The clinical value of predicting lymphovascular invasion in patients with invasive lung adenocarcinoma based on the intratumoral and peritumoral CT radiomics models

The clinical value of predicting lymphovascular invasion in patients with invasive lung adenocarcinoma based on the intratumoral and peritumoral CT radiomics models | 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 The clinical value of predicting lymphovascular invasion in patients with invasive lung adenocarcinoma based on the intratumoral and peritumoral CT radiomics models Miaomiao LIN, Chunli Zhao, haipeng huang, xiang zhao, siyu Yang, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4783280/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 12 Nov, 2025 Read the published version in BMC Cancer → Version 1 posted 11 You are reading this latest preprint version Abstract Purpose: To investigate the clinical value of predicting lymphovascular invasion(LVI) in patients with invasive lung adenocarcinoma(LUAD)based on the intratumoral and peritumoral CT radiomics models. Materials and Methods: The 384 patients with invasive LUAD from Institution 1 were randomly divided into training (n=268) and internal validation (n=116) sets with a ratio of 7:3, and 251 patients from Institution 2 were used as the external validation set. Altogether, 1226 features were extracted from the tumor gross (GT), gross tumor and peritumor (GPT), and peritumor(PT), respectively. Clinical independent predictors for LVI in patients with invasive LUAD were screened using univariate and multivariate logistic regression, a combined model that included clinical predictors and optimal Rad-score was constructed , and a nomogram was drawn. Results: The GPT model showed better predictive efficacy than the GT and PT models, with the area under the curve (AUC) of 0.83, 0.79, and 0.75 in the training, internal validation, and external validation sets, respectively. In the clinical model, the preoperative carcinoembryonic antigen (CEA) level, tumor diameter, and spiculation were the independent predictors. The combined model containing the independent predictors and the GPT-Radscore significantly predicted LVI in patients with invasive LUAD, with AUCs of 0.84, 0.82, and 0.77 in the three cohorts, respectively. Conclusion: The CT scan-based radiomics model which including intratumoral and peritumoral radiomics features can effectively predict LVI in LUAD，and the predictive efficacy is further improved by combining clinically independent predictors. Invasive lung adenocarcinoma Computed tomography Radiomics Lymphovascular invasion Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Lung cancer is the global leading cause of cancer-related deaths [ 1 ], and adenocarcinoma is the predominant histological type of lung cancer [ 2 ]. Surgical resection is the most common treatment modality for lung cancer, but the 5-year postoperative survival of patients with lung cancer remains suboptimal because of postoperative tumor recurrence and metastasis. Tumor recurrence occurs in ≥ 30% of patients, including those with completely resected stage I non-small cell lung cancer (NSCLC) [ 3 ]. Larger tumor size, pleural infiltration, and lymphovascular invasion (LVI)are considered high-risk factors for postoperative recurrence in patients with lung cancer [ 4 – 7 ]. LVI is defined as tumor cell infiltration into the lumen of arteries, veins, or lymphatic vessels [ 8 ]. Cai et al. showed that patients with stage IA and IB NSCLC with LVI had similar survival rates. Therefore, researchers believe that LVI should be an indispensable determinant when deciding the tumor, node, and metastasis (TMN) stage of lung cancer. Consequently, as long as LVI is present in a patient with lung cancer, a patient with stage IA lung cancer should be adjusted to stage IB, showing that LVI is more significant in lung cancer prognosis than tumor size [ 9 ]. Several studies have also confirmed LVI as a pathological factor in the poor prognosis of patients with lung cancer [ 10 , 4 , 11 ]. Previous studies have shown that adjuvant chemotherapy is beneficial for patients with LVI-positive lung cancer to improve their recurrence-free survival and overall survival rates. Furthermore, LVI can be an indicator for preoperative neoadjuvant chemotherapy in patients with NSCLC [ 12 ]. Therefore, an accurate preoperative prediction of LVI would be useful when screening patients with lung cancer who would benefit from neoadjuvant chemotherapy and in developing an individualized surgical plan. Currently, LVI can only be diagnosed using postoperative histopathologic testing, and this hinders the selection of rational treatment options for patients with lung y tcancer. Therefore, an efficient and noninvasive method of detecting the LVI status in lung cancer is urgently needed. Recent studies have shown that certain CT features, such as tumor diameter, spiculation, and nodal component are noninvasive imaging biomarkers for predicting LVI in lung cancer [ 13 , 14 ]. However, diagnostic imaging physicians' experience and understanding of different signs deeply influence human-assessed imaging signs. These assessments are highly subjective and poorly reproducible, limiting the usefulness of conventional imaging for the preoperative prediction of LVI in lung cancer. Radiomics can reveal the biological characteristics of tumors at the molecular level [ 15 – 19 ]. Tumor cell growth through the lymphatic and blood vessels results in reduced homogeneity of the intratumoral and peritumoral textures. Therefore, CT imaging can predict LVI status in lung cancer. Currently, there are a few studies on the prediction of LVI in lung adenocarcinoma (LUAD) using the CT radiomics model, and the majority focus on the changes in the tumor, ignoring the potential value of the peritumor microenvironment [ 20 , 21 , 8 ]. Previous studies on LUAD have shown a range of transition zones between LUAD and normal lung tissues, approximately 1–5 mm, with an average of 3.5 mm. This transition zone contains much information about lymphatic and vascular infiltration, closely associated with tumor metastasis and patient prognosis [ 22 , 23 ]. Therefore, we hypothesized that a peritumor-based radiomics model could effectively predict LVI in patients with invasive LUAD. This study aimed to explore a non-invasive method to predict LVI status in LUAD by establish a combined model based on GPT-Radscore and clinical predictors, then visualizes the model with a nomogram. Materials and methods Patients This retrospective study was conducted following the Declaration of Helsinki. It was approved by our hospital’s Institutional Review Board (approval number: 2024-E109-01), which waived the requirement for informed consent. Patients with invasive LUAD who visited our hospital between January 2016 and May 2023 were selected as the internal cohort, whereas the external validation cohort was obtained from another hospital. This study included patients with pathology-confirmed LUAD, those with pathological reports of definite LVI status, those who underwent unenhanced chest CT-scan within 15 days preoperatively, those with CT images that aligned with diagnostic requirements, and those with solitary lesions. Patients with incomplete clinicopathological data and preoperative CT images,complications with other tumors in the lung, and recurrence of antitumor treatment preoperatively were excluded from the study (Fig. 1 ). In total, this study included 635 patients with invasive LUAD. Furthermore, 384 patients with LUAD from our institution were randomized into training and internal validation sets using a ratio of 7:3. In total, 251 patients with invasive LUAD from the other hospital served as the external validation set. Patients’ clinical information and CT features were collected. CT image acquisition Patients in institution 1 were examined with General GE 256-row spiral and Siemens Somaton Force dual-source CT machines, whereas patients from Institution 2 were examined with Dutch Philips iCT 256 and Siemens Sensation 64-row CT machines. The patients were instructed to lie in a supine position, head first, and a CT scan of the chest was performed. The parameters used were 120 kV tube voltage, 60–120 mA or 100–250 mA tube current, 5 mm layer thickness, 512×512 matrix, 0.625–2.5 mm reconstruction layer thickness, lung window setting 1500 Hu window width, and a window position of -500 Hu. Radiomics analysis Workflow In radiomics analysis, the region of interest(ROI) was outlined, the peritumor region expanded, features were selected and extracted, and the model was constructed, analyzed, and evaluated (Fig. 2 ). Tumor segmentation and peritumor regional extension Tumor segmentation and peritumoral area expansion were performed using the software package "Radiomics" (syngo.via Frontier 1.2.1, version VB10B, Siemens Healthineers, Germany). Lesion segmentation was performed by two radiologists with six (radiologist 1) and 12 (radiologist 2) years of experience diagnosing chest diseases. First, the ROI of the tumor gross (GT) set was outlined along the visible edge of the tumor’s largest level, labeled ROI GT . ROI GPT was obtained by automatically expanding ROI GT outward by 5 mm. Pathologists have found that there is a certain range of transition zone between LUAD tissues and normal lung tissue, with an average range of 3.5 mm, and this transition zone contains a large amount of information that is closely related to the prognosis of patients. Therefore, in the present study, the peritumoral outreach distance was set at 5 mm, with a view to encompassing the lung cancer transition zone region as much as possible. Finally, the area of the two ROIs of the gross tumor and peritumor (GPT) and GT sets was subtracted to obtain the ROI of the peritumor (PT) set labeled as ROI PT . The outlining process attempted to avoid adjacent structures, such as blood vessels, bronchial tubes, chest wall, and mediastinum. CT images of 30 patients were randomly selected to assess the radiomics features consistency radiologist 1 and 2 performed the ROI segmentation and radiomics features extraction of the 30 patients, respectively. During the same period, radiologist 1 and 2 performed GT, GPT, and PT ROI outlines to assess inter-observer agreement. Radiologist 1 performed a second outline 1 month later to assess intra-observer agreement. Both radiologists were blinded to the pathological findings. Consistency was evaluated using the intraclass co-rrelation coefficient (ICC). An ICC > 0.75 indicated good consistency of the extracted radiomics features, and radiologist 1 performed the segmentation of the remaining cases. Feature extraction and screening Radiomics features were extracted using the same software package, "Radiomics". The images were resampled to ensure uniformity in image voxels size, and all image voxels were 1 mm × 1 mm ×1 mm. In total, 1226 radiomics features were extracted from the three ROIs of GT, GPT, and PT, including the first-order, texture, higher-order, and other features obtained by constant transformations. Based on previous studies [ 24 ], the extracted radiomics features were defined according to the Imaging Biomarker Standardization Initiative. First, redundant features were eliminated using minimum redundancy and maximum relevance (mRMR). Then, the least absolute shrinkage and selection operator (LASSO) algorithm was used to reduce the dimensionality of the remaining features and construct the nomogram. The LASSO algorithm adjusts the value of the weight parameter (λ) to suppress the coefficients of some features with less importance to zero. This helps to rapidly screen the features by reducing computational volume. The optimal radiomics features were selected from the screened features to construct a radiomics model and calculate the Radscore. Modele constrution and analysis An radiomics model was constructed using a logistic regression classifier. Clinically independent predictors were screened sequentially by using univariate and multivariate logistic regression algorithm, which combined optimal Radscore to build a combined predictive mode. The combined model was visualized with the nomogram. The predictive efficacy of the model was assessed by using the area under curve (AUC), sensitivity, specificity, and accuracy. The statistic difference in AUCs among the models were evaluated using the Delong test, the nomogram predictive accuracy was assessed by using calibration curves, and the clinical utility of the predictive model was evaluated with the decision curve analysis(DCA). Statistical analysis The R software package (version 4.2.2, http://www.Rproject.org ) was used for data processing and constructing models. Univariate and multivariate logistic regression algorithm were used to determine independent predictors of LVI in patients with LUAD ( P < 0.05). The AUC was calculated to evaluate the diagnostic performance of each model, and the Delong test was used to assess differences in the AUC among models. Results Patient characteristics Table 1 shows the baseline data of patients.Multivariate logistic regression analysis showed that preoperative carcinoembryonic antigen (CEA) [ OR, 2.24; 95% CI:1.21–4.17], spiculation [OR, 3.06; 95% CI:1.74–5.37] and tumor diameter [OR, 1.97; 95% CI:1.50–2.60)]were independent predictors of LVI in patients with invasive LUAD. A clinical model were established by combining the independent predictors above. Table 2 shows the results of the univariate and multivariate logistic regression analysis. Optimal r adiomics f eature s screening and mode ls constructing In total, 1226 radiomics features were extracted from the GT, GPT, and PT ROIs. Redundant features were eliminated by using mRMR and LASSO algorithm. The GT, GPT, and PT models yielded the 3, 7, and 8 best features for constructing the radiomics models finally (Fig. 3). Radiomics models assessment Of the three radiomics models, the AUCs of the GPT model (0.83, 0.79, and 0.75),were higher than those of the GT (AUC: 0.79, 0.76, and 0.72) and PT (AUC: 0.80, 0.75, and 0.73) models. However, Delong's test showed no statistically significant differences in the AUC values of the three models ( P > 0.05). In addition, we comprehensively evaluated the radiomics model’s accuracy, sensitivity, and specificity (Table 3), and the GPT model had a more stable performance. Therefore, we used the radiomics features of GPT and independent clinical predictors to construct a nomogram. Radiomics nomogram construction The combined model was constructed using the GPT-Radscore, preoperative CEA level, spiculation, and tumor diameter，and then visualizes the model with the nomogram. It was constructed based on the training cohort and validated using the internal and external validation cohorts. Fig. 4 illustrates the clinical application of the nomogram. Fig. 5 shows the receiver operating characteristic curve(ROC) of the GPT radiomics model, the clinical model, and the nomogram. The predictive efficacy of clinical, GPT radiomics model and normogram are shown in Table 4. The Delong test showed that in the training cohort, the AUC values of the nomogram were significantly different from the clinical model ( P < 0.05). In the internal validation cohort, the AUC of the nomogram was significantly different from the radiomics model ( P < 0.05). The calibration curve showed that the LVI incidence predicted by the nomogram corresponds well with the actual LVI incidence (Fig. 6). The Hosmer-Lemeshow(H-L) test showed that the nomogram fit well ( P > 0.05). The DCA showed that the nomogram achieved a better net benefit than the clinical model and GPT radiomics in predicting LVI (Fig. 7). Discussion LVI is the first step in inducing tumor spread or recurrence. From a clinical perspective, accurate preoperative prediction of LVI in patients with LUAD is crucial for selecting an appropriate treatment regimen. This study demonstrated that GPT, a radiomics model constructed by combining peritumoral 5 mm features, was superior to GT and PT models in diagnostic performance, confirming the potential value of the peritumoral microenvironment in predicting LVI in patients with invasive LUAD. This study constructed a nomogram using the GPT-Radscore, preoperative CEA level, tumor diameter, and spiculation. The results showed that the radiomics nomogram contributed to the predictive efficacy, with an AUC of 0.84 (95% CI: 0.79-0.89), 0.82 (95% CI: 0.74-0.90), and 0.77 (95% CI: 0.70-0.84) in the training, internal validation and external validation sets respectively. Therefore, the nomogram established in this study can effectively individualize the prediction of LVI status in patients with LUAD. CEA is an acidic glycoprotein with human embryonic antigenic properties and is crucial in the diagnosis and prognosis of lung cancer[25-27]. This study found that high preoperative CEA level was an independent risk factor for LVI-positive in LUAD, and the higher the CEA level, the higher the chance of LVI in postoperative pathology. These findings are similar to that of Li et al. [28]. Previous studies have reported that CEA is a cell adhesion molecule that performs adhesion reactions between cancer cells and stromal collagen and is vital in tumor growth and metastasis[29,30]. Shimada et al.[31]in a study of the relationship between lesion size and LVI in patients with pT1-4N0-2M0 stage NSCLC, found that the probability of LVI varied based on the tumor size. In patients with a maximum diameter of ≤ 5 cm, the larger the tumor’s maximum diameter, the higher the probability of LVI. This study also showed a positive correlation between LVI appearance and tumor size. Tumor size is crucial when determining patients' TMN and affects their prognosis. Furthermore, previous studies have also proved that lymphovascular invasion is a critical factor that promotes the rise of T in TMN staging. Therefore, it can be speculated that tumor diameter and lymphovascular invasion directly affect each other. They can both accelerate the progression of malignant tumors and affect patients’ prognosis. It is generally believed that the pathological basis for the formation of spiculations in malignant tumors is the infiltration and proliferation of tumor cells along blood and lymphatic vessels to the adjacent mesenchyme, which shows short and dense features in CT morphology. Therefore, attention should be paid to the appearance of LVI in patients with spiculation-positive lung cancer. Metastasis is a major cause of death in patients with cancer. Immune cells, such as tumor-associated macrophages in the tumor microenvironment, are involved in the production of immunosuppressive TME in tumors through the production of inflammatory mediators and growth factors, which affect the angiogenic and metastatic behaviors of the tumor cells [32,33]. Lung tissues that appear normal on CT are often invaded by tumor cells, and the invaded areas, when in a transition state, cannot be visualized using imaging systems. Therefore, changes in tumor margins or adjacent tissues on patho-histology can help optimize the predictive performance of radiomics models. Recently, Chen et al. [34], for the first time, constructed a model to predict LVI in patients with NSCLC based on peritumor radiomics features. The results showed that the GPTV 9 model, which included intratumor and peritumor 9 mm regions, had the best prediction performance, proving the radiomics model’s value containing PT V information in disease diagnosis. The transition zone of LUAD has been reported in the literature as 2–5 mm, and this region contains a great deal of information associated with lung cancer’s aggressive biological behavior. Therefore, this study’s maximum value of the peritumor region was placed at 5 mm, expecting the transition zone to be included as much as possible. The findings obtained were similar to those by Chen et al. and confirmed the feasibility of the peritumor microenvironmental radiomics model for the preoperative prediction of LVI in LUAD. In radiomics, the choice of the largest level of a lesion or the entire volume for ROI outlining has been debated [35,36]. In fact, it has not been definitively verified whether three-dimensional (3D)-based models are necessarily superior to the two-dimensional (2D) models in lung cancer diagnostic and treatment applications. Yang et al. [8]. compared the performance of 2D and 3D radiomics models in predicting LVI in lung cancer and other cancers, and their results showed that the 2D model was more effective than the 3D model. Nie et al. [21] reported similar results. Further, 2D features are easier to obtain than 3D features, are more reproducible, and have certain advantages in lung cancer research. Therefore, this study performed lesion ROI outline from the 2D level and obtained a good model prediction performance. For above phenomenon about lung cancer with radiomics, Zhang et al. [37] suggested that the 3D radiomics model is not as effective as the 2D model, because the 3D ROI labeling process generates more "noise,” which interferes with valid information data, thus creating a bias in the experimental results. There are two sources of noise: first, there are certain differences between different operators in recognizing lesions and lesion boundaries, which the 3D ROI labeling process will amplify; second, the thickness of the image layer will affect the amount of noise . The thicker the image layer, the less noise there is, and vice versa. On the Z-axis (thickness direction), the voxel spacing is usually inconsistent with the X- and Y-axes (transverse plane), which may affect the accuracy of the 3D image standard. The 3D image labeling process is far more affected by unclear image boundaries and thickness than 2D images, resulting in less effective data information being obtained from 3D images. This leads to a less effective model than that obtained via a 2D model, although the 3D image provides more information. In future research, it is necessary to improve image histology feature extraction and computation methods to determine the best research method. In addition to the three radiomics models of GT, GPT, and PT, the study also combined age, preoperative CEA level, spiculation, and GPT-Radscore to construct a combined prediction model, and the predictive efficacy of the combined model was improved compared with that of the clinical and radiomics models alone. By integrating the advantages of the clinical and radiomics models, the combined model provides a more stable and reliable disease diagnosis and treatment model. [38,39]. Therefore, a model integrating clinical information and radiomics features is more valuable in predicting LVI in patients with invasive LUAD. In addition, the combined model was visualized using nomogram. A nomogram is a statistical model for individualized predictive analysis of clinical events, quantifying the risk of clinical events according to various risk factors supporting the development of treatment plans. This study has some limitations. First, it is a retrospective study that could not standardize the machine, scanning, and other relevant parameters. Second, previous studies have found that micropapillary and solid LUAD have the highest chance of developing LVI, which shows that the pathological subtypes of invasive LUAD are inextricably linked to the occurrence of LVI. However, the present study did not further subdivide the pathological subtypes of invasive LUAD. To achieve more precise treatment, subsequent studies can further explore the correlation between different pathological subtypes of invasive LUAD and LVI. Third, the present study did not further investigate the effect of LVI on the prognosis of LUAD, which is also a matter of great clinical concern. In conclusion, the GPT radiomics model performed better than the GT and PT radiomics models in predicting the LVI status in patients with invasive LUAD. In addition, the combined model based on GPT radiomics features and clinically independent predictors can further improve prediction efficiency and provide an objective and quantitative reference basis for clinicians to make decisions and assess patient prognosis. Abbreviations LVI Lymphovascular Invasion LUAD Lung Adenocarcinoma GT Tumor Gross GPT Gross Tumor and Peritumor PT Peritumor AUC Area Under the Curve CEA Carcinoembryonic Antigen NSCLC Non-Small Cell Lung Cancer TMN Tumor, Node, and Metastasis ROI Region of Interest ICC Intraclass Co-rrelation Coefficient Mrmr Minimum Redundancy and Maximum Relevance Lasso Least Absolute Shrinkage and Selection Operator DCA Decision Curve Analysis ROC Receiver Operating Characteristic Curve Declarations Acknowledgements We would like to thank doctor Liqun Hu, from the Department of Pathology, The People’s Hospital of Guangxi Zhuang Autonomous Region, Guangxi Academy of Medical Sciences, for her guidance in LVI evaluation during this research. Author contributions Conceptualization: MML, KL. Methodology: MML, HPH, XZ,XXH,SYY,CLZ,KL. Validation: HHP,XZ,SYY. Formal analysis: MML, HPH, XZ,XXH. Data curation: MML, HPH, XZ,XXH,SYY,CLZ. Supervision: KL. Writing-original draft:MML,CLZ. Writing-review & editing: HHP,XZ,SYY. Formal analysis: MML, HPH, XZ,XXH. All authors read and approved the final manuscript. Funding Not applicable. Availability of data and meterials No datasets were generated or analysed during the current study. Ethics approval and consent to participate The study was approved by the The First Affiliated Hospital of Guangxi Medical University (Approval number: 2024-E109-01) with a waived requirement for informed consent (retrospective design). Consent for publication Our research contains no personal data. Competing interests The authors report no competing interests. References Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021; 71 (3):209-49. Siegel RL, Miller KD, Wagle NS, et al. Cancer statistics, 2023. CA Cancer J Clin 2023; 73 (1):17-48. Sung SY, Kwak YK, Lee SW, et al. Lymphovascular Invasion Increases the Risk of Nodal and Distant Recurrence in Node-Negative Stage I-IIA Non-Small-Cell Lung Cancer. Oncology 2018; 95 (3):156-62. Yun JK, Lee GD, Choi S, et al. Comparison of prognostic impact of lymphovascular invasion in stage IA non-small cell lung cancer after lobectomy versus sublobar resection: A propensity score-matched analysis. Lung cancer (Amsterdam, Netherlands) 2020; 146:105-11. Yang H, Mei T. Prognostic significance of visceral pleural invasion in patients with surgically resected small-cell lung cancer: a population-based study. Jpn J Clin Oncol 2022; 52 (9):1045-55. De Giglio A, Di Federico A, Gelsomino F, et al. Prognostic relevance of pleural invasion for resected NSCLC patients undergoing adjuvant treatments: A propensity score-matched analysis of SEER database. Lung cancer (Amsterdam, Netherlands) 2021; 161:18-25. Pathak R, Goldberg SB, Canavan M, et al. Association of Survival With Adjuvant Chemotherapy Among Patients With Early-Stage Non-Small Cell Lung Cancer With vs Without High-Risk Clinicopathologic Features. JAMA Oncol 2020; 6 (11):1741-50. Yang G, Nie P, Zhao L, et al. 2D and 3D texture analysis to predict lymphovascular invasion in lung adenocarcinoma. Eur J Radiol 2020; 129:109111. Cai JS, Wang X, Yang F, et al. Lymphovascular invasion: A non-sized T descriptor for stage IA non-small cell lung cancer. Thorac Cancer 2022; 13 (17):2413-20. Guglielmetti LC, Schneiter D, Hillinger S, et al. Lymphovascular invasion is an independent prognostic factor for survival in pathologically proven N2 non-small cell lung cancer. Swiss Med Wkly 2021; 151:w20385. Suaiti L, Sullivan TB, Rieger-Christ KM, et al. Vascular Invasion Predicts Recurrence in Stage IA2-IB Lung Adenocarcinoma but not Squamous Cell Carcinoma. Clin Lung Cancer 2023; 24 (3):e126-e133. Wu LL, Jiang WM, Qian JY, et al. High-risk characteristics of pathological stage I lung adenocarcinoma after resection: patients for whom adjuvant chemotherapy should be performed. Heliyon 2023; 9 (12):e23207. Choe J, Kim MY, Yun JK, et al. Sublobar Resection in Stage IA Non-Small Cell Lung Cancer: Role of Preoperative CT Features in Predicting Pathologic Lymphovascular Invasion and Postoperative Recurrence. AJR Am J Roentgenol 2021; 217 (4):871-81. Zuo Z, Fan X, Tang Y, et al. Deep learning-powered 3D segmentation derives factors associated with lymphovascular invasion and prognosis in clinical T1 stage non-small cell lung cancer. Heliyon 2023; 9 (4):e15147. Wang X, Xie T, Luo J, et al. Radiomics predicts the prognosis of patients with locally advanced breast cancer by reflecting the heterogeneity of tumor cells and the tumor microenvironment. Breast Cancer Res 2022; 24 (1):20. Li J, Qiu Z, Zhang C, et al. ITHscore: comprehensive quantification of intra-tumor heterogeneity in NSCLC by multi-scale radiomic features. Eur Radiol 2023; 33 (2):893-903. Lv K, Cao X, Du P, et al. Radiomics for the detection of microvascular invasion in hepatocellular carcinoma. World J Gastroenterol 2022; 28 (20):2176-83. Ventura D, Schindler P, Masthoff M, et al. Radiomics of Tumor Heterogeneity in (18)F-FDG-PET-CT for Predicting Response to Immune Checkpoint Inhibition in Therapy-Naïve Patients with Advanced Non-Small-Cell Lung Cancer. Cancers (Basel) 2023; 15 (8):2297. Wu L, Lou X, Kong N, et al. Can quantitative peritumoral CT radiomics features predict the prognosis of patients with non-small cell lung cancer? A systematic review. Eur Radiol 2023; 33 (3):2105-17. Wang J, Zheng Z, Zhang Y, et al. (18)F-FDG PET/CT radiomics for prediction of lymphovascular invasion in patients with early stage non-small cell lung cancer. Front Oncol 2023; 13:1185808. Nie P, Yang G, Wang N, et al. Additional value of metabolic parameters to PET/CT-based radiomics nomogram in predicting lymphovascular invasion and outcome in lung adenocarcinoma. Eur J Nucl Med Mol Imaging 2021; 48 (1):217-30. Zhang L, Yankelevitz DF, Henschke CI, et al. Zone of transition: a potential source of error in tumor volume estimation. Radiology 2010; 256 (2):633-39. Wood SL, Pernemalm M, Crosbie PA, et al. The role of the tumor-microenvironment in lung cancer-metastasis and its relationship to potential therapeutic targets. Cancer Treat Rev 2014; 40 (4):558-66. Van Griethuysen JJM, Fedorov A, Parmar C, et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res 2017; 77 (21):e104-e107. Bi H, Yin L, Fang W, et al. Association of CEA, NSE, CYFRA 21-1, SCC-Ag, and ProGRP with Clinicopathological Characteristics and Chemotherapeutic Outcomes of Lung Cancer. Lab Med 2023；54 (4):372-79. Ren L, Yang Y. Value of dual-source CT dual-energy parameters combined with serum detection of VEGF and CEA in the diagnosis of early lung cancer. Biotechnol Genet Eng Rev 2023; 19:1-12. De Jong C, Deneer VHM, Kelder JC, et al. Association between serum biomarkers CEA and LDH and response in advanced non-small cell lung cancer patients treated with platinum-based chemotherapy. Thorac Cancer 2020; 11 (7):1790-800. Li C, Tian Y, Shen Y, et al. Utility of Volumetric Metabolic Parameters on Preoperative FDG PET/CT for Predicting Tumor Lymphovascular Invasion in Non-Small Cell Lung Cancer. AJR Am J Roentgenol 2021; 217 (6):1433-43. Lee JH, Lee SW. The Roles of Carcinoembryonic Antigen in Liver Metastasis and Therapeutic Approaches. Gastroenterol Res Pract 2017; 2017:7521987. Marshall J. Carcinoembryonic antigen-based vaccines. Semin Oncol 2003; 30 (3 Suppl 8):30-6. Shimada Y, Saji H, Kato Y, et al. The Frequency and Prognostic Impact of Pathological Microscopic Vascular Invasion According to Tumor Size in Non-Small Cell Lung Cancer. Chest 149 (3):775-85. Dallavalasa S, Beeraka NM, Basavaraju CG, et al. The Role of Tumor Associated Macrophages (TAMs) in Cancer Progression, Chemoresistance, Angiogenesis and Metastasis - Current Status. Curr Med Chem 2021; 28 (39):8203-823. Banerjee K, Kerzel T, Bekkhus T, et al. VEGF-C-expressing TAMs rewire the metastatic fate of breast cancer cells. Cell Rep 2023; 42 (12):113507. Chen Q, Shao J, Xue T, et al. Intratumoral and peritumoral radiomics nomograms for the preoperative prediction of lymphovascular invasion and overall survival in non-small cell lung cancer. Eur Radiol 2023; 33 (2):947-58. Meng L, Dong D, Chen X, et al. 2D and 3D CT Radiomic Features Performance Comparison in Characterization of Gastric Cancer: A Multi-Center Study. IEEE J Biomed Health Inform 2021; 25 (3):755-63. Zhu Y, Yao W, Xu BC, et al. Predicting response to immunotherapy plus chemotherapy in patients with esophageal squamous cell carcinoma using non-invasive Radiomic biomarkers. BMC cancer 21 (1):1167. Zhang X, Zhang G, Qiu X, et al. Radiomics under 2D regions, 3D regions, and peritumoral regions reveal tumor heterogeneity in non-small cell lung cancer: a multicenter study. Radiol Med 2023; 128 (9):1079-92. Liu C, Zhao W, Xie J, et al. Development and validation of a radiomics-based nomogram for predicting a major pathological response to neoadjuvant immunochemotherapy for patients with potentially resectable non-small cell lung cancer. Front Immunol 2023; 14:1115291. Jiang L, Zhang Z, Guo S, et al. Clinical-Radiomics Nomogram Based on Contrast-Enhanced Ultrasound for Preoperative Prediction of Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma. Cancers (Basel) 2023; 15 (5):1613. Tables Tables 1 to 4 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Tabel1.docx Table2.docx Table3.docx Tabel4.docx Cite Share Download PDF Status: Published Journal Publication published 12 Nov, 2025 Read the published version in BMC Cancer → Version 1 posted Editorial decision: Revision requested 21 Jun, 2025 Reviews received at journal 04 Jun, 2025 Reviewers agreed at journal 22 May, 2025 Reviewers agreed at journal 27 Oct, 2024 Reviews received at journal 21 Oct, 2024 Reviewers agreed at journal 15 Oct, 2024 Reviewers invited by journal 08 Oct, 2024 Editor invited by journal 31 Jul, 2024 Editor assigned by journal 30 Jul, 2024 Submission checks completed at journal 24 Jul, 2024 First submitted to journal 22 Jul, 2024 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-4783280","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":338307517,"identity":"4ed9a1e6-e1d9-44f9-83a1-43400317f907","order_by":0,"name":"Miaomiao LIN","email":"","orcid":"","institution":"First Affiliated Hospital of GuangXi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Miaomiao","middleName":"","lastName":"LIN","suffix":""},{"id":338307518,"identity":"a7617c03-4749-4c4f-872b-c70c47876680","order_by":1,"name":"Chunli Zhao","email":"","orcid":"","institution":"The People's Hospital of Guangxi Zhuang Autonomous Region","correspondingAuthor":false,"prefix":"","firstName":"Chunli","middleName":"","lastName":"Zhao","suffix":""},{"id":338307519,"identity":"b6dceb2c-910d-4962-8a1e-3b5e33d9f773","order_by":2,"name":"haipeng huang","email":"","orcid":"","institution":"The People's Hospital of Guangxi Zhuang Autonomous Region","correspondingAuthor":false,"prefix":"","firstName":"haipeng","middleName":"","lastName":"huang","suffix":""},{"id":338307520,"identity":"49dfcc97-1d41-4146-b077-3b8fbe7ee697","order_by":3,"name":"xiang zhao","email":"","orcid":"","institution":"First Affiliated Hospital of GuangXi Medical University","correspondingAuthor":false,"prefix":"","firstName":"xiang","middleName":"","lastName":"zhao","suffix":""},{"id":338307521,"identity":"17940622-d7e8-4d31-a5f6-238e65e2f6fa","order_by":4,"name":"siyu Yang","email":"","orcid":"","institution":"The People's Hospital of Guangxi Zhuang Autonomous Region","correspondingAuthor":false,"prefix":"","firstName":"siyu","middleName":"","lastName":"Yang","suffix":""},{"id":338307523,"identity":"0ee13928-5442-43d3-a557-15fc745050f8","order_by":5,"name":"xixin He","email":"","orcid":"","institution":"The People's Hospital of Guangxi Zhuang Autonomous Region","correspondingAuthor":false,"prefix":"","firstName":"xixin","middleName":"","lastName":"He","suffix":""},{"id":338307524,"identity":"5ea0cda7-f02e-4849-9cf3-9f7801ea2766","order_by":6,"name":"Kai Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAy0lEQVRIiWNgGAWjYDCCA2DSgkeCgbHxQUKFDVFaGBsYGCRAWpoNHpxJI14LEDKwST5sO0RYB9/x5ucPPu6RkJFsP9xWkcB2gIG/vTsBrxbJM8cMG2c8k+CR5klsu5HAc4dB4szZDXi1GNzIYWzmOSDBI8cA0iLxjMFAIpdYLfwP2woSDA6ToEVaIrGNISGBCC0gv8ycAdQiOeNhs0TCgTQegn4BhtiDDx8O2NhLnE9/+PHnPxs5/vZe/FowAA9pykfBKBgFo2AUYAUAOxVL72gsHWgAAAAASUVORK5CYII=","orcid":"","institution":"First Affiliated Hospital of GuangXi Medical University","correspondingAuthor":true,"prefix":"","firstName":"Kai","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2024-07-22 16:25:50","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4783280/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4783280/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12885-025-15128-2","type":"published","date":"2025-11-12T15:57:02+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":63312491,"identity":"e025f41c-280b-460d-93c9-49a64be55338","added_by":"auto","created_at":"2024-08-26 20:41:00","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":32052,"visible":true,"origin":"","legend":"\u003cp\u003eThe flowchart of inclusion and exclusion criteria of patients.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/3c2daaadaeba3ba0e0264cd2.png"},{"id":63313093,"identity":"3ba564c3-0607-4122-80ad-46c95f4ec57d","added_by":"auto","created_at":"2024-08-26 20:49:01","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":144202,"visible":true,"origin":"","legend":"\u003cp\u003eStudy fowchart of the radiomics analysis.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/a01464267d9d46c7138a554f.png"},{"id":63313088,"identity":"7773111a-92ee-4815-bd89-5b06bf1cf290","added_by":"auto","created_at":"2024-08-26 20:49:00","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":45085,"visible":true,"origin":"","legend":"\u003cp\u003eRetained optimal radiomics features and corresponding coefficients of different models after dimensionality reduction by LASSO regression analysis. A. GT model, B. GPT model, C. PT model.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/54e5cfe0123707c0a94db374.png"},{"id":63313090,"identity":"33ddd35e-17bf-4f5a-a9be-3c2f491ffab1","added_by":"auto","created_at":"2024-08-26 20:49:01","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":38435,"visible":true,"origin":"","legend":"\u003cp\u003eAn example of the nomogram in clinical application. This was an axial non-enhanced chest CT image of a 64-year-old man patient with LUAD. The factors in the nomogram were analyzed as follows:CEA = “positive,”diameter=3.5cm,spiculation=“positive,”GPT-Radscore =1.73. The total score was 89.5 and the probability of developing LVI was above 0.9.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/2e4022e1f36872617894e0f6.png"},{"id":63312499,"identity":"be8548cc-a08e-4c28-a3cf-b316578f8ecd","added_by":"auto","created_at":"2024-08-26 20:41:01","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":74904,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of GPT radiomic models, clinical models, and nomograms for predicting LVI of LUAD in the (A) training, (B) internal validation, and (C) external validation cohorts, respectively.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/a8217124a7be0ce0a76a086e.png"},{"id":63312501,"identity":"cfc23d9e-75f5-425e-85d9-9f095bd39813","added_by":"auto","created_at":"2024-08-26 20:41:01","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":56413,"visible":true,"origin":"","legend":"\u003cp\u003eNomogram risk prediction model calibration curve. A. Training set; B. test set; C. validation set. The solid line represents the ideal predictive performance, and the dashed line represents the predictive performance of the nomogram. The closer the dotted line is to the solid line, the better the prediction accuracy of the nomogram. Calibration analysis showed that the predicted results of early relapse in the three cohorts were agreed with the actual results.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/0021262ca13fe7ad69ca450b.png"},{"id":63313089,"identity":"e35b4585-6dc9-4834-9dc3-32d6e6043e04","added_by":"auto","created_at":"2024-08-26 20:49:01","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":57901,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curve analysis of clinical model,GPT model and nomogram.The abscissa is the thresh old probability; the ordinate is the net beneft minus the harm. The gray curve indicates that all patients received the intervention; the black horizontal line indicates that all patients did not with a net beneft of 0.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/37c458e66c7f5cbfc10ca3b3.png"},{"id":96104955,"identity":"581d8180-1d33-4ea6-9196-94fc4d264d06","added_by":"auto","created_at":"2025-11-17 16:04:25","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1024436,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/393c4bd7-8972-41a8-85d1-9976b848688d.pdf"},{"id":63312492,"identity":"b610c96f-6960-420e-988e-894d861202f6","added_by":"auto","created_at":"2024-08-26 20:41:01","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":19430,"visible":true,"origin":"","legend":"","description":"","filename":"Tabel1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/5c728997a5cce5307fc31b75.docx"},{"id":63312496,"identity":"60d601b9-5a50-43d9-8e94-ef4589643f06","added_by":"auto","created_at":"2024-08-26 20:41:01","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":13835,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/0d3198992496f950f1df5cf2.docx"},{"id":63312494,"identity":"9a2a87ba-4425-4fc9-8aa4-e8fb31e75848","added_by":"auto","created_at":"2024-08-26 20:41:01","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":14895,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/876f4937bc9cb46e067340f5.docx"},{"id":63313092,"identity":"570283d2-7c85-4522-90b7-f12497f7596f","added_by":"auto","created_at":"2024-08-26 20:49:01","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":14804,"visible":true,"origin":"","legend":"","description":"","filename":"Tabel4.docx","url":"https://assets-eu.researchsquare.com/files/rs-4783280/v1/0b6fa56147c89252dbbf6e3b.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"The clinical value of predicting lymphovascular invasion in patients with invasive lung adenocarcinoma based on the intratumoral and peritumoral CT radiomics models","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLung cancer is the global leading cause of cancer-related deaths [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], and adenocarcinoma is the predominant histological type of lung cancer [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Surgical resection is the most common treatment modality for lung cancer, but the 5-year postoperative survival of patients with lung cancer remains suboptimal because of postoperative tumor recurrence and metastasis. Tumor recurrence occurs in \u0026ge;\u0026thinsp;30% of patients, including those with completely resected stage I non-small cell lung cancer (NSCLC) [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Larger tumor size, pleural infiltration, and lymphovascular invasion (LVI)are considered high-risk factors for postoperative recurrence in patients with lung cancer [\u003cspan additionalcitationids=\"CR5 CR6\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. LVI is defined as tumor cell infiltration into the lumen of arteries, veins, or lymphatic vessels [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Cai et al. showed that patients with stage IA and IB NSCLC with LVI had similar survival rates. Therefore, researchers believe that LVI should be an indispensable determinant when deciding the tumor, node, and metastasis (TMN) stage of lung cancer. Consequently, as long as LVI is present in a patient with lung cancer, a patient with stage IA lung cancer should be adjusted to stage IB, showing that LVI is more significant in lung cancer prognosis than tumor size [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Several studies have also confirmed LVI as a pathological factor in the poor prognosis of patients with lung cancer [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Previous studies have shown that adjuvant chemotherapy is beneficial for patients with LVI-positive lung cancer to improve their recurrence-free survival and overall survival rates. Furthermore, LVI can be an indicator for preoperative neoadjuvant chemotherapy in patients with NSCLC [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Therefore, an accurate preoperative prediction of LVI would be useful when screening patients with lung cancer who would benefit from neoadjuvant chemotherapy and in developing an individualized surgical plan. Currently, LVI can only be diagnosed using postoperative histopathologic testing, and this hinders the selection of rational treatment options for patients with lung y tcancer. Therefore, an efficient and noninvasive method of detecting the LVI status in lung cancer is urgently needed.\u003c/p\u003e \u003cp\u003eRecent studies have shown that certain CT features, such as tumor diameter, spiculation, and nodal component are noninvasive imaging biomarkers for predicting LVI in lung cancer [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. However, diagnostic imaging physicians' experience and understanding of different signs deeply influence human-assessed imaging signs. These assessments are highly subjective and poorly reproducible, limiting the usefulness of conventional imaging for the preoperative prediction of LVI in lung cancer. Radiomics can reveal the biological characteristics of tumors at the molecular level [\u003cspan additionalcitationids=\"CR16 CR17 CR18\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Tumor cell growth through the lymphatic and blood vessels results in reduced homogeneity of the intratumoral and peritumoral textures. Therefore, CT imaging can predict LVI status in lung cancer. Currently, there are a few studies on the prediction of LVI in lung adenocarcinoma (LUAD) using the CT radiomics model, and the majority focus on the changes in the tumor, ignoring the potential value of the peritumor microenvironment [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Previous studies on LUAD have shown a range of transition zones between LUAD and normal lung tissues, approximately 1\u0026ndash;5 mm, with an average of 3.5 mm. This transition zone contains much information about lymphatic and vascular infiltration, closely associated with tumor metastasis and patient prognosis [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Therefore, we hypothesized that a peritumor-based radiomics model could effectively predict LVI in patients with invasive LUAD.\u003c/p\u003e \u003cp\u003eThis study aimed to explore a non-invasive method to predict LVI status in LUAD by establish a combined model based on GPT-Radscore and clinical predictors, then visualizes the model with a nomogram.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatients\u003c/h2\u003e \u003cp\u003eThis retrospective study was conducted following the Declaration of Helsinki. It was approved by our hospital\u0026rsquo;s Institutional Review Board (approval number: 2024-E109-01), which waived the requirement for informed consent. Patients with invasive LUAD who visited our hospital between January 2016 and May 2023 were selected as the internal cohort, whereas the external validation cohort was obtained from another hospital. This study included patients with pathology-confirmed LUAD, those with pathological reports of definite LVI status, those who underwent unenhanced chest CT-scan within 15 days preoperatively, those with CT images that aligned with diagnostic requirements, and those with solitary lesions. Patients with incomplete clinicopathological data and preoperative CT images,complications with other tumors in the lung, and recurrence of antitumor treatment preoperatively were excluded from the study (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn total, this study included 635 patients with invasive LUAD. Furthermore, 384 patients with LUAD from our institution were randomized into training and internal validation sets using a ratio of 7:3. In total, 251 patients with invasive LUAD from the other hospital served as the external validation set. Patients\u0026rsquo; clinical information and CT features were collected.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eCT image acquisition\u003c/h2\u003e \u003cp\u003ePatients in institution 1 were examined with General GE 256-row spiral and Siemens Somaton Force dual-source CT machines, whereas patients from Institution 2 were examined with Dutch Philips iCT 256 and Siemens Sensation 64-row CT machines. The patients were instructed to lie in a supine position, head first, and a CT scan of the chest was performed. The parameters used were 120 kV tube voltage, 60\u0026ndash;120 mA or 100\u0026ndash;250 mA tube current, 5 mm layer thickness, 512\u0026times;512 matrix, 0.625\u0026ndash;2.5 mm reconstruction layer thickness, lung window setting 1500 Hu window width, and a window position of -500 Hu.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eRadiomics analysis\u003c/h2\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003eWorkflow\u003c/h2\u003e \u003cp\u003eIn radiomics analysis, the region of interest(ROI) was outlined, the peritumor region expanded, features were selected and extracted, and the model was constructed, analyzed, and evaluated (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e Tumor segmentation and peritumor regional extension\u003c/p\u003e \u003cp\u003eTumor segmentation and peritumoral area expansion were performed using the software package \"Radiomics\" (syngo.via Frontier 1.2.1, version VB10B, Siemens Healthineers, Germany). Lesion segmentation was performed by two radiologists with six (radiologist 1) and 12 (radiologist 2) years of experience diagnosing chest diseases. First, the ROI of the tumor gross (GT) set was outlined along the visible edge of the tumor\u0026rsquo;s largest level, labeled ROI\u003csub\u003eGT\u003c/sub\u003e. ROI\u003csub\u003eGPT\u003c/sub\u003e was obtained by automatically expanding ROI\u003csub\u003eGT\u003c/sub\u003e outward by 5 mm. Pathologists have found that there is a certain range of transition zone between LUAD tissues and normal lung tissue, with an average range of 3.5 mm, and this transition zone contains a large amount of information that is closely related to the prognosis of patients. Therefore, in the present study, the peritumoral outreach distance was set at 5 mm, with a view to encompassing the lung cancer transition zone region as much as possible. Finally, the area of the two ROIs of the gross tumor and peritumor (GPT) and GT sets was subtracted to obtain the ROI of the peritumor (PT) set labeled as ROI\u003csub\u003ePT\u003c/sub\u003e. The outlining process attempted to avoid adjacent structures, such as blood vessels, bronchial tubes, chest wall, and mediastinum.\u003c/p\u003e \u003cp\u003eCT images of 30 patients were randomly selected to assess the radiomics features consistency radiologist 1 and 2 performed the ROI segmentation and radiomics features extraction of the 30 patients, respectively. During the same period, radiologist 1 and 2 performed GT, GPT, and PT ROI outlines to assess inter-observer agreement. Radiologist 1 performed a second outline 1 month later to assess intra-observer agreement. Both radiologists were blinded to the pathological findings. Consistency was evaluated using the intraclass co-rrelation coefficient (ICC). An ICC\u0026thinsp;\u0026gt;\u0026thinsp;0.75 indicated good consistency of the extracted radiomics features, and radiologist 1 performed the segmentation of the remaining cases.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eFeature extraction and screening\u003c/h2\u003e \u003cp\u003eRadiomics features were extracted using the same software package, \"Radiomics\". The images were resampled to ensure uniformity in image voxels size, and all image voxels were 1 mm \u0026times; 1 mm \u0026times;1 mm. In total, 1226 radiomics features were extracted from the three ROIs of GT, GPT, and PT, including the first-order, texture, higher-order, and other features obtained by constant transformations. Based on previous studies [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e], the extracted radiomics features were defined according to the Imaging Biomarker Standardization Initiative. First, redundant features were eliminated using minimum redundancy and maximum relevance (mRMR). Then, the least absolute shrinkage and selection operator (LASSO) algorithm was used to reduce the dimensionality of the remaining features and construct the nomogram. The LASSO algorithm adjusts the value of the weight parameter (λ) to suppress the coefficients of some features with less importance to zero. This helps to rapidly screen the features by reducing computational volume. The optimal radiomics features were selected from the screened features to construct a radiomics model and calculate the Radscore.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eModele constrution and analysis\u003c/h2\u003e \u003cp\u003eAn radiomics model was constructed using a logistic regression classifier. Clinically independent predictors were screened sequentially by using univariate and multivariate logistic regression algorithm, which combined optimal Radscore to build a combined predictive mode. The combined model was visualized with the nomogram. The predictive efficacy of the model was assessed by using the area under curve (AUC), sensitivity, specificity, and accuracy. The statistic difference in AUCs among the models were evaluated using the Delong test, the nomogram predictive accuracy was assessed by using calibration curves, and the clinical utility of the predictive model was evaluated with the decision curve analysis(DCA).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eThe R software package (version 4.2.2, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.Rproject.org\u003c/span\u003e\u003cspan address=\"http://www.Rproject.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used for data processing and constructing models. Univariate and multivariate logistic regression algorithm were used to determine independent predictors of LVI in patients with LUAD (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The AUC was calculated to evaluate the diagnostic performance of each model, and the Delong test was used to assess differences in the AUC among models.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003ePatient characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTable 1 shows the baseline data\u0026nbsp;of patients.Multivariate logistic regression\u0026nbsp;analysis showed that preoperative carcinoembryonic antigen (CEA)\u0026nbsp;[\u0026nbsp;OR, 2.24; 95% CI:1.21–4.17],\u0026nbsp;spiculation\u0026nbsp;[OR, 3.06; 95% CI:1.74–5.37]\u0026nbsp;and tumor diameter\u0026nbsp;[OR, 1.97; 95% CI:1.50–2.60)]were independent predictors of LVI in patients with invasive LUAD.\u0026nbsp;A\u0026nbsp;clinical model were established by combining the independent predictors above. Table 2 shows the results of the\u0026nbsp;univariate and multivariate logistic regression analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOptimal r\u003c/strong\u003e\u003cstrong\u003eadiomics f\u003c/strong\u003e\u003cstrong\u003eeature\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;screening and mode\u003c/strong\u003e\u003cstrong\u003els constructing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn total, 1226 radiomics\u0026nbsp;features were extracted from the GT, GPT, and PT ROIs. Redundant features were eliminated by using mRMR and LASSO algorithm. The GT, GPT, and PT models yielded the\u0026nbsp;3,\u0026nbsp;7, and\u0026nbsp;8\u0026nbsp;best features for constructing the radiomics models finally\u0026nbsp;(Fig. 3).\u003c/p\u003e\n\u003cp\u003eRadiomics models assessment\u003c/p\u003e\n\u003cp\u003eOf the three\u0026nbsp;radiomics\u0026nbsp;models, the AUCs of the GPT model (0.83, 0.79, and 0.75),were higher than those of the GT (AUC: 0.79, 0.76, and 0.72) and PT (AUC: 0.80, 0.75, and 0.73) models. However, Delong's test showed no statistically significant differences in the AUC values of the three models (\u003cem\u003eP\u0026nbsp;\u003c/em\u003e\u0026gt; 0.05). In addition, we comprehensively evaluated the radiomics model’s accuracy, sensitivity, and specificity (Table 3), and the GPT model had a more stable performance. Therefore, we used the radiomics\u0026nbsp;features of GPT and independent clinical predictors to construct a\u0026nbsp;nomogram.\u003c/p\u003e\n\u003cp\u003eRadiomics nomogram\u0026nbsp;construction\u003c/p\u003e\n\u003cp\u003eThe combined model was constructed using the GPT-Radscore, preoperative CEA level,\u0026nbsp;spiculation, and tumor diameter，and then \u0026nbsp; visualizes the model with the nomogram. It was constructed based on the training cohort and validated using the internal and external validation cohorts. Fig. 4 illustrates the clinical application of the\u0026nbsp;nomogram. Fig. 5 shows the receiver operating characteristic curve(ROC) of the GPT radiomics model, the clinical model, and the\u0026nbsp;nomogram. The predictive efficacy of clinical, GPT radiomics model and normogram are shown in Table 4. The Delong test showed that in the training cohort, the AUC values of the\u0026nbsp;nomogram\u0026nbsp;were significantly different from the clinical model (\u003cem\u003eP\u0026nbsp;\u003c/em\u003e\u0026lt; 0.05). In the internal validation cohort, the AUC of the\u0026nbsp;nomogram\u0026nbsp;was significantly different from\u0026nbsp;the\u0026nbsp;radiomics\u0026nbsp;model (\u003cem\u003eP\u0026nbsp;\u003c/em\u003e\u0026lt; 0.05). The calibration curve showed that the LVI incidence predicted by the nomogram corresponds well with the actual LVI incidence (Fig. 6). The Hosmer-Lemeshow(H-L)\u0026nbsp;test\u0026nbsp;showed\u0026nbsp;that the nomogram fit well\u0026nbsp;(\u003cem\u003eP\u0026nbsp;\u003c/em\u003e\u0026gt; 0.05). The DCA showed that the nomogram achieved a better net benefit than the clinical model and GPT radiomics in predicting LVI (Fig. 7).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eLVI is the first step in inducing tumor spread or recurrence. From a clinical perspective, accurate preoperative prediction of LVI in patients with LUAD is crucial for selecting an appropriate treatment regimen. This study demonstrated that GPT, a radiomics \u0026nbsp;model constructed by combining peritumoral 5 mm features, was superior to GT and PT models in diagnostic performance, confirming the potential value of the peritumoral microenvironment in predicting LVI in patients with invasive LUAD. This study constructed a nomogram using the GPT-Radscore, preoperative CEA level, tumor diameter, and spiculation. The results showed that the radiomics nomogram contributed to the predictive efficacy, with an AUC of 0.84 (95% CI: 0.79-0.89), 0.82 (95% CI: 0.74-0.90), and 0.77 (95% CI: 0.70-0.84) in the training, internal validation and external validation sets respectively. Therefore, the nomogram established in this study can effectively individualize the prediction of LVI status in patients with LUAD.\u003c/p\u003e\n\u003cp\u003eCEA is an acidic glycoprotein with human embryonic antigenic properties and is crucial in the diagnosis and prognosis of lung cancer[25-27]. This study found that high preoperative CEA level was an independent risk factor for LVI-positive in LUAD, and the higher the CEA level, the higher the chance of LVI in postoperative pathology. These findings are similar to that of Li et al.\u0026nbsp;[28]. Previous studies have reported that CEA is a cell adhesion molecule that performs adhesion reactions between cancer cells and stromal collagen and is vital in tumor growth and metastasis[29,30].\u0026nbsp;Shimada et al.[31]in a study of the relationship between lesion size and LVI in patients with pT1-4N0-2M0 stage NSCLC, found that the probability of LVI varied based on the tumor size. In patients with a maximum diameter of ≤ 5 cm, the larger the tumor’s maximum diameter, the higher the probability of LVI. This study also showed a positive correlation between LVI appearance and tumor size. Tumor size is crucial when determining patients' TMN and affects their prognosis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFurthermore, previous studies have also proved that\u0026nbsp;lymphovascular invasion\u0026nbsp;is a critical factor that promotes the rise of T in TMN staging. Therefore, it can be speculated that tumor diameter and\u0026nbsp;lymphovascular invasion\u0026nbsp;directly affect each other. They can both accelerate the progression of malignant tumors and affect patients’ prognosis. It is generally believed that the pathological basis for the formation of spiculations in malignant tumors is the infiltration and proliferation of tumor cells along blood and lymphatic vessels to the adjacent mesenchyme, which shows short and dense features in CT morphology. Therefore, attention should be paid to the appearance of LVI in patients with spiculation-positive lung cancer.\u003c/p\u003e\n\u003cp\u003eMetastasis is a major cause of death in patients with cancer. Immune cells, such as tumor-associated macrophages in the tumor microenvironment, are involved in the production of immunosuppressive TME in tumors through the production of inflammatory mediators and growth factors, which affect the angiogenic and metastatic behaviors of the tumor cells\u0026nbsp;[32,33]. Lung tissues that appear normal on CT are often invaded by tumor cells, and the invaded areas, when in a transition state, cannot be visualized using imaging systems. Therefore, changes in tumor margins or adjacent tissues on patho-histology can help optimize the predictive performance of radiomics models. Recently, Chen et al.\u0026nbsp;[34], for the first time, constructed a model to predict LVI in patients with NSCLC based on peritumor radiomics features. The results showed that the GPTV\u003csub\u003e9\u003c/sub\u003e model, which included intratumor and peritumor 9 mm regions, had the best prediction performance, proving the radiomics model’s value containing PT V information in disease diagnosis. The transition zone of LUAD has been reported in the literature as 2–5 mm, and this region contains a great deal of information associated with lung cancer’s aggressive biological behavior. Therefore, this study’s maximum value of the peritumor region was placed at 5 mm, expecting the transition zone to be included as much as possible. The findings obtained were similar to those by Chen et al. and confirmed the feasibility of the peritumor microenvironmental radiomics model for the preoperative prediction of LVI in LUAD.\u003c/p\u003e\n\u003cp\u003eIn radiomics, the choice of the largest level of a lesion or the entire volume for ROI outlining has been debated\u0026nbsp;[35,36]. In fact, it has not been definitively verified whether three-dimensional (3D)-based models are necessarily superior to the two-dimensional (2D) models in lung cancer diagnostic and treatment applications. Yang et al.\u0026nbsp;[8].\u0026nbsp;compared the performance of 2D and 3D radiomics models in predicting LVI in lung cancer and other cancers, and their results showed that the 2D model was more effective than the 3D model. Nie et al.\u0026nbsp;[21]\u0026nbsp;reported similar results. Further, 2D features are easier to obtain than 3D features, are more reproducible, and have certain advantages in lung cancer research. Therefore, this study performed lesion ROI outline from the 2D level and obtained a good model prediction performance. For above phenomenon about lung cancer with radiomics, Zhang et al. [37] suggested that the 3D radiomics model is not as effective as the 2D model, because the 3D ROI labeling process generates more \"noise,” which interferes with valid information data, thus creating a bias in the experimental results. There are two sources of noise: first, there are certain differences between different operators in recognizing lesions and lesion boundaries, which the 3D ROI labeling process will amplify; second, the thickness of the image layer will affect the \u003cu\u003eamount of noise\u003c/u\u003e. The thicker the image layer, the less noise there is, and vice versa. On the Z-axis (thickness direction), the voxel spacing is usually inconsistent with the X- and Y-axes (transverse plane), which may affect the accuracy of the 3D image standard. The 3D image labeling process is far more affected by unclear image boundaries and thickness than 2D images, resulting in less effective data information being obtained from 3D images. This leads to a less effective model than that obtained via a 2D model, although the 3D image provides more information. In future research, it is necessary to improve image histology feature extraction and computation methods to determine the best research method.\u003c/p\u003e\n\u003cp\u003eIn addition to the three radiomics models of GT, GPT, and PT, the study also combined age, preoperative CEA level, spiculation, and GPT-Radscore to construct a combined prediction model, and the predictive efficacy of the combined model was improved compared with that of the clinical and radiomics models alone. By integrating the advantages of the clinical and radiomics models, the combined model provides a more stable and reliable disease diagnosis and treatment model.\u0026nbsp;[38,39].\u0026nbsp;Therefore, a model integrating clinical information and radiomics features is more valuable in predicting LVI in patients with invasive LUAD. In addition, the combined model was visualized using nomogram. A nomogram is a statistical model for individualized predictive analysis of clinical events, quantifying the risk of clinical events according to various risk factors supporting the development of treatment plans.\u003c/p\u003e\n\u003cp\u003eThis study has some limitations. First, it is a retrospective study that could not standardize the machine, scanning, and other relevant parameters. Second, previous studies have found that micropapillary and solid LUAD have the highest chance of developing LVI, which shows that the pathological subtypes of invasive LUAD are inextricably linked to the occurrence of LVI. However, the present study did not further subdivide the pathological subtypes of invasive LUAD. To achieve more precise treatment, subsequent studies can further explore the correlation between different pathological subtypes of invasive LUAD and LVI. Third, the present study did not further investigate the effect of LVI on the prognosis of LUAD, which is also a matter of great clinical concern.\u003c/p\u003e\n\u003cp\u003eIn conclusion, the GPT radiomics model performed better than the GT and PT radiomics models in predicting the LVI status in patients with invasive LUAD. In addition, the combined model based on GPT radiomics features and clinically independent predictors can further improve prediction efficiency and provide an objective and quantitative reference basis for clinicians to make decisions and assess patient prognosis.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eLVI \u0026nbsp; \u0026nbsp; \u0026nbsp;Lymphovascular Invasion\u003c/p\u003e\n\u003cp\u003eLUAD \u0026nbsp; Lung Adenocarcinoma\u003c/p\u003e\n\u003cp\u003eGT \u0026nbsp; \u0026nbsp; Tumor Gross\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGPT \u0026nbsp; \u0026nbsp;Gross Tumor and Peritumor\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePT \u0026nbsp; \u0026nbsp;Peritumor\u003c/p\u003e\n\u003cp\u003eAUC \u0026nbsp;Area Under the Curve\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCEA \u0026nbsp; Carcinoembryonic Antigen\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNSCLC \u0026nbsp; Non-Small Cell Lung Cancer\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTMN \u0026nbsp; \u0026nbsp; Tumor, Node, and Metastasis\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eROI \u0026nbsp; \u0026nbsp; Region of Interest\u003c/p\u003e\n\u003cp\u003eICC \u0026nbsp; Intraclass Co-rrelation Coefficient\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMrmr \u0026nbsp;Minimum Redundancy and Maximum Relevance\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eLasso \u0026nbsp;Least Absolute Shrinkage and Selection Operator\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDCA \u0026nbsp; Decision Curve Analysis\u003c/p\u003e\n\u003cp\u003eROC \u0026nbsp; Receiver Operating Characteristic Curve\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to thank doctor Liqun Hu, from the Department of Pathology,\u0026nbsp;The People\u0026rsquo;s Hospital of Guangxi Zhuang Autonomous Region, Guangxi Academy of Medical Sciences, for her guidance in LVI evaluation during this research.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: MML, KL. Methodology: MML, HPH, XZ,XXH,SYY,CLZ,KL. Validation:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHHP,XZ,SYY. Formal analysis: MML, HPH, XZ,XXH. Data curation: MML, HPH, XZ,XXH,SYY,CLZ. Supervision: KL. Writing-original draft:MML,CLZ. Writing-review \u0026amp; editing: HHP,XZ,SYY. Formal analysis: MML, HPH, XZ,XXH. All authors read and approved the\u0026nbsp;final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and meterials\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo datasets were generated or analysed during the current study.\u003c/p\u003e\n\u003cp\u003eEthics approval and consent to participate\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe study was approved by the The First Affiliated Hospital of Guangxi Medical University (Approval number: 2024-E109-01) with a waived requirement for informed consent (retrospective design).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur research contains no personal data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors report no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021; 71 (3):209-49. \u003c/li\u003e\n\u003cli\u003eSiegel RL, Miller KD, Wagle NS, et al. Cancer statistics, 2023. CA Cancer J Clin 2023; 73 (1):17-48. \u003c/li\u003e\n\u003cli\u003eSung SY, Kwak YK, Lee SW, et al. Lymphovascular Invasion Increases the Risk of Nodal and Distant Recurrence in Node-Negative Stage I-IIA Non-Small-Cell Lung Cancer. Oncology 2018; 95 (3):156-62. \u003c/li\u003e\n\u003cli\u003eYun JK, Lee GD, Choi S, et al. Comparison of prognostic impact of lymphovascular invasion in stage IA non-small cell lung cancer after lobectomy versus sublobar resection: A propensity score-matched analysis. Lung cancer (Amsterdam, Netherlands) 2020; 146:105-11. \u003c/li\u003e\n\u003cli\u003eYang H, Mei T. Prognostic significance of visceral pleural invasion in patients with surgically resected small-cell lung cancer: a population-based study. Jpn J Clin Oncol 2022; 52 (9):1045-55. \u003c/li\u003e\n\u003cli\u003eDe Giglio A, Di Federico A, Gelsomino F, et al. Prognostic relevance of pleural invasion for resected NSCLC patients undergoing adjuvant treatments: A propensity score-matched analysis of SEER database. Lung cancer (Amsterdam, Netherlands) 2021; 161:18-25. \u003c/li\u003e\n\u003cli\u003ePathak R, Goldberg SB, Canavan M, et al. Association of Survival With Adjuvant Chemotherapy Among Patients With Early-Stage Non-Small Cell Lung Cancer With vs Without High-Risk Clinicopathologic Features. JAMA Oncol 2020; 6 (11):1741-50.\u003c/li\u003e\n\u003cli\u003eYang G, Nie P, Zhao L, et al. 2D and 3D texture analysis to predict lymphovascular invasion in lung adenocarcinoma. Eur J Radiol 2020; 129:109111. \u003c/li\u003e\n\u003cli\u003eCai JS, Wang X, Yang F, et al. Lymphovascular invasion: A non-sized T descriptor for stage IA non-small cell lung cancer. Thorac Cancer 2022; 13 (17):2413-20.\u003c/li\u003e\n\u003cli\u003eGuglielmetti LC, Schneiter D, Hillinger S, et al. Lymphovascular invasion is an independent prognostic factor for survival in pathologically proven N2 non-small cell lung cancer. Swiss Med Wkly 2021; 151:w20385. \u003c/li\u003e\n\u003cli\u003eSuaiti L, Sullivan TB, Rieger-Christ KM, et al. Vascular Invasion Predicts Recurrence in Stage IA2-IB Lung Adenocarcinoma but not Squamous Cell Carcinoma. Clin Lung Cancer 2023; 24 (3):e126-e133. \u003c/li\u003e\n\u003cli\u003eWu LL, Jiang WM, Qian JY, et al. High-risk characteristics of pathological stage I lung adenocarcinoma after resection: patients for whom adjuvant chemotherapy should be performed. Heliyon 2023; 9 (12):e23207. \u003c/li\u003e\n\u003cli\u003eChoe J, Kim MY, Yun JK, et al. Sublobar Resection in Stage IA Non-Small Cell Lung Cancer: Role of Preoperative CT Features in Predicting Pathologic Lymphovascular Invasion and Postoperative Recurrence. AJR Am J Roentgenol 2021; 217 (4):871-81. \u003c/li\u003e\n\u003cli\u003eZuo Z, Fan X, Tang Y, et al. Deep learning-powered 3D segmentation derives factors associated with lymphovascular invasion and prognosis in clinical T1 stage non-small cell lung cancer. Heliyon 2023; 9 (4):e15147. \u003c/li\u003e\n\u003cli\u003eWang X, Xie T, Luo J, et al. Radiomics predicts the prognosis of patients with locally advanced breast cancer by reflecting the heterogeneity of tumor cells and the tumor microenvironment. Breast Cancer Res 2022; 24 (1):20. \u003c/li\u003e\n\u003cli\u003eLi J, Qiu Z, Zhang C, et al. ITHscore: comprehensive quantification of intra-tumor heterogeneity in NSCLC by multi-scale radiomic features. Eur Radiol 2023; 33 (2):893-903. \u003c/li\u003e\n\u003cli\u003eLv K, Cao X, Du P, et al. Radiomics for the detection of microvascular invasion in hepatocellular carcinoma. World J Gastroenterol 2022; 28 (20):2176-83.\u003c/li\u003e\n\u003cli\u003eVentura D, Schindler P, Masthoff M, et al. Radiomics of Tumor Heterogeneity in (18)F-FDG-PET-CT for Predicting Response to Immune Checkpoint Inhibition in Therapy-Na\u0026iuml;ve Patients with Advanced Non-Small-Cell Lung Cancer. Cancers (Basel) 2023; 15 (8):2297. \u003c/li\u003e\n\u003cli\u003eWu L, Lou X, Kong N, et al. Can quantitative peritumoral CT radiomics features predict the prognosis of patients with non-small cell lung cancer? A systematic review. Eur Radiol 2023; 33 (3):2105-17. \u003c/li\u003e\n\u003cli\u003eWang J, Zheng Z, Zhang Y, et al. (18)F-FDG PET/CT radiomics for prediction of lymphovascular invasion in patients with early stage non-small cell lung cancer. Front Oncol 2023; 13:1185808. \u003c/li\u003e\n\u003cli\u003eNie P, Yang G, Wang N, et al. Additional value of metabolic parameters to PET/CT-based radiomics nomogram in predicting lymphovascular invasion and outcome in lung adenocarcinoma. Eur J Nucl Med Mol Imaging 2021; 48 (1):217-30. \u003c/li\u003e\n\u003cli\u003eZhang L, Yankelevitz DF, Henschke CI, et al. Zone of transition: a potential source of error in tumor volume estimation. Radiology 2010; 256 (2):633-39. \u003c/li\u003e\n\u003cli\u003eWood SL, Pernemalm M, Crosbie PA, et al. The role of the tumor-microenvironment in lung cancer-metastasis and its relationship to potential therapeutic targets. Cancer Treat Rev 2014; 40 (4):558-66. \u003c/li\u003e\n\u003cli\u003eVan Griethuysen JJM, Fedorov A, Parmar C, et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res 2017; 77 (21):e104-e107.\u003c/li\u003e\n\u003cli\u003eBi H, Yin L, Fang W, et al. Association of CEA, NSE, CYFRA 21-1, SCC-Ag, and ProGRP with Clinicopathological Characteristics and Chemotherapeutic Outcomes of Lung Cancer. Lab Med 2023；54 (4):372-79. \u003c/li\u003e\n\u003cli\u003eRen L, Yang Y. Value of dual-source CT dual-energy parameters combined with serum detection of VEGF and CEA in the diagnosis of early lung cancer. Biotechnol Genet Eng Rev 2023; 19:1-12. \u003c/li\u003e\n\u003cli\u003eDe Jong C, Deneer VHM, Kelder JC, et al. Association between serum biomarkers CEA and LDH and response in advanced non-small cell lung cancer patients treated with platinum-based chemotherapy. Thorac Cancer 2020; 11 (7):1790-800. \u003c/li\u003e\n\u003cli\u003eLi C, Tian Y, Shen Y, et al. Utility of Volumetric Metabolic Parameters on Preoperative FDG PET/CT for Predicting Tumor Lymphovascular Invasion in Non-Small Cell Lung Cancer. AJR Am J Roentgenol 2021; 217 (6):1433-43. \u003c/li\u003e\n\u003cli\u003eLee JH, Lee SW. The Roles of Carcinoembryonic Antigen in Liver Metastasis and Therapeutic Approaches. Gastroenterol Res Pract 2017; 2017:7521987. \u003c/li\u003e\n\u003cli\u003eMarshall J. Carcinoembryonic antigen-based vaccines. Semin Oncol 2003; 30 (3 Suppl 8):30-6. \u003c/li\u003e\n\u003cli\u003eShimada Y, Saji H, Kato Y, et al. The Frequency and Prognostic Impact of Pathological Microscopic Vascular Invasion According to Tumor Size in Non-Small Cell Lung Cancer. Chest 149 (3):775-85. \u003c/li\u003e\n\u003cli\u003eDallavalasa S, Beeraka NM, Basavaraju CG, et al. The Role of Tumor Associated Macrophages (TAMs) in Cancer Progression, Chemoresistance, Angiogenesis and Metastasis - Current Status. Curr Med Chem 2021; 28 (39):8203-823. \u003c/li\u003e\n\u003cli\u003eBanerjee K, Kerzel T, Bekkhus T, et al. VEGF-C-expressing TAMs rewire the metastatic fate of breast cancer cells. Cell Rep 2023; 42 (12):113507. \u003c/li\u003e\n\u003cli\u003eChen Q, Shao J, Xue T, et al. Intratumoral and peritumoral radiomics nomograms for the preoperative prediction of lymphovascular invasion and overall survival in non-small cell lung cancer. Eur Radiol 2023; 33 (2):947-58. \u003c/li\u003e\n\u003cli\u003eMeng L, Dong D, Chen X, et al. 2D and 3D CT Radiomic Features Performance Comparison in Characterization of Gastric Cancer: A Multi-Center Study. IEEE J Biomed Health Inform 2021; 25 (3):755-63. \u003c/li\u003e\n\u003cli\u003eZhu Y, Yao W, Xu BC, et al. Predicting response to immunotherapy plus chemotherapy in patients with esophageal squamous cell carcinoma using non-invasive Radiomic biomarkers. BMC cancer 21 (1):1167. \u003c/li\u003e\n\u003cli\u003eZhang X, Zhang G, Qiu X, et al. Radiomics under 2D regions, 3D regions, and peritumoral regions reveal tumor heterogeneity in non-small cell lung cancer: a multicenter study. Radiol Med 2023; 128 (9):1079-92.\u003c/li\u003e\n\u003cli\u003eLiu C, Zhao W, Xie J, et al. Development and validation of a radiomics-based nomogram for predicting a major pathological response to neoadjuvant immunochemotherapy for patients with potentially resectable non-small cell lung cancer. Front Immunol 2023; 14:1115291. \u003c/li\u003e\n\u003cli\u003eJiang L, Zhang Z, Guo S, et al. Clinical-Radiomics Nomogram Based on Contrast-Enhanced Ultrasound for Preoperative Prediction of Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma. Cancers (Basel) 2023; 15 (5):1613. \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 4 are available in the Supplementary Files section.\u003c/p\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-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Invasive lung adenocarcinoma, Computed tomography, Radiomics, Lymphovascular invasion","lastPublishedDoi":"10.21203/rs.3.rs-4783280/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4783280/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose: \u003c/strong\u003eTo investigate the clinical value of predicting lymphovascular invasion(LVI) in patients with invasive lung adenocarcinoma(LUAD)based on the intratumoral and peritumoral CT radiomics models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMaterials and Methods: \u003c/strong\u003eThe\u003cstrong\u003e \u003c/strong\u003e384 patients with invasive LUAD from Institution 1 were randomly divided into training (n=268) and internal validation (n=116) sets with a ratio of 7:3, and 251 patients from Institution 2 were used as the external validation set. Altogether, 1226 features were extracted from the tumor gross (GT), gross tumor and peritumor (GPT), and peritumor(PT), respectively. Clinical independent predictors for LVI in patients with invasive LUAD were screened using univariate and multivariate logistic regression, a combined model that included clinical predictors and optimal Rad-score was constructed , and a nomogram was drawn.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e The GPT model showed better predictive efficacy than the GT and PT models, with the area under the curve (AUC) of 0.83, 0.79, and 0.75 in the training, internal validation, and external validation sets, respectively. In the clinical model, the preoperative carcinoembryonic antigen (CEA) level, tumor diameter, and spiculation were the independent predictors. The combined model containing the independent predictors and the GPT-Radscore significantly predicted LVI in patients with invasive LUAD, with AUCs of 0.84, 0.82, and 0.77 in the three cohorts, respectively.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e The CT scan-based radiomics model which including intratumoral and peritumoral radiomics features can effectively predict LVI in LUAD，and the predictive efficacy is further improved by combining clinically independent predictors.\u003c/p\u003e","manuscriptTitle":"The clinical value of predicting lymphovascular invasion in patients with invasive lung adenocarcinoma based on the intratumoral and peritumoral CT radiomics models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-08-26 20:40:56","doi":"10.21203/rs.3.rs-4783280/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-21T06:42:42+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-04T13:30:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"81912269874204221927656585530688268482","date":"2025-05-22T13:52:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"114621910198272654005235961257056092305","date":"2024-10-27T14:18:12+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-10-21T14:58:16+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"338022773368545004846202030173053212279","date":"2024-10-15T09:11:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-10-08T13:37:52+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-07-31T19:03:39+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-07-30T11:20:45+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-07-24T13:50:43+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Cancer","date":"2024-07-22T16:24:27+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f37fdff0-4de1-4bcf-87a6-30b8161ca242","owner":[],"postedDate":"August 26th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-11-17T15:59:31+00:00","versionOfRecord":{"articleIdentity":"rs-4783280","link":"https://doi.org/10.1186/s12885-025-15128-2","journal":{"identity":"bmc-cancer","isVorOnly":false,"title":"BMC Cancer"},"publishedOn":"2025-11-12 15:57:02","publishedOnDateReadable":"November 12th, 2025"},"versionCreatedAt":"2024-08-26 20:40:56","video":"","vorDoi":"10.1186/s12885-025-15128-2","vorDoiUrl":"https://doi.org/10.1186/s12885-025-15128-2","workflowStages":[]},"version":"v1","identity":"rs-4783280","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4783280","identity":"rs-4783280","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}