Computed tomography-based texture analysis for predicting adjuvant therapy response in postoperative patients with EGFR-mutant non-small cell lung cancer

Introduction

Background and objectives

With the advancement of genetic testing technology, first-line treatment with epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKIs) has demonstrated significant efficacy in patients with EGFR-mutant advanced non-small cell lung cancer (NSCLC). Large-scale clinical trials have reported that both EGFR-TKI monotherapy and combination with radiotherapy or chemotherapy can significantly prolong overall survival and improve patient outcomes (1-3). In recent years, the registration and subsequent publication of several clinical trials evaluating the efficacy of postoperative targeted adjuvant therapies in resectable NSCLC—such as a phase III randomized controlled trial investigating icotinib as adjuvant therapy for patients with completely resected stage II–IIIA EGFR-mutant NSCLC—have opened new avenues for postoperative adjuvant treatment in this patient population (4). However, emerging evidence suggests that EGFR-TKI agents are not universally effective in patients with EGFR-mutant NSCLC, with many developing varying degrees of acquired resistance (5). Studies have found that such resistance is significantly associated with certain imaging features. For instance, heterogeneous metabolic activity and structural irregularities observed on positron emission tomography-computed tomography (PET-CT) and magnetic resonance imaging (MRI) may indicate a higher risk of drug resistance, reflecting a close correlation between the internal heterogeneity of the tumor and sensitivity to drugs (6-8).

Texture analysis offers a noninvasive method for quantitatively analyzing medical images, enabling the extraction of intratumoral texture features that reflect tumor heterogeneity and microenvironment characteristics, and holds promise as a tool for the noninvasive prediction of tumor biological behavior (9,10). CT remains one of the most commonly used imaging modalities in clinical practice for evaluating treatment responses in patients with NSCLC (11). Based on this, we conducted a study in which CT-based texture analysis was used to predict the efficacy of adjuvant EGFR-TKI therapy in patients with EGFR-mutant NSCLC following surgery. The aim of this study was to enable the pretreatment prediction of therapeutic response and to facilitate the early identification of patients with NSCLC who would benefit from targeted therapy. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1554/rc).

Methods

Participants

Data from patients with EGFR-mutant NSCLC admitted to the First Affiliated Hospital of Hebei North University from January 2019 to September 2024 who underwent complete surgical resection and received first-generation EGFR-TKIs as adjuvant targeted therapy were retrospectively collected. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Institutional Review Board of the First Affiliated Hospital of Hebei North University (approval No. K2020273). The requirement for individual consent was waived due to the retrospective nature of the analysis.

The inclusion criteria were as follows: (I) complete surgical resection of lung cancer with negative resection margins confirmed by pathology (R0 resection); (II) histologically confirmed lung adenocarcinoma with either exon 19 deletion or exon 21 point mutation in the EGFR gene; (III) pathological staging of stage III NSCLC according to the American Joint Committee on Cancer (AJCC) eighth edition tumor node metastasis (TNM) staging system (12); (IV) preoperative chest CT examination performed within 1 week before surgery; and (V) administration of only postoperative EGFR-TKI therapy (those who received chemotherapy were excluded).

Meanwhile, the exclusion criteria were as follows: (I) a history of prior radiotherapy, chemotherapy, or EGFR-TKI targeted therapy; (II) no history of other types of malignant tumors; (III) severe cardiac, hepatic, renal, or hematologic functional abnormalities; (IV) no periodic review after treatment to assess efficacy, poor treatment compliance, or incomplete images or clinical data; and (V) age <18 years (Figure 1).

Figure 1 Flowchart of patient inclusion and exclusion. EGFR, epidermal growth factor receptor; NSCLC, non-small cell lung cancer; TKI, tyrosine kinase inhibitor.

The data collected included basic information (age, gender, smoking history, drinking history, history of hypertension, history of diabetes mellitus, and history of coronary heart disease) from the medical record query system at the time of patient admission, initial relevant CT imaging signs (spiculation), and postoperative pathology (tumor component type, lymph node metastasis, and pleural invasion).

All patients were started on targeted therapy with oral first-generation EGFR-TKIs, which included administration of 150 mg of icotinib three times a day for 1 month after operation. Therapeutic efficacy was assessed after 6 months of targeted therapy, and patients were divided into a good prognosis group and a poor prognosis group based on the presence of local recurrence or distant metastasis (as determined by CT or MRI, biopsy pathology, or abnormal tumor markers).

CT scanning parameters and methods

All chest CT scans were performed with a 320-slice Aquilion ONE CT scanner (Canon Medical Systems, Otawara, Japan). For scanning, patients were in the supine position with arms raised to hold their head and entered the scanner head first. The scanning parameters were as follows: a tube voltage of 120 kV, a tube current adjusted automatically with an automatic exposure control, and a matrix size of 512×512, and a reconstruction slice thickness of 5 mm with a standard soft tissue reconstruction algorithm. The scan started from the lung apices to the lower edge of the posterior rib-diaphragm angle, covering the entire lung region, including the bilateral thoracic cavities; to avoid artifacts produced by respiratory movements, patients were instructed to inspire deeply and then begin breath-holding before the initiation of the scan.

Region of interest (ROI) sketching and texture analysis feature extraction

Nonenhanced chest CT images in this study were exported in Digital Imaging and Communications in Medicine (DICOM) format. 3D Slicer v. 5.6.2 software (http://www.slicer.org) was used to manually delineate ROIs layer by layer along tumor margins under lung window settings [window width, 1,500 Hounsfield units (HU); window shift, −600 HU], with areas of liquefactive necrosis being excluded to preserve active tumor components (Figure 2). To ensure reliability, double-blind processing was applied throughout ROI delineation, feature extraction, and clinical indicator collection (operators were unaware of patient names, prognosis, or other predictive factors). Segmentation was initially performed by radiologists with over 3 years of experience and then reviewed by senior physicians with over a decade of expertise. Significant discrepancies were resolved through discussion until a consensus was reached. Interobserver reproducibility validation yielded intraclass correlation coefficient values of 0.944 for single measurements and 0.971 for mean measurements, both exceeding 0.75 and indicating excellent consistency. Texture features were extracted with the SlicerRadiomics plugin in 3D Slicer software. Image preprocessing and feature enhancement were performed within 3D Slicer software under the following core parameters: (I) voxel resampling coefficient, 1×1×1 mm³; (II) Laplacian of Gaussian filter coefficient, 1×1×1 mm³; and (III) gray-value range, −150 to 2,047.25. Texture features included five descriptors: gray-level co-occurrence matrix (GLCM), gray-level run-length matrix (GLRLM), gray-level size-zone matrix (GLSZM), gray-level dependence matrix (GLDM), and neighboring gray-tone difference matrix (NGTDM). A complete case analysis strategy was adopted, with available data from 48 patients being used without imputation of missing values.

Figure 2 Methodology for the outlining of lung tumor ROIs (green). ROI, region of interest.

Screening of texture analysis features

The extracted texture feature parameters were standardized via Z-score normalized with R software (The R Foundation for Statistical Computing; https://www.r-project.org). Dimensionality reduction was performed to refine the feature set. Correlation analysis was conducted to evaluate interfeature relationships, with one feature retained from each pair exhibiting a correlation coefficients >0.9. Subsequently, key features were selected via the least absolute shrinkage and selection operator (LASSO) algorithm, with the optimal λ value determined through 10-fold cross-validation. A Hosmer-Lemeshow (H-L) test was applied to the constructed model to evaluate the model’s fit, and calibration curves and decision curves were used to demonstrate the predictive performance of each model.

The steps for using the prediction model are as follows: first, the patient’s preoperative images, pathology, and basic information are collected. The texture features of images first taken upon admission before surgery are extracted. Next, the extracted texture features, image findings, pathology information, and basic information are input into the model. Finally, the prediction model is used to predict whether the patient will experience recurrence 6 months after surgery.

Statistical analysis

Measurement data were evaluated for normality with the Shapiro-Wilk test in SPSS 27.0 (IBM Corp., Armonk, NY, USA). Normally distributed variables are presented as the mean ± standard deviation, while nonnormally distributed data are expressed as the median with interquartile range. Between-group comparisons were performed with the independent two-sample t-test or nonparametric test as appropriate. Categorical variables are reported as proportions and were compared with the Chi-squared test or the Fisher exact test. Logistic regression was employed to evaluate the associations between each index parameter and disease outcomes in patients with EGFR-mutant advanced-stage NSCLC. Predictive performance was assessed via receiver operating characteristic (ROC) curve analysis. Correlation was analyzed with the Pearson or Spearman test. P<0.05 was considered statistically significant.

Results

Patients’ basic information and pathological and CT imaging features

A total of 48 patients were included in the study, comprising 30 (62.5%) patients with a favorable postoperative prognosis and 18 (37.5%) with a poor postoperative prognosis (Table 1).

Preliminary screening of texture features

Initially, 150 texture features were extracted. Dimension reduction was performed with the LASSO algorithm (Figure 3), yielding three candidate features at the optimal l-value of 0.0753: original-glrlm-long-run-high-gray-level-emphasis, log-sigma-1-0-mm-3D-GLCM-inverse-variance, log-sigma-1-0-mm-3D-GLDM-small-dependence-high-gray-level-emphasis.

Figure 3 Texture feature downscaling and screening process to derive the optimal λ value and key features. (A) Binomial deviance plot: the best hyperparameter λ value was selected by 10-fold cross-validation in the LASSO regression model, with the lowest value being the feature that best matches the true result. (B) Coefficients plot: the LASSO regression model identified texture features with nonzero coefficients. LASSO, least absolute shrinkage and selection operator.

Identification of the independent risk factors

One-way logistic regression analysis showed that among the clinical indicators, a history of smoking was significantly associated with poor prognosis (P<0.05), whereas there was no statistically significant difference between the good prognosis group and poor prognosis group in terms of gender, age, hypertension, diabetes mellitus, coronary heart disease, smoking history, alcohol consumption, burr, composition, lymph node metastasis, or pleural invasion (P>0.05). Multivariate logistic regression analysis further confirmed that smoking history was an independent risk factor for poor prognosis in patients with EGFR-mutant advanced-stage NSCLC (P<0.05) (Table 2).

For the final screening of texture features, single-factor and multifactor logistic regression analyses were further conducted on the three candidate texture features obtained from preliminary dimensionality reduction. The results indicated that long-run high-gray-level emphasis (LRHGE) was the sole independent predictor of poor prognosis in patients with EGFR-mutant advanced-stage NSCLC (P<0.05) (Table 2).

Comparison of predictive efficacy of the models

The combined model incorporating both clinical variables and texture features demonstrated superior efficacy in predicting outcomes of postoperative adjuvant targeted therapy in patients with EGFR-mutant NSCLC [area under the curve (AUC) =0.90; 95% confidence interval (CI): 0.810–0.990] compared to the clinical model alone (AUC =0.756; 95% CI: 0.627–0.884) and the texture-analysis model alone (AUC =0.771; 95% CI: 0.634–0.908) (Table 3 and Figure 4). The DeLong test confirmed that the AUC difference between the clinical model and the combined model was statistically significant (Z =−2.95, P<0.05), and the AUC difference between the texture analysis model and the combined model was statistically significant (Z =−1.20, P<0.05).

Figure 4 ROC curves for each model. CI, confidence interval; ROC, receiver operating characteristic.

The H-L test for the texture-analysis model and combined model yielded P values of 0.802 and 0.579, respectively, with a P value >0.10 indicating that both models had good goodness of fit. Decision curve analysis showed that when the threshold probability exceeded 7%, the combined model assessment of postoperative targeted therapy efficacy in patients with EGFR-mutated advanced-stage NSCLC provides a high net benefit and is most effective in assessing the risk of poor prognosis in these patients (Figure 5). Calibration curve analysis showed that the estimates of the combined model were highly consistent with the actual observations (Figure 6). These results demonstrated that the combined model had good agreement with the actual occurrence of poor prognosis and the risk of actual occurrence of postoperative targeted therapy in patients with EGFR-mutant advanced-stage NSCLC. Additionally, a nomogram was developed to visualize and integrate the predictive models, offering a practical tool for individualized efficacy assessment in this patient population (Figure 7).

Figure 5 Decision curves for each model.

Figure 6 Calibration curves of the combined model.

Figure 7 Column line diagram. Smoking: 0, non-smoking patients; 1, smoking patients.

Discussion

This study comprehensively analyzed the texture features of tumor lesions in chest CT scans of patients with EGFR-mutant NSCLC. By integrating clinical factors, we constructed a predictive model for the efficacy of postoperative targeted adjuvant therapy. The results indicated that the clinical model, texture-analysis model, and combined model each possess distinct advantages across different metrics, providing valuable assessment capabilities for evaluating the efficacy of postoperative targeted adjuvant therapy in patients with EGFR-mutated NSCLC. Among the models, the combined model demonstrated the most stable performance with a more pronounced predictive advantage. Therefore, texture analysis of chest CT scans can effectively predict the efficacy of postoperative targeted therapy in patients with EGFR-mutant NSCLC. Building upon this foundation, we also developed a straightforward prognostic assessment tool that provides clinicians with a convenient method for the early identification of patients who may potentially benefit from targeted therapy.

Due to the limitations of traditional clinical approaches—including the limited sensitivity and specificity of tumor markers and genetic testing, susceptibility to nontumor factors, and the psychological burden imposed on patients by repeated biopsies—developing a noninvasive and efficient method for evaluating treatment efficacy in patients with NSCLC undergoing postoperative targeted therapy has become a critical challenge (13,14). Texture feature analysis, as a noninvasive assessment method, offers a novel pathway for the dynamic monitoring of disease progression in patients with NSCLC. Previous studies on CT textural features have clearly demonstrated their precision in the pathological classification, staging evaluation, genomic feature identification, and prognostic determination in cancer, providing crucial support for subsequent research (15,16). In recent years, a growing body of evidence has demonstrated that the textural characteristics of malignant tumors are closely associated with patient survival rates and treatment response (17,18). Wu et al. (19) constructed a model to evaluate the efficacy of EGFR-TKI treatment in patients with advanced-stage EGFR-mutant NSCLC by combining pretreatment CT radiomics features with clinical predictors. The model achieved an AUC value of 0.977, further validating the value of CT-based texture features in assessing the efficacy of targeted therapy in patients with advanced-stage EGFR-mutant NSCLC. Imageomics techniques based on tumor texture features have demonstrated efficacy in characterizing tumor evolution, evaluating treatment response, and predicting the survival outcomes for patients with EGFR-mutant advanced-stage NSCLC receiving various chemotherapy regimens (20,21). These studies confirm that compared to traditional methods, texture analysis exhibits significant advantages in assessing prognosis and treatment efficacy for patients with EGFR-mutant advanced-stage NSCLC, with enhanced accuracy and reliability. Therefore, we innovatively applied texture analysis technology to evaluate the efficacy of postoperative targeted adjuvant therapy in patients with EGFR-mutant NSCLC. In our study, extracting preoperative texture features from tumors of patients with EGFR-mutant NSCLC yielded an AUC value of 0.771 (95% CI: 0.634–0.908), indicating excellent predictive performance. Consequently, CT texture analysis holds significant predictive value for evaluating the efficacy of postoperative targeted adjuvant therapy in patients with EGFR-mutant NSCLC.

The results of this study indicate that the texture feature LRHGE, belonging to the GLRLM category, is an independent risk factor for poor prognosis in patients with EGFR-mutant advanced-stage NSCLC (P<0.05). In texture analysis, GLRLM is highly sensitive to regional-scale heterogeneity by reflecting distribution patterns throughout the entire course of the tumor. It uniquely excels at evaluating large-scale texture variations within tumors, providing an intuitive representation of the high heterogeneity within tumor regions. Tumor heterogeneity is a primary cause of drug resistance (22,23). Therefore, this textural feature may, to some extent, suggest that certain tumor cells possess drug-resistant properties, which could potentially be a contributing factor to the suboptimal treatment outcomes observed in certain patients. EGFR gene alterations shape the patterns and characteristics of gene expression, leading to divergent responses in the microenvironment and treatment outcomes along the progression trajectory. This process drives tumor progression and malignancy by establishing a favorable ecosystem (24). Organizational genomics analyses have established the interpretability of texture features. These features (GLRLM) are closely associated with biological pathways relevant to tumor development, and their gene expression may alter texture distribution by influencing tumors and their microenvironments (25,26). This aligns with the core findings of this study: the EGFR gene may influence tumor progression by regulating expression in NSCLC tumor cells and affecting the tumor microenvironment, thereby altering the distribution of GLRLM-type texture features. This provides support for accurately assessing the efficacy of targeted therapies in patients. There may be a deeper indirect connection between these texture-related parameters and tumor drug resistance, warranting further validation in subsequent studies.

Smoking is one of the major risk factors for lung cancer, but its impact in patients with EGFR-mutant NSCLC is relatively complex. Song et al. (27) reported that EGFR mutations are more prevalent among nonsmokers, and smoking status also significantly impacts prognosis and treatment response in patients with EGFR-mutant NSCLC. Smokers may exhibit poorer outcomes and respond less effectively to EGFR-TKI therapy as compared to nonsmokers. Koo et al. have also shown that gas trapping and lung-marking changes observed on expiratory-phase CT images in smokers may be associated with a significant decline in lung function (28) and that these features interact with the biological behavior of tumors. In our study, among the various clinical predictors examined, smoking was the sole independent risk factor associated with poor postoperative prognosis following targeted therapy in patients with NSCLC (P<0.05), with the related clinical model achieving an AUC of 0.756. This finding reveals the critical role of smoking as a clinical factor in patient treatment outcomes, providing essential guidance for clinicians in more accurately assessing prognostic risks and optimizing individualized treatment decisions.

Furthermore, we confirmed that the combined model integrating clinical indicators and texture analysis features achieved an AUC value of 0.900 (95% CI: 0.810–0.990), which is comparable to the findings reported by Wu et al. (19). The AUC values for the single-texture feature model and the single-clinical factor model were 0.771 (95% CI: 0.634–0.908) and 0.756 (95% CI: 0.627–0.884), respectively, with the texture-analysis model demonstrating superior performance to the clinical model. Therefore, the combined model serves as a noninvasive and effective imaging method for evaluating the efficacy of postoperative targeted adjuvant therapy in patients with EGFR-mutant advanced-stage NSCLC. It not only addresses the limitations of traditional single-feature models but also provides a novel reference for clinical treatment decisions. Compared to traditional clinical assessment methods, this approach allows for the early prediction of disease progression and targeted decision-making while also effectively reducing patients’ financial burden and the psychological stress associated with efficacy evaluation.

This study involved several limitations that should be mentioned. Given that this study exclusively targeted a postoperative subgroup of patients with stage III EGFR-mutant NSCLC and that efficacy assessment was limited to first-generation erlotinib, the relatively small cohort size and lack of a validation cohort constrains the generalizability of the findings. Nevertheless, this study established a foundation for the preliminary, systematic exploration of CT-based texture analysis in this molecularly defined population and further paves the way for future in-depth investigations with larger sample sizes. We intend to expand the sample size through multicenter collaborative studies with multiple peer institutions. Furthermore, integrating multiomics data—including genomics and metabolomics with clinical variables—may allow for a more holistic evaluation of patient outcomes. It is also important to note that this study focused exclusively on lung adenocarcinoma, and its conclusions may have limited applicability to other types of lung cancer. Future research will aim to broaden the scope of investigation and to examine the potential value and effectiveness of these texture features in other lung cancer subtypes.

Conclusions

CT-based texture analysis offers unique advantages in quantifying tumor heterogeneity—a feature often undetectable through conventional imaging methods—thereby providing a more comprehensive perspective for assessing patient prognosis and treatment efficacy. This study demonstrated the potential of texture feature analysis in evaluating the prognosis of patients with EGFR-mutant advanced-stage NSCLC. However, further research is still needed to validate these findings and translate them into routine clinical practice.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1554/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1554/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1554/prf

Funding: This study was supported by Hebei Provincial Medical Science Research Project Program of Hebei Provincial Health Commission (No. 20210597).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1554/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of the First Affiliated Hospital of Hebei North University (approval No. K2020273) and individual consent for this retrospective analysis was waived.

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

References

1. Xue J, Li B, Wang Y, et al. Efficacy and Safety of Epidermal Growth Factor Receptor (EGFR)-Tyrosine Kinase Inhibitor Combination Therapy as First-Line Treatment for Patients with Advanced EGFR-Mutated, Non-Small Cell Lung Cancer: A Systematic Review and Bayesian Network Meta-Analysis. Cancers (Basel) 2022;14:4894. [Crossref] [PubMed]
2. Yang F, Zhang W, Shang X, et al. Comparison of the efficacy and safety of first-line treatments based on clinicopathological characteristics for patients with advanced epidermal growth factor receptor mutated non-small-cell lung cancer: A systematic review and network meta-analysis. Crit Rev Oncol Hematol 2022;177:103760. [Crossref] [PubMed]
3. Ahn MJ, Tsai CM, Shepherd FA, et al. Osimertinib in patients with T790M mutation-positive, advanced non-small cell lung cancer: Long-term follow-up from a pooled analysis of 2 phase 2 studies. Cancer 2019;125:892-901. [Crossref] [PubMed]
4. He J, Su C, Liang W, et al. Icotinib versus chemotherapy as adjuvant treatment for stage II-IIIA EGFR-mutant non-small-cell lung cancer (EVIDENCE): a randomised, open-label, phase 3 trial. Lancet Respir Med 2021;9:1021-9. [Crossref] [PubMed]
5. Lee CY, Lee SW, Hsu YC. Drug Resistance in Late-Stage Epidermal Growth Factor Receptor (EGFR)-Mutant Non-Small Cell Lung Cancer Patients After First-Line Treatment with Tyrosine Kinase Inhibitors. Int J Mol Sci 2025;26:2042. [Crossref] [PubMed]
6. Zhang G, Zhao Z, Cao Y, et al. Relationship between epidermal growth factor receptor mutations and CT features in patients with lung adenocarcinoma. Clin Radiol 2021;76:473.e17-24. [Crossref] [PubMed]
7. Hughes DJ, Chand G, Johnson J, et al. PD-L1 imaging with [99mTc]NM-01 SPECT/CT is associated with metabolic response to pembrolizumab with/without chemotherapy in advanced lung cancer. Br J Cancer 2025;132:913-21.
8. Yang H, Liu Z, Wang H, et al. Relationship between EGFR, ALK Gene Mutation and Imaging and Pathological Features in Invasive Lung Adenocarcinoma. Zhongguo Fei Ai Za Zhi 2022;25:147-55. [Crossref] [PubMed]
9. Pizurica M, Zheng Y, Carrillo-Perez F, et al. Digital profiling of gene expression from histology images with linearized attention. Nat Commun 2024;15:9886. [Crossref] [PubMed]
10. Bortolotto C, Pinto A, Brero F, et al. CT and MRI radiomic features of lung cancer (NSCLC): comparison and software consistency. Eur Radiol Exp 2024;8:71. [Crossref] [PubMed]
11. Rivière A, Lecuyer AI, Laurent E, et al. 1-year mortality in lung cancer in France according to key timepoints of care pathways. ERJ Open Res 2022;8:00157-2022. [Crossref] [PubMed]
12. Goldstraw P, Chansky K, Crowley J, et al. The IASLC Lung Cancer Staging Project: Proposals for Revision of the TNM Stage Groupings in the Forthcoming (Eighth) Edition of the TNM Classification for Lung Cancer. J Thorac Oncol 2016;11:39-51. [Crossref] [PubMed]
13. Pascual J, Attard G, Bidard FC, et al. ESMO recommendations on the use of circulating tumour DNA assays for patients with cancer: a report from the ESMO Precision Medicine Working Group. Ann Oncol 2022;33:750-68. [Crossref] [PubMed]
14. Ma L, Guo H, Zhao Y, et al. Liquid biopsy in cancer current: status, challenges and future prospects. Signal Transduct Target Ther 2024;9:336. [Crossref] [PubMed]
15. Wu K, Wu P, Yang K, et al. A comprehensive texture feature analysis framework of renal cell carcinoma: pathological, prognostic, and genomic evaluation based on CT images. Eur Radiol 2022;32:2255-65. [Crossref] [PubMed]
16. Yardimci AH, Sel I, Bektas CT, et al. Computed tomography texture analysis in patients with gastric cancer: a quantitative imaging biomarker for preoperative evaluation before neoadjuvant chemotherapy treatment. Jpn J Radiol 2020;38:553-60. [Crossref] [PubMed]
17. Bellini D, Carbone I, Rengo M, et al. Performance of Machine Learning and Texture Analysis for Predicting Response to Neoadjuvant Chemoradiotherapy in Locally Advanced Rectal Cancer with 3T MRI. Tomography 2022;8:2059-72. [Crossref] [PubMed]
18. Kolios C, Sannachi L, Dasgupta A, et al. MRI texture features from tumor core and margin in the prediction of response to neoadjuvant chemotherapy in patients with locally advanced breast cancer. Oncotarget 2021;12:1354-65. [Crossref] [PubMed]
19. Wu Z, Hua P, Chen X, et al. A study on the prediction of targeted therapy efficacy in advanced lung adenocarcinoma patients with EGFR mutations using CT-based delta-radiomics model. Front Med (Lausanne) 2025;12:1599206. [Crossref] [PubMed]
20. Collin A, Groza V, Missenard L, et al. A Model-Strengthened Imaging Biomarker for Survival Prediction in EGFR-Mutated Non-small-cell Lung Carcinoma Patients Treated with Tyrosine Kinase Inhibitors. Bull Math Biol 2021;83:68. [Crossref] [PubMed]
21. Shi L, Zhou XL, Sun JJ, et al. Whole-tumor perfusion CT using texture analysis in unresectable stage IIIA/B non-small cell lung cancer treated with recombinant human endostatin. Quant Imaging Med Surg 2019;9:968-75. [Crossref] [PubMed]
22. Kim HS, Kim YJ, Kim KG, et al. Preoperative CT texture features predict prognosis after curative resection in pancreatic cancer. Sci Rep 2019;9:17389. [Crossref] [PubMed]
23. Zhang A, Miao K, Sun H, et al. Tumor heterogeneity reshapes the tumor microenvironment to influence drug resistance. Int J Biol Sci 2022;18:3019-33. [Crossref] [PubMed]
24. Zhang J, Zhou W, Li N, et al. Multi-omics analysis unveils immunosuppressive microenvironment in the occurrence and development of multiple pulmonary lung cancers. NPJ Precis Oncol 2024;8:155. [Crossref] [PubMed]
25. Wang Y, Pan X, Lin H, et al. Multi-scale pathology image texture signature is a prognostic factor for resectable lung adenocarcinoma: a multi-center, retrospective study. J Transl Med 2022;20:595. [Crossref] [PubMed]
26. Wang S, Rong R, Yang DM, et al. Computational Staining of Pathology Images to Study the Tumor Microenvironment in Lung Cancer. Cancer Res 2020;80:2056-66. [Crossref] [PubMed]
27. Song J, Shi J, Dong D, et al. A New Approach to Predict Progression-free Survival in Stage IV EGFR-mutant NSCLC Patients with EGFR-TKI Therapy. Clin Cancer Res 2018;24:3583-92. [Crossref] [PubMed]
28. Koo MC, Au R, Hague CJ, et al. Expiration CT Gas Trapping Measures with Texture-Based Radiomics Improves Association with Lung Function and Lung Function Decline in COPD. Acad Radiol 2025;32:3711-21. [Crossref] [PubMed]

(English Language Editor: J. Gray)