Prediction of EGFR Mutation Status in Non-small Cell Lung Cancer Based on Multiparametric MRI Radiomics

Prediction of EGFR Mutation Status in Non-small Cell Lung Cancer Based on Multiparametric MRI Radiomics | 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 Prediction of EGFR Mutation Status in Non-small Cell Lung Cancer Based on Multiparametric MRI Radiomics Yubo Wang, Hao Hu, Yadan Yin, Jiyun Zhang, Yang Fu, Jiageng Li, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7663043/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 12 Dec, 2025 Read the published version in BMC Medical Imaging → Version 1 posted 10 You are reading this latest preprint version Abstract Background: To establish and validate a predictive model based on magnetic resonance imaging (MRI) radiomics features combined with clinicopathological factors to predict the mutation status of epidermal growth factor receptor (EGFR) in non-small cell lung cancer（NSCLC). Patients and methods: A total of 91 NSCLC patient (72 in the training cohort and 19 in the validation cohort) were included in this study. A total of 1708 radiomics features were extracted from the MRI (T2w and CET1w) sequences. The variance threshold method combined with the univariate selection method, least absolute shrinkage and selection operator (LASSO) regression were used to screen important radiomics features, calculate radiomics scores, and construct a radiomics model. The combination of radiomics scores (Rad scores) and independent predictive factors was based on multivariate logistic regression analysis to construct a radiomics nomogram to predict EGFR mutation status. The predictive performance and clinical practicality of the model were evaluated using the area under the curve (AUC), calibration curve, and clinical decision curve. Result: EGFR mutations were identified in 30.8 % (28/91) of patients. Thirteen important radiomic features were selected from 1708 radiomics features. The radiomics model effectively classified EGFR mutants and wild-type, with AUCs of 0.846 and 0.808 for the training and validation cohorts, respectively, and had a higher diagnostic efficiency, with AUCs of 0.880 and 0.859, respectively. The calibration curve showed that the model had a good predictive performance, and the decision curve indicated that the radiomic nomogram had high clinical benefits. Conclusion: The predictive model based on MRI radiomics has good diagnostic efficacy for EGFR mutation status in NSCLC and can provide guidance for individualized targeted therapies. magnetic resonance imaging radiomics machine learning non-small cell lung cancer epidermal growth factor receptor Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 1.Introduction Lung cancer is one of the most common malignant tumors worldwide, with the highest mortality rate worldwide [ 1 ] . In China, approximately 733000 new cases of lung cancer occur annually, and more than 610000 people die from lung cancer [ 2 ] . According to the latest national cancer statistics released by the National Cancer Center in 2022, the incidence and mortality rates of lung cancer was highest in China [ 3 ] , accounting for 37% of new cases and 39% of deaths worldwide. Non-small cell lung cancer (NSCLC) accounts for 80% -90% of all primary lung cancers. Approximately 50% of NSCLC patients are diagnosed at an advanced stage (stage III or IV), and the 5-year survival rate is only 18% [ 4 ] . In recent years, the treatment of patients with advanced NSCLC has made breakthroughs, and the application of molecular-targeted drugs has significantly improved the survival time of patients with gene mutations [ 5 ] . Research has shown that continuous activation of the tyrosine kinase domain of the epidermal growth factor receptor (EGFR) in the lung cancer tumor tissue is caused by mutations in the EGFR gene. Therefore, EGFR mutation status is a key factor in determining the therapeutic efficacy of EGFR tyrosine kinase inhibitors (EGFR-TKIs) [ 6 ] . Riely et al. [ 7 ] found that the response rate to EGFR-TKIs in EGFR-mutant patients (60% -80%) was significantly higher than TEGFR wild-type or unknown EGFR mutations (10% -20%). In addition, a large number of clinical trials have shown that, compared to EGFR wild-type patients, EGFR mutant NSCLC patients who receive EGFR TKIs treatment have improved two-year progression-free survival (PFS) with few relapses, and the patients who receive EGFR TKIs again after recurrence can also benefit sustainably [ 8 ] . However, current research on EGFR-TKI treatment mainly focuses on stage II and III patients [ 9 – 11 ] , and early application of EGFR-TKI treatment can achieve better clinical benefits [ 12 ] . Therefore, early and accurate differentiation between EGFR-mutant and-wild-type patients is crucial for achieving early, combined, and personalized precision treatment. Currently, EGFR mutations in NSCLC are mostly detected via puncture or postoperative histopathological biopsy. However, owing to the extensive heterogeneity of tumors, biopsies for detecting EGFR mutations must accurately locate the affected tissue area, which may increase the risk of cancer metastasis. Difficulties, such as repeated sampling, poor tissue sample acquisition, and high costs, limit the applicability of pathological biopsies [ 13 ] . Circulating tumor DNA (ctDNA) analysis is an emerging tool for assessing EGFR mutation status [ 14 ] . However, its high false negative rate and high cost limit its widespread clinical application [ 15 ] . Therefore, there is an urgent need to develop a non-invasive, simple, rapid, and reliable detection method to accurately diagnose EGFR. In recent years, with the advent of the big data era, artificial intelligence (AI) has shown unparalleled advantages in mining medical image information. It is expected to extract key imaging markers and dominant features related to EGFR mutations from multimodal and multi parameter imaging information, which can be used to predict EGFR mutation status in NSCLC patients. Radiomics is an emerging non-invasive technology that utilizes medical imaging analysis and data mining methods. It extracts a large amount of high-throughput quantitative information, combines clinical data to establish predictive models through machine learning and other methods, and ultimately, guides clinical decision making. It has been widely used for tumor diagnosis [ 16 ] . Radiomics mines quantitative imaging features, such as intensity, shape, texture, and wavelets from medical imaging images. By combining medical statistics and machine learning, key imaging radiomics can be screened to evaluate tumor heterogeneity and provide valuable information for tumor grading, staging, treatment, efficacy evaluation, and survival prediction [ 17 ] . Many studies [ 18 – 22 ] have reported a potential relationship between EGFR mutation status and PET/CT and CT radiomic features. One study evaluated the quality and performance of a CT radiomics-based model in predicting EGFR mutation status in patients with NSCLC [ 21 ] , and the results showed that the model had high accuracy in predicting EGFR mutation status, with an area under the curve (AUC) of 0.801. Another analysis on the value of 18 F-FDG PET/CT radiomics in predicting EGFR gene mutations in NSCLC also confirmed [ 22 ] that the AUC for predicting EGFR mutations in NSCLC using 18 F-FDG PET/CT radiomics in the training cohort was 0.84; the AUC in validation cohort were 0.82. However, there are currently few reports involving MRI, and these have mainly focused on patients with advanced lung cancer. Therefore, the purpose of this study was to establish and validate a predictive model based on multiparametric MRI radiomics features combined with clinical pathological factors to predict EGFR mutation status in early NSCLC patients, thereby guiding personalized targeted therapy. 2.Materials and methods 2.1 Patient cohort and image data The Institutional Review Board of the First Affiliated Hospital of Nanjing Medical University approved this retrospective study (No. 2022-NT-11). The requirement for informed consent was patients. A retrospective study was conducted on 92 patients with histopathologically confirmed malignant pulmonary nodules at the First Affiliated Hospital of Nanjing Medical University between August 2019 and May 2021. Among these, 69 were confirmed by surgical pathology, and 23 were confirmed by CT-guided biopsy pathology. Among 92 cases of malignant nodules, the pathological subtypes were 61 cases of adenocarcinoma, four cases of squamous cell carcinoma, and 27 cases that were not otherwise specified (NOS). The inclusion criteria for this study were as follows: (a) patients who underwent MRI examination within one month before surgery or biopsy, (b) patients who did not receive antitumor treatment before MRI examination, (c) patients confirmed by surgery or biopsy pathology, and (d) EGFR mutation detection results. Patients with the following conditions were excluded: (a) no EGFR mutation test results (n = 1), and (b) poor MRI image quality or missing image data (n = 0). Ultimately, 91 patients were included in the study and randomly divided into the training cohort (n = 72) and the validation cohort (n = 19) with an 8:2 ratio (Fig. 1 ). Baseline clinical and pathological information of the patients were collected from the medical record system, including age, sex, and lymph node metastasis (LNM) (Table 1 ). Table 1 Demographic and Clinical Pathological Characteristics of Training and Validation Cohorts Demography and Clinicopathological characteristics training cohort N = 72 validation cohort N = 19 p value EGFR Mutation Status -N(%) 1.000 WT 50 (69.4%) 13 (68.4%) Mutant 22 (30.6%) 6 (31.6%) Age-N(%) 0.315 Female 30 (41.7%) 11 (57.9%) Male 42 (58.3%) 8 (42.1%) Age/year, Median [upper and lower quartiles] 64.00 [56.50, 70.00] 62.00 [51.00, 66.50] 0.366 Position-N(%) 0.203 Upper left 18 (25.0%) 2 (10.5%) Lower left 11 (15.3%) 6 (31.6%) Upper right 23 (31.9%) 8 (42.1%) Right middle 6 ( 8.3%) 2 (10.5%) Lower right 14 (19.4%) 1 ( 5.3%) Maximum diameter/cm, Median [upper and lower quartiles] 2.00 [1.40, 3.10] 2.00 [1.45, 3.25] 0.903 Lymph node metastasis-N(%) 0.679 No-metastasis 55 (76.4%) 13 (68.4%) Metastasis 17 (23.6%) 6 (31.6%) 2.2 MRI image acquisition and analysis The MRI scans of all patients were performed using a Siemens 3.0 T MR scanner (Verio Tim) and a 16 channel body phased array coil. Conventional MR scans include axial T1 weighted imaging with a repetition time (TR) and echo time (TE) of 140/2.5ms and axial free-breathing BLADE T2 weighted imaging (TR/TE, 1200/93 ms). The dynamic contrast-enhanced (DCE) scan included 3D volumetric interpolated breath-hold examination (VIBE) sequence, TR/TE = 3.19/1.13 ms, layer thickness = 3 mm, field of view = 400 mm 2 , matrix = 160 × 224, and flip angle = 15 °. Through elbow vein puncture, a high-pressure injector was used to inject 0.1 mmol/kg of gadopentosamide (GE Health Care) at a flow rate of 4.0 mL/s. Subsequently, 20 mL saline was injected at the same rate. DCE MR consists of four baseline and 31 enhanced images. The time resolution was 8.8 seconds, and the total acquisition time was 5 min 33 s. After completing DCE MR, another set of enhanced T1 weighted images was acquired. 2.3 EGFR gene testing EGFR mutation detection results were obtained from the tumor tissue specimens that had been surgically removed or biopsied. The amplification blockade mutation system polymerase chain reaction (ARMS-PCR) method was used to detect mutation sites in four exons (exons 18–21) of the EGFR coding region. The results were determined based on the interpretation principles provided by the detection kit. If any exon mutation was detected, the tumor was identified as an EGFR mutant; otherwise, it was identified as a wild-type EGFR. 2.4 Image segmentation and radiomics feature extraction The extraction of radiomic features followed the Image Biomarker Standardization Initiative (IBSI), and Syngo via Frontier 1.2.1, VB10B version (Siemens Healthineers, Germany) complied with IBSI. In the absence of knowledge of the patient's clinical information from the two chest imaging radiologists, a radiologist with 10 years of experience in chest diagnosis first semi-automatically delineated the Volume of Interest (VOI) around the tumor, and another radiologist with four years of experience in chest diagnosis confirmed it. First, enhanced T1WI (contrast enhanced T1 weighted, CET1w) and T2WI (T2 weighted, T2w) images were imported into the Radiomics prototype software (Radiomics, Frontier, Siemens), and several segmentation tools were used in the radiomics segmentation module for the semi-automatic segmentation of tumors. Using semi-automatic segmentation, the tumor boundary line is manually drawn; then, based on the random walk algorithm, adjacent voxels with the same grayscale as the drawn boundary line can be automatically found in 3D space, ultimately generating segmentation results for the solid or subsolid lung lesions. The most effective method for obtaining a voxel set for a portion of a lesion is to implement a 3D region-growing algorithm starting from the center point of the region of interest (ROI). Subsequently, based on the density distribution of the grayscale values in the ROI, the threshold of the lesion was determined or adaptively determined. The region growing algorithm obtains the complete lesion area and additional partial vascular system and finally, uses morphological segmentation algorithms to remove the blood vessels from the segmentation results. If the segmentation is incorrect, the operator can manually correct the sketched results in 3D using a radiomics software prototype. A linear interpolation algorithm was used to resample all MRI images to an isotropic voxel size of 1.00x1.00x1.00 mm 3 , and the interpolator used for resampling was B-spline interpolation. After image preprocessing, 854 radiomic features were extracted from the tumor volume of interest of two MR sequences, CET1w and T2W images, based on the radiology library in Python. The extracted radiomics features mainly included the following four categories: a) intensity statistical features, which quantitatively describe the distribution of voxel intensity in MRI images using commonly used measures; b) the shape and size features that reflect the shape- and size-related information of the ROI; c) texture features, which can quantify the heterogeneity of the ROI, were obtained according to the gray run length and gray-level co-occurrence texture matrix; and d) features of higher-order statistics using exponents, logarithms, square roots, wavelets (including wavelet-LHL, wavelet-LHH, wavelet-HLL, wavelet-LLH, wavelet-HLH, wavelet-HHH, wavelet-HHL, and wavelet-LLL) and other filters to recalculate the intensity, texture, and other features of the transformed image. For the detailed characteristic calculation formula, please refer to the website https://pyradiomics.Readthedocs.io/en/latest/.html (Fig. 2 ). Siemens Radiology prototype software was used for semi-automatic segmentation of the MRI image data. A total of 1708 radiomic features were extracted from the tumor volume of interest in the MRI images, including shape and size, texture, and wavelet features. The variance threshold method combined with the univariate selection method and least absolute shrinkage and selection operator (LASSO) regression were used to select 13 key radiomics features, establish a radiomics signature, calculate the Rad score, and construct radiomics models. The Rad-score was combined with independent predictors to construct a radiomics nomogram to predict the EGFR mutation status. 2.5 Selection of radiomics features A random number seed generated by the computer was used to allocate 80% of the dataset to the training set and 20% of the dataset to the verification set. The most relevant radiomics features of the mutation types were selected from the training set. Before feature selection, all radiomic features were standardized; that is, the mean value was removed, divided by its standard deviation, and each group of eigenvalues was converted into standardized data with a mean value of zero and a variance of one. Due to the large number of radiomics features, the following three methods were used to reduce the dimension of the extracted MRI radiomics features and avoid the problems of model over-fitting and multicollinearity: first, the variance threshold method was used to reduce the dimension of the features, and the features with variance less than 0.8 were deleted; second, the univariate selection method was used to screen the features that were not significant (p > 0.05 was deleted); and finally, the LASSO regression algorithm was used to fit the most relevant indicators with the research objectives, and the weights of these indicators were obtained. After feature dimensionality reduction, the Rad-score of each patient (T2WI_Rad-score and CET1w_Rad-score) was calculated. The Rad score was considered as the comprehensive embodiment of radiomics features and included in the subsequent model construction. The calculation formula of Rad-score is: Radscore = intercept + feature_1 × coefficient_1+...+feature_n × coefficient_n (1) The intercept and feature[i] are the eigenvalues and fitting intercept terms generated in the LASSO regression, respectively, and the coefficient[i] is the feature weight coefficient generated in the LASSO regression. 2.6 Construction and training of radiomics prediction model The T2w_ Rad-Score and CET1w_Rad-Score were used as inputs to construct the differential diagnostic model of EGFR mutation status in the prediction of the radiomics model of EGFR mutation status. All the data were randomly divided into training and validation sets (the proportion was 8:2). A training set was used to train the model and a validation set was used to evaluate its generalization ability. The differential diagnosis model of EGFR mutations was constructed using logistic regression, including the T2W radiomics model, CET1w radiomics model, and T2W and CET1w combined radiomics model. Receiver operating characteristic curves (ROC) and AUC were plotted to evaluate the accuracy, precision, sensitivity, and specificity of the model. 2.7 Analysis of clinicopathological factors and construction of prediction model First, clinical factors were analyzed using univariate analysis to screen for significant clinical factors related to the EGFR mutation status. Bilateral p < 0.05 was considered statistically significant. All data were randomly divided into training and verification cohorts (division ratio, 8:2). Clinical factors significantly related to EGFR mutation status were selected as inputs to build a clinical model for the diagnosis of EGFR mutation status. A clinical model was constructed using logistic regression algorithm. An ROC curve was drawn and the AUC value was calculated to evaluate the accuracy, precision, sensitivity, and specificity of the model. 2.8 Establishment of radiomics nomogram A radiomics nomogram combined with T2w_ Rad-Score, CET1w_Rad-Score, and significant clinicopathological factors was established based on a multivariate logistic regression model. ROC curves and AUC values were drawn for the nomogram model. At the same time, the calibration curve of nomogram was drawn to evaluate the calibration and identification ability of nomogram. Decision curve analysis (DCA) was performed to quantify the net benefit of the nomogram model to evaluate the clinical practicability of the model, and the accuracy, precision, sensitivity, and specificity of the comprehensive model were evaluated. 2.9 Statistical analysis In the analysis of clinical data, the classified variables were described by frequency and constituent ratios. The description and comparison methods for continuous variables depend on whether the data follow a normal distribution. The Kolmogorov-Smirnov test was used to test the normality of the data. If the data obey normal distribution, the mean and standard deviation (± s) are used to express, and the independent sample t-test is used for comparison. If it did not follow a normal distribution, it was expressed as the median [interquartile range] (IQR]) and compared using the Mann-Whitney U test. The AUC and 95% confidence interval (95% CI) were used to evaluate the effectiveness of the model. In the statistical test, bilateral p < 0.05 was considered statistically significant. All statistical analyses were performed using R software (version 4.3.0; https://www.r-project.org ) and Python (version 3.8.0). 3.Results 3.1 Clinical features The demographic and clinicopathological characteristics of the patients in the training and validation cohorts are presented in Table 1 . The patients were randomly divided into cohorts, with 72 and 19 patients in the training and validation cohorts, respectively. There were no significant differences in sex or age between the training and validation cohorts (p > 0.05). The other clinicopathological features are shown in Table 1 . This indicated that the baseline data between the training and validation cohorts were balanced. 3.2 Selection of important radiomics features and establishment of radiomics signature A total of 854 radiomics features were extracted using T2WI. First, the variance threshold method was used to screen out features with a variance of less than 0.8 to obtain 851 features. The univariate selection method was used to screen out features with an insignificant difference (p < 0.05) to obtain 49 features. Finally, the LASSO algorithm was used to fit all the features based on gene mutation types (Figure. 2a, 2b). Finally, seven important radiomics features were screened, and the Rad-score was calculated. The seven important radiomic features were as follows (Fig. 3 ). A total of 854 radiomic features were extracted from CET1w. First, the variance threshold method was used to screen out features with a variance of less than 0.8 to obtain 851 features. The univariate selection method was used to screen out features with an insignificant difference (p < 0.05) to obtain 14 features. Finally, the LASSO algorithm was used to fit all features based on the type of gene mutation (Figure. 4a, 4b). Finally, four important radiomics features were screened and the Rad-score was calculated. The four important radiomics features are as follows (Fig. 5 ): wavelet-LHH_firstorder_ kurtosis, wavelet-LLL _glcm_Difference Entropy, wavelet-LLL _glcm_Imc2,wavelet-LLH_glcm_In verseVariance. A total of 1708radiomic features were extracted from the T2WI and CET1w images. First, the variance threshold method was used to screen out the features with variance less than 0.8 to obtain 1702 features. The univariate selection method was used to screen out features with an insignificant difference (p < 0.05) to obtain 43 features. Finally, the LASSO algorithm was used to fit all features based on the type of gene mutation (Figure. 6a, 6b). Finally, 13 important radiomics features were screened, and the Rad-score was calculated. The 13 important radiomics features are as follows (Fig. 7 ). 3.3 Analysis of clinicopathological factors Multivariate logistic regression analysis of the clinicopathological factors showed that sex (0.155) and maximum diameter ( 0.716) were independent predictors. 3.4 Construction and prediction efficiency of multimodal prediction model The AUC of the training and validation cohorts for the prediction model based on the T2WI radiomic signature were 0.780 and 0.632, respectively. The AUC of the training and validation cohorts for the prediction model based on the CET1w radiomic signature were 0.725 and 0.684, respectively. The AUCs of the training and validation cohorts of the prediction model based on T2WI combined with the CET1w radiomic signature were 0.846 and 0.808, respectively. A clinical model was established based on the independent predictive factors (sex and maximum diameter) with AUC values of 0.720 and 0.679 for the training and validation cohorts, respectively. The combined model was constructed by multivariate logistic regression analysis based on T2WI combined with the CET1w radiomics signature and independent predictors (sex and maximum diameter); the AUCs of the training and validation cohorts were 0.880 and 0.859, respectively (Table 2 ). Figure. 8a, b, c, d, e show the AUC values of different prediction models. Based on the multivariate logistic regression analysis, independent predictors (Rad-Score, sex, and maximum diameter) were screened to construct a radiomics nomogram (Figure. 9). The calibration curve showed that the nomogram prediction results were in good agreement with the pathological results (Figure. 10a, b). Decision curve analysis showed that, within a reasonable threshold range, the combined model had a higher overall net benefit than the T2WI, CET1w, T2WI, CET1w, and clinical models (Fig. 11 ). Table 2 Comparison of Prediction Efficiency of Multimodal Prediction Models Prediction model Data set AUC Accuracy Precision Sensitivity Specificity T2WI model training cohort 0.780(0.653–0.871) 0.746 0.560 0.737 0.750 validation cohort 0.632(0.417–0.904) 0.643 0.455 0.556 0.684 CET1w model training cohort 0.725(0.591–0.826) 0.667 0.467 0.737 0.636 validation cohort 0.684(0.478–0.857) 0.643 0.462 0.667 0.632 T2WI & CET1wmodel training cohort 0.846(0.742–0.919) 0.750 0.571 0.727 0.760 validation cohort 0.808(0.571–0.941) 0.737 0.556 0.833 0.692 Clinical model training cohort 0.720(0.595–0.819) 0.736 0.552 0.727 0.740 validation cohort 0.679(0.450–0.871) 0.684 0.500 0.833 0.615 Combined model training cohort 0.880(0.790–0.948) 0.847 0.703 0.864 0.840 validation cohort 0.859(0.657–0.983) 0.789 0.625 0.833 0.769 4.Discussion In this study, we established five prediction models based on multiparametric MRI to extract radiomics features combined with clinicopathological factors, classified the EGFR mutation status of NSCLC, and compared the classification efficiencies of the five prediction models. The results showed that the five prediction models could classify the EGFR mutation status of NSCLC well, and the combined prediction model constructed by score combined with independent predictors had the best classification efficiency. The AUC of the training and validation cohorts were 0.880 and 0.859, respectively. This showed that our model can effectively diagnose and classify the EGFR mutation status in NSCLC. After EGFR TKIs treatment, the median survival time of EGFR mutation-positive patients can reach 25 months, and the quality of life can improve [ 23 ] . The objective remission rate is as high as 70%-80%, and the PFS can reach 9–14 months. Therefore, it has become the standard first-line treatment for EGFR-mutant NSCLC [ 24 ] . However, its clinical efficacy in patients with wild-type EGFR is poor, and it is prone to drug resistance and related adverse reactions [ 25 ] , resulting in poor clinical applicability. Currently, the diagnosis of EGFR mutation status mainly depends on gene detection, which is expensive and time-consuming. Early and accurate prediction of EGFR mutation status in patients with early stage NSCLC can guide the clinical selection of targeted therapy populations more accurately. Therefore, mining and screening of more comprehensive and effective lesion heterogeneity and microenvironment characteristics through noninvasive examination and achieving accurate prediction of EGFR mutation status in NSCLC are the key scientific problems that was solved in this study. Emerging AI technologies, including deep learning feature extraction and segmentation technology, deep survival analysis, and radiomics analysis, can quantitatively extract and accurately segment the depth features of the lung tumor lesion area, simulate the nonlinear risk score function for survival analysis modeling, extract the high-throughput features contained in the image, quantitatively describe the heterogeneity of the lesion, achieve accurate prediction of disease recurrence and prognosis, non-invasively and comprehensively quantify the heterogeneity and microenvironment of the lesion, and provide guidance for the in-depth mining and application of medical images, which will aid in selecting a more appropriate, effective, and personalized treatment scheme. However, currently, the application of radiomics and deep learning technology to the study of lung lesions is mostly focused on CT or positron emission tomography/CT images. A meta-analysis by Felfli [ 21 ] and Ma [ 22 ] found that most CT radiomics only extracted plain scan image features of the lesions, which has certain limitations. The latest meta-analysis [ 26 ] on the diagnostic accuracy of MRI radiomics features for predicting the EGFR mutation status in patients with NSCLC brain metastases showed that the aggregate sensitivity and specificity of MRI radiomics features for EGFR mutation detection were 0.86 and 0.83%, respectively, which is consistent with CT and 18 F-FDG PET/CT radiomics. The most important reason is that MRI has the advantage of multiple sequences and parameters. Different sequences (such as T1WI, T2WI, and DWI) can reflect the shape, structure, metabolism, and function of the tumors from different angles and comprehensively capture their characteristics. The combination of multiple sequences can capture the heterogeneity of tumors more comprehensively and improve the ability of the model to describe the characteristics of the tumors, which can integrate the complementary information of multiple sequences and reduce the limitations of a single sequence. Li et al. [ 27 ] also confirmed these findings. In the experiment used to distinguish between T790M resistance and no T790M mutation in patients with NSCLC brain metastasis, a prediction model was established from the radiomics features extracted from T2WI, T2 fluid attenuation inversion recovery (T2-FLAIR), and DWI and T1-CE sequences; the AUC values in the training and validation cohorts were 0.886 and 0.850, respectively. A multicenter study [ 28 ] showed that the combination of CT and MRI dual-mode radiomics to predict EGFR status in patients with brain metastases from NSCLC showed good calibration and differentiation abilities in both the internal (AUC = 0.866) and external test cohorts (AUC = 0.818). At the same time, compared with the advantages of multi-modal and multi sequence MRI, extracting different modal features from CT images can also achieve the corresponding effect. Zhang X [ 29 ] combined CT images and clinical pathological data, and the "deep radscore" learning radiomics model built on the data sources of the 3D tumor region achieved impressive AUC values in predicting EGFR mutation (AUC = 0.884), PD-1/PD-L1 positive (AUC = 0.893), lymphatic vascular invasion (AUC = 0.889), and pleural invasion (AUC = 0.903); Zhang G et al. [ 30 ] constructed the radiomics model from the radiomics features extracted from multi-phase CT (non-enhanced and enhanced CT, including arterial phase and venous phase CT), which showed good performance in identifying EGFR mutation status in patients with lung adenocarcinoma (AUC = 0.925) and had good consistency with the prediction performance of the clinical radiomics comprehensive model (AUC = 0.927). Compared with the above studies, this study extracted the T2WI plain MRI scan images and also extracted the T1 enhanced scan features combined with clinical pathological factors. Therefore, the final joint prediction model performed the best. The AUC of the training and validation cohorts were 0.880 and 0.859, respectively, which were better than those of the single radiomics or clinical models. At the same time, this study further explored the value of multiparametric MRI radiomics in predicting EGFR mutation status in patients with early NSCLC. The above studies have proven that multimodal and fused features can provide more comprehensive information, improve diagnostic accuracy, guide personalized treatment programs, and enhance the robustness and generalization ability of the model compared with single image features. Multimodal feature fusion enables the model to learn the characteristics of data from multiple perspectives, reducing dependence on single-mode data, which helps to improve the adaptability of the model to different datasets and clinical scenes and reduce the risk of overfitting. MRI performs well in distinguishing the different pathological features of the tissues. Multiparametric MRI can completely display the tumor heterogeneity of the lesions, provide more sensitive quantitative and qualitative imaging markers, and help predict the EGFR mutation status of lung cancer to guide clinical treatments. Currently, research on the EGFR mutation status of NSCLC based on MRI radiomics is mainly focused on patients with advanced lung cancer metastasis, whereas research on the EGFR mutation status of patients with early lung cancer is relatively limited. Ying et al. [ 31 ] studied the EGFR mutation status of 230 patients with lung cancer complicated by brain metastasis, which had good predictive efficiency, and the AUC in the training, validation, and external validation cohorts were 0.896, 0.856, and 0.889, respectively. A multicenter study by Cao et al. [ 32 ] showed that MRI radiomics has good predictive value for EGFR mutation status and subtypes in patients with lung cancer spinal metastasis, and T1WI has a higher predictive efficiency than T2WI. The combined model integrating the two sequences and clinicopathological factors showed the best predictive efficiency for predicting EGFR mutation status, 19 site mutations, and 21 site mutations: training cohort (0.829 vs. 0.885 vs. 0.919), validation cohort (0.760 vs. 0.777 vs. 0.811), and external validation (0.780 vs. 0.846 vs. 0.818). Park et al. [ 33 ] predicted the EGFR mutation status in patients with brain metastasis based on DTI imaging and T1 enhanced sequence of MRI sequences, and the results showed that the prediction efficiency of integrating DWI, T1WI, and DTI sequences was significantly higher than that of single sequences, and the AUC, accuracy, sensitivity, and specificity in the test cohort were 0.73, 78.6%, 81.3%, and 76.9%, respectively. Wang et al. [ 34 ] predicted the EGFR mutation status of NSCLC based on multiparametric MRI radiomics, wherein the AUC of the ADC prediction model was 0.805, whereas the multi-sequence prediction model showed better prediction efficiency, and its nomogram AUC reached 0.925 in the training cohort and 0.727 in the validation cohort. The above research showed that multiparametric MRI radiomics can be used as a non-invasive examination method that can objectively and scientifically describe the morphological and internal structural characteristics of lung cancer. The integration of clinical and pathological features is expected to replace biomarkers and achieve early, sensitive, and accurate prediction of EGFR mutation status in lung cancer, which can provide guidance for early targeted treatment of NSCLC and improve prognosis. However, current research mainly focuses on patients with advanced lung cancer, and patients with early stage lung cancer benefit less from treatment. To achieve an early and accurate diagnosis and treatment of lung cancer, it is particularly important to realize the early prediction of EGFR mutation status. Therefore, this study built a prediction model based on multiparametric MRI radiomics features combined with clinicopathological features to predict EGFR mutation status in early lung cancer. Our results showed that the combined prediction model has high prediction efficiency, which is a small breakthrough compared with previous studies. The main advantage of this study is that, at present, research on radiomics in predicting the EGFR mutation status of lung cancer is mainly focused on CT and PET/CT, while research based on MRI is relatively rare. MRI uses no radiation, and multiparametric MRI can non-invasively observe various metabolic and pathological changes in the living tissues in the early stage, providing functional imaging, which plays a key role in identifying the EGFR mutation status. Therefore, the prediction model established in this study has a high prediction efficiency. Second, current research on MRI radiomics to predict the EGFR mutation status in NSCLC mainly focuses on patients with advanced lung cancer. This study mainly focused on patients with early stage NSCLC, which is of great significance for the early diagnosis and treatment of patients with EGFR mutations in early lung cancer. This can provide a more scientific diagnosis and treatment scheme for clinical practice, and targeted therapy can be used before surgery to improve the long-term prognosis of patients. Third, all patients were scanned using the standard scheme of the same MRI device, avoiding heterogeneity caused by images with different scanning and reconstruction parameters, thus making the results more stable and reliable. In addition, semiautomatic segmentation tools were used in our radiomics research, which largely limited individual differences in manual rendering. The main limitations of this study were as follows: first, this was a retrospective study, which may have introduced a selection bias; and second, this was a single-center study with a small sample size and lack of external validation. Therefore, deep learning algorithms could not be used to improve the efficiency of the model. Cao et al. [ 35 ] confirmed that a deep learning-combined radiomics model based on multi-sequence MRI showed high efficiency in predicting the EGFR mutation status of NSCLC patients with brain metastasis (0.943 and 0.938 in the training data cohort). Therefore, in the future, it is necessary to conduct multi center, large-sample prospective research and joint deep learning algorithms to further verify and optimize our prediction model and increase its applicability, stability, and generalization ability. Third, mutation sites were not studied because of the small sample size. The prognosis of different mutation sites and the benefits of targeted therapy vary. Therefore, it is important to identify mutation sites before targeted therapy. The next step will include the study of the mutation sites. 5. Conclusion Predicting the EGFR mutation status in patients with early NSCLC using multiparametric MRI radiomics combined with clinicopathological features has better predictive diagnostic value and can provide guidance for individualized targeted therapy. Abbreviations MRI, magnetic resonance imaging; AUC, area under the curve; EGFR, epidermal growth factor receptor; NSCLC, non-small cell lung cancer; LASSO, least absolute shrinkage and selection operator; Rad scores, radiomics scores; EGFR-TKIs, EGFR tyrosine kinase inhibitors; PFS, progression-free survival; ctDNA, Circulating tumor DNA; AI, artificial intelligence; NOS, not otherwise specified; LNM, lymph node metastasis; TR, repetition time; TE, echo time; DCE, dynamic contrast-enhanced; VIBE, volumetric interpolated breath-hold examination; IBSI，Image Biomarker Standardization Initiative；VOI，Volume of Interest ；CET1w，contrast enhanced T1 weighted；T2w，T2 weighted； ROI，region of interest ；ROC, receiver operating characteristic curve；DCA, decision curve analysis; T2-FLAIR，T2 fluid attenuation inversion recovery. Declarations Authors contributions Conception and design: B. Yang, H. Xu. Acquisition of data: H. Xu. Analysis and interpretation of data: Y.D. Yin, J.Y. Zhang, Y. Fu, J.G. Li. Statistical analysis: X.Q. Sun, B.S. Xie, M.X. Kong. Drafting the article: Y.B. Wang, H. Hu. Critically revising the article: all authors. Reviewed and approved the submitted version of the manuscript: all authors. Study supervision: B. Yang, H. Xu. Funding National Natural Science Foundation of China Project (No. 82160348)；Yunnan Province Major Special Plan (No. 202302AA310018-D-8); Yunnan Province's "Xingdian Talent Support Program" Youth Talent Project (No. XDYC-QNRC-2022-0608);2024 Senior Health Technology and Medical Discipline Leader of Yunnan Provinc ial Health Commission(No. D-2024056). Data availability The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request. Ethics approval and consent to participate The Institutional Review Board of the First Affiliated Hospital of Nanjing Medical University approved this retrospective study (No. 2022-NT-11) and waived the requirement of informed consent from patients. All methods were carried out in accordance with relevant guidelines and regulations. Our study adhered to the principles of the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. References Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209–49. Chen W, Zheng R, Baade PD, et al. Cancer statistics in China, 2015. CA Cancer J Clin. 2016;66(2):115–32. Zheng R, Zhang S, Zeng H et al. Cancer incidence and mortality in China, 2016. J Natl Cancer Cent 2022,2(1): 1–9. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209–49. Khorrami M, Prasanna P, Gupta A, et al. Changes in CT Radiomic Features Associated with Lymphocyte Distribution Predict Overall Survival and Response to Immunotherapy in Non-Small Cell Lung Cancer. Cancer Immunol Res. 2020;8(1):108–19. Huang YH, Hsu KH, Chin CS, et al. The Clinical Outcomes of Different First-Line EGFR-TKIs Plus Bevacizumab in Advanced EGFR-Mutant Lung Adenocarcinoma. Cancer Res Treat. 2022;54(2):434–44. Riely GJ, Pao W, Pham D, et al. Clinical course of patients with non-small cell lung cancer and epidermal growth factor receptor exon 19 and exon 21 mutations treated with gefitinib or erlotinib. Clin Cancer Res. 2006;12(3 Pt 1):839–44. Pennell NA, Neal JW, Chaft JE, et al. SELECT: A Phase II Trial of Adjuvant Erlotinib in Patients With Resected Epidermal Growth Factor Receptor-Mutant Non-Small-Cell Lung Cancer. J Clin Oncol. 2019;37(2):97–104. Yue D, Xu S, Wang Q, et al. Erlotinib versus vinorelbine plus cisplatin as adjuvant therapy in Chinese patients with stage IIIA EGFR mutation-positive non-small-cell lung cancer (EVAN): a randomised, open-label, phase 2 trial. Lancet Respir Med. 2018;6(11):863–73. Zhong WZ, Wang Q, Mao WM, et al. Gefitinib Versus Vinorelbine Plus Cisplatin as Adjuvant Treatment for Stage II-IIIA (N1-N2) EGFR-Mutant NSCLC: Final Overall Survival Analysis of CTONG1104 Phase III Trial. J Clin Oncol. 2021;39(7):713–22. Tada H. Mitsudomi Tetsuya,Misumi Toshihiro EGFR RandomizedPhase III Study of Gefitinib Versus Cisplatin Plus Vinorelbine for Patients With Resected Stage II-IIIA Non-Small-Cell Lung Cancer WithMutation (IMPACT).J Clin Oncol, 2022, 40: 231–41. Govindan R, Mandrekar SJ, Gerber DE, et al. ALCHEMIST Trials: A Golden Opportunity to Transform Outcomes in Early-Stage Non-Small Cell Lung Cancer. Clin Cancer Res. 2015;21(24):5439–44. Rios Velazquez E, Parmar C, Liu Y, et al. Somatic Mutations Drive Distinct Imaging Phenotypes in Lung Cancer. Cancer Res. 2017;77(14):3922–30. Zhang Y, Chang L, Yang Y, et al. Intratumor heterogeneity comparison among differentsubtypes of non-small-cell lung cancer through multi-region tissue and matched ct DNA sequencing. Mol Cancer. 2019;18(1):7. Hur JY, Kim HJ, Lee JS, et al. Extracellular vesicle-derived DNA for performing EGFR genotyping of NSCLC patients. Mol Cancer. 2018;17(1):15. Chetan MR, Gleeson FV. Radiomics in predicting treatment response in non-small-cell lung cancer: current status, challenges and future perspectives. Eur Radiol. 2021;31(2):1049–58. Bera K, Braman N, Gupta A, et al. Predicting cancer outcomes with radiomics and artificial intelligence in radiology. Nat Rev Clin Oncol. 2022;19(2):132–46. Hou S, Fan Y, Wang X, et al. Radiomics for Detection of the EGFR Mutation in Liver Metastatic NSCLC. Acad Radiol. 2023;30(6):1039–46. Yang L, Xu P, Li M, et al. PET/CT Radiomic Features: A Potential Biomarker for EGFR Mutation Status and Survival Outcome Prediction in NSCLC Patients Treated With TKIs. Front Oncol. 2022;12:894323. Yang X, Liu M, Ren Y, et al. Using contrast-enhanced CT and non-contrast-enhanced CT to predict EGFR mutation status in NSCLC patients-a radiomics nomogram analysis. Eur Radiol. 2022;32(4):2693–703. Felfli M, Liu Y, Zerka F, et al. Systematic Review, Meta-Analysis and Radiomics Quality Score Assessment of CT Radiomics-Based Models Predicting Tumor EGFR Mutation Status in Patients with Non-Small-Cell Lung Cancer. Int J Mol Sci. 2023;24(14):11433. Ma N, Yang W, Wang Q, et al. Predictive value of 18F-FDG PET/CT radiomics for EGFR mutation status in non-small cell lung cancer: a systematic review and meta-analysis. Front Oncol. 2024;14:1281572. La Salvia A, Rossi A, Galetta D, et al. Intercalated Chemotherapy and Epidermal Growth Factor Receptor Inhibitors for Patients With Advanced Non-Small-cell Lung Cancer: A Systematic Review and Meta-analysis. Clin Lung Cancer. 2017;18(1):23–33. Rosell R, Carcereny E, Gervais R, et al. Erlotinib versus standard chemotherapy as first-line treatment for European patients with advanced EGFR mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial. Lancet Oncol. 2012;13(3):239–46. Mok TS, Cheng Y, Zhou X, et al. Updated Overall Survival in a Randomized Study Comparing Dacomitinib with Gefitinib as First-Line Treatment in Patients with Advanced Non-Small-Cell Lung Cancer and EGFR-Activating Mutations. Drugs. 2021;81(2):257–66. Long Y, Zhao R, Du X. Diagnostic accuracy of MRI-based radiomic features for EGFR mutation status in non-small cell lung cancer patients with brain metastases: a meta-analysis. Front Oncol. 2025;14:1428929. Li Y, Lv X, Wang B, et al. Predicting EGFR T790M Mutation in Brain Metastases Using Multisequence MRI-Based Radiomics Signature. Acad Radiol. 2023;30(9):1887–95. Ouyang Z, Zhang G, He S, et al. CT and MRI bimodal radiomics for predicting EGFR status in NSCLC patients with brain metastases: A multicenter study. Eur J Radiol. 2025;183:111853. Zhang X, Zhang G, Qiu X, et al. Exploring non-invasive precision treatment in non-small cell lung cancer patients through deep learning radiomics across imaging features and molecular phenotypes. Biomark Res. 2024;12(1):12. Zhang G, Man Q, Shang L, et al. Using Multi-phase CT Radiomics Features to Predict EGFR Mutation Status in Lung Adenocarcinoma Patients. Acad Radiol. 2024;31(6):2591–600. Fan Y, Wang X, Dong Y et al. Multiregional radiomics of brain metastasis can predict response to EGFR-TKI in metastatic NSCLC. Eur Radiol. 2023;10.1007/s00330-023-09709-7. Cao R, Chen H, Wang H, et al. Comprehensive analysis of prediction of the EGFR mutation and subtypes based on the spinal metastasis from primary lung adenocarcinoma. Front Oncol. 2023;13:1154327. Park YW, An C, Lee J, et al. Diffusion tensor and postcontrast T1-weighted imaging radiomics to differentiate the epidermal growth factor receptor mutation status of brain metastases from non-small cell lung cancer. Neuroradiology. 2021;63(3):343–52. Wang Y, Wan Q, Xia X, et al. Value of radiomics model based on multi-parametric magnetic resonance imaging in predicting epidermal growth factor receptor mutation status in patients with lung adenocarcinoma. J Thorac Dis. 2021;13(6):3497–508. Cao P, Jia X, Wang X, et al. Deep learning radiomics for the prediction of epidermal growth factor receptor mutation status based on MRI in brain metastasis from lung adenocarcinoma patients. BMC Cancer. 2025;25(1):443. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 12 Dec, 2025 Read the published version in BMC Medical Imaging → Version 1 posted Editorial decision: Revision requested 29 Sep, 2025 Reviews received at journal 27 Sep, 2025 Reviewers agreed at journal 24 Sep, 2025 Reviews received at journal 23 Sep, 2025 Reviewers agreed at journal 23 Sep, 2025 Reviewers invited by journal 23 Sep, 2025 Editor invited by journal 23 Sep, 2025 Editor assigned by journal 23 Sep, 2025 Submission checks completed at journal 23 Sep, 2025 First submitted to journal 20 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7663043","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":522272527,"identity":"9408940a-4f29-442a-9749-305d76869aad","order_by":0,"name":"Yubo Wang","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yubo","middleName":"","lastName":"Wang","suffix":""},{"id":522272530,"identity":"ae01a95e-b993-40f7-aeee-4308c533c796","order_by":1,"name":"Hao Hu","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Hu","suffix":""},{"id":522272531,"identity":"362bc9d2-efb8-42d7-b34b-7ca78e7e0d89","order_by":2,"name":"Yadan Yin","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yadan","middleName":"","lastName":"Yin","suffix":""},{"id":522272532,"identity":"e2591264-d7f9-4318-b6aa-c5c7c0e16e17","order_by":3,"name":"Jiyun Zhang","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jiyun","middleName":"","lastName":"Zhang","suffix":""},{"id":522272533,"identity":"2c4a988a-3abf-47a8-856d-fb50e58a1a0d","order_by":4,"name":"Yang Fu","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Fu","suffix":""},{"id":522272535,"identity":"69ca8464-2b40-48a0-8f1c-3aa4e2fe4987","order_by":5,"name":"Jiageng Li","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jiageng","middleName":"","lastName":"Li","suffix":""},{"id":522272539,"identity":"54c232fe-d21a-4d0c-aa61-66fc510b8f1a","order_by":6,"name":"Xueqing Sun","email":"","orcid":"","institution":"The First Hospital of Mile，Honghe","correspondingAuthor":false,"prefix":"","firstName":"Xueqing","middleName":"","lastName":"Sun","suffix":""},{"id":522272541,"identity":"09ee796f-4d8b-4634-b0d8-10fe84254469","order_by":7,"name":"Mengxue Kong","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Mengxue","middleName":"","lastName":"Kong","suffix":""},{"id":522272543,"identity":"f0df2b14-d597-419a-9793-89c071ef20ec","order_by":8,"name":"Bosen Xie","email":"","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bosen","middleName":"","lastName":"Xie","suffix":""},{"id":522272544,"identity":"fc50330c-9646-4be2-812d-e04fcb98073d","order_by":9,"name":"Hai Xu","email":"","orcid":"","institution":"First Affiliated Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hai","middleName":"","lastName":"Xu","suffix":""},{"id":522272545,"identity":"31c6ff12-9242-4964-aac2-0e7807121ac8","order_by":10,"name":"Bin Yang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIiWNgGAWjYJACZgYDMNXA8MHAxo4ULYwNjDMK0pKJ1MIA0cLM8+EQYwMh5QbHzx5+XVBwR95curHxsY3BAWYG9sNHN+DVciYvzXqGwTPDnXMONhvnGNzhY+BJS7uBV8uBHDNjHoPDjBtuJLZJ5xg8Y2aQ4DHDr+X8G7AWe6CW9t8WQL0NBLXcyDF+DNSSCLIFGHREaJG88caMeYbB4WSQXyR7DNKS2Qj5he98jvHngj+HbbdLNx/88OOPjR0/++FjeLUoHGBgkwC7UAIqwoZPOQjINzAwf0DRMgpGwSgYBaMAHQAAGjVRO/3g9UQAAAAASUVORK5CYII=","orcid":"","institution":"Kunming Medical University Affiliated Calmette Hospital","correspondingAuthor":true,"prefix":"","firstName":"Bin","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2025-09-20 07:08:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7663043/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7663043/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12880-025-02029-w","type":"published","date":"2025-12-12T15:59:13+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":92843674,"identity":"1ac16cd7-0091-45d6-a50d-03c1e0c5f3ea","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1486090,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/54b36f6a26bd42f397c1a726.docx"},{"id":92845836,"identity":"45862dae-74dd-4012-b193-03fc49c516f9","added_by":"auto","created_at":"2025-10-06 09:35:14","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11386,"visible":true,"origin":"","legend":"","description":"","filename":"67ba770ef3664bff8a66276fd704cd3e.json","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/228e213ab8b4381ab122cf0d.json"},{"id":92845838,"identity":"f3b08c97-447c-4159-8ee4-c53fa7313784","added_by":"auto","created_at":"2025-10-06 09:35:15","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":126584,"visible":true,"origin":"","legend":"","description":"","filename":"67ba770ef3664bff8a66276fd704cd3e1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/f14d9e2cebbc164d8118ab9a.xml"},{"id":92847099,"identity":"7ab357ee-cb69-46d7-8112-ee0aae3a026c","added_by":"auto","created_at":"2025-10-06 09:43:14","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":169146,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/f86ec0c46d1c4005f315b696.png"},{"id":92843672,"identity":"4d4bf8d6-aa59-4c64-995d-320c1cb9b27e","added_by":"auto","created_at":"2025-10-06 09:19:14","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44271,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/ab864fe55033655a23ef75a7.png"},{"id":92843676,"identity":"10822e4a-1bda-41b8-8fe6-ec3ce7bf24f0","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43658,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/1855157150d9440ca556853a.png"},{"id":92845470,"identity":"1bb13b9c-94d3-4d35-9a65-93e9c1ad35b0","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47110,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/fb312293c49c7696013dab60.png"},{"id":92843681,"identity":"1a7a9a68-c6f3-494c-8f75-3037d3c56835","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43493,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/38ec4bb43741fda6a80411b9.png"},{"id":92845469,"identity":"6f559140-6fcd-437d-9bcd-8cf1cb151a99","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68445,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/8a92cb16c1e7bb180388350f.png"},{"id":92843687,"identity":"1891d120-5fd5-4ca0-870c-34734f3601d3","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":150436,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage15.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/0b089a2637c9be22ef2ce900.jpeg"},{"id":92845841,"identity":"20a9ed27-b888-4721-9e99-0ff1ef582f2c","added_by":"auto","created_at":"2025-10-06 09:35:15","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151019,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage16.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/8b7afd17049fef2ad49da0b5.png"},{"id":92845480,"identity":"c4d1658e-eb85-422c-9bec-ff9e6548f305","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":217806,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/683a98c66e2d1624b858a199.png"},{"id":92843691,"identity":"c3eda84d-1d4a-40e4-807d-7a77a542c05a","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"jpeg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":188298,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/8b287cebd83e0eed97d7fe0c.jpeg"},{"id":92845478,"identity":"b1b9eb20-f2cf-49ab-b4eb-15d2b55a1a77","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":62741,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/ccdceca55606b9f1cf40927f.png"},{"id":92843685,"identity":"dd5b59bc-e7ea-4a5e-89b7-3d50ae284522","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"jpeg","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":220807,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/65fa681160b068349541a0e7.jpeg"},{"id":92845471,"identity":"bd10dd7c-f83d-46be-a76b-ce3a9edf0398","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"jpeg","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":131353,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/4e1cded5c8f0f38f00951fc7.jpeg"},{"id":92843689,"identity":"50e371f4-517e-48fc-aae5-fd37023b7b18","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"jpeg","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":215321,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/ff3a417f396bd9e870da5600.jpeg"},{"id":92843697,"identity":"9a9aa352-0b17-40c7-bb39-b703781c9f7d","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108158,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/0f63725e4c4a08eee9827586.png"},{"id":92843692,"identity":"d38b8450-b19c-4436-b306-662406ba6100","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43554,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/d85d07490579a6627fd48f6b.png"},{"id":92843693,"identity":"1b51014e-fdbd-4a3c-ac80-4edd0fddc07f","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52251,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/bddafcfef4424743e65ffa58.png"},{"id":92845475,"identity":"a2d92dc0-e6dd-4f72-9717-b4d3b8761401","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11078,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/1bad12707ae5de7bdb310c31.png"},{"id":92843713,"identity":"9d7c593e-fafe-45f3-b76c-fdf44b0cd979","added_by":"auto","created_at":"2025-10-06 09:19:16","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10756,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/2954fefbfb0556be21ec0cab.png"},{"id":92845477,"identity":"e37537db-62df-477b-bcd9-57677a1dd588","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12331,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/54454decd6f323219960b4af.png"},{"id":92843690,"identity":"49632798-2e84-4086-91b2-00a5f7e0aea7","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10668,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/c4ef945b3e4563af41972462.png"},{"id":92843708,"identity":"67fb3dc1-e29c-49a9-8c15-5bc724f414b3","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31321,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/14b87c1ae36f0fbb0fb67821.png"},{"id":92845839,"identity":"46ac3ad0-a623-4064-ae2a-ea44c47f16fb","added_by":"auto","created_at":"2025-10-06 09:35:15","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":74690,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage15.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/788d8930e9d86c3cf5e7a2d5.png"},{"id":92843714,"identity":"75148d86-235e-4a2a-bcc1-68f3d52fe452","added_by":"auto","created_at":"2025-10-06 09:19:16","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38091,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage16.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/8e4b416806b9c51541952b44.png"},{"id":92843707,"identity":"384e1e8d-f46f-48a5-ac09-231672a80890","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44637,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/6372c57fb93e6a79098ce6a3.png"},{"id":92843704,"identity":"392fb9f2-45ec-4866-875e-99c63aa4588c","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43350,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/c5bf6c8e4949b2dbb123bc15.png"},{"id":92843710,"identity":"e41e941e-2f57-4a40-8f30-4cf6786e4ebf","added_by":"auto","created_at":"2025-10-06 09:19:16","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18204,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/f45389b301dba9acb4b514b3.png"},{"id":92843698,"identity":"465da095-e549-4c52-be4f-ecddf3475f08","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45520,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/5e95be0abf745cab677f1111.png"},{"id":92843709,"identity":"57c41d69-1c36-43ec-a8ef-43d0bf6fa9ce","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":32152,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/ea3341c7a5f0418d76063fe9.png"},{"id":92845479,"identity":"efe0adc2-f73d-44e9-a2de-aaffe4baf708","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43234,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/09d0be5e5c922d47085a4ecf.png"},{"id":92843703,"identity":"337c80e2-6706-4796-b423-8a5ac1756759","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29334,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/d5432b5f91459bdbe1db4bb7.png"},{"id":92843695,"identity":"3d6dac22-ef72-4799-b240-607fe5511157","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10593,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/08fcc767298aef5abdd10879.png"},{"id":92843700,"identity":"cd629c73-8824-4ed7-9664-0d41d04b720f","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"xml","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":122857,"visible":true,"origin":"","legend":"","description":"","filename":"67ba770ef3664bff8a66276fd704cd3e1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/b4dede8d613974dc40a07dd8.xml"},{"id":92845481,"identity":"d047ab3e-3ebb-493e-966f-f0db634a178b","added_by":"auto","created_at":"2025-10-06 09:27:16","extension":"html","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136609,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/fe2305bf7170432185a196e9.html"},{"id":92843666,"identity":"599f716c-44b4-49b2-bef0-dc87dc48ca9b","added_by":"auto","created_at":"2025-10-06 09:19:14","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":96632,"visible":true,"origin":"","legend":"\u003cp\u003ePatient's admission criteria and flow chart.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/56feb32625ba2fa2904baef1.png"},{"id":92843665,"identity":"563f74f5-674b-4a31-9d7d-29cc1df96992","added_by":"auto","created_at":"2025-10-06 09:19:14","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":71060,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow of establishing and verifying the nomogram of radiomics.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/d413cb7b7a356c38f89f7ebd.png"},{"id":92845467,"identity":"78a7f52a-8eac-4cf2-8e96-7d2ab901a097","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":157367,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 2a \u003c/strong\u003eIn LASSO regression model, 10 fold cross validation was used to select the optimal tuning parameter log (α)=1.6；\u003cstrong\u003eFigure 2b \u003c/strong\u003eSeven non-zero coefficient radiomicsfeatures (each color line represents the change of its coefficient) were obtained by λ.\u003c/p\u003e","description":"","filename":"21.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/b1257b5f4e4df297d9b6b845.png"},{"id":92845461,"identity":"6b842e11-5ac0-44fe-bf39-07a2ce8fd463","added_by":"auto","created_at":"2025-10-06 09:27:14","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":38581,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure. 3 \u003c/strong\u003e7 radiomics features with significant predictive value and their weights screened out.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/c3522cc6d1650a7a7179a7a0.png"},{"id":92843667,"identity":"a0720221-f64a-4fd7-ac96-f15346aa2a43","added_by":"auto","created_at":"2025-10-06 09:19:14","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":70628,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 4a\u003c/strong\u003e In LASSO regression model, 10 fold cross validation was used to select the optimal tuning parameter log (α)=1.78. \u003cstrong\u003eFigure. 4b\u003c/strong\u003e Four non-zero coefficient imageomics features are obtained by using -log (α) (each color line represents the change of its coefficient).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/6fc8a813d7dd1ed7c514b56b.png"},{"id":92845462,"identity":"43552e22-42b0-4d9b-b9f2-cbe4d08058b6","added_by":"auto","created_at":"2025-10-06 09:27:14","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":27681,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 5 \u003c/strong\u003eFour radiomics features with significant predictive value and their weights were selected.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/447dd6944264ee77ef211a98.png"},{"id":92843669,"identity":"80b0b9b6-b834-4b05-b5ca-6b6d999b835f","added_by":"auto","created_at":"2025-10-06 09:19:14","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":62165,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 6a \u003c/strong\u003eIn LASSO regression model, 10 fold cross validation was used to select the optimal tuning parameter log (α)=1.65；\u003cstrong\u003eFigure 6b \u003c/strong\u003eUsing -log (α) to obtain 13 non-zero coefficient radiomics features (each color line represents the change of its coefficient).\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/9f9625ccd80a0d8f4718cc4a.png"},{"id":92845465,"identity":"bd52090c-b1f4-403b-ac54-05648450c17f","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":56159,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 7 \u003c/strong\u003e13 radiomics features with significant predictive value and their weights were selected.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/9f06f1029754c9896b5ec8b3.png"},{"id":92845468,"identity":"e5af54fc-d95e-47d3-adec-8359850d2e8b","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":41878,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure8a,b,c,d,e \u003c/strong\u003eT2WI model, CET1w model, T2WI\u0026amp;CET1w model, clinical model and combined model were used to predict the receiver operating characteristic curve of EGFR mutation.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/e514b98cb16c568b073717bf.png"},{"id":92843683,"identity":"7d81338f-6a33-439e-b870-bcafd5c403ab","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":26422,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 9 \u003c/strong\u003eRadiomics nomogram\u003c/p\u003e\n\u003cp\u003eRadiomics scores combined with independent predictors (gender and maximum diameter) were used to construct an radiomics nomogram to predict EGFR mutation status based on multivariate logistic regression analysis.\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/2ceab4dc0674335c7525cd73.png"},{"id":92843679,"identity":"d784e997-9a29-4b3e-85f2-d3ecf6d7906b","added_by":"auto","created_at":"2025-10-06 09:19:15","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":37178,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 10a,b \u003c/strong\u003eCalibration curve of training cohort and validation cohort.\u003c/p\u003e\n\u003cp\u003eCalibration curve analysis, the abscissa represents the prediction probability of radiomics nomogram, and the ordinate represents the actual observation results. The diagonal dotted line is the reference line, indicating the perfect fitting between the prediction result and the actual result. The thin dotted line indicates the prediction efficiency of the prediction model. The closer the diagonal dotted line is, the better the prediction efficiency is.\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/d0dab24979f017d311c96f83.png"},{"id":92845473,"identity":"bf59a10e-2bc3-435e-a4a8-89c57a65ce81","added_by":"auto","created_at":"2025-10-06 09:27:15","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":53486,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 11 \u003c/strong\u003eDecision curve analysis of training cohort and validation cohort.\u003c/p\u003e\n\u003cp\u003eIn the analysis of decision curve, the ordinate is to measure the net benefit calculated from the true positive and false positive results. Among the positive prediction (marked as \"all\"), negative prediction (marked as \"None\") and all prediction models (marked as \"T2WI model, CET1w model, T2WI\u0026amp;CET1w model, clinical model and combined model\"), the combined model within a reasonable threshold range guides clinical decision-making, and the net benefit of patients is the largest.\u003c/p\u003e","description":"","filename":"11.png","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/6ff35784fa4b1352d2875fc6.png"},{"id":98244777,"identity":"43843460-7732-4e39-8454-246c9b98b8bb","added_by":"auto","created_at":"2025-12-15 16:15:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1756313,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7663043/v1/c1322a8d-0c17-49e1-8e37-dbce9a798a99.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Prediction of EGFR Mutation Status in Non-small Cell Lung Cancer Based on Multiparametric MRI Radiomics","fulltext":[{"header":"1.Introduction","content":"\u003cp\u003eLung cancer is one of the most common malignant tumors worldwide, with the highest mortality rate worldwide\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. In China, approximately 733000 new cases of lung cancer occur annually, and more than 610000 people die from lung cancer\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. According to the latest national cancer statistics released by the National Cancer Center in 2022, the incidence and mortality rates of lung cancer was highest in China\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e, accounting for 37% of new cases and 39% of deaths worldwide. Non-small cell lung cancer (NSCLC) accounts for 80% -90% of all primary lung cancers. Approximately 50% of NSCLC patients are diagnosed at an advanced stage (stage III or IV), and the 5-year survival rate is only 18%\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. In recent years, the treatment of patients with advanced NSCLC has made breakthroughs, and the application of molecular-targeted drugs has significantly improved the survival time of patients with gene mutations\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Research has shown that continuous activation of the tyrosine kinase domain of the epidermal growth factor receptor (EGFR) in the lung cancer tumor tissue is caused by mutations in the EGFR gene. Therefore, EGFR mutation status is a key factor in determining the therapeutic efficacy of EGFR tyrosine kinase inhibitors (EGFR-TKIs)\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. Riely et al.\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e found that the response rate to EGFR-TKIs in EGFR-mutant patients (60% -80%) was significantly higher than TEGFR wild-type or unknown EGFR mutations (10% -20%). In addition, a large number of clinical trials have shown that, compared to EGFR wild-type patients, EGFR mutant NSCLC patients who receive EGFR TKIs treatment have improved two-year progression-free survival (PFS) with few relapses, and the patients who receive EGFR TKIs again after recurrence can also benefit sustainably\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. However, current research on EGFR-TKI treatment mainly focuses on stage II and III patients\u003csup\u003e[\u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e, and early application of EGFR-TKI treatment can achieve better clinical benefits\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. Therefore, early and accurate differentiation between EGFR-mutant and-wild-type patients is crucial for achieving early, combined, and personalized precision treatment.\u003c/p\u003e\u003cp\u003eCurrently, EGFR mutations in NSCLC are mostly detected via puncture or postoperative histopathological biopsy. However, owing to the extensive heterogeneity of tumors, biopsies for detecting EGFR mutations must accurately locate the affected tissue area, which may increase the risk of cancer metastasis. Difficulties, such as repeated sampling, poor tissue sample acquisition, and high costs, limit the applicability of pathological biopsies\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. Circulating tumor DNA (ctDNA) analysis is an emerging tool for assessing EGFR mutation status\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. However, its high false negative rate and high cost limit its widespread clinical application\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. Therefore, there is an urgent need to develop a non-invasive, simple, rapid, and reliable detection method to accurately diagnose EGFR. In recent years, with the advent of the big data era, artificial intelligence (AI) has shown unparalleled advantages in mining medical image information. It is expected to extract key imaging markers and dominant features related to EGFR mutations from multimodal and multi parameter imaging information, which can be used to predict EGFR mutation status in NSCLC patients. Radiomics is an emerging non-invasive technology that utilizes medical imaging analysis and data mining methods. It extracts a large amount of high-throughput quantitative information, combines clinical data to establish predictive models through machine learning and other methods, and ultimately, guides clinical decision making. It has been widely used for tumor diagnosis\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. Radiomics mines quantitative imaging features, such as intensity, shape, texture, and wavelets from medical imaging images. By combining medical statistics and machine learning, key imaging radiomics can be screened to evaluate tumor heterogeneity and provide valuable information for tumor grading, staging, treatment, efficacy evaluation, and survival prediction\u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. Many studies\u003csup\u003e[\u003cspan additionalcitationids=\"CR19 CR20 CR21\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e have reported a potential relationship between EGFR mutation status and PET/CT and CT radiomic features. One study evaluated the quality and performance of a CT radiomics-based model in predicting EGFR mutation status in patients with NSCLC\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e, and the results showed that the model had high accuracy in predicting EGFR mutation status, with an area under the curve (AUC) of 0.801. Another analysis on the value of \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT radiomics in predicting EGFR gene mutations in NSCLC also confirmed\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e that the AUC for predicting EGFR mutations in NSCLC using \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT radiomics in the training cohort was 0.84; the AUC in validation cohort were 0.82. However, there are currently few reports involving MRI, and these have mainly focused on patients with advanced lung cancer.\u003c/p\u003e\u003cp\u003eTherefore, the purpose of this study was to establish and validate a predictive model based on multiparametric MRI radiomics features combined with clinical pathological factors to predict EGFR mutation status in early NSCLC patients, thereby guiding personalized targeted therapy.\u003c/p\u003e"},{"header":"2.Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Patient cohort and image data\u003c/h2\u003e\u003cp\u003e The Institutional Review Board of the First Affiliated Hospital of Nanjing Medical University approved this retrospective study (No. 2022-NT-11). The requirement for informed consent was patients. A retrospective study was conducted on 92 patients with histopathologically confirmed malignant pulmonary nodules at the First Affiliated Hospital of Nanjing Medical University between August 2019 and May 2021. Among these, 69 were confirmed by surgical pathology, and 23 were confirmed by CT-guided biopsy pathology. Among 92 cases of malignant nodules, the pathological subtypes were 61 cases of adenocarcinoma, four cases of squamous cell carcinoma, and 27 cases that were not otherwise specified (NOS). The inclusion criteria for this study were as follows: (a) patients who underwent MRI examination within one month before surgery or biopsy, (b) patients who did not receive antitumor treatment before MRI examination, (c) patients confirmed by surgery or biopsy pathology, and (d) EGFR mutation detection results. Patients with the following conditions were excluded: (a) no EGFR mutation test results (n\u0026thinsp;=\u0026thinsp;1), and (b) poor MRI image quality or missing image data (n\u0026thinsp;=\u0026thinsp;0). Ultimately, 91 patients were included in the study and randomly divided into the training cohort (n\u0026thinsp;=\u0026thinsp;72) and the validation cohort (n\u0026thinsp;=\u0026thinsp;19) with an 8:2 ratio (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Baseline clinical and pathological information of the patients were collected from the medical record system, including age, sex, and lymph node metastasis (LNM) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\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\u003eDemographic and Clinical Pathological Characteristics of Training and Validation Cohorts\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDemography and\u003c/p\u003e\u003cp\u003eClinicopathological characteristics\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003etraining cohort\u003c/p\u003e\u003cp\u003e\u003cem\u003eN\u0026thinsp;=\u0026thinsp;72\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003evalidation cohort\u003c/p\u003e\u003cp\u003e\u003cem\u003eN\u0026thinsp;=\u0026thinsp;19\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEGFR Mutation Status\u003c/p\u003e\u003cp\u003e-N(%)\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=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e50 (69.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13 (68.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMutant\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e22 (30.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6 (31.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge-N(%)\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=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.315\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e30 (41.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e11 (57.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e42 (58.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8 (42.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge/year,\u003c/p\u003e\u003cp\u003eMedian [upper and lower quartiles]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e64.00\u003c/p\u003e\u003cp\u003e[56.50, 70.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e62.00\u003c/p\u003e\u003cp\u003e[51.00, 66.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.366\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePosition-N(%)\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=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.203\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUpper left\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e18 (25.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2 (10.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLower left\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e11 (15.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6 (31.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUpper right\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e23 (31.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8 (42.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRight middle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e6 ( 8.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2 (10.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLower right\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e14 (19.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1 ( 5.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMaximum diameter/cm,\u003c/p\u003e\u003cp\u003eMedian [upper and lower quartiles]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.00 [1.40, 3.10]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2.00 [1.45, 3.25]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.903\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLymph node metastasis-N(%)\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=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.679\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo-metastasis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e55 (76.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13 (68.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetastasis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e17 (23.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6 (31.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\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=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 MRI image acquisition and analysis\u003c/h2\u003e\u003cp\u003eThe MRI scans of all patients were performed using a Siemens 3.0 T MR scanner (Verio Tim) and a 16 channel body phased array coil. Conventional MR scans include axial T1 weighted imaging with a repetition time (TR) and echo time (TE) of 140/2.5ms and axial free-breathing BLADE T2 weighted imaging (TR/TE, 1200/93 ms). The dynamic contrast-enhanced (DCE) scan included 3D volumetric interpolated breath-hold examination (VIBE) sequence, TR/TE\u0026thinsp;=\u0026thinsp;3.19/1.13 ms, layer thickness\u0026thinsp;=\u0026thinsp;3 mm, field of view\u0026thinsp;=\u0026thinsp;400 mm\u003csup\u003e2\u003c/sup\u003e, matrix\u0026thinsp;=\u0026thinsp;160 \u0026times; 224, and flip angle\u0026thinsp;=\u0026thinsp;15 \u0026deg;. Through elbow vein puncture, a high-pressure injector was used to inject 0.1 mmol/kg of gadopentosamide (GE Health Care) at a flow rate of 4.0 mL/s. Subsequently, 20 mL saline was injected at the same rate. DCE MR consists of four baseline and 31 enhanced images. The time resolution was 8.8 seconds, and the total acquisition time was 5 min 33 s. After completing DCE MR, another set of enhanced T1 weighted images was acquired.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 EGFR gene testing\u003c/h2\u003e\u003cp\u003eEGFR mutation detection results were obtained from the tumor tissue specimens that had been surgically removed or biopsied. The amplification blockade mutation system polymerase chain reaction (ARMS-PCR) method was used to detect mutation sites in four exons (exons 18\u0026ndash;21) of the EGFR coding region. The results were determined based on the interpretation principles provided by the detection kit. If any exon mutation was detected, the tumor was identified as an EGFR mutant; otherwise, it was identified as a wild-type EGFR.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Image segmentation and radiomics feature extraction\u003c/h2\u003e\u003cp\u003eThe extraction of radiomic features followed the Image Biomarker Standardization Initiative (IBSI), and Syngo via Frontier 1.2.1, VB10B version (Siemens Healthineers, Germany) complied with IBSI. In the absence of knowledge of the patient's clinical information from the two chest imaging radiologists, a radiologist with 10 years of experience in chest diagnosis first semi-automatically delineated the Volume of Interest (VOI) around the tumor, and another radiologist with four years of experience in chest diagnosis confirmed it. First, enhanced T1WI (contrast enhanced T1 weighted, CET1w) and T2WI (T2 weighted, T2w) images were imported into the Radiomics prototype software (Radiomics, Frontier, Siemens), and several segmentation tools were used in the radiomics segmentation module for the semi-automatic segmentation of tumors. Using semi-automatic segmentation, the tumor boundary line is manually drawn; then, based on the random walk algorithm, adjacent voxels with the same grayscale as the drawn boundary line can be automatically found in 3D space, ultimately generating segmentation results for the solid or subsolid lung lesions. The most effective method for obtaining a voxel set for a portion of a lesion is to implement a 3D region-growing algorithm starting from the center point of the region of interest (ROI). Subsequently, based on the density distribution of the grayscale values in the ROI, the threshold of the lesion was determined or adaptively determined. The region growing algorithm obtains the complete lesion area and additional partial vascular system and finally, uses morphological segmentation algorithms to remove the blood vessels from the segmentation results. If the segmentation is incorrect, the operator can manually correct the sketched results in 3D using a radiomics software prototype. A linear interpolation algorithm was used to resample all MRI images to an isotropic voxel size of 1.00x1.00x1.00 mm\u003csup\u003e3\u003c/sup\u003e, and the interpolator used for resampling was B-spline interpolation. After image preprocessing, 854 radiomic features were extracted from the tumor volume of interest of two MR sequences, CET1w and T2W images, based on the radiology library in Python. The extracted radiomics features mainly included the following four categories: a) intensity statistical features, which quantitatively describe the distribution of voxel intensity in MRI images using commonly used measures; b) the shape and size features that reflect the shape- and size-related information of the ROI; c) texture features, which can quantify the heterogeneity of the ROI, were obtained according to the gray run length and gray-level co-occurrence texture matrix; and d) features of higher-order statistics using exponents, logarithms, square roots, wavelets (including wavelet-LHL, wavelet-LHH, wavelet-HLL, wavelet-LLH, wavelet-HLH, wavelet-HHH, wavelet-HHL, and wavelet-LLL) and other filters to recalculate the intensity, texture, and other features of the transformed image. For the detailed characteristic calculation formula, please refer to the website \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pyradiomics.Readthedocs.io/en/latest/.html\u003c/span\u003e\u003cspan address=\"https://pyradiomics.Readthedocs.io/en/latest/.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eSiemens Radiology prototype software was used for semi-automatic segmentation of the MRI image data. A total of 1708 radiomic features were extracted from the tumor volume of interest in the MRI images, including shape and size, texture, and wavelet features. The variance threshold method combined with the univariate selection method and least absolute shrinkage and selection operator (LASSO) regression were used to select 13 key radiomics features, establish a radiomics signature, calculate the Rad score, and construct radiomics models. The Rad-score was combined with independent predictors to construct a radiomics nomogram to predict the EGFR mutation status.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Selection of radiomics features\u003c/h2\u003e\u003cp\u003eA random number seed generated by the computer was used to allocate 80% of the dataset to the training set and 20% of the dataset to the verification set. The most relevant radiomics features of the mutation types were selected from the training set. Before feature selection, all radiomic features were standardized; that is, the mean value was removed, divided by its standard deviation, and each group of eigenvalues was converted into standardized data with a mean value of zero and a variance of one. Due to the large number of radiomics features, the following three methods were used to reduce the dimension of the extracted MRI radiomics features and avoid the problems of model over-fitting and multicollinearity: first, the variance threshold method was used to reduce the dimension of the features, and the features with variance less than 0.8 were deleted; second, the univariate selection method was used to screen the features that were not significant (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05 was deleted); and finally, the LASSO regression algorithm was used to fit the most relevant indicators with the research objectives, and the weights of these indicators were obtained. After feature dimensionality reduction, the Rad-score of each patient (T2WI_Rad-score and CET1w_Rad-score) was calculated. The Rad score was considered as the comprehensive embodiment of radiomics features and included in the subsequent model construction.\u003c/p\u003e\u003cp\u003eThe calculation formula of Rad-score is:\u003c/p\u003e\u003cp\u003eRadscore\u0026thinsp;=\u0026thinsp;intercept\u0026thinsp;+\u0026thinsp;feature_1 \u0026times; coefficient_1+...+feature_n \u0026times; coefficient_n (1)\u003c/p\u003e\u003cp\u003eThe intercept and feature[i] are the eigenvalues and fitting intercept terms generated in the LASSO regression, respectively, and the coefficient[i] is the feature weight coefficient generated in the LASSO regression.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Construction and training of radiomics prediction model\u003c/h2\u003e\u003cp\u003eThe T2w_ Rad-Score and CET1w_Rad-Score were used as inputs to construct the differential diagnostic model of EGFR mutation status in the prediction of the radiomics model of EGFR mutation status. All the data were randomly divided into training and validation sets (the proportion was 8:2). A training set was used to train the model and a validation set was used to evaluate its generalization ability. The differential diagnosis model of EGFR mutations was constructed using logistic regression, including the T2W radiomics model, CET1w radiomics model, and T2W and CET1w combined radiomics model. Receiver operating characteristic curves (ROC) and AUC were plotted to evaluate the accuracy, precision, sensitivity, and specificity of the model.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.7 Analysis of clinicopathological factors and construction of prediction model\u003c/h2\u003e\u003cp\u003eFirst, clinical factors were analyzed using univariate analysis to screen for significant clinical factors related to the EGFR mutation status. Bilateral p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant. All data were randomly divided into training and verification cohorts (division ratio, 8:2). Clinical factors significantly related to EGFR mutation status were selected as inputs to build a clinical model for the diagnosis of EGFR mutation status. A clinical model was constructed using logistic regression algorithm. An ROC curve was drawn and the AUC value was calculated to evaluate the accuracy, precision, sensitivity, and specificity of the model.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.8 Establishment of radiomics nomogram\u003c/h2\u003e\u003cp\u003eA radiomics nomogram combined with T2w_ Rad-Score, CET1w_Rad-Score, and significant clinicopathological factors was established based on a multivariate logistic regression model. ROC curves and AUC values were drawn for the nomogram model. At the same time, the calibration curve of nomogram was drawn to evaluate the calibration and identification ability of nomogram. Decision curve analysis (DCA) was performed to quantify the net benefit of the nomogram model to evaluate the clinical practicability of the model, and the accuracy, precision, sensitivity, and specificity of the comprehensive model were evaluated.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e2.9 Statistical analysis\u003c/h2\u003e\u003cp\u003eIn the analysis of clinical data, the classified variables were described by frequency and constituent ratios. The description and comparison methods for continuous variables depend on whether the data follow a normal distribution. The Kolmogorov-Smirnov test was used to test the normality of the data. If the data obey normal distribution, the mean and standard deviation (\u0026plusmn;\u0026thinsp;s) are used to express, and the independent sample t-test is used for comparison. If it did not follow a normal distribution, it was expressed as the median [interquartile range] (IQR]) and compared using the Mann-Whitney U test. The AUC and 95% confidence interval (95% CI) were used to evaluate the effectiveness of the model. In the statistical test, bilateral p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant. All statistical analyses were performed using R software (version 4.3.0; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.r-project.org\u003c/span\u003e\u003cspan address=\"https://www.r-project.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and Python (version 3.8.0).\u003c/p\u003e\u003c/div\u003e"},{"header":"3.Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Clinical features\u003c/h2\u003e\u003cp\u003eThe demographic and clinicopathological characteristics of the patients in the training and validation cohorts are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The patients were randomly divided into cohorts, with 72 and 19 patients in the training and validation cohorts, respectively. There were no significant differences in sex or age between the training and validation cohorts (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). The other clinicopathological features are shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. This indicated that the baseline data between the training and validation cohorts were balanced.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Selection of important radiomics features and establishment of radiomics signature\u003c/h2\u003e\u003cp\u003eA total of 854 radiomics features were extracted using T2WI. First, the variance threshold method was used to screen out features with a variance of less than 0.8 to obtain 851 features. The univariate selection method was used to screen out features with an insignificant difference (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) to obtain 49 features. Finally, the LASSO algorithm was used to fit all the features based on gene mutation types (Figure. 2a, 2b). Finally, seven important radiomics features were screened, and the Rad-score was calculated. The seven important radiomic features were as follows (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e). A total of 854 radiomic features were extracted from CET1w. First, the variance threshold method was used to screen out features with a variance of less than 0.8 to obtain 851 features. The univariate selection method was used to screen out features with an insignificant difference (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) to obtain 14 features. Finally, the LASSO algorithm was used to fit all features based on the type of gene mutation (Figure. 4a, 4b). Finally, four important radiomics features were screened and the Rad-score was calculated. The four important radiomics features are as follows (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003e): wavelet-LHH_firstorder_ kurtosis, wavelet-LLL _glcm_Difference Entropy, wavelet-LLL _glcm_Imc2,wavelet-LLH_glcm_In verseVariance. A total of 1708radiomic features were extracted from the T2WI and CET1w images. First, the variance threshold method was used to screen out the features with variance less than 0.8 to obtain 1702 features. The univariate selection method was used to screen out features with an insignificant difference (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) to obtain 43 features. Finally, the LASSO algorithm was used to fit all features based on the type of gene mutation (Figure. 6a, 6b). Finally, 13 important radiomics features were screened, and the Rad-score was calculated. The 13 important radiomics features are as follows (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Analysis of clinicopathological factors\u003c/h2\u003e\u003cp\u003eMultivariate logistic regression analysis of the clinicopathological factors showed that sex (0.155) and maximum diameter ( 0.716) were independent predictors.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Construction and prediction efficiency of multimodal prediction model\u003c/h2\u003e\u003cp\u003eThe AUC of the training and validation cohorts for the prediction model based on the T2WI radiomic signature were 0.780 and 0.632, respectively. The AUC of the training and validation cohorts for the prediction model based on the CET1w radiomic signature were 0.725 and 0.684, respectively. The AUCs of the training and validation cohorts of the prediction model based on T2WI combined with the CET1w radiomic signature were 0.846 and 0.808, respectively. A clinical model was established based on the independent predictive factors (sex and maximum diameter) with AUC values of 0.720 and 0.679 for the training and validation cohorts, respectively. The combined model was constructed by multivariate logistic regression analysis based on T2WI combined with the CET1w radiomics signature and independent predictors (sex and maximum diameter); the AUCs of the training and validation cohorts were 0.880 and 0.859, respectively (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Figure. 8a, b, c, d, e show the AUC values of different prediction models. Based on the multivariate logistic regression analysis, independent predictors (Rad-Score, sex, and maximum diameter) were screened to construct a radiomics nomogram (Figure. 9). The calibration curve showed that the nomogram prediction results were in good agreement with the pathological results (Figure. 10a, b). Decision curve analysis showed that, within a reasonable threshold range, the combined model had a higher overall net benefit than the T2WI, CET1w, T2WI, CET1w, and clinical models (Fig.\u0026nbsp;\u003cspan refid=\"Fig12\" class=\"InternalRef\"\u003e11\u003c/span\u003e).\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\u003eComparison of Prediction Efficiency of Multimodal Prediction Models\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"7\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePrediction model\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eData set\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eT2WI model\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003etraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.780(0.653\u0026ndash;0.871)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.746\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.560\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003evalidation cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.632(0.417\u0026ndash;0.904)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.643\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.455\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.556\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.684\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCET1w model\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003etraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.725(0.591\u0026ndash;0.826)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.467\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.636\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003evalidation cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.684(0.478\u0026ndash;0.857)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.643\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.462\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.632\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eT2WI \u0026amp; CET1wmodel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003etraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.846(0.742\u0026ndash;0.919)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.571\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.727\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.760\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003evalidation cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.808(0.571\u0026ndash;0.941)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.556\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.692\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eClinical model\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003etraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.720(0.595\u0026ndash;0.819)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.736\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.552\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.727\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.740\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003evalidation cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.679(0.450\u0026ndash;0.871)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.684\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.615\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCombined model\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003etraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.880(0.790\u0026ndash;0.948)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.847\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.703\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.864\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.840\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003evalidation cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.859(0.657\u0026ndash;0.983)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.789\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.625\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.769\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"},{"header":"4.Discussion","content":"\u003cp\u003eIn this study, we established five prediction models based on multiparametric MRI to extract radiomics features combined with clinicopathological factors, classified the EGFR mutation status of NSCLC, and compared the classification efficiencies of the five prediction models. The results showed that the five prediction models could classify the EGFR mutation status of NSCLC well, and the combined prediction model constructed by score combined with independent predictors had the best classification efficiency. The AUC of the training and validation cohorts were 0.880 and 0.859, respectively. This showed that our model can effectively diagnose and classify the EGFR mutation status in NSCLC.\u003c/p\u003e\u003cp\u003eAfter EGFR TKIs treatment, the median survival time of EGFR mutation-positive patients can reach 25 months, and the quality of life can improve\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. The objective remission rate is as high as 70%-80%, and the PFS can reach 9\u0026ndash;14 months. Therefore, it has become the standard first-line treatment for EGFR-mutant NSCLC\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. However, its clinical efficacy in patients with wild-type EGFR is poor, and it is prone to drug resistance and related adverse reactions\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e, resulting in poor clinical applicability. Currently, the diagnosis of EGFR mutation status mainly depends on gene detection, which is expensive and time-consuming. Early and accurate prediction of EGFR mutation status in patients with early stage NSCLC can guide the clinical selection of targeted therapy populations more accurately. Therefore, mining and screening of more comprehensive and effective lesion heterogeneity and microenvironment characteristics through noninvasive examination and achieving accurate prediction of EGFR mutation status in NSCLC are the key scientific problems that was solved in this study.\u003c/p\u003e\u003cp\u003eEmerging AI technologies, including deep learning feature extraction and segmentation technology, deep survival analysis, and radiomics analysis, can quantitatively extract and accurately segment the depth features of the lung tumor lesion area, simulate the nonlinear risk score function for survival analysis modeling, extract the high-throughput features contained in the image, quantitatively describe the heterogeneity of the lesion, achieve accurate prediction of disease recurrence and prognosis, non-invasively and comprehensively quantify the heterogeneity and microenvironment of the lesion, and provide guidance for the in-depth mining and application of medical images, which will aid in selecting a more appropriate, effective, and personalized treatment scheme. However, currently, the application of radiomics and deep learning technology to the study of lung lesions is mostly focused on CT or positron emission tomography/CT images. A meta-analysis by Felfli\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e and Ma\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003efound that most CT radiomics only extracted plain scan image features of the lesions, which has certain limitations. The latest meta-analysis\u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e on the diagnostic accuracy of MRI radiomics features for predicting the EGFR mutation status in patients with NSCLC brain metastases showed that the aggregate sensitivity and specificity of MRI radiomics features for EGFR mutation detection were 0.86 and 0.83%, respectively, which is consistent with CT and \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT radiomics. The most important reason is that MRI has the advantage of multiple sequences and parameters. Different sequences (such as T1WI, T2WI, and DWI) can reflect the shape, structure, metabolism, and function of the tumors from different angles and comprehensively capture their characteristics. The combination of multiple sequences can capture the heterogeneity of tumors more comprehensively and improve the ability of the model to describe the characteristics of the tumors, which can integrate the complementary information of multiple sequences and reduce the limitations of a single sequence. Li et al.\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e also confirmed these findings. In the experiment used to distinguish between T790M resistance and no T790M mutation in patients with NSCLC brain metastasis, a prediction model was established from the radiomics features extracted from T2WI, T2 fluid attenuation inversion recovery (T2-FLAIR), and DWI and T1-CE sequences; the AUC values in the training and validation cohorts were 0.886 and 0.850, respectively. A multicenter study\u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e showed that the combination of CT and MRI dual-mode radiomics to predict EGFR status in patients with brain metastases from NSCLC showed good calibration and differentiation abilities in both the internal (AUC\u0026thinsp;=\u0026thinsp;0.866) and external test cohorts (AUC\u0026thinsp;=\u0026thinsp;0.818). At the same time, compared with the advantages of multi-modal and multi sequence MRI, extracting different modal features from CT images can also achieve the corresponding effect. Zhang X\u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e combined CT images and clinical pathological data, and the \"deep radscore\" learning radiomics model built on the data sources of the 3D tumor region achieved impressive AUC values in predicting EGFR mutation (AUC\u0026thinsp;=\u0026thinsp;0.884), PD-1/PD-L1 positive (AUC\u0026thinsp;=\u0026thinsp;0.893), lymphatic vascular invasion (AUC\u0026thinsp;=\u0026thinsp;0.889), and pleural invasion (AUC\u0026thinsp;=\u0026thinsp;0.903); Zhang G et al.\u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003econstructed the radiomics model from the radiomics features extracted from multi-phase CT (non-enhanced and enhanced CT, including arterial phase and venous phase CT), which showed good performance in identifying EGFR mutation status in patients with lung adenocarcinoma (AUC\u0026thinsp;=\u0026thinsp;0.925) and had good consistency with the prediction performance of the clinical radiomics comprehensive model (AUC\u0026thinsp;=\u0026thinsp;0.927). Compared with the above studies, this study extracted the T2WI plain MRI scan images and also extracted the T1 enhanced scan features combined with clinical pathological factors. Therefore, the final joint prediction model performed the best. The AUC of the training and validation cohorts were 0.880 and 0.859, respectively, which were better than those of the single radiomics or clinical models. At the same time, this study further explored the value of multiparametric MRI radiomics in predicting EGFR mutation status in patients with early NSCLC. The above studies have proven that multimodal and fused features can provide more comprehensive information, improve diagnostic accuracy, guide personalized treatment programs, and enhance the robustness and generalization ability of the model compared with single image features. Multimodal feature fusion enables the model to learn the characteristics of data from multiple perspectives, reducing dependence on single-mode data, which helps to improve the adaptability of the model to different datasets and clinical scenes and reduce the risk of overfitting.\u003c/p\u003e\u003cp\u003eMRI performs well in distinguishing the different pathological features of the tissues. Multiparametric MRI can completely display the tumor heterogeneity of the lesions, provide more sensitive quantitative and qualitative imaging markers, and help predict the EGFR mutation status of lung cancer to guide clinical treatments. Currently, research on the EGFR mutation status of NSCLC based on MRI radiomics is mainly focused on patients with advanced lung cancer metastasis, whereas research on the EGFR mutation status of patients with early lung cancer is relatively limited. Ying et al.\u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e studied the EGFR mutation status of 230 patients with lung cancer complicated by brain metastasis, which had good predictive efficiency, and the AUC in the training, validation, and external validation cohorts were 0.896, 0.856, and 0.889, respectively. A multicenter study by Cao et al.\u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e showed that MRI radiomics has good predictive value for EGFR mutation status and subtypes in patients with lung cancer spinal metastasis, and T1WI has a higher predictive efficiency than T2WI. The combined model integrating the two sequences and clinicopathological factors showed the best predictive efficiency for predicting EGFR mutation status, 19 site mutations, and 21 site mutations: training cohort (0.829 vs. 0.885 vs. 0.919), validation cohort (0.760 vs. 0.777 vs. 0.811), and external validation (0.780 vs. 0.846 vs. 0.818). Park et al.\u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e predicted the EGFR mutation status in patients with brain metastasis based on DTI imaging and T1 enhanced sequence of MRI sequences, and the results showed that the prediction efficiency of integrating DWI, T1WI, and DTI sequences was significantly higher than that of single sequences, and the AUC, accuracy, sensitivity, and specificity in the test cohort were 0.73, 78.6%, 81.3%, and 76.9%, respectively. Wang et al.\u003csup\u003e[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/sup\u003e predicted the EGFR mutation status of NSCLC based on multiparametric MRI radiomics, wherein the AUC of the ADC prediction model was 0.805, whereas the multi-sequence prediction model showed better prediction efficiency, and its nomogram AUC reached 0.925 in the training cohort and 0.727 in the validation cohort. The above research showed that multiparametric MRI radiomics can be used as a non-invasive examination method that can objectively and scientifically describe the morphological and internal structural characteristics of lung cancer. The integration of clinical and pathological features is expected to replace biomarkers and achieve early, sensitive, and accurate prediction of EGFR mutation status in lung cancer, which can provide guidance for early targeted treatment of NSCLC and improve prognosis. However, current research mainly focuses on patients with advanced lung cancer, and patients with early stage lung cancer benefit less from treatment. To achieve an early and accurate diagnosis and treatment of lung cancer, it is particularly important to realize the early prediction of EGFR mutation status. Therefore, this study built a prediction model based on multiparametric MRI radiomics features combined with clinicopathological features to predict EGFR mutation status in early lung cancer. Our results showed that the combined prediction model has high prediction efficiency, which is a small breakthrough compared with previous studies.\u003c/p\u003e\u003cp\u003eThe main advantage of this study is that, at present, research on radiomics in predicting the EGFR mutation status of lung cancer is mainly focused on CT and PET/CT, while research based on MRI is relatively rare. MRI uses no radiation, and multiparametric MRI can non-invasively observe various metabolic and pathological changes in the living tissues in the early stage, providing functional imaging, which plays a key role in identifying the EGFR mutation status. Therefore, the prediction model established in this study has a high prediction efficiency. Second, current research on MRI radiomics to predict the EGFR mutation status in NSCLC mainly focuses on patients with advanced lung cancer. This study mainly focused on patients with early stage NSCLC, which is of great significance for the early diagnosis and treatment of patients with EGFR mutations in early lung cancer. This can provide a more scientific diagnosis and treatment scheme for clinical practice, and targeted therapy can be used before surgery to improve the long-term prognosis of patients. Third, all patients were scanned using the standard scheme of the same MRI device, avoiding heterogeneity caused by images with different scanning and reconstruction parameters, thus making the results more stable and reliable. In addition, semiautomatic segmentation tools were used in our radiomics research, which largely limited individual differences in manual rendering.\u003c/p\u003e\u003cp\u003eThe main limitations of this study were as follows: first, this was a retrospective study, which may have introduced a selection bias; and second, this was a single-center study with a small sample size and lack of external validation. Therefore, deep learning algorithms could not be used to improve the efficiency of the model. Cao et al.\u003csup\u003e[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e confirmed that a deep learning-combined radiomics model based on multi-sequence MRI showed high efficiency in predicting the EGFR mutation status of NSCLC patients with brain metastasis (0.943 and 0.938 in the training data cohort). Therefore, in the future, it is necessary to conduct multi center, large-sample prospective research and joint deep learning algorithms to further verify and optimize our prediction model and increase its applicability, stability, and generalization ability. Third, mutation sites were not studied because of the small sample size. The prognosis of different mutation sites and the benefits of targeted therapy vary. Therefore, it is important to identify mutation sites before targeted therapy. The next step will include the study of the mutation sites.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003ePredicting the EGFR mutation status in patients with early NSCLC using multiparametric MRI radiomics combined with clinicopathological features has better predictive diagnostic value and can provide guidance for individualized targeted therapy.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eMRI, magnetic resonance imaging; AUC, area under the curve; EGFR, epidermal growth factor receptor; NSCLC, non-small cell lung cancer; LASSO, least absolute shrinkage and selection operator; Rad scores, radiomics scores; EGFR-TKIs, EGFR tyrosine kinase inhibitors; PFS, progression-free survival; ctDNA, Circulating tumor DNA; AI, artificial intelligence; NOS, not otherwise specified; LNM, lymph node metastasis; TR, repetition time; TE, echo time; DCE, dynamic contrast-enhanced; VIBE, volumetric interpolated breath-hold examination; IBSI，Image Biomarker Standardization Initiative；VOI，Volume of Interest ；CET1w，contrast enhanced T1 weighted；T2w，T2 weighted； ROI，region of interest ；ROC, receiver operating characteristic curve；DCA, decision curve analysis; T2-FLAIR，T2 fluid attenuation inversion recovery.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConception and design: B. Yang, H. Xu. Acquisition of data: H. Xu. Analysis and interpretation of data: Y.D. Yin, J.Y. Zhang, Y. Fu, J.G. Li. Statistical analysis: X.Q. Sun, B.S. Xie, M.X. Kong. Drafting the article: Y.B. Wang, H. Hu. Critically revising the article: all authors. Reviewed and approved the submitted version of the manuscript: all authors. Study supervision: B. Yang, H. Xu.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNational Natural Science Foundation of China Project (No. 82160348)；Yunnan Province Major Special Plan (No. 202302AA310018-D-8); Yunnan Province\u0026apos;s \u0026quot;Xingdian Talent Support Program\u0026quot; Youth Talent Project (No. XDYC-QNRC-2022-0608);2024 Senior Health Technology and Medical Discipline Leader of Yunnan Provinc ial Health Commission(No. D-2024056).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Institutional Review Board of the First Affiliated Hospital of Nanjing Medical University approved this retrospective study (No. 2022-NT-11) and waived the requirement of informed consent from patients. All methods were carried out in accordance with relevant guidelines and regulations.\u0026nbsp;Our study adhered to the principles of the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen W, Zheng R, Baade PD, et al. Cancer statistics in China, 2015. CA Cancer J Clin. 2016;66(2):115\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZheng R, Zhang S, Zeng H et al. Cancer incidence and mortality in China, 2016. J Natl Cancer Cent 2022,2(1): 1\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKhorrami M, Prasanna P, Gupta A, et al. Changes in CT Radiomic Features Associated with Lymphocyte Distribution Predict Overall Survival and Response to Immunotherapy in Non-Small Cell Lung Cancer. Cancer Immunol Res. 2020;8(1):108\u0026ndash;19.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang YH, Hsu KH, Chin CS, et al. The Clinical Outcomes of Different First-Line EGFR-TKIs Plus Bevacizumab in Advanced EGFR-Mutant Lung Adenocarcinoma. Cancer Res Treat. 2022;54(2):434\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRiely GJ, Pao W, Pham D, et al. Clinical course of patients with non-small cell lung cancer and epidermal growth factor receptor exon 19 and exon 21 mutations treated with gefitinib or erlotinib. Clin Cancer Res. 2006;12(3 Pt 1):839\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePennell NA, Neal JW, Chaft JE, et al. SELECT: A Phase II Trial of Adjuvant Erlotinib in Patients With Resected Epidermal Growth Factor Receptor-Mutant Non-Small-Cell Lung Cancer. J Clin Oncol. 2019;37(2):97\u0026ndash;104.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYue D, Xu S, Wang Q, et al. Erlotinib versus vinorelbine plus cisplatin as adjuvant therapy in Chinese patients with stage IIIA EGFR mutation-positive non-small-cell lung cancer (EVAN): a randomised, open-label, phase 2 trial. Lancet Respir Med. 2018;6(11):863\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhong WZ, Wang Q, Mao WM, et al. Gefitinib Versus Vinorelbine Plus Cisplatin as Adjuvant Treatment for Stage II-IIIA (N1-N2) EGFR-Mutant NSCLC: Final Overall Survival Analysis of CTONG1104 Phase III Trial. J Clin Oncol. 2021;39(7):713\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTada H. Mitsudomi Tetsuya,Misumi Toshihiro EGFR RandomizedPhase III Study of Gefitinib Versus Cisplatin Plus Vinorelbine for Patients With Resected Stage II-IIIA Non-Small-Cell Lung Cancer WithMutation (IMPACT).J Clin Oncol, 2022, 40: 231\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGovindan R, Mandrekar SJ, Gerber DE, et al. ALCHEMIST Trials: A Golden Opportunity to Transform Outcomes in Early-Stage Non-Small Cell Lung Cancer. Clin Cancer Res. 2015;21(24):5439\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRios Velazquez E, Parmar C, Liu Y, et al. Somatic Mutations Drive Distinct Imaging Phenotypes in Lung Cancer. Cancer Res. 2017;77(14):3922\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Chang L, Yang Y, et al. Intratumor heterogeneity comparison among differentsubtypes of non-small-cell lung cancer through multi-region tissue and matched ct DNA sequencing. Mol Cancer. 2019;18(1):7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHur JY, Kim HJ, Lee JS, et al. Extracellular vesicle-derived DNA for performing EGFR genotyping of NSCLC patients. Mol Cancer. 2018;17(1):15.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChetan MR, Gleeson FV. Radiomics in predicting treatment response in non-small-cell lung cancer: current status, challenges and future perspectives. Eur Radiol. 2021;31(2):1049\u0026ndash;58.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBera K, Braman N, Gupta A, et al. Predicting cancer outcomes with radiomics and artificial intelligence in radiology. Nat Rev Clin Oncol. 2022;19(2):132\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHou S, Fan Y, Wang X, et al. Radiomics for Detection of the EGFR Mutation in Liver Metastatic NSCLC. Acad Radiol. 2023;30(6):1039\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang L, Xu P, Li M, et al. PET/CT Radiomic Features: A Potential Biomarker for EGFR Mutation Status and Survival Outcome Prediction in NSCLC Patients Treated With TKIs. Front Oncol. 2022;12:894323.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang X, Liu M, Ren Y, et al. Using contrast-enhanced CT and non-contrast-enhanced CT to predict EGFR mutation status in NSCLC patients-a radiomics nomogram analysis. Eur Radiol. 2022;32(4):2693\u0026ndash;703.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFelfli M, Liu Y, Zerka F, et al. Systematic Review, Meta-Analysis and Radiomics Quality Score Assessment of CT Radiomics-Based Models Predicting Tumor EGFR Mutation Status in Patients with Non-Small-Cell Lung Cancer. Int J Mol Sci. 2023;24(14):11433.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMa N, Yang W, Wang Q, et al. Predictive value of 18F-FDG PET/CT radiomics for EGFR mutation status in non-small cell lung cancer: a systematic review and meta-analysis. Front Oncol. 2024;14:1281572.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLa Salvia A, Rossi A, Galetta D, et al. Intercalated Chemotherapy and Epidermal Growth Factor Receptor Inhibitors for Patients With Advanced Non-Small-cell Lung Cancer: A Systematic Review and Meta-analysis. Clin Lung Cancer. 2017;18(1):23\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRosell R, Carcereny E, Gervais R, et al. Erlotinib versus standard chemotherapy as first-line treatment for European patients with advanced EGFR mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial. Lancet Oncol. 2012;13(3):239\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMok TS, Cheng Y, Zhou X, et al. Updated Overall Survival in a Randomized Study Comparing Dacomitinib with Gefitinib as First-Line Treatment in Patients with Advanced Non-Small-Cell Lung Cancer and EGFR-Activating Mutations. Drugs. 2021;81(2):257\u0026ndash;66.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLong Y, Zhao R, Du X. Diagnostic accuracy of MRI-based radiomic features for EGFR mutation status in non-small cell lung cancer patients with brain metastases: a meta-analysis. Front Oncol. 2025;14:1428929.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi Y, Lv X, Wang B, et al. Predicting EGFR T790M Mutation in Brain Metastases Using Multisequence MRI-Based Radiomics Signature. Acad Radiol. 2023;30(9):1887\u0026ndash;95.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOuyang Z, Zhang G, He S, et al. CT and MRI bimodal radiomics for predicting EGFR status in NSCLC patients with brain metastases: A multicenter study. Eur J Radiol. 2025;183:111853.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang X, Zhang G, Qiu X, et al. Exploring non-invasive precision treatment in non-small cell lung cancer patients through deep learning radiomics across imaging features and molecular phenotypes. Biomark Res. 2024;12(1):12.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang G, Man Q, Shang L, et al. Using Multi-phase CT Radiomics Features to Predict EGFR Mutation Status in Lung Adenocarcinoma Patients. Acad Radiol. 2024;31(6):2591\u0026ndash;600.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFan Y, Wang X, Dong Y et al. Multiregional radiomics of brain metastasis can predict response to EGFR-TKI in metastatic NSCLC. Eur Radiol. 2023;10.1007/s00330-023-09709-7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCao R, Chen H, Wang H, et al. Comprehensive analysis of prediction of the EGFR mutation and subtypes based on the spinal metastasis from primary lung adenocarcinoma. Front Oncol. 2023;13:1154327.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePark YW, An C, Lee J, et al. Diffusion tensor and postcontrast T1-weighted imaging radiomics to differentiate the epidermal growth factor receptor mutation status of brain metastases from non-small cell lung cancer. Neuroradiology. 2021;63(3):343\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Y, Wan Q, Xia X, et al. Value of radiomics model based on multi-parametric magnetic resonance imaging in predicting epidermal growth factor receptor mutation status in patients with lung adenocarcinoma. J Thorac Dis. 2021;13(6):3497\u0026ndash;508.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCao P, Jia X, Wang X, et al. Deep learning radiomics for the prediction of epidermal growth factor receptor mutation status based on MRI in brain metastasis from lung adenocarcinoma patients. BMC Cancer. 2025;25(1):443.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"magnetic resonance imaging, radiomics, machine learning, non-small cell lung cancer, epidermal growth factor receptor","lastPublishedDoi":"10.21203/rs.3.rs-7663043/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7663043/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground: \u003c/strong\u003eTo establish and validate a predictive model based on magnetic resonance imaging (MRI) radiomics features combined with clinicopathological factors to predict the mutation status of epidermal growth factor receptor (EGFR) in non-small cell lung cancer（NSCLC).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatients and methods: \u003c/strong\u003eA total of 91 NSCLC patient (72 in the training cohort and 19 in the validation cohort) were included in this study. A total of 1708 radiomics features were extracted from the MRI (T2w and CET1w) sequences. The variance threshold method combined with the univariate selection method, least absolute shrinkage and selection operator (LASSO) regression were used to screen important radiomics features, calculate radiomics scores, and construct a radiomics model. The combination of radiomics scores (Rad scores) and independent predictive factors was based on multivariate logistic regression analysis to construct a radiomics nomogram to predict EGFR mutation status. The predictive performance and clinical practicality of the model were evaluated using the area under the curve (AUC), calibration curve, and clinical decision curve.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResult: \u003c/strong\u003eEGFR mutations were identified in 30.8 % (28/91) of patients. Thirteen important radiomic features were selected from 1708 radiomics features. The radiomics model effectively classified EGFR mutants and wild-type, with AUCs of 0.846 and 0.808 for the training and validation cohorts, respectively, and had a higher diagnostic efficiency, with AUCs of 0.880 and 0.859, respectively. The calibration curve showed that the model had a good predictive performance, and the decision curve indicated that the radiomic nomogram had high clinical benefits.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion: \u003c/strong\u003eThe predictive model based on MRI radiomics has good diagnostic efficacy for EGFR mutation status in NSCLC and can provide guidance for individualized targeted therapies.\u003c/p\u003e","manuscriptTitle":"Prediction of EGFR Mutation Status in Non-small Cell Lung Cancer Based on Multiparametric MRI Radiomics","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-06 09:19:09","doi":"10.21203/rs.3.rs-7663043/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-29T13:04:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-27T14:46:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"164628154211788470613106707831703084627","date":"2025-09-24T11:28:27+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-24T00:38:29+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"31452103197475840237131644533074536098","date":"2025-09-23T14:57:08+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-23T14:54:48+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-09-23T11:25:39+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-23T09:50:43+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-23T09:50:12+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-09-20T06:52:46+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"10c173b3-fbfb-4910-9eb9-a17f7b3ec867","owner":[],"postedDate":"October 6th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-12-15T16:08:57+00:00","versionOfRecord":{"articleIdentity":"rs-7663043","link":"https://doi.org/10.1186/s12880-025-02029-w","journal":{"identity":"bmc-medical-imaging","isVorOnly":false,"title":"BMC Medical Imaging"},"publishedOn":"2025-12-12 15:59:13","publishedOnDateReadable":"December 12th, 2025"},"versionCreatedAt":"2025-10-06 09:19:09","video":"","vorDoi":"10.1186/s12880-025-02029-w","vorDoiUrl":"https://doi.org/10.1186/s12880-025-02029-w","workflowStages":[]},"version":"v1","identity":"rs-7663043","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7663043","identity":"rs-7663043","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}