hyuna S, Jacques. F,Rebecca LSl.Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.CA Cancer. J Clin 2024 May-Jun;74(3):229\u0026ndash;63.\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3322/caac.21834\u003c/span\u003e\u003cspan address=\"10.3322/caac.21834\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Li ZJ, Chen YX, Xiao H, Zhou YK, Du SS, et al. Outcomes of Image-Guided Moderately Hypofractionated Radiotherapy for Stage III Non-Small-Cell Lung Cancer. J Oncol. 2021;2021(0):2721261. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1155/2021/2721261\u003c/span\u003e\u003cspan address=\"10.1155/2021/2721261\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWanyin L, Carole AR,Andrew GN,Saeed M. The 8(th) lung cancer TNM classification and clinical staging system: review of the changes and clinical implications. Quant Imaging Med Surg. 2018;8(7):709\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/qims.2018.08.02\u003c/span\u003e\u003cspan address=\"10.21037/qims.2018.08.02\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZiv GM. Staging in Non-Small Cell Lung Cancer.Surg Oncol Clin. N Am. 2016;25(3):493\u0026ndash;502. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.soc.2016.02.004\u003c/span\u003e\u003cspan address=\"10.1016/j.soc.2016.02.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTian Y, He Y, Li X. Liu XW.Novel nomograms to predict lymph node metastasis and distant metastasis in resected patients with early-stage non-small cell lung cancer. Ann Palliat Med. 2021;10(3):2548\u0026ndash;66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/apm-20-1756\u003c/span\u003e\u003cspan address=\"10.21037/apm-20-1756\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang ZR, Feng HX, Zhao H, Hu J, Liu LX, Liu Y, et al. Sublobar resection is associated with better perioperative outcomes in elderly patients with clinical stage I non-small cell lung cancer: a multicenter retrospective cohort study. J Thorac Dis. 2019;11(5):1838\u0026ndash;48. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/jtd.2019.05.20\u003c/span\u003e\u003cspan address=\"10.21037/jtd.2019.05.20\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBabatunde AYK, Chi-Fu JY, Brian CG, Li XC, Michael SM,Gu L, et al. A national analysis of wedge resection versus stereotactic body radiation therapy for stage IA non-small cell lung cancer. J Thorac Cardiovasc Surg. 2017;154(2):675\u0026ndash;e6864. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jtcvs.2017.02.065\u003c/span\u003e\u003cspan address=\"10.1016/j.jtcvs.2017.02.065\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMegan ED, Navneet S, Nofisat I,Mara BA, Douglas. AA,JeffreyB,et al. Management of Stage III Non-Small-Cell Lung Cancer: ASCO Guideline. J Clin Oncol. 2022;40(12):1356\u0026ndash;84. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1200/JCO.21.02528\u003c/span\u003e\u003cspan address=\"10.1200/JCO.21.02528\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePostmus PE, Kerr KM, Oudkerk M, Waller SSDA, Vansteenkiste J, et al. Early and locally advanced non-small-cell lung cancer (NSCLC): ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28(0):iv1\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/annonc/mdx222\u003c/span\u003e\u003cspan address=\"10.1093/annonc/mdx222\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHwayoung S, Sung H Y,Junghoon K, Jihang K, Kyoung WL, Woojoo L, et al. Application of N Descriptors Proposed by the International Association for the Study of Lung Cancer in Clinical Staging. Radiology. 2021;300(2):450\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1148/radiol.2021204461\u003c/span\u003e\u003cspan address=\"10.1148/radiol.2021204461\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNoeul K, Sun HS, HongseokY,Byung WJ, Kyungjong L, Sang-Won U, et al. Infectious complications of EBUS-TBNA: A nested case-control study using 10-year registry data. Lung Cancer. 2021;161(0):1\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.lungcan.2021.08.016\u003c/span\u003e\u003cspan address=\"10.1016/j.lungcan.2021.08.016\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAnnalori P, Paolo S. Lung Cancer and Radiological Imaging. Curr Radiopharm. 2020;13(3):238\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2174/1874471013666200523161849\u003c/span\u003e\u003cspan address=\"10.2174/1874471013666200523161849\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCong MD, Feng H, Ren JL, Xu Q, Cong LL,Hou ZZ, Wang YY, et al. Development of a predictive radiomics model for lymph node metastases in pre-surgical CT-based stage IA non-small cell lung cancer. Lung Cancer. 2020;139(0):73\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.lungcan.2019.11.003\u003c/span\u003e\u003cspan address=\"10.1016/j.lungcan.2019.11.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFieke H, David JH, Johannes MD, Naomi B, Rob A E M T, Michel WJM, W, et al. Variation Between Multidisciplinary Tumor Boards in Clinical Staging and Treatment Recommendations for Patients With Locally Advanced Non-small Cell Lung Cancer. Chest. 2020;158(6):2675\u0026ndash;87. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.chest.2020.07.054\u003c/span\u003e\u003cspan address=\"10.1016/j.chest.2020.07.054\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMichael KG, Ware GK, Chara ER, Courtney CM, Anita ND, Hidenobu S, et al. Test performance of positron emission tomography and computed tomography for mediastinal staging in patients with non-small-cell lung cancer: a meta-analysis. Ann Intern Med. 2003;139(11):879\u0026ndash;92. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7326/0003-4819-139-11-200311180-00013\u003c/span\u003e\u003cspan address=\"10.7326/0003-4819-139-11-200311180-00013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSobic-Saranovic D, Petrusic I, Artiko V, Pavlovic S, Subotic D, Saranovic D, Nagorni-Obradovic L, Petrovic N, et al. Comparison of 18F-FDG PET/CT and MDCT for staging/restaging of non-small cell lung cancer. Neoplasma. 2015;62(2):295\u0026ndash;301. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4149/neo_2015_035\u003c/span\u003e\u003cspan address=\"10.4149/neo_2015_035\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSang Woo Lee,Seong-Jang Kim. Is Delayed Image of 18F-FDG PET/CT Necessary for Mediastinal Lymph Node Staging in Non-Small Cell Lung Cancer Patients? Clin Nucl Med. 2022;47(5):414\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/RLU.0000000000004110\u003c/span\u003e\u003cspan address=\"10.1097/RLU.0000000000004110\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang M, Li SL, Li NLYM, Zhang SY,Lu FL,Yang Y, et al. Occult mediastinal lymph node metastasis in FDG-PET/CT node-negative lung adenocarcinoma patients: Risk factors and histopathological study. Thorac Cancer. 2019;10(6):1453\u0026ndash;60. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/1759-7714.13093\u003c/span\u003e\u003cspan address=\"10.1111/1759-7714.13093\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSarah JG, Anthony W K,Jonathan TP, Kyle B, Frank CD, Daniel JB, et al. Indications for invasive mediastinal staging in patients with early non-small cell lung cancer staged with PET-CT. Lung Cancer. 2017;109(0):36\u0026ndash;41. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.lungcan.2017.04.018\u003c/span\u003e\u003cspan address=\"10.1016/j.lungcan.2017.04.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYi Y, Jiang YH,BaiBB,Zhang CF. Emerging role of deep learning-based artificial intelligence in tumor pathology. Cancer Commun (Lond). 2020;40(4):154\u0026ndash;66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/cac2.12012\u003c/span\u003e\u003cspan address=\"10.1002/cac2.12012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePhilippe L, Ralph T H L,Timo MD, Jurgen P, Evelyn EC, van de Jong T et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol 2017;14(12):749\u0026ndash;762. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nrclinonc.2017.141\u003c/span\u003e\u003cspan address=\"10.1038/nrclinonc.2017.141\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllan FF, A,S\u0026eacute;rgio AS, Raul L Jr, Tarc\u0026iacute;sio A R,Agl\u0026aacute;ia M G X,Erica R. Combining machine learning and texture analysis to differentiate mediastinal lymph nodes in lung cancer patients. Phys Eng Sci Med. 2021;44(2):387\u0026ndash;94. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s13246-021-00988-2\u003c/span\u003e\u003cspan address=\"10.1007/s13246-021-00988-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFabian CL, Piotr W, Thorsten AB, Mirjam S, Miriam R, Mathilda W, et al. Transfer-Learning Deep Radiomics and Hand-Crafted Radiomics for Classifying Lymph Nodes from Contrast-Enhanced Computed Tomography in Lung Cancer. Cancers (Basel). 2023;15(10). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/cancers15102850\u003c/span\u003e\u003cspan address=\"10.3390/cancers15102850\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKensuke K, Hironobu S, Kyoichi O, Toshiteru T, Hyungeun Y, Shinji A. Impact of the number of dissected lymph nodes on machine learning-based prediction of postoperative lung cancer recurrence: a single-hospital retrospective cohort study. BMJ Open Respir Res. 2024;11(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmjresp-2023-001926\u003c/span\u003e\u003cspan address=\"10.1136/bmjresp-2023-001926\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEvan R, Brian C, Mousumi B. Classif J Appl Stat. 2019;46(16):2987\u0026ndash;3007. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/02664763.2019.1625876\u003c/span\u003e\u003cspan address=\"10.1080/02664763.2019.1625876\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen B, Xia WJ, Wang ZQ, Zhao H, Li XF, Liu LX, Liu Y, Hu J, et al. Risk analyses of N2 lymph-node metastases in patients with T1 non-small cell lung cancer: a multi-center real-world observational study in China. J Cancer Res Clin Oncol. 2019;145(11):2771\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00432-019-03006-x\u003c/span\u003e\u003cspan address=\"10.1007/s00432-019-03006-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShan LY, Zhang L, Zhu XL,Wang ZL,Fang SH,Lin JF, et al. Chinese never smokers with adenocarcinoma of the lung are younger and have fewer lymph node metastases than smokers. Respir Res. 2022;23(1):293. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12931-022-02199-z\u003c/span\u003e\u003cspan address=\"10.1186/s12931-022-02199-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRyuichi Y, Yoshitaka E, Takuya A, Hiroyuki D,Makoto T, Wataru S, et al. Diagnostic artificial intelligence model predicts lymph node status in non-small cell lung cancer using simplified examination. J Thorac Dis. 2024;16(11):7320\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/jtd-24-1067\u003c/span\u003e\u003cspan address=\"10.21037/jtd-24-1067\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu SL, Luo M, Li YL. Machine Learning for the Prediction of Lymph Nodes Micrometastasis in Patients with Non-Small Cell Lung Cancer: A Comparative Analysis of Two Practical Prediction Models for Gross Target Volume Delineatio. Cancer Manag Res. 2021;13(0):4811\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/CMAR.S313941\u003c/span\u003e\u003cspan address=\"10.2147/CMAR.S313941\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTang YBA, Garu,Chen X, Guan ZY,Cai XD,Huang HX, et al. Risk factors for mediastinal lymph node metastases in early-stage non-small-cell lung cancer and prediction model establishment. Am J Cancer Res. 2024;14(12):5952\u0026ndash;64. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.62347/DIZG4944\u003c/span\u003e\u003cspan address=\"10.62347/DIZG4944\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRobert JG, Paul EK, Hedvig H. Radiomics: Images Are More than Pictures. They Are Data[J] Radiol. 2016;278(2):563\u0026ndash;77. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1148/radiol.2015151169\u003c/span\u003e\u003cspan address=\"10.1148/radiol.2015151169\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2015 Nov 18.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang R, Zhang RR,Luan T,Liu BW,Zhang YM, Xu YP, et al. A Radiomics Nomogram for Preoperative Prediction of Clinical Occult Lymph Node Metastasis in cT1-2N0M0 Solid Lung Adenocarcinoma. Cancer Manag Res. 2021;13(0):8157\u0026ndash;67. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/CMAR.S330824\u003c/span\u003e\u003cspan address=\"10.2147/CMAR.S330824\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGeng MF, Geng. MS,WeiR,Chen MW.Artificial intelligence neural network analysis and application of CT imaging features to predict lymph node metastasis in non-small cell lung cancer. J Thorac Dis. 2022;14(11):4384\u0026ndash;94. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/jtd-22-1511\u003c/span\u003e\u003cspan address=\"10.21037/jtd-22-1511\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKyongmin SB, Bomi G, Sae J, Na,Ji HH, Sang HC, Ho JA, et al. DeepCUBIT: Predicting Lymphovascular Invasion or Pathological Lymph Node Involvement of Clinical T1 Stage Non-Small Cell Lung Cancer on Chest CT Scan Using Deep Cubical Nodule Transfer Learning Algorithm. Front Oncol. 2021;11(0):661244. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2021.661244\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2021.661244\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBoris G, Philipp H B,Alexander DJ, B,Alexander H,Julian MM, R,Christian F, et al. Inferring FDG-PET-positivity of lymph node metastases in proven lung cancer from contrast-enhanced CT using radiomics and machine learning. Eur Radiol Exp. 2022;6(1):44. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s41747-022-00296-8\u003c/span\u003e\u003cspan address=\"10.1186/s41747-022-00296-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu SL, Luo M, Li YLM. Learning for the Prediction of Lymph Nodes Micrometastasis in Patients with Non-Small Cell Lung Cancer: A Comparative Analysis of Two Practical Prediction Models for Gross Target Volume Delineation. Cancer Manag Res. 2021;13(0):4811\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/CMAR.S313941\u003c/span\u003e\u003cspan address=\"10.2147/CMAR.S313941\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"lung cancer, lymph node metastasis, computed tomography, radiomics","lastPublishedDoi":"10.21203/rs.3.rs-7815825/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7815825/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eThis study aimed to develop a predictive model for lymph node metastasis in patients with lung cancer using non-contrast computed tomography (CT).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eA total of 403 patients with lung cancer who met the inclusion criteria were randomly divided into training (n\u0026thinsp;=\u0026thinsp;282) and test (n\u0026thinsp;=\u0026thinsp;121) sets. Clinical information was collected, and radiomic features were extracted from non-contrast chest CT images using the \u0026ldquo;Radiomics\u0026rdquo; toolkit in 3D Slicer software. Subsequently, least absolute shrinkage and selection operator regression analysis was employed to reduce the number of variables and establish a prediction model for lymph node metastasis in patients with lung cancer based on non-contrast CT scans. The predictive performance and clinical utility of the model were evaluated using the area under the curve (AUC) and decision curve analysis, and Shapley additive explanations analysis was applied to enhance interpretability.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eLymph node metastasis was present in 35.5% (143/403) of patients. Two clinical features and 16 radiomic features most strongly associated with lymph node metastasis were identified, and nine models were constructed. The receiver operating characteristic curves of the combined clinical\u0026ndash;radiomic model demonstrated favorable predictive performance. The clinical\u0026ndash;radiomic SVM model demonstrated the best performance in predicting lymph node status (AUC\u0026thinsp;=\u0026thinsp;0.927 in the training set, 0.852 in the internal test set, and 0.812 in the external test set).Decision curve analysis indicated that the prediction model provided substantial clinical benefits.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe radiomics model based on non-contrast CT demonstrated good diagnostic performance in predicting lymph node metastasis in patients with lung cancer and may provide guidance for individualized targeted therapy.\u003c/p\u003e","manuscriptTitle":"Combining Computed Tomography Radiomics and Clinical Features to Predict Lymph Node Metastasis in Patients with Lung Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-31 12:34:43","doi":"10.21203/rs.3.rs-7815825/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-12T08:56:27+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-11T23:06:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"301643796259394781446611084288664810380","date":"2025-11-06T21:51:57+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-03T14:29:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"130671268606224020307649489833963842639","date":"2025-10-22T09:26:36+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-20T13:16:07+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-20T07:12:33+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-10-17T13:18:46+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-17T12:43:10+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-10-17T12:39:29+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"10c173b3-fbfb-4910-9eb9-a17f7b3ec867","owner":[],"postedDate":"October 31st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-23T16:03:03+00:00","versionOfRecord":{"articleIdentity":"rs-7815825","link":"https://doi.org/10.1186/s12880-026-02262-x","journal":{"identity":"bmc-medical-imaging","isVorOnly":false,"title":"BMC Medical Imaging"},"publishedOn":"2026-03-16 15:59:22","publishedOnDateReadable":"March 16th, 2026"},"versionCreatedAt":"2025-10-31 12:34:43","video":"","vorDoi":"10.1186/s12880-026-02262-x","vorDoiUrl":"https://doi.org/10.1186/s12880-026-02262-x","workflowStages":[]},"version":"v1","identity":"rs-7815825","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7815825","identity":"rs-7815825","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}