United Predictability of CT radiomics on invasive pathological features in clinical stage IA-IIA non-small cell lung cancer: a double-center study

United Predictability of CT radiomics on invasive pathological features in clinical stage IA-IIA non-small cell lung cancer: a double-center study | 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 United Predictability of CT radiomics on invasive pathological features in clinical stage IA-IIA non-small cell lung cancer: a double-center study Fengnian Zhao, Wang Jiang, Xiaoxue Wang, Yunqing Zhao, Qingna yan, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4488259/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objectives Limited surgery has received increasing attention to minimize damage and preserve more functional lung tissue. However, invasive pathological features including occult lymph node metastasis, visceral pleural invasion, lymphovascular invasion and tumor spread through air spaces may become risk factors for prognosis after limited surgery. The aim of this study was to unitedly predict these invasive pathological features based on computed tomography (CT) radiomics in patients with early stage non-small cell lung cancer (NSCLC). Methods From January 2016 to February 2023, 910 patients with clinical stage IA-IIA NSCLC underwent resection and were divided into training and validation group based on different institution. Radiomics features were extracted by the PyRadiomics software after tumor lesion segmentation and screened by spearman correlation analysis, minimum redundancy maximum relevance and the least absolute shrinkage and selection operator regression analysis. Univariate analysis followed by multivariable logistic regression were performed to estimate the independent predictors. A predictive model was established with visual nomogram and external validation, and evaluated in terms of area under the receiver operating characteristic curve (AUC). Results 225 patients had invasive pathological features (33.2%), and four independent predictors were identified: larger consolidation diameter (p = 0.032), pleural attachment (p = 0.013), texture (p < 0.001) and Rad-score (p < 0.001). The combined model showed good calibration with an AUC of 0.815, compared with 0.778 and 0.691 when radiomics or traditional CT features were used alone. For the validation group, the AUC was 0.792, compared with 0.745 and 0.701 in radiomics or traditional CT features model. Conclusion Our predictive model can non-invasively assess the risk of invasive pathological features in patients with clinical stage IA-IIA NSCLC, enable surgeons perform more reasonable and individualized treatment choices. Computed Tomography Radiomics Invasive Pathological Features Non-small Cell Lung Cancer Predictive Model Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Highlights 1. We discovered that the traditional CT features, including larger consolidation diameter, pleural attachment, solid texture and Rad-score were the independent predictors of invasive pathological features. 2. Radiomics could improve the predictive performance of models, the AUC significantly increased in combined model, compared with that when traditional clinicoradiological feature was used alone. 1. Introduction Lung cancer is the leading cause of cancer-related mortality worldwide, of which approximately 80-85% is non-small cell lung cancer (NSCLC). The standard treatment strategy for early NSCLC is lobectomy combined with systematic lymph node (LN) dissection to obtain the best chance of cure. In the past decade, with the promotion of early screening for lung cancer and the gradual maturity of low-dose thin-layer computed tomography (CT) technology, the proportion of early and small volume lung cancer has also increased. Hence, there has been a trend in surgical practice towards employing limited procedures such as sublobectomy (segmentectomy or wedge resection) and selective lymph node dissection to minimize tissue damage and preserve functional lung tissue. Multiple studies have indicated that there is no significant disparity in perioperative safety or long-term survival outcomes between lobectomy and sublobar resection for early NSCLC [4-6] . Nevertheless, invasive pathological characteristics including occult lymph node metastasis (OLM), visceral pleural invasion (VPI), lymphovascular invasion (LVI) and tumor spread through air spaces (STAS) may serve as potential risk factors for local recurrence and distant metastasis following limited surgery for early NSCLC [7-10] . Accurate preoperative identification of early-stage NSCLC patients remains a crucial and challenging task. This underscores the importance of precise preoperative prediction utilizing clinical and radiological features to identify suitable candidates for limited surgery. 18 F-fludeoxyglucose positron emission computed tomography (PET-CT) demonstrates high sensitivity and specificity in LN staging and assessment of tumor invasiveness [11] , due to the reliable reproducibility of maximum standardized uptake values (SUV max ), which reflect the metabolic activity and invasiveness of the tumor. Previous research has indicated that SUV max serves as a predictive factor for invasive pathological characteristics [12-14] . Despite its efficacy, the widespread adoption of SUV max in preoperative assessments for early NSCLC is hindered by its high cost. Alternative studies have utilized CT imaging to make preoperative predictions, highlighting the significance of primary tumor characteristics as independent predictors of invasive pathological features [15-20] . However, the predictive accuracy of traditional CT features alone remains limited. Radiomics, a method that involves converting images into diggable data through machine-learning methods [21] , offers quantifiable and objective characteristics that may enhance predictive capabilities in medical imaging analysis [22] . Prior researches have examined the radiomics features of primary tumors on CT or PET-CT scans and found that they can predict invasive pathological features [23-26] . However, the vast majority of studies have focused solely on a single risk factor, lacking clinical practicality. In this study, the authors conducted a retrospective, dual-center clinical study to noninvasively and collectively predict invasive pathological features in NSCLC patients with clinical stage IA to IIA using radiomics analysis of CT scans. This information aims to assist surgeons and patients in selecting appropriate treatment strategies and surgical option. 2. Materials and methods 2.1. Patients A total of 1080 patients were initially enrolled in the study, with surgery dates ranging from January 2016 to February 2023, based on specific inclusion and exclusion criteria outlined in Fig. 1 . Ultimately, 910 patients were included in the analysis, with 678 from ( ) assigned to the training group and 232 from ( ) to the external validation group. Clinical data, such as age, sex, smoking status, family history, genetic mutation status, and tumor markers, were collected from the clinical database. NSCLC was categorized according to the 2015 WHO classification system [ 27 ] , tumor lymph nodes metastasis (TNM) classification and tumor staging were performed according to the 8th edition of the staging system published by the Union for International Cancer Control and the American Joint Committee on Cancer [ 28 ] . Because this was a retrospective nonintervention study, approval of the Medical Research Ethics Committee and Institutional Review Board was waived. 2.2. Definition of OLM, VPI, LVI and STAS All enrolled patients underwent preoperative chest contrast-enhanced CT scanning to evaluate the status of LN. The criteria for determining cN0 on CT included all LNs having a short-axis diameter of less than 10 mm without obvious heterogeneous enhancement. Invasive pathological features were obtained from the pathological database and reviewed by a senior pathologist who was unaware of the patients' clinical and radiological outcomes. Information on dissected LN from pathological reports was collected, with involvement of either hilar or mediastinal LN identified as OLM. At least 10 regional LNs were requested to be removed and pathologically examined as American College of Surgeons Commission on Cancer recommended [ 29 ] . VPI was denoted as the invasion of tumor reaches beyond the pleural elastic layer or the surface of visceral pleura. LVI was defined as the infiltration of tumour cells into lymphatic, arterial or venous lumens at the periphery of carcinoma. STAS was considered to exist when the micropapillary clusters, solid nests, or single cells beyond the edge of tumor extending into the air spaces in surrounding lung parenchyma according to the 2015 WHO classification [ 27 ] . Patients with either invasive pathological feature would be categorized into risk group. 2.3. CT scanning protocol Chest CT examinations were performed using five multidetector CT systems of three types: Lightspeed16, GE Healthcare, Milwaukee, WI, USA; Somatom Sensation 64, Siemens, Erlangen, Germany; Discovery CT750 HD, GE Healthcare. The scanning parameters were: (a) 120 kVp with the automatic regulation of the tube current and 1.5-mm reconstruction thickness and intervals for the 64-detector scanner and (b) 120 kVp, 150–200 mAs, and 1.25-mm reconstruction thickness intervals for the other two types of scanners. All of the 910 patients underwent contrastenhanced CT. Non-ionic iodinated contrast material (300 mg of iodine per millilitre, Ul-travist; Bayer Pharma, Berlin, Germany) was injected at a dose of 1.3–1.5 ml/kg body weight at a rate of 2.5 ml/s using an automated injector. CTenhanced scanning was performed with a 70-second delay. 2.4. CT image interpretation and preprocessing Two experienced clinical radiologists, one with 9 years of experience in CT imaging of thoracic malignancies and the other with 6 years of experience, independently analyzed and confirmed the cN0 stage of the CT images after training. Both radiologists were blinded to clinical and pathological information, and any discrepancies in image interpretation were resolved through negotiation and discussion. The CT descriptors and scoring criteria utilized in the analysis were detailed in Table 1 . The images were viewed with a lung window width of 1500 HU and window level of -600 HU, as well as a mediastinal window width of 350 HU at level of 40 HU. The CT descriptors were evaluated on multiplanar reconstructed images and reported with a standardized scoring sheet. Image preprocessing was completed by a linear interpolation algorithm to resample the thickness of CT images to 1 mm. Gaussian filtering was used to preprocess CT images. Table 1 CT Characteristics for NSCLC. Characteristic Definition Scoring and Definition Location Lung nodule located in the outer third of the lung was defined as peripheral tumor, while others were located centrally 1, central; 2, peripheral Maximum diameter The greatest dimension on the multiplanar reconstructed images with a lung window cm Consolidation diameter The greatest dimension on the multiplanar reconstructed images with a mediastinal window cm Contour The overall shape of roundness 1, round; 2, oval; 3, somewhat irregular; 4, irregular Lobulation A wavy or scalloped configuration of tumor’s surface 0, absence; 1, presence Spiculation Lines radiating from the margins of the tumor 0, absence; 1, presence Texture Solid or GGO 0, pure GGO; 1, mixed GGO with solid part 50%; 3, solid Calcification Any patterns of calcification in the tumor 0, absence; 1, presence Air bronchogram Tubelike or branched air structure within the tumor 0, absence; 1, presence Bubble-like lucency Air space in the tumor with diameter ≤ 5mm at the time of diagnosis prior to biopsy or treatment 0, absence; 1, presence Cavity Air space in the tumor with diameter > 5mm at the time of diagnosis prior to biopsy or treatment 0, absence; 1, presence Enhancement degree Enhancement degree = A post - A pre , where A pre and A post was unenhanced and contrast-enhanced CT attenuation of tumor, respectively HU Enhancement heterogeneity Heterogeneity of tumor on contrast-enhanced images 1, homogeneous; 2, slight or moderate heterogeneous; 3, marked heterogeneous Relative enhancement Relative enhancement E rel = (A post - A pre )/E art , where E art was enhancement attenuation of the artery on the same section % Pleural attachment Tumor attaches to the fissure/Pleura 0, absence; 1, presence Pleural retraction Retraction of the pleura toward the tumor 0, absence; 1, presence Bronchovascular bundle thickening Convergence of vessels to the tumor 0, no significant thickening; 1, obvious thickening Obstructive change Consolidation shadow caused by obstructive pneumonia or atelectasis at the edge of tumor 0, absence; 1, presence Peripheral emphysema Peripheral emphysema caused by the tumor or preexisting emphysema 0, absence; 1, slight or moderate; 2, severe Peripheral fibrosis Peripheral fibrosis caused by the tumor or preexisting fibrosis 0, absence; 1, slight or moderate; 2, severe CT, computed tomography; NSCLC, non-small cell lung cancer; GGO, ground-glass opacity. 2.5. Tumor segmentation and features extraction Tumor segmentation was conducted utilizing a manual approach of delineating regions of interest (ROI) on processed CT enhanced scan images with the assistance of ITK-snap 3.6.0 by a radiologist possessing 6 years of experience in image segmentation. The radiologist was provided with information regarding the tumor location, while remaining blinded to additional details. Subsolid tumors were segmented at the lung window, while solid tumors were segmented at the mediastinal window in order to enhance the identification of blood vessels, LNs, and lung consolidation surrounding the tumors. The accuracy of the segmentation results was to be verified or adjusted by another senior radiologist. Radiomics features were extracted using the PyRadiomics 3.0 open-source software program ( http://www.radiomics.io/pyradiomics.html ) [ 30 ] . A total of 1316 radiomics features were extracted from the 3D ROI of CT images (Table S1 ). 2.6. Feature selection and establishment of radiomics signature The Z-score was utilized to standardize the radiomics parameters across all patients. The Spearman pairwise correlation analysis was utilized to determine the strength of correlations between features, with features exhibiting an absolute correlation value exceeding 0.9 being eliminated. Following this, the top 100 features were chosen through the application of the minimum redundancy maximum relevance (mRMR) method. These selected features were then inputted into least absolute shrinkage and selection operator (LASSO) models to identify the optimal subsets for assessing invasive pathological characteristics. Subsequently, the identified features were utilized to construct a logistic regression model with cross-validation method by a backward step-wise selection to eliminate non-significant variables. The radiomics score (Rad-score) formula was derived through a linear combination of the selected features weighted by their corresponding coefficients with the following formula: Radscore = b + Ci × Xi, where b represents a constant term, Xi denotes the value of the selected feature, and Ci represents the regression coefficient associated with the selected feature. Each patient’s Rad-score was calculated by this formula to compare the difference between risk group and non-risk group. 2.7. Statistical analysis All statistical analyses were performed with the statistical software R 4.3.0 and SPSS 26.0. The continuous variables were expressed as mean values and standard deviations, and categorical variables as frequency. Agreement between two readers were analyzed by the ĸ index and Kendall coefficient of concordance. Non-parametric two-sample Wilcoxon test was used for ranked or continuous variables, and chi-square or Fisher’s test for categorical variables in univariate analysis. Subsequently, multivariate logistic regression analysis was performed to test the ability of combining Rad-score, radiological and clinical features to identify risk group. Models were assessed for predicative accuracy using the ten-fold cross-validation approach and results over the ten models were averaged. The performance of prediction model was evaluated with calibration curve and Hosmer–Lemeshow test, then visualized by nomogram. Area under the receiver operating characteristic (ROC) curve (AUC) was used to assess the predictive efficacy, and AUC of different models was compared by the Delong test. P values < 0.05 were regarded as statistically significant. 3. Results 3.1. Reader reproducibility Agreement among the two readers was good (Table S2). The intraclass correlation coefficient for maximum diameter, consolidation diameter, degree of enhancement and relative enhancement was 0.91 (range, 0.88–0.94), 0.92 (range, 0.88–0.95), 0.85 (range, 0.82–0.89) and 0.86 (range, 0.83–0.88), respectively. 3.2. Patient demographics Clinical characteristics and histological subtypes of training group were shown in Table S3. 501, 117 and 60 patients were diagnosed with clinical stage IA, IB and IIA, respectively. 225 (33.2%) cases among these patients had invasive pathological features, including 83, 95, 88, and 122 cases of OLM, VPI, LVI, and STAS confirmed by postoperative pathology. No significant difference of either feature was observed between training and validation group. 3.3. Correlation of invasive pathological features with clinical and radiological features The association between clinical features with invasive pathological features was presented in Table 2 . Significantly, patients with clinical stage IB and IIA [50/117 (43.9%), 40/60 (66.7%) vs. 135/501 (26.9%)] developed regional LN involvement, VPI, LVI or STAS more frequently than IA patients (odds ratio (OR) = 2.81, 95% confidence intervals (CI): 1.97, 4.00 for IB and OR = 4.68, 95% CI: 2.66, 8.23 for IIA). No significant association was noted between other clinical features with invasive pathological features. Table 2 Association between Clinical Characteristics with Risk Factors. Variable Negative Group Positive Group P Value Univariate OR (95% CI) Number Age (years) Sex Male Female Smoking history Yes No Family history Yes No Histological subtype Adenocarcinoma Squamous cell carcinoma Other* EGFR Mutation Wild KRAS Mutation Wild ALK Positive Negative Clinical stage IA IB IIA Tumor markers CEA (ng/ml) CA125 (U/ml) NSE (ng/ml) SCC (ng/ml) Cyfra 21 − 1 (ng/ml) 453 59.49 (± 8.63) 204 249 212 241 108 345 374 70 9 133 88 16 171 7 137 366 67 20 5.73 (± 7.26) 11.22 (± 8.91) 7.52 (± 4.03) 1.63 (± 0.78) 3.25 (± 1.95) 225 59.65 (± 8.63) 110 115 112 113 54 171 192 24 9 75 41 5 88 8 89 135 50 40 7.14 (± 6.25) 12.86 (± 7.55) 7.19 (± 4.05) 1.88 (± 0.86) 3.46 (± 2.07) 0.824 0.343 0.465 0.927 0.863 0.422 0.341 0.286 < 0.001 0.084 0.244 0.672 0.327 0.589 Reference 2.81 (1.97, 4.00) 4.68 (2.66, 8.23) Data for age and tumor markers are mean ± standard deviation. OR, odds ratio; CI, confidence interval; EGFR, epidermal growth factor receptor; KRAS, kirsten rat sarcoma viral oncogene; ALK, anaplastic lymphoma kinase; CEA, carcinoembryonic antigen; CA125, carbohydrate antigen 125; NSE, neuron-specific enolase; SCC, squamous cell carcinoma antigen; Cyfra 21 − 1, cytokeratin 19 fragment. *Other histologic subtype contains adenosquamous carcinoma, large cell lung cancer and sarcomatoid carcinoma. Table 3 Association between Traditional CT Features with Risk Factors. Variable Negative Group Positive Group P Value Univariate OR(95% CI) Number 453 225 Maximum diameter (cm) 2.48 (± 1.03) 3.02 (± 0.97) < 0.001 Consolidation diameter (cm) 2.00 (± 1.32) 2.91 (± 1.04) < 0.001 Bronchovascular bundle thickening 0 388 174 0.007 Reference 1 65 51 1.75 (1.16, 2.63) Lobulation 0 310 75 < 0.001 Reference 1 143 150 4.34 (3.08, 6.10) Spiculation 0 368 166 0.025 Reference 1 85 59 1.54 (1.05, 2.25) Texture 0 80 1 < 0.001 Reference 1 108 12 Reference 2 83 37 11.57 (6.41, 20.88) 3 182 175 5.21 (3.61, 7.52) Obstructive change 0 438 208 0.014 Reference 1 15 17 2.39 (1.17, 4.87) Pleural attachment 0 260 77 < 0.001 Reference 1 193 148 2.59 (1.86, 3.61) Pleural retraction 0 172 61 0.005 Reference 1 281 164 1.65 (1.16, 2.34) Data for maximum diameter and consolidation diameter are mean ± standard deviation. CT, computed tomography; OR, odds ratio; CI, confidence interval. Univariate analysis showed that tumors with a larger overall size (p < 0.001) and solid component size (p < 0.001), bronchovascular bundle thickening (OR = 1.75, 95% CI: 1.16, 2.63; p = 0.007), lobulation (OR = 4.34, 95% CI: 3.08, 6.10; p < 0.001), spiculation (OR = 1.54, 95% CI: 1.05, 2.25; p = 0.025), solid texture (OR = 11.57, 95% CI: 6.41, 20.88 for score 2; OR = 5.21, 95% CI: 3.61, 7.52 for score 3; p < 0.001), pleural attachment (OR = 2.59, 95% CI: 1.86, 3.61; p < 0.001), pleural retraction (OR = 1.65, 95% CI: 1.16, 2.34; p = 0.005) and obstructive change (OR = 2.39, 95% CI: 1.17, 4.87; p = 0.014) were more likely to develop invasive pathological features. While, other radiological features were not statistically significant (Table S4). 3.4. Screening and integration of radiomics features A total of 1316 radiomics features were extracted from the 3D ROI of each CT image and Spearman pairwise correlation analysis and the mRMR method were utilized to remove highly correlated features to reduce redundancy. The top 100 features were screened through LASSO algorithm to avoid overfitting and leaving 11 radiomics features (Table S1 ; Fig. 2 ). Six features were excluded after logistic regression in order to remove the non-significant variables, finally leaving five radiomics features to construct radiomics model and generate the Rad-score formula weighted by their respective coefficients (Table 4 ). The predictive accuracy of the radiomics model was evaluated through the cross-validation method. A ten-fold random sample of the dataset was employed to construct a predictive model, which was iteratively repeated to generate ten models. The average results of these models yielded an AUC value of 0.778 for the best predictive accuracy. The Rad-score for each patient was visualized in a waterfall plot (Fig. 3 ) and significant difference was observed between risk group and negative group (p < 0.001). Table 4 The Radiomics Features in Rad-score Formula* after Logistic Regression Analysis. Radiomics Features Risk Factor P Value Odds Ratio (95% CI) original_ngtdm_Strength exponential_glszm_SmallAreaEmphasis gradient_firstorder_Minimum wavelet-LLL_glcm_JointAverage wavelet-HHH_ngtdm_Strength original_glcm_Imc2 exponential_firstorder_InterquartileRange logarithm_glcm_Idn square_gldm_DependenceVariance squareroot_firstorder_RootMeanSquared wavelet-HHL_glcm_Imc2 0.014 < 0.001 0.004 0.024 0.049 NA NA NA NA NA NA 0.69 (0.52, 0.93) 2.05 (1.45, 2.90) 0.71 (0.56, 0.90) 1.51 (1.06, 2.15) 0.22 (0.05, 0.99) Rad-score, radiomics score; CI, confidence interval; NA, not applicable (variables that were not included in the equation of multivariate logistic regression analysis with backward stepwise selection). *Rad-score = -1.294-0.37*original_ngtdm_Strength + 0.717*exponential_glszm_SmallAreaEmphasis -0.342*gradient_firstorder_Minimum + 0.409*wavelet-LLL_glcm_JointAverage-1.525* wavelet-HHH_ngtdm_Strength 3.5. Predictive model and validation test The multivariable logistic regression model of clinicoradiological features and radiomics features revealed that larger consolidation diameter (OR = 1.26, 95% CI: 1.02, 1.55, p = 0.032), pleural attachment (OR = 1.70, 95% CI: 1.12, 2.59, p = 0.013), texture (OR = 1.56, 95% CI: 1.33, 1.83, p < 0.001 for score 2; OR = 1.85, 95% CI: 1.09, 3.13, p = 0.032 for score 3) and Rad-score (OR = 4.61, 95% CI: 1.99, 10.64, p < 0.001) were significant independent predictors (Table 5 ). The result of Hosmer-Lemeshow goodnessoffit test was not statistically significant (p = 0.833), indicating a high concordance between the predicted and observed probabilities. Calibration curves, ROC curves and nomogram were drawn in Fig. 4 – 6 . Table 5 Multivariable Logistic Regression Analysis of Radiological Features Combined with Rad-score Predicting the Presence of Risk Factors. Variable Risk Factors P Value Odds Ratio (95% CI) Consolidation diameter Pleural attachment 0 1 Texture 0 1 2 3 Rad-score Maximum diameter Bronchovascular bundle thickening Lobulation Spiculation Obstructive change Pleural retraction Clinical stage 0.032 1.26 (1.02, 1.55) 0.013 Reference 1.70 (1.12, 2.59) < 0.001 Reference Reference < 0.001 1.56 (1.33, 1.83) 0.032 1.85 (1.09, 3.13) < 0.001 4.61 (1.99, 10.64) NA NA NA NA NA NA NA NA NA NA NA NA NA NA Rad-score, radiomics score; CI, confidence interval; NA, not applicable (variables that were not included in the equation of multivariate logistic regression analysis with backward stepwise selection). Formula: ex / (1 + ex ), x = -0.516 + 0.229 × consolidation diameter + 0.532 × pleural attachment + 1.526 × texture (solid part > 50%) + 0.446 × Rad-score The AUC of combined model increased to 0.815 (95% CI: 0.779, 0.850), compared with 0.778 (95% CI: 0.740, 0.817; p = 0.275) and 0.691 (95% CI: 0.646, 0.736; p < 0.001) when Rad-score or traditional radiological feature was used alone (Fig. 5 A). In the validation group, the accuracy of the combined prediction model was reasonable with an AUC of 0.792 (95% CI: 0.756, 0.828), compared with 0.745 (95% CI: 0.716, 0.783; p = 0.038) and 0.701 (95% CI: 0.661, 0.739; p < 0.001) when utilizing either the Rad-score or traditional radiological feature in isolation (Fig. 5 B). The combined model exhibited a sensitivity of 86.56%, specificity of 81.63%, and accuracy of 81.30%, and in the external validation set, achieved a sensitivity of 74.32%, specificity of 80.49%, and accuracy of 78.65% when the cutoff was determined at the maximum Youden index. 4. Discussion The efficacy of limited surgery in early NSCLC patients has recently been demonstrated as significant, resulting in the preservation of more functional lung tissue and reducing perioperative mortality compared to standard lobectomy [31, 32] . Although the National Comprehensive Cancer Network guidelines recommend sublobar resection for peripheral nodules ≤2 cm with ≥50% ground-glass appearance on CT or a long doubling time (≥400 days), there remains a lack of definitive selection criteria for limited resection [33] . In order to determine the suitability of limited resection for early-stage NSCLC patients, precise LN staging is essential to verify the absence of LN involvement. Additionally, the invasive pathological characteristics of the primary tumor, such as VPI, LVI, and STAS, serve as significant prognostic indicators for both local recurrence and distant metastasis, ultimately impacting the overall survival rate of NSCLC patients who undergo limited resection [7-10] . In cases where the N0 status is uncertain or there is a high likelihood of pathological invasiveness, lobectomy and systematic LN dissection should be performed instead of limited surgery. We conducted a comprehensive analysis of 11 retrospective studies and meta-analyses to determine the average incidence rates of OLM, VPI, LVI, and STAS in lung cancer, which were found to be 14.0% (range 9.8% to 15.9%), 21.4% (range 11.8% to 23.0%), 23.1% (range 12.4% to 33.6%), and 31.3% (range 15.5% to 43.3%), respectively [9-12, 16, 17, 34-38] . These findings align with our own results, underscoring the significance of accurately predicting invasive pathological factors non-invasively prior to surgery. CT emerges as a pivotal tool for providing detailed imaging information and is therefore routinely employed in clinical practice. The findings from the reports indicate that CT features of the primary tumor, such as tumor size and proportion of solid component, were identified as independent predictors of invasive pathological features [16, 18, 39, 40] . It was observed that the maximum consolidation diameter served as an independent prognostic factor, as opposed to the overall size of the tumor. In accordance with the guidelines set forth by the Union for International Cancer Control, the diameter of the invasive component within the tumor should be considered as a criterion for T staging, as it was found to be more predictive of prognosis [41, 42] . Ground-glass opacity (GGO) predominant pulmonary nodules in CT scans are commonly believed to be associated with a low likelihood of invasive pathology [43] . Patients with GGO predominant NSCLC have been shown to have a more favorable prognosis following surgical resection [44] , suggesting that they may be suitable candidates for less extensive surgical procedures. Suzuk et al. (2011) demonstrated that a consolidation-to-tumor ratio of less than 0.25 or 0.5 on CT scans could accurately predict the absence of LVI or LN involvement [18] , with similar results of other researches on predicting VPI and STAS [16, 40] . These results align with our own research findings. Multiple studies have demonstrated that STAS is not typically present in pure ground glass opacity nodules (pGGN) [17, 45, 46] . In our research, only one patient with pGGN was pathologically confirmed to have VPI with a lepidic pattern adenocarcinoma (ADC). Although 44 out of 93 cases (47.3%) of pGGN were found to have microinvasive adenocarcinoma (MIA) upon postoperative pathology examination, the existence of invasive pathological characteristics in pGGN should not be disregarded. Pleural attachment is commonly associated with a poor prognosis in NSCLC and is widely acknowledged as an independent predictor of VPI [20, 40, 47] . In our research, pleural attachment demonstrated a robust predictive capability in the collective prediction of invasive pathological characteristics, despite the lack of significant associations between pleural attachment and OLM, LVI, and STAS, in contrast to the findings of Lin et al. (2021), who suggested that pleural attachment exhibited strong diagnostic accuracy for STAS [16] . This discrepancy may be attributed to variations in study populations and inclusion criteria. Qi et al. (2016) posited that pleural retraction served as an independent risk factor for VPI [48] , although there remains ongoing debate regarding the potential association between this peritumoral feature with the malignancy of the lesion. Gallagher et al. (1990) posited that pleural retraction may be attributed to elastasis, inflammatory invasion, and thick fibrous proliferation, suggesting it is merely indicative of pleural fiber tension [49] . Our univariate analysis indicated a correlation between pleural retraction and invasive pathological characteristics, yet it did not emerge as a standalone predictive factor. Numerous studies have identified lobulation and spiculation as independent predictors of OLM, LVI and STAS [16, 37] , which were highly related to invasive pathological features in our study, but not independent predictors. The relationship between ADC and squamous cell carcinoma (SCC) subtypes of NSCLC and invasive pathological features was found to be significant in our study, although they were not identified as independent predictors. ADC and SCC are the predominant subtypes of NSCLC, with ADC representing the majority of cases in our study. However, 94 cases of SCC were excluded from the study due to difficulties in lesion identification on chest CT scans. These lesions were primarily endobronchial in nature, with some being obscured by distal atelectasis or pneumonia on imaging. The classification of lung ADC by the 2021 WHO guidelines led to the exclusion of carcinoma in situ (CIS) from the study sample (n=18) [50] . In order to reduce selective bias as well as the inability to confirm the pathology of MIA prior to surgery, MIA was not excluded from the study, though none of these cases exhibited invasive pathological features. Previous studies have identified LVI and STAS as significant risk factors for OLM, leading to upstaging of the N category [35, 51] . Vaghjiani et al. (2020) found that VPI was more prevalent in tumors with STAS positivity [12] , contributing to T-stage progression from T1 to T2 in lung cancer with a diameter less than 3 cm [28] . There is ample evidence to suggest that the presence of LVI or STAS may serve as an initial mechanism in the progression of tumor development by disseminating tumor cells beyond the confines of the primary carcinoma into lymphatic, vessel, or air spaces, ultimately leading to the development of OLM or VPI. It is recognized that certain pathological characteristics, specifically micropapillary and solid patterns, are strongly associated with a negative prognosis in NSCLC [52] . However, when these features are present in a small proportion, they may not necessarily pose a significant risk for limited resection, making prediction based on imaging features more challenging. Additionally, extracapsular extension was reported to occur more frequently in advanced stages of lung cancer [53] , and the detection of micro-metastases may vary among different institutions due to differences in testing methods, leading to a lack of uniformity in diagnostic accuracy [54] . Therefore, the above invasive pathological features were not included in this study of early-stage NSCLC based on CT radiomics. Radiomics, the process of converting radiographic images into quantifiable data, plays a crucial role in developing predictive models and potentially enhancing diagnostic precision [55] . In a radiomics study conducted by Jiang et al. (2020), the Random Forest algorithm was utilized to develop a predictive model that demonstrated efficacy in forecasting STAS positive tumors with an AUC of 0.7 [56] . Zhong et al. (2023) reported significant predictive value in the deep learning signature of primary tumors based on PET-CT scans for OLM of lung cancer [26] . In this research, Spearman pairwise correlation analysis and the mRMR method were employed to eliminate highly correlated features and minimize redundancy. Subsequently, the LASSO algorithm and logistic regression analysis were employed to streamline the selection of radiomics features and construct the prediction model. This method is commonly utilized in the feature selection process, particularly for high dimensional data, and has shown promise in reducing the multilinearity between features [57] . We incorporated the relevant radiomics features into the Rad-score and developed a combined prediction model represented by a nomogram with the AUC of 0.815. This integration led to a significant enhancement in predictive accuracy compared to the traditional CT features model (p<0.001). Consequently, individuals at high risk for limited surgery in early-stage NSCLC can be identified, while sublobectomy may be considered a viable option for those in the low-risk group. Our study encountered several limitations. Firstly, it was a retrospective study focusing solely on clinical IA-IIA stage NSCLC cases to enhance the precision of the predictive model, thereby restricting its generalizability. Additionally, manual tumor segmentation proved to be time-consuming, requiring frequent corrections to delineate the boundaries of adjacent structures such as blood vessels, bronchi, chest wall, and mediastinum, particularly on CT images with a 1 mm slice thickness. Furthermore, the validation cohort from another medical center was inadequately sized, highlighting the need for additional prospective data from multiple sources. Conclusion In conclusion, this study provided a noninvasive prediction tool that combined traditional radiological characteristics and radiomics features based on the preoperative CT enhanced image. The prediction model can be incorporated into the specific treatment strategy decision-making process of NSCLC patients with clinical stage IA-IIA, and determine the subgroup of patients appropriate to limited surgery. Declarations Author Contribution 1 guarantor of integrity of the entire study --------- Fengnian Zhao, Wang jiang2 study concepts and design --------------------------- Fengnian Zhao, Wang jiang3 literature research ------------------------------------- Yunqing Zhao, Xiaoxue Wang4 clinical studies ----------------------------------------- Qingna Yan, Yunqing Zhao5 experimental studies / data analysis --------------- Fengnian Zhao, Dong LI6 statistical analysis ------------------------------------- Yunqing Zhao7 manuscript preparation ------------------------------ Wang Jiang, Xiaoxue Wang8 manuscript editing ------------------------------------ Dong Li, Guiming Zhou Data Availability Data is provided within the manuscript and supplementary information files. References Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7-33. Miller KD, Nogueira L, Devasia T, Mariotto AB, Yabroff KR, Jemal A, et al. Cancer treatment and survivorship statistics, 2022. CA Cancer J Clin. 2022;72(5):409-36. Hoy H, Lynch T, Beck M. Surgical Treatment of Lung Cancer. Crit Care Nurs Clin North Am. 2019;31(3):303-+. Suzuki K, Saji H, Aokage K, Watanabe S-i, Okada M, Mizusawa J, et al. Comparison of pulmonary segmentectomy and lobectomy: Safety results of a randomized trial. J Thorac Cardiovasc Surg. 2019;158(3):895-907. Wen Z, Zhao Y, Fu F, Hu H, Sun Y, Zhang Y, et al. Comparison of outcomes following segmentectomy or lobectomy for patients with clinical N0 invasive lung adenocarcinoma of 2 cm or less in diameter. J Cancer Res Clin Oncol. 2020;146(6):1603-13. Beck KS, Gil B, Na SJ, Hong JH, Chun SH, An HJ, 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. Adachi H, Sakamaki K, Nishii T, Yamamoto T, Nagashima T, Ishikawa Y, et al. Lobe-Specific Lymph Node Dissection as a Standard Procedure in Surgery for Non-Small Cell Lung Cancer: A Propensity Score Matching Study. J Thorac Oncol. 2017;12(1):85-93. Blaauwgeers H, Flieder D, Warth A, Harms A, Monkhorst K, Witte B, et al. A Prospective Study of Loose Tissue Fragments in Non-Small Cell Lung Cancer Resection Specimens: An Alternative View to “Spread Through Air Spaces”. Am J Surg Pathol (2017) 41:1226. Liu QX, Deng XF, Zhou D, Li JM, Min JX, Dai JG. Visceral pleural invasion impacts the prognosis of non-small cell lung cancer: a meta-analysis. Eur J Surg Oncol 2016; 42(11):1707–1713. Shiono S, Abiko M, Sato T. Positron emission tomography/computed tomography and lymphovascular invasion predict recurrence in stage I lung cancers. J Thorac Oncol 2011;6:43–47. Zhong YF, Cai C, Chen T, Gui H, Chen C, Deng JJ, et al. PET/CT-based deep learning grading signature to optimize surgical decisions for clinical stage I invasive lung adenocarcinoma and biologic basis under its prediction: a multicenter study. Eur J Nucl Med Mol Imaging. 2024;51(2):521-34. Vaghjiani RG, Takahashi Y, Eguchi T, Lu SH, Kameda K, Tano Z, et al. Tumor Spread Through Air Spaces Is a Predictor of Occult Lymph Node Metastasis in Clinical Stage IA Lung Adenocarcinoma. J Thorac Oncol. 2020;15(5):792-802. Park HK, Jeon K, Koh W-J, Suh GY, Kim H, Kwon OJ, et al. Occult nodal metastasis in patients with non-small cell lung cancer at clinical stage IA by PET/CT. Respirology. 2010;15(8):1179-84. Tanaka T, Shinya T, Sato S, Mitsuhashi T, Ichimura K, Soh J, et al. Predicting pleural invasion using HRCT and 18F-FDG PET/CT in lung adenocarcinoma with pleural contact. Ann Nucl Med 2015; 29(9):757–765. Naidich DP. Is spread of tumor through air spaces a concern for interpreting lung nodules on CT images? Radiology (2018) 289:841–2. Qi L, Xue K, Cai YJ, Lu JJ, Li XH, Li M. Predictors of CT Morphologic Features to Identify Spread Through Air Spaces Preoperatively in Small-Sized Lung Adenocarcinoma. Front Oncol. 2021;10. Kim SK, Kim TJ, Chung MJ, Kim TS, Lee KS, Zo JI, et al. Lung Adenocarcinoma: CT Features Associated with Spread through Air Spaces. Radiology. 2018;289(3):831-40. Suzuki K, Koike T, Asakawa T, Kusumoto M, Asamura H, Nagai K, et al. A prospective radiological study of thin-section computed tomography to predict pathological noninvasiveness in peripheral clinical IA lung cancer (Japan Clinical Oncology Group 0201). J Thorac Oncol (2011) 6(4):751–6. Zhao LL, Xie HK, Zhang LP, Zha JY, Zhou FY, Jiang GN, et al. Visceral pleural invasion in lung adenocarcinoma <=3 cm with ground-glass opacity: a clinical, pathological and radiological study. J Thorac Dis 2016; 8(7):1788–1797. Seok Y, Lee E. Visceral pleural invasion is a significant prognostic factor in patients with partly solid lung adenocarcinoma sized 30 mm or smaller. Thorac Cardiovasc Surg 2018; 66(2):150–155. Aerts HJ, Velazquez ER, Leijenaar RT, Parmar C, Grossmann P, Carvalho S, et al. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat Commun 2014;5:4006. Huang Y, Liu Z, He L, Chen X, Pan D, Ma Z, et al. Radiomics signature: A potential biomarker for the prediction of disease-free survival in early-stage (I or II) non-small cell lung cancer. Radiology 2016;281:947-57. Christie J, Romine P, Nair V, Mattonen S. Prediction of pathologic lymphovascular invasion in non-small cell lung cancer using multi-modality tumour and peri-tumoural radiomics. Med Phys. 2022;49(8):5650-1. Zha XY, Liu YQ, Ping XX, Bao JY, Wu Q, Hu S, et al. A Nomogram Combined Radiomics and Clinical Features as Imaging Biomarkers for Prediction of Visceral Pleural Invasion in Lung Adenocarcinoma. Front Oncol. 2022;12. Bassi M, Russomando A, Vannucci J, Ciardiello A, Dolciami M, Ricci P, et al. Role of radiomics in predicting lung cancer spread through air spaces in a heterogeneous dataset. Transl Lung Cancer R. 2022;11(4):560-71. Zhong YF, Cai C, Chen T, Gui H, Deng JJ, Yang ML, et al. PET/CT based cross-modal deep learning signature to predict occult nodal metastasis in lung cancer. Nat Commun. 2023;14(1). Travis WD, Brambilla E, Nicholson AG, Yatabe Y, Austin JHM, Beasley MB, et al. The 2015 World Health Organization Classification of Lung Tumors Impact of Genetic, Clinical and Radiologic Advances Since the 2004 Classification. J Thorac Oncol. 2015;10(9):1243-60. Rami-Porta R, Bolejack V, Crowley J, Ball D, Kim J, Lyons G, et al. The IASLC Lung Cancer Staging Project Proposals for the Revisions of the T Descriptors in the Forthcoming Eighth Edition of the TNM Classification for Lung Cancer. J Thorac Oncol. 2015;10(7):990-1003. Liu JB, Huffman KM, Palis BE, Shulman LN, Winchester DP, Ko CY, et al. Reliability of the American College of Surgeons Commission on Cancer's Quality of Care Measures for Hospital and Surgeon Profiling. J Am Coll Surg. 2017;224(2):180-+. Van Griethuysen JJM, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res. 2017;77(21):E104-E7. Cao JL, Yuan P, Wang YQ, Xu JM, Yuan XS, Wang ZT, et al. Survival Rates After Lobectomy, Segmentectomy, and Wedge Resection for Non-Small Cell Lung Cancer. Ann Thorac Surg. 2018;105(5):1483-91. Okada M. Radical Sublobar Resection for Small-Diameter Lung Cancers. Thorac Surg Clin. 2013;23(3):301-+. Ettinger DS, Wood DE, Aisner DL, Akerley W, Bauman JR, Bharat A, et al. Non-Small Cell Lung Cancer, Version 2.2021 Featured Updates to the NCCN Guidelines. J Natl Compr Canc Ne. 2021;19(3):254-66. Moon Y, Choi SY, Park JK, Lee KY. Risk Factors for Occult Lymph Node Metastasis in Peripheral Non-Small Cell Lung Cancer with Invasive Component Size 3 cm or Less. World J. Surg. 2020;44(5):1658-65. Seto K, Kuroda H, Yoshida T, Sakata S, Mizuno T, Sakakura N, et al. Higher frequency of occult lymph node metastasis in clinical N0 pulmonary adenocarcinoma with ALK rearrangement. Cancer Manag. Res. 2018;10:2117-24. Dai CY, Xie HK, Su H, She YL, Zhu EJ, Fan ZW, et al. Tumor Spread through Air Spaces Affects the Recurrence and Overall Survival in Patients with Lung Adenocarcinoma >2 to 3 cm. J Thorac Oncol (2017) 12:1052–60. Yang GJ, Nie P, Zhao LZ, Guo J, Xue W, Yan L, et al. 2D and 3D texture analysis to predict lymphovascular invasion in lung adenocarcinoma. Eur J Radiol. 2020;129. Onozato Y, Nakajima T, Yokota H, Morimoto J, Nishiyama A, Toyoda T, et al. Radiomics is feasible for prediction of spread through air spaces in patients with nonsmall cell lung cancer. Sci. Rep. 2021;11(1). Zang RC, Qiu B, Gao SG, He J. A Model Predicting Lymph Node Status for Patients with Clinical Stage T1aN0-2M0 Nonsmall Cell Lung Cancer. Chin Med J. 2017;130(4):398-403. Chen ZF, Jiang SX, Li ZL, Rao LJ, Zhang XS. Clinical Value of F-18-FDG PET/CT in Prediction of Visceral Pleural Invasion of Subsolid Nodule Stage I Lung Adenocarcinoma. Acad Radiol. 2020;27(12):1691-9. Travis WD, Asamura H, Bankier AA, Beasley MB, Detterbeck F, Flieder DB, et al. The IASLC Lung Cancer Staging Project: Proposals for Coding T Categories for Subsolid Nodules and Assessment of Tumor Size in Part-Solid Tumors in the Forthcoming Eighth Edition of the TNM Classification of Lung Cancer. J Thorac Oncol. 2016;11(8):1204-23. Rami-Porta R, Asamura H, Travis WD, Rusch VW. Lung Cancer-Major Changes in the American Joint Committee on Cancer Eighth Edition Cancer Staging Manual. Ca-Cancer J Clin. 2017;67(2):138-55. Taylor ML, Carmona F, Thiagarajan RR, Westgate L, Ferguson MA, del Nido PJ, et al. Mild Postoperative Acute Kidney Injury and Outcomes After Surgery for Congenital Heart Disease. J Thorac Cardiovasc Surg (2013) 146(1):146–52. Ohde Y, Nagai K, Yoshida J, Nishimura M, Takahashi K, Suzuki K, et al. The Proportion of Consolidation to Ground-Glass Opacity on High Resolution CT Is a Good Predictor for Distinguishing the Population of Non-Invasive Peripheral Adenocarcinoma. Lung Cancer (2003) 42(3):303–10. Ledda RE, Milanese G, Gnetti L, Borghesi A, Sverzellati N, Silva M, et al. Spread through air spaces in lung adenocarcinoma: is radiology reliable yet? J Thorac Dis (2019) 11:S256–61. Toyokawa G, Yamada Y, Tagawa T, Kamitani T, Yamasaki Y, Shimokawa M, et al. Computed tomography features of resected lung adenocarcinomas with spread through air spaces. J Thorac Cardiovasc Surg (2018) 156:1670–6. Tang EK, Chen CS, Wu CC, Wu MT, Yang TL, Liang HL, et al. Natural history of persistent pulmonary subsolid nodules: long-term observation of different interval growth. Heart Lung Circ 2019; 28(11):1747–1754. Qi LP, Li XT, Yang Y, Chen JF, Wang J, Chen ML, et al. Multivariate analysis of pleural invasion of peripheral non-small cell lung cancer-based computed tomography features. J Comput Assist Tomogr. 2016; 40(5):757–762. Gallagher B, Urbanski SJ. The signifificance of pleural elastica invasion by lung carcinomas. Hum Pathol 1990; 21(5):512–517. Nicholson AG, Tsao MS, Beasley MB, Borczuk AC, Brambilla E, Cooper WA, et al. The 2021 WHO Classification of Lung Tumors: Impact of Advances Since 2015. J Thorac Oncol. 2022;17(3):362-87. Decaluwe H, Moons J, Fieuws S, De Wever W, Deroose C, Stanzi A, et al. Is central lung tumour location really predictive for occult mediastinal nodal disease in (suspected) non-small-cell lung cancer staged cN0 on F-18-fluorodeoxyglucose positron emission tomography-computed tomography? Eur J Cardiothorac Surg. 2018;54(1):134-40. Motono N, Mizoguchi T, Ishikawa M, Iwai S, Iijima Y, Uramoto H. Predictive value of recurrence of solid and micropapillary subtype in lung adenocarcinoma. Oncology. 2023. Chen DL, Ding QF, Wang W, Chen C, Chen YB. ASO Author Reflections: Old Song, New Sung-Extracapsular Extension in Lung Cancer in the Era of Eighth-Edition N Classification. Ann Surg Oncol. 2021;28(4):2099-100. Belanger AR, Hollyfield J, Yacovone G, Ceppe AS, Akulian JA, Burks AC, et al. Incidence and clinical relevance of non-small cell lung cancer lymph node micro-metastasis detected by staging endobronchial ultrasound-guided transbronchial needle aspiration. J Thorac Dis. 2019;11(8):3649-57. Gillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016;278(2):563-77. Jiang CS, Luo Y, Yuan JL, You SY, Chen ZQ, Wu MX, et al. CT-based radiomics and machine learning to predict spread through air space in lung adenocarcinoma. Eur Radiol. 2020;30(7):4050-7. Arya M, Sastry GH, Motwani A, Kumar S, Zaguia A. A novel extra tree ensemble optimized dl framework (Eteodl) for early detection of diabetes. Front Public Health (2021) 9:797877. Additional Declarations No competing interests reported. Supplementary Files Supplementarymaterial.docx Cite Share Download PDF Status: Posted Version 1 posted 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-4488259","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":313655885,"identity":"7a02e38b-7c8c-4e6b-9af5-807ddf567633","order_by":0,"name":"Fengnian Zhao","email":"","orcid":"","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Fengnian","middleName":"","lastName":"Zhao","suffix":""},{"id":313655887,"identity":"b4c84c02-f210-4c30-b6da-8f7a0518ac38","order_by":1,"name":"Wang Jiang","email":"","orcid":"","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wang","middleName":"","lastName":"Jiang","suffix":""},{"id":313655888,"identity":"d10da91a-89c1-4148-8c9c-fe3302fb2ec7","order_by":2,"name":"Xiaoxue Wang","email":"","orcid":"","institution":"Institute of Hematology \u0026 Blood Diseases Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaoxue","middleName":"","lastName":"Wang","suffix":""},{"id":313655889,"identity":"241de13b-0618-4d5e-9848-ea54bd14f87b","order_by":3,"name":"Yunqing Zhao","email":"","orcid":"","institution":"Institute of Hematology \u0026 Blood Diseases Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yunqing","middleName":"","lastName":"Zhao","suffix":""},{"id":313655890,"identity":"37dc90c3-59d1-4d66-a773-32e129ca32ce","order_by":4,"name":"Qingna yan","email":"","orcid":"","institution":"Tianjin Medical University Cancer Institute and Hospital","correspondingAuthor":false,"prefix":"","firstName":"Qingna","middleName":"","lastName":"yan","suffix":""},{"id":313655891,"identity":"2f07a5db-92c7-4035-81bf-f1a312131ade","order_by":5,"name":"Dong Li","email":"","orcid":"","institution":"Tianjin Medical University General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Dong","middleName":"","lastName":"Li","suffix":""},{"id":313655892,"identity":"de235ad2-20a4-4085-ba2f-c21d2ab42137","order_by":6,"name":"Guiming Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYDACCQjFDEQHDnz4QbwWA6AWtsSDM3tI0ALEPMaHOdiI0CE/u/mY5NecP+y6M3I+HGbgYZDnFzuAXwvjnGNp0rLbDJjNbuRuOFxgwWA4c3YCfi3MEjlm0pIwLTN4GBIMbhPQwobQkvPgMA8bEVp4gFokP0K0MBCnRUIiLdmacZsxs9mZZwbAQJYg7Bf5GckHb/7cJpdsdjz58YcPP2zk+aUJaAEBZh4GhmSYrYSVgwAjMJnYEad0FIyCUTAKRiQAAJRDQbJ0ywh4AAAAAElFTkSuQmCC","orcid":"","institution":"Tianjin Medical University General Hospital","correspondingAuthor":true,"prefix":"","firstName":"Guiming","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2024-05-28 05:36:55","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4488259/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4488259/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":58385145,"identity":"590873ac-ed65-431a-b08f-9f6fea810468","added_by":"auto","created_at":"2024-06-14 18:38:43","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":45934,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of patients selection and exclusion. NSCLC, non-small cell lung cancer; CT, computed tomography; PACS, picture archiving and communication system; CIS, \u003ca href=\"http://www.baidu.com/link?url=uKoshn-10jg0oHprdZdP-ixtfO0KL4r7Hyv-mXXwr4NvkW5AbT6wWf2OCovod5suCkaf5L0Uejc2c4jc9t2yZjwY25sAEFV-7qy0W0e3yJhWijTSjJLLJu-QT1ZLrcZM\" target=\"https://www.baidu.com/_blank\"\u003ecarcinoma in situ\u003c/a\u003e; LN, lymph node; OLM, occult lymph node metastasis; VPI, visceral pleural invasion; LVI, lymphovascular invasion; STAS, tumor spread through air spaces.\u003c/p\u003e","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/112bffcbe7dd1e771bd14163.png"},{"id":58387034,"identity":"d381ce45-99f4-461d-a10c-6b160be64035","added_by":"auto","created_at":"2024-06-14 18:46:43","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":46845,"visible":true,"origin":"","legend":"\u003cp\u003e(A) Radiomics features screened by the LASSO regression model. The horizontal axis represents the log lambda of each radiomics, and the vertical axis represents the coefficient of each radiomics. (B) The mean squared error of radiomics features displayed by the Lasso regression analysis. Two vertical lines represented the lambda values when number of variables decreased to two lowest levels.\u003c/p\u003e","description":"","filename":"OnlineFigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/77cf8a81700cb116e1350f53.png"},{"id":58385150,"identity":"5042b29a-9dce-43bb-b8d7-5671ac601970","added_by":"auto","created_at":"2024-06-14 18:38:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":20856,"visible":true,"origin":"","legend":"\u003cp\u003eWaterfall plots of Rad-score for patients in training (A) and validation cohort (B).\u003c/p\u003e","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/7a6d95136c5e61584ce28faf.png"},{"id":58385149,"identity":"ffce8fe0-d731-42dd-a8c2-e24d37006389","added_by":"auto","created_at":"2024-06-14 18:38:43","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":34394,"visible":true,"origin":"","legend":"\u003cp\u003e(A) Calibration curve of the combined model in training group with a C-index of 0.913 (n=678). (B) Calibration curve of the combined model in validation\u003cem\u003e \u003c/em\u003egroup with a C-index of 0.887 (n=232).\u003c/p\u003e","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/a4621fd66f346446994430f8.png"},{"id":58387036,"identity":"db865397-7d7a-40fc-9535-70840dfa38ca","added_by":"auto","created_at":"2024-06-14 18:46:43","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":43007,"visible":true,"origin":"","legend":"\u003cp\u003e(A) The ROC curve for training group. The AUC of combined model increased to 0.815 (95% CI: 0.779, 0.850), compared with 0.778 (95% CI: 0.740, 0.817) and 0.691 (95% CI: 0.646, 0.736) when Rad-score or traditional CT feature was used alone. (B) The ROC curve for validation group. The AUC of combined model increased to 0.792 (95% CI: 0.756, 0.828), compared with 0.745 (95% CI: 0.705, 0.785) and 0.701 (95% CI: 0.660, 0.742) in Rad-score and traditional CT features model.\u003c/p\u003e","description":"","filename":"OnlineFigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/2cacd2a20c72209d3941fb15.png"},{"id":58385147,"identity":"b199df0b-6af3-45d0-bced-67ff90564d42","added_by":"auto","created_at":"2024-06-14 18:38:43","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":15719,"visible":true,"origin":"","legend":"\u003cp\u003eNomogram predicting the likelihood of risk factors with limited surgery in clinical stage IA-IIA NSCLC. According to the location of value on the second to the fifth axis, we can get the vertically corresponding points on the first axis. Summing up the four points together, we can get the total points and the vertically corresponding predicted value on the last axis.\u003c/p\u003e","description":"","filename":"OnlineFigure6.png","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/d6ba73874cb5051a78810794.png"},{"id":63631853,"identity":"acc95be0-c98d-4222-b8ff-94daec6c49fe","added_by":"auto","created_at":"2024-08-30 10:51:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1149234,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/56760e37-fb98-4219-a576-f5f4bbc446aa.pdf"},{"id":58385144,"identity":"c6ebc562-ad73-4d3b-8840-0a556072b1c3","added_by":"auto","created_at":"2024-06-14 18:38:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":26871,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-4488259/v1/be1c997c57717b4778d8aec6.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"United Predictability of CT radiomics on invasive pathological features in clinical stage IA-IIA non-small cell lung cancer: a double-center study","fulltext":[{"header":"Highlights","content":"\u003cp\u003e1. We discovered that the traditional CT features, including larger consolidation diameter, pleural attachment, solid texture and Rad-score were the independent predictors of invasive pathological features.\u003c/p\u003e\n\u003cp\u003e2. Radiomics could improve the predictive performance of models, the AUC significantly increased in combined model, compared with that when traditional clinicoradiological feature was used alone.\u0026nbsp;\u003c/p\u003e"},{"header":"1. Introduction","content":"\u003cp\u003eLung cancer is the leading cause of cancer-related mortality worldwide, of which approximately 80-85% is non-small cell lung cancer (NSCLC). The standard treatment strategy for early NSCLC is lobectomy combined with systematic lymph node (LN) dissection to obtain the best chance of cure. In the past decade, with the promotion of early screening for lung cancer and the gradual maturity of low-dose thin-layer computed tomography (CT) technology, the proportion of early and small volume lung cancer has also increased. Hence, there has been a trend in surgical practice towards employing limited procedures such as sublobectomy (segmentectomy or wedge resection) and selective lymph node dissection to minimize tissue damage and preserve functional lung tissue. Multiple studies have indicated that there is no significant disparity in perioperative safety or long-term survival outcomes between lobectomy and sublobar resection for early NSCLC \u003csup\u003e[4-6]\u003c/sup\u003e. Nevertheless, invasive pathological characteristics including occult lymph node metastasis (OLM), visceral pleural invasion (VPI), lymphovascular invasion (LVI) and tumor spread through air spaces (STAS) may serve as potential risk factors for local recurrence and distant metastasis following limited surgery for early NSCLC \u003csup\u003e[7-10]\u003c/sup\u003e. Accurate preoperative identification of early-stage NSCLC patients remains a crucial and challenging task. This underscores the importance of precise preoperative prediction utilizing clinical and radiological features to identify suitable candidates for limited surgery.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e18\u003c/sup\u003eF-fludeoxyglucose positron emission computed tomography (PET-CT) demonstrates high sensitivity and specificity in LN staging and assessment of tumor invasiveness \u003csup\u003e[11]\u003c/sup\u003e, due to the reliable reproducibility of maximum standardized uptake values (SUV\u003csub\u003emax\u003c/sub\u003e), which reflect the metabolic activity and invasiveness of the tumor. Previous research has indicated that SUV\u003csub\u003emax\u003c/sub\u003e serves as a predictive factor for invasive pathological characteristics \u003csup\u003e[12-14]\u003c/sup\u003e. Despite its efficacy, the widespread adoption of SUV\u003csub\u003emax\u003c/sub\u003e in preoperative assessments for early NSCLC is hindered by its high cost. Alternative studies have utilized CT imaging to make preoperative predictions, highlighting the significance of primary tumor characteristics as independent predictors of invasive pathological features \u003csup\u003e[15-20]\u003c/sup\u003e. However, the predictive accuracy of traditional CT features alone remains limited. Radiomics, a method that involves converting images into diggable data through machine-learning methods \u003csup\u003e[21]\u003c/sup\u003e, offers quantifiable and objective characteristics that may enhance predictive capabilities in medical imaging analysis \u003csup\u003e[22]\u003c/sup\u003e. Prior researches have examined the radiomics features of primary tumors on CT or PET-CT scans and found that they can predict invasive pathological features \u003csup\u003e[23-26]\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHowever, the vast majority of studies have focused solely on a single risk factor, lacking clinical practicality. In this study, the authors conducted a retrospective, dual-center clinical study to noninvasively and collectively predict invasive pathological features in NSCLC patients with clinical stage IA to IIA using radiomics analysis of CT scans. This information aims to assist surgeons and patients in selecting appropriate treatment strategies and surgical option.\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Patients\u003c/h2\u003e \u003cp\u003eA total of 1080 patients were initially enrolled in the study, with surgery dates ranging from January 2016 to February 2023, based on specific inclusion and exclusion criteria outlined in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Ultimately, 910 patients were included in the analysis, with 678 from ( ) assigned to the training group and 232 from ( ) to the external validation group. Clinical data, such as age, sex, smoking status, family history, genetic mutation status, and tumor markers, were collected from the clinical database. NSCLC was categorized according to the 2015 WHO classification system \u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e, tumor lymph nodes metastasis (TNM) classification and tumor staging were performed according to the 8th edition of the staging system published by the Union for International Cancer Control and the American Joint Committee on Cancer \u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e. Because this was a retrospective nonintervention study, approval of the Medical Research Ethics Committee and Institutional Review Board was waived.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Definition of OLM, VPI, LVI and STAS\u003c/h2\u003e \u003cp\u003eAll enrolled patients underwent preoperative chest contrast-enhanced CT scanning to evaluate the status of LN. The criteria for determining cN0 on CT included all LNs having a short-axis diameter of less than 10 mm without obvious heterogeneous enhancement. Invasive pathological features were obtained from the pathological database and reviewed by a senior pathologist who was unaware of the patients' clinical and radiological outcomes. Information on dissected LN from pathological reports was collected, with involvement of either hilar or mediastinal LN identified as OLM. At least 10 regional LNs were requested to be removed and pathologically examined as American College of Surgeons Commission on Cancer recommended \u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e. VPI was denoted as the invasion of tumor reaches beyond the pleural elastic layer or the surface of visceral pleura. LVI was defined as the infiltration of tumour cells into lymphatic, arterial or venous lumens at the periphery of carcinoma. STAS was considered to exist when the micropapillary clusters, solid nests, or single cells beyond the edge of tumor extending into the air spaces in surrounding lung parenchyma according to the 2015 WHO classification \u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e. Patients with either invasive pathological feature would be categorized into risk group.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. CT scanning protocol\u003c/h2\u003e \u003cp\u003eChest CT examinations were performed using five multidetector CT systems of three types: Lightspeed16, GE Healthcare, Milwaukee, WI, USA; Somatom Sensation 64, Siemens, Erlangen, Germany; Discovery CT750 HD, GE Healthcare. The scanning parameters were: (a) 120 kVp with the automatic regulation of the tube current and 1.5-mm reconstruction thickness and intervals for the 64-detector scanner and (b) 120 kVp, 150\u0026ndash;200 mAs, and 1.25-mm reconstruction thickness intervals for the other two types of scanners. All of the 910 patients underwent contrastenhanced CT. Non-ionic iodinated contrast material (300 mg of iodine per millilitre, Ul-travist; Bayer Pharma, Berlin, Germany) was injected at a dose of 1.3\u0026ndash;1.5 ml/kg body weight at a rate of 2.5 ml/s using an automated injector. CTenhanced scanning was performed with a 70-second delay.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. CT image interpretation and preprocessing\u003c/h2\u003e \u003cp\u003eTwo experienced clinical radiologists, one with 9 years of experience in CT imaging of thoracic malignancies and the other with 6 years of experience, independently analyzed and confirmed the cN0 stage of the CT images after training. Both radiologists were blinded to clinical and pathological information, and any discrepancies in image interpretation were resolved through negotiation and discussion. The CT descriptors and scoring criteria utilized in the analysis were detailed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The images were viewed with a lung window width of 1500 HU and window level of -600 HU, as well as a mediastinal window width of 350 HU at level of 40 HU. The CT descriptors were evaluated on multiplanar reconstructed images and reported with a standardized scoring sheet. Image preprocessing was completed by a linear interpolation algorithm to resample the thickness of CT images to 1 mm. Gaussian filtering was used to preprocess CT images.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCT Characteristics for NSCLC.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDefinition\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eScoring and Definition\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLocation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLung nodule located in the outer third of the lung was defined as peripheral tumor, while others were located centrally\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1, central; 2, peripheral\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMaximum diameter\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe greatest dimension on the multiplanar reconstructed images with a lung window\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ecm\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConsolidation diameter\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe greatest dimension on the multiplanar reconstructed images with a mediastinal window\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ecm\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eContour\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe overall shape of roundness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1, round; 2, oval; 3, somewhat irregular; 4, irregular\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLobulation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA wavy or scalloped configuration of tumor\u0026rsquo;s surface\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpiculation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLines radiating from the margins of the tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTexture\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSolid or GGO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, pure GGO; 1, mixed GGO with solid part\u0026thinsp;\u0026lt;\u0026thinsp;50%;\u003c/p\u003e \u003cp\u003e2, mixed GGO with solid part\u0026thinsp;\u0026gt;\u0026thinsp;50%; 3, solid\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCalcification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAny patterns of calcification in the tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAir bronchogram\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTubelike or branched air structure within the tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBubble-like lucency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAir space in the tumor with diameter\u0026thinsp;\u0026le;\u0026thinsp;5mm at the time of diagnosis prior to biopsy or treatment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCavity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAir space in the tumor with diameter\u0026thinsp;\u0026gt;\u0026thinsp;5mm at the time of diagnosis prior to biopsy or treatment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEnhancement degree\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEnhancement degree\u0026thinsp;=\u0026thinsp;A\u003csub\u003epost\u003c/sub\u003e - A\u003csub\u003epre\u003c/sub\u003e, where A\u003csub\u003epre\u003c/sub\u003e and\u003c/p\u003e \u003cp\u003eA\u003csub\u003epost\u003c/sub\u003e was unenhanced and contrast-enhanced CT attenuation of tumor, respectively\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHU\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEnhancement heterogeneity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHeterogeneity of tumor on contrast-enhanced images\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1, homogeneous; 2, slight or moderate heterogeneous; 3, marked heterogeneous\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRelative enhancement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRelative enhancement E\u003csub\u003erel\u003c/sub\u003e = (A\u003csub\u003epost\u003c/sub\u003e - A\u003csub\u003epre\u003c/sub\u003e )/E\u003csub\u003eart\u003c/sub\u003e, where E\u003csub\u003eart\u003c/sub\u003e was enhancement attenuation of the artery on the same section\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePleural attachment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTumor attaches to the fissure/Pleura\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePleural retraction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetraction of the pleura toward the tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBronchovascular bundle thickening\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eConvergence of vessels to the tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, no significant thickening; 1, obvious thickening\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObstructive change\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eConsolidation shadow caused by obstructive pneumonia or atelectasis at the edge of tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, presence\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePeripheral emphysema\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePeripheral emphysema caused by the tumor or preexisting emphysema\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, slight or moderate; 2, severe\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePeripheral fibrosis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePeripheral fibrosis caused by the tumor or preexisting fibrosis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0, absence; 1, slight or moderate; 2, severe\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eCT, computed tomography; NSCLC, non-small cell lung cancer; GGO, ground-glass opacity.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Tumor segmentation and features extraction\u003c/h2\u003e \u003cp\u003eTumor segmentation was conducted utilizing a manual approach of delineating regions of interest (ROI) on processed CT enhanced scan images with the assistance of ITK-snap 3.6.0 by a radiologist possessing 6 years of experience in image segmentation. The radiologist was provided with information regarding the tumor location, while remaining blinded to additional details. Subsolid tumors were segmented at the lung window, while solid tumors were segmented at the mediastinal window in order to enhance the identification of blood vessels, LNs, and lung consolidation surrounding the tumors. The accuracy of the segmentation results was to be verified or adjusted by another senior radiologist. Radiomics features were extracted using the PyRadiomics 3.0 open-source software program (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.radiomics.io/pyradiomics.html\u003c/span\u003e\u003cspan address=\"http://www.radiomics.io/pyradiomics.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) \u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. A total of 1316 radiomics features were extracted from the 3D ROI of CT images (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6. Feature selection and establishment of radiomics signature\u003c/h2\u003e \u003cp\u003eThe Z-score was utilized to standardize the radiomics parameters across all patients. The Spearman pairwise correlation analysis was utilized to determine the strength of correlations between features, with features exhibiting an absolute correlation value exceeding 0.9 being eliminated. Following this, the top 100 features were chosen through the application of the minimum redundancy maximum relevance (mRMR) method. These selected features were then inputted into least absolute shrinkage and selection operator (LASSO) models to identify the optimal subsets for assessing invasive pathological characteristics. Subsequently, the identified features were utilized to construct a logistic regression model with cross-validation method by a backward step-wise selection to eliminate non-significant variables. The radiomics score (Rad-score) formula was derived through a linear combination of the selected features weighted by their corresponding coefficients with the following formula:\u003c/p\u003e \u003cp\u003eRadscore\u0026thinsp;=\u0026thinsp;b\u0026thinsp;+\u0026thinsp;Ci \u0026times; Xi,\u003c/p\u003e \u003cp\u003ewhere b represents a constant term, Xi denotes the value of the selected feature, and Ci represents the regression coefficient associated with the selected feature. Each patient\u0026rsquo;s Rad-score was calculated by this formula to compare the difference between risk group and non-risk group.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.7. Statistical analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were performed with the statistical software R 4.3.0 and SPSS 26.0. The continuous variables were expressed as mean values and standard deviations, and categorical variables as frequency. Agreement between two readers were analyzed by the ĸ index and Kendall coefficient of concordance. Non-parametric two-sample Wilcoxon test was used for ranked or continuous variables, and chi-square or Fisher\u0026rsquo;s test for categorical variables in univariate analysis. Subsequently, multivariate logistic regression analysis was performed to test the ability of combining Rad-score, radiological and clinical features to identify risk group. Models were assessed for predicative accuracy using the ten-fold cross-validation approach and results over the ten models were averaged. The performance of prediction model was evaluated with calibration curve and Hosmer\u0026ndash;Lemeshow test, then visualized by nomogram. Area under the receiver operating characteristic (ROC) curve (AUC) was used to assess the predictive efficacy, and AUC of different models was compared by the Delong test. P values\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were regarded as statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Reader reproducibility\u003c/h2\u003e \u003cp\u003eAgreement among the two readers was good (Table S2). The intraclass correlation coefficient for maximum diameter, consolidation diameter, degree of enhancement and relative enhancement was 0.91 (range, 0.88\u0026ndash;0.94), 0.92 (range, 0.88\u0026ndash;0.95), 0.85 (range, 0.82\u0026ndash;0.89) and 0.86 (range, 0.83\u0026ndash;0.88), respectively.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Patient demographics\u003c/h2\u003e \u003cp\u003eClinical characteristics and histological subtypes of training group were shown in Table S3. 501, 117 and 60 patients were diagnosed with clinical stage IA, IB and IIA, respectively. 225 (33.2%) cases among these patients had invasive pathological features, including 83, 95, 88, and 122 cases of OLM, VPI, LVI, and STAS confirmed by postoperative pathology. No significant difference of either feature was observed between training and validation group.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Correlation of invasive pathological features with clinical and radiological features\u003c/h2\u003e \u003cp\u003eThe association between clinical features with invasive pathological features was presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Significantly, patients with clinical stage IB and IIA [50/117 (43.9%), 40/60 (66.7%) vs. 135/501 (26.9%)] developed regional LN involvement, VPI, LVI or STAS more frequently than IA patients (odds ratio (OR)\u0026thinsp;=\u0026thinsp;2.81, 95% confidence intervals (CI): 1.97, 4.00 for IB and OR\u0026thinsp;=\u0026thinsp;4.68, 95% CI: 2.66, 8.23 for IIA). No significant association was noted between other clinical features with invasive pathological features.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAssociation between Clinical Characteristics with Risk Factors.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNegative Group\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003cp\u003eGroup\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP Value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUnivariate OR\u003c/p\u003e \u003cp\u003e(95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber\u003c/p\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003cp\u003eSex\u003c/p\u003e \u003cp\u003eMale\u003c/p\u003e \u003cp\u003eFemale\u003c/p\u003e \u003cp\u003eSmoking history\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003cp\u003eNo\u003c/p\u003e \u003cp\u003eFamily history\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003cp\u003eNo\u003c/p\u003e \u003cp\u003eHistological subtype\u003c/p\u003e \u003cp\u003eAdenocarcinoma\u003c/p\u003e \u003cp\u003eSquamous cell carcinoma\u003c/p\u003e \u003cp\u003eOther*\u003c/p\u003e \u003cp\u003eEGFR\u003c/p\u003e \u003cp\u003eMutation\u003c/p\u003e \u003cp\u003eWild\u003c/p\u003e \u003cp\u003eKRAS\u003c/p\u003e \u003cp\u003eMutation\u003c/p\u003e \u003cp\u003eWild\u003c/p\u003e \u003cp\u003eALK\u003c/p\u003e \u003cp\u003ePositive\u003c/p\u003e \u003cp\u003eNegative\u003c/p\u003e \u003cp\u003eClinical stage\u003c/p\u003e \u003cp\u003eIA\u003c/p\u003e \u003cp\u003eIB\u003c/p\u003e \u003cp\u003eIIA\u003c/p\u003e \u003cp\u003eTumor markers\u003c/p\u003e \u003cp\u003eCEA (ng/ml)\u003c/p\u003e \u003cp\u003eCA125 (U/ml)\u003c/p\u003e \u003cp\u003eNSE (ng/ml)\u003c/p\u003e \u003cp\u003eSCC (ng/ml)\u003c/p\u003e \u003cp\u003eCyfra 21\u0026thinsp;\u0026minus;\u0026thinsp;1 (ng/ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e453\u003c/p\u003e \u003cp\u003e59.49 (\u0026plusmn;\u0026thinsp;8.63)\u003c/p\u003e \u003cp\u003e204\u003c/p\u003e \u003cp\u003e249\u003c/p\u003e \u003cp\u003e212\u003c/p\u003e \u003cp\u003e241\u003c/p\u003e \u003cp\u003e108\u003c/p\u003e \u003cp\u003e345\u003c/p\u003e \u003cp\u003e374\u003c/p\u003e \u003cp\u003e70\u003c/p\u003e \u003cp\u003e9\u003c/p\u003e \u003cp\u003e133\u003c/p\u003e \u003cp\u003e88\u003c/p\u003e \u003cp\u003e16\u003c/p\u003e \u003cp\u003e171\u003c/p\u003e \u003cp\u003e7\u003c/p\u003e \u003cp\u003e137\u003c/p\u003e \u003cp\u003e366\u003c/p\u003e \u003cp\u003e67\u003c/p\u003e \u003cp\u003e20\u003c/p\u003e \u003cp\u003e5.73 (\u0026plusmn;\u0026thinsp;7.26)\u003c/p\u003e \u003cp\u003e11.22 (\u0026plusmn;\u0026thinsp;8.91)\u003c/p\u003e \u003cp\u003e7.52 (\u0026plusmn;\u0026thinsp;4.03)\u003c/p\u003e \u003cp\u003e1.63 (\u0026plusmn;\u0026thinsp;0.78)\u003c/p\u003e \u003cp\u003e3.25 (\u0026plusmn;\u0026thinsp;1.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e225\u003c/p\u003e \u003cp\u003e59.65 (\u0026plusmn;\u0026thinsp;8.63)\u003c/p\u003e \u003cp\u003e110\u003c/p\u003e \u003cp\u003e115\u003c/p\u003e \u003cp\u003e112\u003c/p\u003e \u003cp\u003e113\u003c/p\u003e \u003cp\u003e54\u003c/p\u003e \u003cp\u003e171\u003c/p\u003e \u003cp\u003e192\u003c/p\u003e \u003cp\u003e24\u003c/p\u003e \u003cp\u003e9\u003c/p\u003e \u003cp\u003e75\u003c/p\u003e \u003cp\u003e41\u003c/p\u003e \u003cp\u003e5\u003c/p\u003e \u003cp\u003e88\u003c/p\u003e \u003cp\u003e8\u003c/p\u003e \u003cp\u003e89\u003c/p\u003e \u003cp\u003e135\u003c/p\u003e \u003cp\u003e50\u003c/p\u003e \u003cp\u003e40\u003c/p\u003e \u003cp\u003e7.14 (\u0026plusmn;\u0026thinsp;6.25)\u003c/p\u003e \u003cp\u003e12.86 (\u0026plusmn;\u0026thinsp;7.55)\u003c/p\u003e \u003cp\u003e7.19 (\u0026plusmn;\u0026thinsp;4.05)\u003c/p\u003e \u003cp\u003e1.88 (\u0026plusmn;\u0026thinsp;0.86)\u003c/p\u003e \u003cp\u003e3.46 (\u0026plusmn;\u0026thinsp;2.07)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003cp\u003e0.343\u003c/p\u003e \u003cp\u003e0.465\u003c/p\u003e \u003cp\u003e0.927\u003c/p\u003e \u003cp\u003e0.863\u003c/p\u003e \u003cp\u003e0.422\u003c/p\u003e \u003cp\u003e0.341\u003c/p\u003e \u003cp\u003e0.286\u003c/p\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003cp\u003e0.084\u003c/p\u003e \u003cp\u003e0.244\u003c/p\u003e \u003cp\u003e0.672\u003c/p\u003e \u003cp\u003e0.327\u003c/p\u003e \u003cp\u003e0.589\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003cp\u003e2.81 (1.97, 4.00)\u003c/p\u003e \u003cp\u003e4.68 (2.66, 8.23)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eData for age and tumor markers are mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation.\u003c/p\u003e \u003cp\u003eOR, odds ratio; CI, confidence interval; EGFR, epidermal growth factor receptor; KRAS, kirsten rat sarcoma viral oncogene; ALK, anaplastic lymphoma kinase; CEA, carcinoembryonic antigen; CA125, carbohydrate antigen 125; NSE, neuron-specific enolase; SCC, squamous cell carcinoma antigen; Cyfra 21\u0026thinsp;\u0026minus;\u0026thinsp;1, cytokeratin 19 fragment.\u003c/p\u003e \u003cp\u003e*Other histologic subtype contains adenosquamous carcinoma, large cell lung cancer and sarcomatoid carcinoma.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAssociation between Traditional CT Features with Risk Factors.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNegative Group\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003cp\u003eGroup\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP Value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUnivariate OR(95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e453\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e225\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMaximum diameter (cm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.48 (\u0026plusmn;\u0026thinsp;1.03)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.02 (\u0026plusmn;\u0026thinsp;0.97)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConsolidation diameter (cm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.00 (\u0026plusmn;\u0026thinsp;1.32)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.91 (\u0026plusmn;\u0026thinsp;1.04)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBronchovascular bundle\u003c/p\u003e \u003cp\u003ethickening\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e388\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e174\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.007\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.75 (1.16, 2.63)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLobulation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e310\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e143\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e150\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4.34 (3.08, 6.10)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpiculation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e368\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e166\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.54 (1.05, 2.25)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTexture\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e108\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11.57 (6.41, 20.88)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e182\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e175\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5.21 (3.61, 7.52)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObstructive change\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e438\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e208\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.39 (1.17, 4.87)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePleural attachment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e260\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e193\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e148\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.59 (1.86, 3.61)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePleural retraction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e172\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e281\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e164\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.65 (1.16, 2.34)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eData for maximum diameter and consolidation diameter are mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation.\u003c/p\u003e \u003cp\u003eCT, computed tomography; OR, odds ratio; CI, confidence interval.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eUnivariate analysis showed that tumors with a larger overall size (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and solid component size (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), bronchovascular bundle thickening (OR\u0026thinsp;=\u0026thinsp;1.75, 95% CI: 1.16, 2.63; p\u0026thinsp;=\u0026thinsp;0.007), lobulation (OR\u0026thinsp;=\u0026thinsp;4.34, 95% CI: 3.08, 6.10; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), spiculation (OR\u0026thinsp;=\u0026thinsp;1.54, 95% CI: 1.05, 2.25; p\u0026thinsp;=\u0026thinsp;0.025), solid texture (OR\u0026thinsp;=\u0026thinsp;11.57, 95% CI: 6.41, 20.88 for score 2; OR\u0026thinsp;=\u0026thinsp;5.21, 95% CI: 3.61, 7.52 for score 3; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), pleural attachment (OR\u0026thinsp;=\u0026thinsp;2.59, 95% CI: 1.86, 3.61; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), pleural retraction (OR\u0026thinsp;=\u0026thinsp;1.65, 95% CI: 1.16, 2.34; p\u0026thinsp;=\u0026thinsp;0.005) and obstructive change (OR\u0026thinsp;=\u0026thinsp;2.39, 95% CI: 1.17, 4.87; p\u0026thinsp;=\u0026thinsp;0.014) were more likely to develop invasive pathological features. While, other radiological features were not statistically significant (Table S4).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.4. Screening and integration of radiomics features\u003c/h2\u003e \u003cp\u003eA total of 1316 radiomics features were extracted from the 3D ROI of each CT image and Spearman pairwise correlation analysis and the mRMR method were utilized to remove highly correlated features to reduce redundancy. The top 100 features were screened through LASSO algorithm to avoid overfitting and leaving 11 radiomics features (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Six features were excluded after logistic regression in order to remove the non-significant variables, finally leaving five radiomics features to construct radiomics model and generate the Rad-score formula weighted by their respective coefficients (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The predictive accuracy of the radiomics model was evaluated through the cross-validation method. A ten-fold random sample of the dataset was employed to construct a predictive model, which was iteratively repeated to generate ten models. The average results of these models yielded an AUC value of 0.778 for the best predictive accuracy. The Rad-score for each patient was visualized in a waterfall plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) and significant difference was observed between risk group and negative group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe Radiomics Features in Rad-score Formula* after Logistic Regression Analysis.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eRadiomics Features\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eRisk Factor\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eP Value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOdds Ratio\u003c/p\u003e \u003cp\u003e(95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eoriginal_ngtdm_Strength\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eexponential_glszm_SmallAreaEmphasis\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003egradient_firstorder_Minimum\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003ewavelet-LLL_glcm_JointAverage\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003ewavelet-HHH_ngtdm_Strength\u003c/b\u003e\u003c/p\u003e \u003cp\u003eoriginal_glcm_Imc2\u003c/p\u003e \u003cp\u003eexponential_firstorder_InterquartileRange\u003c/p\u003e \u003cp\u003elogarithm_glcm_Idn\u003c/p\u003e \u003cp\u003esquare_gldm_DependenceVariance\u003c/p\u003e \u003cp\u003esquareroot_firstorder_RootMeanSquared\u003c/p\u003e \u003cp\u003ewavelet-HHL_glcm_Imc2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.014\u003c/p\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003cp\u003e0.004 0.024\u003c/p\u003e \u003cp\u003e0.049\u003c/p\u003e \u003cp\u003eNA\u003c/p\u003e \u003cp\u003eNA\u003c/p\u003e \u003cp\u003eNA\u003c/p\u003e \u003cp\u003eNA\u003c/p\u003e \u003cp\u003eNA\u003c/p\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.69 (0.52, 0.93)\u003c/p\u003e \u003cp\u003e2.05 (1.45, 2.90)\u003c/p\u003e \u003cp\u003e0.71 (0.56, 0.90)\u003c/p\u003e \u003cp\u003e1.51 (1.06, 2.15)\u003c/p\u003e \u003cp\u003e0.22 (0.05, 0.99)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eRad-score, radiomics score; CI, confidence interval; NA, not applicable (variables that were not included in the equation of multivariate logistic regression analysis with backward stepwise selection).\u003c/p\u003e \u003cp\u003e*Rad-score = -1.294-0.37*original_ngtdm_Strength\u0026thinsp;+\u0026thinsp;0.717*exponential_glszm_SmallAreaEmphasis\u003c/p\u003e \u003cp\u003e-0.342*gradient_firstorder_Minimum\u0026thinsp;+\u0026thinsp;0.409*wavelet-LLL_glcm_JointAverage-1.525*\u003c/p\u003e \u003cp\u003ewavelet-HHH_ngtdm_Strength\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.5. Predictive model and validation test\u003c/h2\u003e \u003cp\u003eThe multivariable logistic regression model of clinicoradiological features and radiomics features revealed that larger consolidation diameter (OR\u0026thinsp;=\u0026thinsp;1.26, 95% CI: 1.02, 1.55, p\u0026thinsp;=\u0026thinsp;0.032), pleural attachment (OR\u0026thinsp;=\u0026thinsp;1.70, 95% CI: 1.12, 2.59, p\u0026thinsp;=\u0026thinsp;0.013), texture (OR\u0026thinsp;=\u0026thinsp;1.56, 95% CI: 1.33, 1.83, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 for score 2; OR\u0026thinsp;=\u0026thinsp;1.85, 95% CI: 1.09, 3.13, p\u0026thinsp;=\u0026thinsp;0.032 for score 3) and Rad-score (OR\u0026thinsp;=\u0026thinsp;4.61, 95% CI: 1.99, 10.64, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) were significant independent predictors (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The result of Hosmer-Lemeshow goodnessoffit test was not statistically significant (p\u0026thinsp;=\u0026thinsp;0.833), indicating a high concordance between the predicted and observed probabilities. Calibration curves, ROC curves and nomogram were drawn in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMultivariable Logistic Regression Analysis of Radiological Features Combined with Rad-score Predicting the Presence of Risk Factors.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eRisk Factors\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eP Value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOdds Ratio\u003c/p\u003e \u003cp\u003e(95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"16\" rowspan=\"17\"\u003e \u003cp\u003e\u003cb\u003eConsolidation diameter\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003ePleural attachment\u003c/b\u003e\u003c/p\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e1\u003c/p\u003e \u003cp\u003e\u003cb\u003eTexture\u003c/b\u003e\u003c/p\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e1\u003c/p\u003e \u003cp\u003e2\u003c/p\u003e \u003cp\u003e3\u003c/p\u003e \u003cp\u003e\u003cb\u003eRad-score\u003c/b\u003e\u003c/p\u003e \u003cp\u003eMaximum diameter\u003c/p\u003e \u003cp\u003eBronchovascular bundle thickening\u003c/p\u003e \u003cp\u003eLobulation\u003c/p\u003e \u003cp\u003eSpiculation\u003c/p\u003e \u003cp\u003eObstructive change\u003c/p\u003e \u003cp\u003ePleural retraction\u003c/p\u003e \u003cp\u003eClinical stage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.032\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.26 (1.02, 1.55)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.70 (1.12, 2.59)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.56 (1.33, 1.83)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.032\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.85 (1.09, 3.13)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.61 (1.99, 10.64)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eRad-score, radiomics score; CI, confidence interval; NA, not applicable (variables that were not included in the equation of multivariate logistic regression analysis with backward stepwise selection).\u003c/p\u003e \u003cp\u003eFormula: ex / (1\u0026thinsp;+\u0026thinsp;ex ), x = -0.516\u0026thinsp;+\u0026thinsp;0.229 \u0026times; consolidation diameter\u0026thinsp;+\u0026thinsp;0.532 \u0026times; pleural attachment\u0026thinsp;+\u0026thinsp;1.526 \u0026times; texture (solid part\u0026thinsp;\u0026gt;\u0026thinsp;50%)\u0026thinsp;+\u0026thinsp;0.446 \u0026times; Rad-score\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe AUC of combined model increased to 0.815 (95% CI: 0.779, 0.850), compared with 0.778 (95% CI: 0.740, 0.817; p\u0026thinsp;=\u0026thinsp;0.275) and 0.691 (95% CI: 0.646, 0.736; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) when Rad-score or traditional radiological feature was used alone (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). In the validation group, the accuracy of the combined prediction model was reasonable with an AUC of 0.792 (95% CI: 0.756, 0.828), compared with 0.745 (95% CI: 0.716, 0.783; p\u0026thinsp;=\u0026thinsp;0.038) and 0.701 (95% CI: 0.661, 0.739; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) when utilizing either the Rad-score or traditional radiological feature in isolation (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003eB). The combined model exhibited a sensitivity of 86.56%, specificity of 81.63%, and accuracy of 81.30%, and in the external validation set, achieved a sensitivity of 74.32%, specificity of 80.49%, and accuracy of 78.65% when the cutoff was determined at the maximum Youden index.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThe efficacy of limited surgery in early NSCLC patients has recently been demonstrated as significant, resulting in the preservation of more functional lung tissue and reducing perioperative mortality compared to standard lobectomy \u003csup\u003e[31, 32]\u003c/sup\u003e. Although the National Comprehensive Cancer Network guidelines recommend sublobar resection for peripheral nodules\u0026nbsp;≤2 cm with\u0026nbsp;≥50% ground-glass appearance on CT or a long doubling time (≥400 days), there remains a lack of definitive selection criteria for limited resection \u003csup\u003e[33]\u003c/sup\u003e. In order to determine the suitability of limited resection for early-stage NSCLC patients, precise LN staging is essential to verify the absence of LN involvement. Additionally, the invasive pathological characteristics of the primary tumor, such as VPI, LVI, and STAS, serve as significant prognostic indicators for both local recurrence and distant metastasis, ultimately impacting the overall survival rate of NSCLC patients who undergo limited resection \u003csup\u003e[7-10]\u003c/sup\u003e. In cases where the N0 status is uncertain or there is a high likelihood of pathological invasiveness, lobectomy and systematic LN dissection should be performed instead of limited surgery. We conducted a comprehensive analysis of 11 retrospective studies and meta-analyses to determine the average incidence rates of OLM, VPI, LVI, and STAS in lung cancer, which were found to be 14.0% (range 9.8% to 15.9%), 21.4% (range 11.8% to 23.0%), 23.1% (range 12.4% to 33.6%), and 31.3% (range 15.5% to 43.3%), respectively \u003csup\u003e[9-12, 16, 17, 34-38]\u003c/sup\u003e. These findings align with our own results, underscoring the significance of accurately predicting invasive pathological factors non-invasively prior to surgery.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCT emerges as a pivotal tool for providing detailed imaging information and is therefore routinely employed in clinical practice. The findings from the reports indicate that CT features of the primary tumor, such as tumor size and proportion of solid component, were identified as independent predictors of invasive pathological features \u003csup\u003e[16, 18, 39, 40]\u003c/sup\u003e. It was observed that the maximum consolidation diameter served as an independent prognostic factor, as opposed to the overall size of the tumor. In accordance with the guidelines set forth by the Union for International Cancer Control, the diameter of the invasive component within the tumor should be considered as a criterion for T staging, as it was found to be more predictive of prognosis \u003csup\u003e[41, 42]\u003c/sup\u003e. Ground-glass opacity (GGO) predominant pulmonary nodules in CT scans are commonly believed to be associated with a low likelihood of invasive pathology \u003csup\u003e[43]\u003c/sup\u003e. Patients with GGO predominant NSCLC have been shown to have a more favorable prognosis following surgical resection \u003csup\u003e[44]\u003c/sup\u003e, suggesting that they may be suitable candidates for less extensive surgical procedures. Suzuk et al. (2011) demonstrated that a consolidation-to-tumor ratio of less than 0.25 or 0.5 on CT scans could accurately predict the absence of LVI or LN involvement \u003csup\u003e[18]\u003c/sup\u003e, with similar results of other researches on predicting VPI and STAS \u003csup\u003e[16, 40]\u003c/sup\u003e. These results align with our own research findings. Multiple studies have demonstrated that STAS is not typically present in pure ground glass opacity nodules (pGGN) \u003csup\u003e[17, 45, 46]\u003c/sup\u003e. In our research, only one patient with pGGN was pathologically confirmed to have VPI with a lepidic pattern adenocarcinoma (ADC). Although 44 out of 93 cases (47.3%) of pGGN were found to have microinvasive adenocarcinoma (MIA) upon postoperative pathology examination, the existence of invasive pathological characteristics in pGGN should not be disregarded.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePleural attachment is commonly associated with a poor prognosis in NSCLC and is widely acknowledged as an independent predictor of VPI \u003csup\u003e[20, 40, 47]\u003c/sup\u003e. In our research, pleural attachment demonstrated a robust predictive capability in the collective prediction of invasive pathological characteristics, despite the lack of significant associations between pleural attachment and OLM, LVI, and STAS, in contrast to the findings of Lin et al. (2021), who suggested that pleural attachment exhibited strong diagnostic accuracy for STAS\u003csup\u003e\u0026nbsp;[16]\u003c/sup\u003e. This discrepancy may be attributed to variations in study populations and inclusion criteria. Qi et al. (2016) posited that pleural retraction served as an independent risk factor for VPI\u003csup\u003e\u0026nbsp;[48]\u003c/sup\u003e, although there remains ongoing debate regarding the potential association between this peritumoral feature with the malignancy of the lesion. Gallagher et al. (1990) posited that pleural retraction may be attributed to elastasis, inflammatory invasion, and thick fibrous proliferation, suggesting it is merely indicative of pleural fiber tension \u003csup\u003e[49]\u003c/sup\u003e. Our univariate analysis indicated a correlation between pleural retraction and invasive pathological characteristics, yet it did not emerge as a standalone predictive factor. Numerous studies have identified lobulation and spiculation as independent predictors of OLM, LVI and STAS\u003csup\u003e\u0026nbsp;[16, 37]\u003c/sup\u003e, which were highly related to invasive pathological features in our study, but not independent predictors.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe relationship between ADC and squamous cell carcinoma (SCC) subtypes of NSCLC and invasive pathological features was found to be significant in our study, although they were not identified as independent predictors. ADC and SCC are the predominant subtypes of NSCLC, with ADC representing the majority of cases in our study. However, 94 cases of SCC were excluded from the study due to difficulties in lesion identification on chest CT scans. These lesions were primarily endobronchial in nature, with some being obscured by distal atelectasis or pneumonia on imaging. The classification of lung ADC by the 2021 WHO guidelines led to the exclusion of carcinoma in situ (CIS) from the study sample (n=18) \u003csup\u003e[50]\u003c/sup\u003e. In order to reduce selective bias as well as the inability to confirm the pathology of MIA prior to surgery, MIA was not excluded from the study, though none of these cases exhibited invasive pathological features. Previous studies have identified LVI and STAS as significant risk factors for OLM, leading to upstaging of the N category \u003csup\u003e[35, 51]\u003c/sup\u003e. Vaghjiani et al. (2020) found that VPI was more prevalent in tumors with STAS positivity \u003csup\u003e[12]\u003c/sup\u003e, contributing to T-stage progression from T1 to T2 in lung cancer with a diameter less than 3 cm \u003csup\u003e[28]\u003c/sup\u003e. There is ample evidence to suggest that the presence of LVI or STAS may serve as an initial mechanism in the progression of tumor development by disseminating tumor cells beyond the confines of the primary carcinoma into lymphatic, vessel, or air spaces, ultimately leading to the development of OLM or VPI.\u0026nbsp;It is recognized that certain pathological characteristics, specifically micropapillary and solid patterns, are strongly associated with a negative prognosis in NSCLC\u003csup\u003e\u0026nbsp;[52]\u003c/sup\u003e. However, when these features are present in a small proportion, they may not necessarily pose a significant risk for limited resection, making prediction based on imaging features more challenging. Additionally, extracapsular extension was reported to occur more frequently in advanced stages of lung cancer\u003csup\u003e\u0026nbsp;[53]\u003c/sup\u003e, and the detection of micro-metastases may vary among different institutions due to differences in testing methods, leading to a lack of uniformity in diagnostic accuracy\u003csup\u003e\u0026nbsp;[54]\u003c/sup\u003e.\u0026nbsp;Therefore, the above invasive pathological features were not included in this study of early-stage NSCLC based on CT radiomics.\u003c/p\u003e\n\u003cp\u003eRadiomics, the process of converting radiographic images into quantifiable data, plays a crucial role in developing predictive models and potentially enhancing diagnostic precision \u003csup\u003e[55]\u003c/sup\u003e. In a radiomics study conducted by Jiang et al. (2020), the Random Forest algorithm was utilized to develop a predictive model that demonstrated efficacy in forecasting STAS positive tumors with an AUC of 0.7\u003csup\u003e\u0026nbsp;[56]\u003c/sup\u003e. Zhong et al. (2023) reported significant predictive value in the deep learning signature of primary tumors based on PET-CT scans for OLM of lung cancer\u003csup\u003e\u0026nbsp;[26]\u003c/sup\u003e. In this research, Spearman pairwise correlation analysis and the mRMR method were employed to eliminate highly correlated features and minimize redundancy. Subsequently, the LASSO algorithm and logistic regression analysis were employed to streamline the selection of radiomics features and construct the prediction model. This method is commonly utilized in the feature selection process, particularly for high dimensional data, and has shown promise in reducing the multilinearity between features \u003csup\u003e[57]\u003c/sup\u003e. We incorporated the relevant radiomics features into the Rad-score and developed a combined prediction model represented by a nomogram with the AUC of 0.815. This integration led to a significant enhancement in predictive accuracy compared to the traditional CT features model (p\u0026lt;0.001). Consequently, individuals at high risk for limited surgery in early-stage NSCLC can be identified, while sublobectomy may be considered a viable option for those in the low-risk group.\u003c/p\u003e\n\u003cp\u003eOur study encountered several limitations. Firstly, it was a retrospective study focusing solely on clinical IA-IIA stage NSCLC cases to enhance the precision of the predictive model, thereby restricting its generalizability. Additionally, manual tumor segmentation proved to be time-consuming, requiring frequent corrections to delineate the boundaries of adjacent structures such as blood vessels, bronchi, chest wall, and mediastinum, particularly on CT images with a 1 mm slice thickness. Furthermore, the validation cohort from another medical center was inadequately sized, highlighting the need for additional prospective data from multiple sources.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, this study provided a noninvasive prediction tool that combined traditional radiological characteristics and radiomics features based on the preoperative CT enhanced image. The prediction model can be incorporated into the specific treatment strategy decision-making process of NSCLC patients with clinical stage IA-IIA, and determine the subgroup of patients appropriate to limited surgery.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003e1 guarantor of integrity of the entire study --------- Fengnian Zhao, Wang jiang2 study concepts and design --------------------------- Fengnian Zhao, Wang jiang3 literature research ------------------------------------- Yunqing Zhao, Xiaoxue Wang4 clinical studies ----------------------------------------- Qingna Yan, Yunqing Zhao5 experimental studies / data analysis --------------- Fengnian Zhao, Dong LI6 statistical analysis ------------------------------------- Yunqing Zhao7 manuscript preparation ------------------------------ Wang Jiang, Xiaoxue Wang8 manuscript editing ------------------------------------ Dong Li, Guiming Zhou\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData is provided within the manuscript and supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSiegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7-33.\u003c/li\u003e\n\u003cli\u003eMiller KD, Nogueira L, Devasia T, Mariotto AB, Yabroff KR, Jemal A, et al. Cancer treatment and survivorship statistics, 2022. CA Cancer J Clin. 2022;72(5):409-36.\u003c/li\u003e\n\u003cli\u003eHoy H, Lynch T, Beck M. Surgical Treatment of Lung Cancer. Crit Care Nurs Clin North Am. 2019;31(3):303-+.\u003c/li\u003e\n\u003cli\u003eSuzuki K, Saji H, Aokage K, Watanabe S-i, Okada M, Mizusawa J, et al. Comparison of pulmonary segmentectomy and lobectomy: Safety results of a randomized trial. J Thorac Cardiovasc Surg. 2019;158(3):895-907.\u003c/li\u003e\n\u003cli\u003eWen Z, Zhao Y, Fu F, Hu H, Sun Y, Zhang Y, et al. Comparison of outcomes following segmentectomy or lobectomy for patients with clinical N0 invasive lung adenocarcinoma of 2 cm or less in diameter. J Cancer Res Clin Oncol. 2020;146(6):1603-13.\u003c/li\u003e\n\u003cli\u003eBeck KS, Gil B, Na SJ, Hong JH, Chun SH, An HJ, 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.\u003c/li\u003e\n\u003cli\u003eAdachi H, Sakamaki K, Nishii T, Yamamoto T, Nagashima T, Ishikawa Y, et al. Lobe-Specific Lymph Node Dissection as a Standard Procedure in Surgery for Non-Small Cell Lung Cancer: A Propensity Score Matching Study. J Thorac Oncol. 2017;12(1):85-93.\u003c/li\u003e\n\u003cli\u003eBlaauwgeers H, Flieder D, Warth A, Harms A, Monkhorst K, Witte B, et al. A Prospective Study of Loose Tissue Fragments in Non-Small Cell Lung Cancer Resection Specimens: An Alternative View to \u0026ldquo;Spread Through Air Spaces\u0026rdquo;. Am J Surg Pathol (2017) 41:1226. \u003c/li\u003e\n\u003cli\u003eLiu QX, Deng XF, Zhou D, Li JM, Min JX, Dai JG. Visceral pleural invasion impacts the prognosis of non-small cell lung cancer: a meta-analysis. Eur J Surg Oncol 2016; 42(11):1707\u0026ndash;1713. \u003c/li\u003e\n\u003cli\u003eShiono S, Abiko M, Sato T. Positron emission tomography/computed tomography and lymphovascular invasion predict recurrence in stage I lung cancers. J Thorac Oncol 2011;6:43\u0026ndash;47.\u003c/li\u003e\n\u003cli\u003eZhong YF, Cai C, Chen T, Gui H, Chen C, Deng JJ, et al. PET/CT-based deep learning grading signature to optimize surgical decisions for clinical stage I invasive lung adenocarcinoma and biologic basis under its prediction: a multicenter study. Eur J Nucl Med Mol Imaging. 2024;51(2):521-34.\u003c/li\u003e\n\u003cli\u003eVaghjiani RG, Takahashi Y, Eguchi T, Lu SH, Kameda K, Tano Z, et al. Tumor Spread Through Air Spaces Is a Predictor of Occult Lymph Node Metastasis in Clinical Stage IA Lung Adenocarcinoma. J Thorac Oncol. 2020;15(5):792-802.\u003c/li\u003e\n\u003cli\u003ePark HK, Jeon K, Koh W-J, Suh GY, Kim H, Kwon OJ, et al. Occult nodal metastasis in patients with non-small cell lung cancer at clinical stage IA by PET/CT. Respirology. 2010;15(8):1179-84.\u003c/li\u003e\n\u003cli\u003eTanaka T, Shinya T, Sato S, Mitsuhashi T, Ichimura K, Soh J, et al. Predicting pleural invasion using HRCT and 18F-FDG PET/CT in lung adenocarcinoma with pleural contact. Ann Nucl Med 2015; 29(9):757\u0026ndash;765. \u003c/li\u003e\n\u003cli\u003eNaidich DP. Is spread of tumor through air spaces a concern for interpreting lung nodules on CT images? Radiology (2018) 289:841\u0026ndash;2. \u003c/li\u003e\n\u003cli\u003eQi L, Xue K, Cai YJ, Lu JJ, Li XH, Li M. Predictors of CT Morphologic Features to Identify Spread Through Air Spaces Preoperatively in Small-Sized Lung Adenocarcinoma. Front Oncol. 2021;10.\u003c/li\u003e\n\u003cli\u003eKim SK, Kim TJ, Chung MJ, Kim TS, Lee KS, Zo JI, et al. Lung Adenocarcinoma: CT Features Associated with Spread through Air Spaces. Radiology. 2018;289(3):831-40.\u003c/li\u003e\n\u003cli\u003eSuzuki K, Koike T, Asakawa T, Kusumoto M, Asamura H, Nagai K, et al. A prospective radiological study of thin-section computed tomography to predict pathological noninvasiveness in peripheral clinical IA lung cancer (Japan Clinical Oncology Group 0201). J Thorac Oncol (2011) 6(4):751\u0026ndash;6. \u003c/li\u003e\n\u003cli\u003eZhao LL, Xie HK, Zhang LP, Zha JY, Zhou FY, Jiang GN, et al. Visceral pleural invasion in lung adenocarcinoma \u0026lt;=3 cm with ground-glass opacity: a clinical, pathological and radiological study. J Thorac Dis 2016; 8(7):1788\u0026ndash;1797.\u003c/li\u003e\n\u003cli\u003eSeok Y, Lee E. Visceral pleural invasion is a significant prognostic factor in patients with partly solid lung adenocarcinoma sized 30 mm or smaller. Thorac Cardiovasc Surg 2018; 66(2):150\u0026ndash;155. \u003c/li\u003e\n\u003cli\u003eAerts HJ, Velazquez ER, Leijenaar RT, Parmar C, Grossmann P, Carvalho S, et al. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat Commun 2014;5:4006.\u003c/li\u003e\n\u003cli\u003eHuang Y, Liu Z, He L, Chen X, Pan D, Ma Z, et al. Radiomics signature: A potential biomarker for the prediction of disease-free survival in early-stage (I or II) non-small cell lung cancer. Radiology 2016;281:947-57.\u003c/li\u003e\n\u003cli\u003eChristie J, Romine P, Nair V, Mattonen S. Prediction of pathologic lymphovascular invasion in non-small cell lung cancer using multi-modality tumour and peri-tumoural radiomics. Med Phys. 2022;49(8):5650-1.\u003c/li\u003e\n\u003cli\u003eZha XY, Liu YQ, Ping XX, Bao JY, Wu Q, Hu S, et al. A Nomogram Combined Radiomics and Clinical Features as Imaging Biomarkers for Prediction of Visceral Pleural Invasion in Lung Adenocarcinoma. Front Oncol. 2022;12.\u003c/li\u003e\n\u003cli\u003eBassi M, Russomando A, Vannucci J, Ciardiello A, Dolciami M, Ricci P, et al. Role of radiomics in predicting lung cancer spread through air spaces in a heterogeneous dataset. Transl Lung Cancer R. 2022;11(4):560-71.\u003c/li\u003e\n\u003cli\u003eZhong YF, Cai C, Chen T, Gui H, Deng JJ, Yang ML, et al. PET/CT based cross-modal deep learning signature to predict occult nodal metastasis in lung cancer. Nat Commun. 2023;14(1).\u003c/li\u003e\n\u003cli\u003eTravis WD, Brambilla E, Nicholson AG, Yatabe Y, Austin JHM, Beasley MB, et al. The 2015 World Health Organization Classification of Lung Tumors Impact of Genetic, Clinical and Radiologic Advances Since the 2004 Classification. J Thorac Oncol. 2015;10(9):1243-60.\u003c/li\u003e\n\u003cli\u003eRami-Porta R, Bolejack V, Crowley J, Ball D, Kim J, Lyons G, et al. The IASLC Lung Cancer Staging Project Proposals for the Revisions of the T Descriptors in the Forthcoming Eighth Edition of the TNM Classification for Lung Cancer. J Thorac Oncol. 2015;10(7):990-1003.\u003c/li\u003e\n\u003cli\u003eLiu JB, Huffman KM, Palis BE, Shulman LN, Winchester DP, Ko CY, et al. Reliability of the American College of Surgeons Commission on Cancer\u0026apos;s Quality of Care Measures for Hospital and Surgeon Profiling. J Am Coll Surg. 2017;224(2):180-+.\u003c/li\u003e\n\u003cli\u003eVan Griethuysen JJM, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res. 2017;77(21):E104-E7.\u003c/li\u003e\n\u003cli\u003eCao JL, Yuan P, Wang YQ, Xu JM, Yuan XS, Wang ZT, et al. Survival Rates After Lobectomy, Segmentectomy, and Wedge Resection for Non-Small Cell Lung Cancer. Ann Thorac Surg. 2018;105(5):1483-91.\u003c/li\u003e\n\u003cli\u003eOkada M. Radical Sublobar Resection for Small-Diameter Lung Cancers. Thorac Surg Clin. 2013;23(3):301-+.\u003c/li\u003e\n\u003cli\u003eEttinger DS, Wood DE, Aisner DL, Akerley W, Bauman JR, Bharat A, et al. Non-Small Cell Lung Cancer, Version 2.2021 Featured Updates to the NCCN Guidelines. J Natl Compr Canc Ne. 2021;19(3):254-66.\u003c/li\u003e\n\u003cli\u003eMoon Y, Choi SY, Park JK, Lee KY. Risk Factors for Occult Lymph Node Metastasis in Peripheral Non-Small Cell Lung Cancer with Invasive Component Size 3 cm or Less. World J. Surg. 2020;44(5):1658-65.\u003c/li\u003e\n\u003cli\u003eSeto K, Kuroda H, Yoshida T, Sakata S, Mizuno T, Sakakura N, et al. Higher frequency of occult lymph node metastasis in clinical N0 pulmonary adenocarcinoma with ALK rearrangement. Cancer Manag. Res. 2018;10:2117-24.\u003c/li\u003e\n\u003cli\u003eDai CY, Xie HK, Su H, She YL, Zhu EJ, Fan ZW, et al. Tumor Spread through Air Spaces Affects the Recurrence and Overall Survival in Patients with Lung Adenocarcinoma \u0026gt;2 to 3 cm. J Thorac Oncol (2017) 12:1052\u0026ndash;60.\u003c/li\u003e\n\u003cli\u003eYang GJ, Nie P, Zhao LZ, Guo J, Xue W, Yan L, et al. 2D and 3D texture analysis to predict lymphovascular invasion in lung adenocarcinoma. Eur J Radiol. 2020;129.\u003c/li\u003e\n\u003cli\u003eOnozato Y, Nakajima T, Yokota H, Morimoto J, Nishiyama A, Toyoda T, et al. Radiomics is feasible for prediction of spread through air spaces in patients with nonsmall cell lung cancer. Sci. Rep. 2021;11(1).\u003c/li\u003e\n\u003cli\u003eZang RC, Qiu B, Gao SG, He J. A Model Predicting Lymph Node Status for Patients with Clinical Stage T1aN0-2M0 Nonsmall Cell Lung Cancer. Chin Med J. 2017;130(4):398-403.\u003c/li\u003e\n\u003cli\u003eChen ZF, Jiang SX, Li ZL, Rao LJ, Zhang XS. Clinical Value of F-18-FDG PET/CT in Prediction of Visceral Pleural Invasion of Subsolid Nodule Stage I Lung Adenocarcinoma. Acad Radiol. 2020;27(12):1691-9.\u003c/li\u003e\n\u003cli\u003eTravis WD, Asamura H, Bankier AA, Beasley MB, Detterbeck F, Flieder DB, et al. The IASLC Lung Cancer Staging Project: Proposals for Coding T Categories for Subsolid Nodules and Assessment of Tumor Size in Part-Solid Tumors in the Forthcoming Eighth Edition of the TNM Classification of Lung Cancer. J Thorac Oncol. 2016;11(8):1204-23.\u003c/li\u003e\n\u003cli\u003eRami-Porta R, Asamura H, Travis WD, Rusch VW. Lung Cancer-Major Changes in the American Joint Committee on Cancer Eighth Edition Cancer Staging Manual. Ca-Cancer J Clin. 2017;67(2):138-55.\u003c/li\u003e\n\u003cli\u003eTaylor ML, Carmona F, Thiagarajan RR, Westgate L, Ferguson MA, del Nido PJ, et al. Mild Postoperative Acute Kidney Injury and Outcomes After Surgery for Congenital Heart Disease. J Thorac Cardiovasc Surg (2013) 146(1):146\u0026ndash;52. \u003c/li\u003e\n\u003cli\u003eOhde Y, Nagai K, Yoshida J, Nishimura M, Takahashi K, Suzuki K, et al. The Proportion of Consolidation to Ground-Glass Opacity on High Resolution CT Is a Good Predictor for Distinguishing the Population of Non-Invasive Peripheral Adenocarcinoma. Lung Cancer (2003) 42(3):303\u0026ndash;10. \u003c/li\u003e\n\u003cli\u003eLedda RE, Milanese G, Gnetti L, Borghesi A, Sverzellati N, Silva M, et al. Spread through air spaces in lung adenocarcinoma: is radiology reliable yet? J Thorac Dis (2019) 11:S256\u0026ndash;61. \u003c/li\u003e\n\u003cli\u003eToyokawa G, Yamada Y, Tagawa T, Kamitani T, Yamasaki Y, Shimokawa M, et al. Computed tomography features of resected lung adenocarcinomas with spread through air spaces. J Thorac Cardiovasc Surg (2018) 156:1670\u0026ndash;6.\u003c/li\u003e\n\u003cli\u003eTang EK, Chen CS, Wu CC, Wu MT, Yang TL, Liang HL, et al. Natural history of persistent pulmonary subsolid nodules: long-term observation of different interval growth. Heart Lung Circ 2019; 28(11):1747\u0026ndash;1754. \u003c/li\u003e\n\u003cli\u003eQi LP, Li XT, Yang Y, Chen JF, Wang J, Chen ML, et al. Multivariate analysis of pleural invasion of peripheral non-small cell lung cancer-based computed tomography features. J Comput Assist Tomogr. 2016; 40(5):757\u0026ndash;762. \u003c/li\u003e\n\u003cli\u003eGallagher B, Urbanski SJ. The signifificance of pleural elastica invasion by lung carcinomas. Hum Pathol 1990; 21(5):512\u0026ndash;517.\u003c/li\u003e\n\u003cli\u003eNicholson AG, Tsao MS, Beasley MB, Borczuk AC, Brambilla E, Cooper WA, et al. The 2021 WHO Classification of Lung Tumors: Impact of Advances Since 2015. J Thorac Oncol. 2022;17(3):362-87.\u003c/li\u003e\n\u003cli\u003eDecaluwe H, Moons J, Fieuws S, De Wever W, Deroose C, Stanzi A, et al. Is central lung tumour location really predictive for occult mediastinal nodal disease in (suspected) non-small-cell lung cancer staged cN0 on F-18-fluorodeoxyglucose positron emission tomography-computed tomography? Eur J Cardiothorac Surg. 2018;54(1):134-40.\u003c/li\u003e\n\u003cli\u003eMotono N, Mizoguchi T, Ishikawa M, Iwai S, Iijima Y, Uramoto H. Predictive value of recurrence of solid and micropapillary subtype in lung adenocarcinoma. Oncology. 2023.\u003c/li\u003e\n\u003cli\u003eChen DL, Ding QF, Wang W, Chen C, Chen YB. ASO Author Reflections: Old Song, New Sung-Extracapsular Extension in Lung Cancer in the Era of Eighth-Edition N Classification. Ann Surg Oncol. 2021;28(4):2099-100.\u003c/li\u003e\n\u003cli\u003eBelanger AR, Hollyfield J, Yacovone G, Ceppe AS, Akulian JA, Burks AC, et al. Incidence and clinical relevance of non-small cell lung cancer lymph node micro-metastasis detected by staging endobronchial ultrasound-guided transbronchial needle aspiration. J Thorac Dis. 2019;11(8):3649-57.\u003c/li\u003e\n\u003cli\u003eGillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016;278(2):563-77.\u003c/li\u003e\n\u003cli\u003eJiang CS, Luo Y, Yuan JL, You SY, Chen ZQ, Wu MX, et al. CT-based radiomics and machine learning to predict spread through air space in lung adenocarcinoma. Eur Radiol. 2020;30(7):4050-7.\u003c/li\u003e\n\u003cli\u003eArya M, Sastry GH, Motwani A, Kumar S, Zaguia A. A novel extra tree ensemble optimized dl framework (Eteodl) for early detection of diabetes. Front Public Health (2021) 9:797877.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Computed Tomography, Radiomics, Invasive Pathological Features, Non-small Cell Lung Cancer, Predictive Model","lastPublishedDoi":"10.21203/rs.3.rs-4488259/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4488259/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e \u003cp\u003eLimited surgery has received increasing attention to minimize damage and preserve more functional lung tissue. However, invasive pathological features including occult lymph node metastasis, visceral pleural invasion, lymphovascular invasion and tumor spread through air spaces may become risk factors for prognosis after limited surgery. The aim of this study was to unitedly predict these invasive pathological features based on computed tomography (CT) radiomics in patients with early stage non-small cell lung cancer (NSCLC).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eFrom January 2016 to February 2023, 910 patients with clinical stage IA-IIA NSCLC underwent resection and were divided into training and validation group based on different institution. Radiomics features were extracted by the PyRadiomics software after tumor lesion segmentation and screened by spearman correlation analysis, minimum redundancy maximum relevance and the least absolute shrinkage and selection operator regression analysis. Univariate analysis followed by multivariable logistic regression were performed to estimate the independent predictors. A predictive model was established with visual nomogram and external validation, and evaluated in terms of area under the receiver operating characteristic curve (AUC).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003e225 patients had invasive pathological features (33.2%), and four independent predictors were identified: larger consolidation diameter (p\u0026thinsp;=\u0026thinsp;0.032), pleural attachment (p\u0026thinsp;=\u0026thinsp;0.013), texture (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and Rad-score (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The combined model showed good calibration with an AUC of 0.815, compared with 0.778 and 0.691 when radiomics or traditional CT features were used alone. For the validation group, the AUC was 0.792, compared with 0.745 and 0.701 in radiomics or traditional CT features model.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eOur predictive model can non-invasively assess the risk of invasive pathological features in patients with clinical stage IA-IIA NSCLC, enable surgeons perform more reasonable and individualized treatment choices.\u003c/p\u003e","manuscriptTitle":"United Predictability of CT radiomics on invasive pathological features in clinical stage IA-IIA non-small cell lung cancer: a double-center study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-14 18:38:38","doi":"10.21203/rs.3.rs-4488259/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"17af4845-a1d9-47a1-963d-1e71379d5b13","owner":[],"postedDate":"June 14th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-08-30T10:43:48+00:00","versionOfRecord":[],"versionCreatedAt":"2024-06-14 18:38:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4488259","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4488259","identity":"rs-4488259","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}