Super-Resolution CT-Based Intratumoral and Peritumoral Radiomics Analysis for Trinary Classification of PD-L1 Status in Lung Adenocarcinoma ≥ 1cm

Super-Resolution CT-Based Intratumoral and Peritumoral Radiomics Analysis for Trinary Classification of PD-L1 Status in Lung Adenocarcinoma ≥ 1cm | 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 Super-Resolution CT-Based Intratumoral and Peritumoral Radiomics Analysis for Trinary Classification of PD-L1 Status in Lung Adenocarcinoma ≥ 1cm Yanqing Ma, Huizhi Ni, Zheng Guan, Yi Lin, Wenjie Liang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7629216/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Objectives To explore the potential of super-resolution CT-based intratumoral and peritumoral radiomics analysis for the trinary classification of PD-L1 status in lung adenocarcinoma with a diameter of ≥ 1cm, aiming to enhance the accuracy of PD-L1 status determination and support the development of personalized treatment strategies. Materials and methods Between 2016 and 2024, 949 lung adenocarcinoma patients was divided into three PD-L1 status groups, including TPS- (n = 519), TPS+ (n = 324), and TPS++ (n = 106). Clinical data were collected, and radiomics features were extracted from intratumoral and peritumoral regions using super-resolution CT images. After feature selection, Intra-Radio model, Peri-Radio model, and Intra/Peri-Radio model was developed using machine learning algorithms. Results Significant differences were found in gender ( P < 0.05), smoke ( P = 0.008), age ( P = 0.035), and diameter ( P < 0.05). The multi-layer perceptron (MLP) algorithm showed the highest accuracy (0.651 and 0.512 in the training and testing groups). The Intra/Peri-Radio model using MLP achieved micro AUC values of 0.826 (95%CI: 0.808 to 0.845) in the training group and 0.713 (95%CI: 0.678 to 0.748) in the testing group, and macro AUC values of 0.818 (95%CI: 0.781 to 0.853) in the training group and 0.660 (95%CI: 0.587 to 0.728) in the testing group. The Intra-Radio and Peri-Radio models based on MLP showed similar micro and macro AUC values, which were lower than those of Intra/Peri-Radio model. Conclusions Super-resolution CT-based intratumoral and peritumoral radiomics analysis effectively classified PD-L1 status in lung adenocarcinoma ≥ 1cm. MLP-based Intra/Peri-Radio model demonstrated strong performance, highlighting the potential of radiomics and machine learning for personalized treatments. adenocarcinoma lung PD-L1 radiomics computed tomography Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Lung adenocarcinoma is acknowledged as one of the most common malignant neoplasms, contributing significantly to the global healthcare burden due to its elevated mortality rates [ 1 ] . The Global Cancer Statistics reveal that lung cancer remains the leading cause of cancer-related deaths, emphasizing the urgent requirement for effective diagnostic and therapeutic measures [ 2 ] . Despite advancements in targeted therapies and immunotherapy have improved patient survival rates [ 3 ] , precise evaluation of programmed death-ligand 1(PD-L1) status remains challenging [ 4 ] . Conventional imaging techniques frequently fail to differentiate the various biological traits of tumors [ 5 ] , highlighting the necessity for innovative diagnostic approaches that can improve the accuracy of PD-L1 status evaluation [ 6 ] . In recent years, radiomics has emerged as a promising discipline, utilizing advanced imaging technologies to extract quantitative features from medical images [ 7 ] , thereby facilitating a deeper understanding of tumor heterogeneity and its micro-environment [ 8 ] . This study aims to employ super-resolution computed tomography (CT) to perform a radiomics analysis of both the intratumoral and peritumoral tissues in patients with lung adenocarcinoma. Previous studies have suggested that radiomics features can uncover vital insights into the tumor micro-environment and their correlation with PD-L1 expression levels [ 9 ] . This creates an opportunity to enhance patient stratification and guide personalized treatment plans based on imaging-derived biomarkers. The methodological approach of this research integrates super-resolution CT imaging to achieve superior image resolution and enhanced analytical precision [ 10 ] . By utilizing machine learning models alongside radiomics feature extraction, the study aspires to categorize PD-L1 status into three distinct classes. This strategy not only addresses the shortcomings associated with traditional imaging modalities but also offers a robust framework for managing intricate datasets. Lung adenocarcinoma with a diameter of ≥ 1cm typically exhibit higher invasiveness and metastatic potential [ 11 ] , and accurate evaluation of PD-L1 status can assist physicians in devising more tailored treatment plans prior to surgery [ 12 ] . The primary objective of the research is to harness super-resolution CT and machine learning to provide a comprehensive analysis of PD-L1 status through radiomics profiling, thereby establishing a foundation for customized therapeutic interventions. Materials and methods This retrospective study received approval from the Medical Ethics Committee of Zhejiang Provincial People’s Hospital (Approval No.2020QT108), and requirement for informed consent was waived. All procedures adhered to the principles outlined in the Declaration of Helsinki and complied with relevant regulatory guidelines. Patient population Between January 2016 and December 2024, a cohort of 949 patients with pathologically confirmed lung adenocarcinoma was enrolled. The inclusion criteria mandated that CT scans be performed within two weeks prior to surgery, CT images be of high quality and free from motion or respiratory artifacts, and patients have pathological confirmation of invasive adenocarcinoma (IAC) with a diameter exceeding 10 mm, as well as undergo PD-L1 immunohistochemical detection. Exclusion criteria encompassed individuals who underwent preoperative radiotherapy or chemotherapy, those with a previous diagnosis of primary cancers posing a considerable risk for metastasis, and patients with active infections or a documented history of particular infections. Ultimately, 949 patients were categorized into three distinct groups based on the TPS derived from PD-L1 immunohistochemical analysis. The classifications were as follows: TPS < 1% (negative/-, with less than 1% of tumor cells demonstrating positive PD-L1 staining), 1% ≤ TPS < 50% (low-positive/+, with 1% to 49% of tumor cells positive for PD-L1), and TPS ≥ 50% (high-positive/++, with 50% or more of tumor cells exhibiting positive PD-L1 staining). The patient enrollment process is depicted in Fig. 1 . CT imaging and super-resolution reconstruction The CT imaging data were obtained using 64 or 128-row detector CT scanners, specifically the Somatom Definition AS/AS + and Philips Incisive CT models. The chest image, captured during the patient's breath-hold status after a deep inhalation, is a volumetric scan extending from the lung apices to the adrenal glands, with the patient positioned supine and arms elevated to minimize motion and beam-hardening artifacts. The scanning parameters are as follows: 120 kVp tube voltage, automatic current regulation, standard algorithm reconstruction, and slice thickness/interval of 1–2 mm. Super-resolution reconstruction entails resampling isotropic voxels to achieve a resolution of 1×1×1 mm 3 , alongside the implementation of optimized lung window settings, specifically a width of 1500 HU and a level of -600 HU, to standardize CT images. Meanwhile, a hybrid architecture integrating generative adversarial networks (GANs) with deep transfer learning algorithms is employed to enhance the in-plane resolution fourfold into 1×1×0.25 mm 3 . In this process, GANs operate on a competitive learning paradigm where the generator network synthesizes realistic images approximating true data distributions. The comprehensive workflow of the super-reconstruction analysis is shown in Fig. 2 . Radiomics feature extraction and selection Radiomics features are categorized into geometric properties depicting the 3D shape and structure of tumors, first-order statistics measuring voxel intensity distributions within lesions to reflect internal uniformity and variability, and texture characteristics analyzing the spatial arrangement of intensity values using methods like gray-level co-occurrence matrix (GLCM), gray-level dependence matrix (GLDM), gray-level run-length matrix (GLRLM), gray-level size zone matrix (GLSZM) ,and neighborhood gray-tone difference matrix (NGTDM) to capture complex voxel relationships. Following a consensus reached through discussion, two radiologists (10 and 15 years of experience) delineated the intratumoral volume of interest (VOI) layer by layer, and subsequently used Python software to automatically expand the VOI outward by 3 mm to generate peritumoral VOI. Finally, 1561 intratumoral and peritumoral radiomics features including 306 first-order, 14 shape, 374 GLCM, 272 GLRLM, 238 GLDM, 272 GLSZM, and 85 NGTDM features, were extracted using in-house software based on PyRadiomics ( http://PyRadiomics.readthedocs.io ), respectively. After calculating radiomics features and normalizing them via Z-score standardization, synthetic minority oversampling technique (SMOTE), univariate analysis including ANOVA (analysis of variance), correlation analysis (to remove features with a correlation coefficient exceeding 0.9), the mRMR (maximal redundancy minimal relevance) algorithm (for filtering unnecessary/redundant features), and LASSO (least absolute shrinkage and selection operator) regression (to identify the most relevant features by shrinking coefficients of less important ones to zero) is conducted, with the selected features further validated through 5-fold cross-validation. Model construction and evaluation The 949 cases were randomly divided into training and testing groups at a 7:3 ratio. Various machine learning methods, including support vector machine(SVM), RandomForest, Logistic Regression, LightGBM, ExtraTrees, XGBoost, and multi-layer perceptron (MLP), were used to construct Trinary Classification models for predicting PD-L1 status based on the selected intratumoral and peritumoral radiomics features. Ultimately, three radiomics classification models were developed: the intratumoral radiomics (Intral-Radio) model, the peritumoral radiomics (Peri-Radio) model, and the combined intratumoral-peritumoral radiomics (Intra/Peri-Radio) model. The performance of these three-class machine learning models was rigorously evaluated across training and testing cohorts using metrics including accuracy, precision, sensitivity, specificity, PPV, NPV, recall, F1 score, and ROC-AUC (receiver operating characteristic curve-area under the curve). Calibration curves assessed model consistency, while decision curve analysis evaluated net clinical benefit. The study’s framework is illustrated in Fig. 3 . Statistical analysis Data analysis was performed with IBM SPSS (Version 24.0.0), R (Version 3.4.1), and Python (Version 3.7.12). Quantitative data are presented as mean ± SD and categorical data shown as counts (percentages). Clinical baseline data of patients were compared using one-way ANOVA if they met normality and homogeneity assumptions; otherwise, the Kruskal-Wallis test was applied. Categorical variables were assessed via the chi-square test. Statistical significance was set at P < 0.05 (two-tailed). Results Clinical baseline of patients A cohort of 949 patients were divided into three groups, with 519 in the TPS- group, 324 in the TPS + group and 106 in the TPS + + group. There was statistical differences in the clinical data regarding gender (χ 2 = 18.966, P < 0.05) and smoke (χ 2 =117.018, P = 0.008), but no significant statistical differences were observed in age (H = 6.710, < 0.05), tumor history (χ 2 =0.357, P = 0.837), drink (χ 2 =2.683, P = 0.261), hypertension (χ 2 =0.208, P = 0.901), and diameter (H = 81.563, P < 0.05) between three groups. The cohort was divided into the training group (TPS-, TPS+, and TPS++) and testing group (TPS-, TPS+, and TPS++) with a random proportion of 7:3 for model reconstruction and validation. The specific clinical data was shown in Table 1 . Table 1 The clinical information of patients. TPS- TPS+ TPS++ Cohort Gender(female/male) 263(27.7%)/ 256(27.0%) 170(17.9%)/ 154(16.2) 78(8.2%)/ 28(3.0%) 511(53.8%)/ 438(46.2%) Age(mean ± SD) 63.561 ± 0.443 64.595 ± 0.522 65.962 ± 0.842 64.182 ± 9.735 Diameter(mean ± SD) 1.722 ± 0.034 1.966 ± 0.054 2.879 ± 0.144 1.934 ± 1.010 Smoke(%) 35(3.7%) 84(8.9%) 49(5.2%) 168(17.7%) Drink(%) 58(6.1%) 43(4.5%) 8(0.8%) 109(11.5%) Tumor history(%) 72(7.6%) 49(5.2%) 14(1.5%) 135(14.2%) Hypertension(%) 102(10.7%) 61(6.4%) 22(2.3%) 185(19.5%) The comparison of models Following the feature selection, 19 intratumoral and peritumoral radiomics features were retained to develop trinary classification models using SVM, RandomForest, Logistic Regression, LightGBM, ExtraTrees, XGBoost, and MLP algorithms. The accuracy of MLP algorithms (0.651 in the training group and 0.512 in the testing group) was the highest compared with other methods. The Intra/Peri-Radio model using the MLP algorithm had micro AUC values of 0.826 (95%CI: 0.808 to 0.845) in the training group and 0.713 (95%CI: 0.678 to 0.748) in the testing group, and macro AUC values of 0.818 (95%CI: 0.781 to 0.853) in the training group and 0.660 (95%CI: 0.587 to 0.728) in the testing group. The AUC values for this model’s standalone classification for category TPS- were 0.787 and 0.680, for category TPS + were 0.769 and 0.498, for category TPS + + were 0.895 and 0.794, in the training and testing groups respectively (Fig. 4 ). While, the Intra-Radio and Peri-Radio model based on MLP machine learning method show similar micro and macro AUC values in the training group (micro AUC: 0.751 vs. 0.738 ,macro AUC: 0.738 vs. 0.734) and testing group (micro AUC: 0.654 vs. 0.596,macro AUC: 0.633 vs. 0.520). The detail results of three models were illustrated in Table 2 . Table 2 The detail performance of three models. Micro-AUC Macro-AUC TPS- AUC TPS + AUC TPS + + AUC Intra-Radio model Training group 0.751 (0.730–0.773) 0.738 (0.696–0.778) 0.723 (0.684–0.761) 0.632 (0.587–0.677) 0.856 (0.817–0.895) Testing group 0.654 (0.618–0.691) 0.633 (0.555–0.705) 0.625 (0.561–0.690) 0.476 (0.408–0.544) 0.789 (0.698–0.881) Peri-Radio model Training group 0.738 (0.715–0.760) 0.734 (0.689–0.777) 0.708 (0.669–0.747) 0.697 (0.655–0.738) 0.794 (0.743–0.846) Testing group 0.596 (0.557–0.635) 0.520 (0.438–0.597) 0.562 (0.495–0.630) 0.491 (0.419–0.562) 0.500 (0.401–0.598) Intra/Peri-Radio model Training group 0.826 (0.808–0.845) 0.818 (0.781–0.853) 0.787 (0.753–0.821) 0.769 (0.732–0.807) 0.895 (0.859–0.930) Testing group 0.713 (0.678–0.748) 0.660 (0.587–0.728) 0.680 (0.616–0.744) 0.498 (0.423–0.573) 0.794 (0.721–0.868) Discussion The increasing incidence of lung adenocarcinoma poses a significant public health challenge [ 13 ] , as it not only impacts patients' survival rates but also imposes substantial economic burdens on healthcare systems [ 14 ] . Despite advancements in targeted therapies and immunotherapies, accurate assessment of PD-L1 expression status, an important biomarker for predicting treatment efficacy, remains a challenge [ 15 ] . Given the limitations of traditional imaging techniques in discriminating the diverse biological characteristics of tumors [ 16 ] , this study explores the potential of super-resolution CT to analyze radiomics features of lung adenocarcinoma and classify PD-L1 status into three distinct categories, aiming to enhance decision-making in personalized medicine. The Intra-Radio and Peri-Radio models based on the MLP machine learning method exhibited comparable micro (trainging group: 0.751 vs. 0.738, testing group: 0.654 vs. 0.596) and macro AUC (trainging group: 0.738 vs. 0.734, testing group: 0.633 vs. 0.520) values in both the training and testing groups. The comparable performance underscores the significant value of peritumoral tissue in predicting PD-L1 status and highlights the necessity of incorporating peritumoral information into future diagnostic strategies to enhance prediction accuracy [ 17 ] . By leveraging advanced imaging technology and machine learning methodologies, we investigate the relationship between intratumoral radiomics characteristics of tumors and their peritumoral micro-environments and PD-L1 expression levels. The methodology utilized in this research modestly contributes to the field by offering a three-class classification of PD-L1 expression, an area that has not been extensively investigated in prior literature. While previous studies have focused primarily on the correlation between PD-L1 expression and clinical outcomes or treatment responses [ 15 ] , it has been noted that PD-L1 expression is more commonly found in male smokers with adenocarcinoma histology who had EGFR mutation and EML4-ALK fusion protein [ 18 ] . Our research shows that radiomics features from super-resolution CT images can effectively differentiate PD-L1 status in lung adenocarcinoma patients, echoing Wang et al. (2021), who stressed the importance of integrating imaging biomarkers into clinical decision-making to improve cancer immunotherapy efficacy [ 19 ] . For the Intra/Peri-Radio model in classifying status of PD-L1, the micro and macro AUC values in both training and testing groups exceed 0.5, demonstrating that this model’s ability to rank the expression levels of PD-L1 is significantly better than random chance. However, the fact that the macro AUCs (0.818 in training group and 0.660 in testing group) were lower than their micro AUCs (0.826 in training and 0.713 in testing group), reflecting an imbalance sample size and suggesting that some classes have weaker distinguishing capabilities, which in turn lower the average performance. To address the issue of imbalanced sample sizes, we utilized the SMOTE technique to create additional synthetic samples from the existing minority class instances, thereby increasing their quantity and facilitating the model's ability to effectively learn characteristics [ 20 ] . Further verification of Intra/Peri-Radio model's ability to distinguish different TPS statuses in the testing group revealed that category TPS + + had an AUC of 0.794 (95%CI, 0.721–0.868), indicating robust differentiation from Categories TPS- and TPS+. And category TPS-, with an AUC of 0.680 (95%CI, 0.616–0.744), exhibited a moderate level of classification. In contrast, category TPS+'s AUC of 0.498 (95%CI, 0.423–0.573) revealed weaker identification and insufficient distinction from the other two categories. This results implied that during the assessing of TPS + status, it is crucial to heed the subtle indicators of potential limitations or the underlying complexities in category differentiation. The limitations of this study include the use of single-center data, which may impact the statistical power and limit the generalizability of the findings. Additionally, the lack of a longitudinal study design limits our capacity to evaluate the long-term predictive value of the radiomics features for patient outcomes. Future research should include extended follow-up and investigate combining these imaging biomarkers with other clinical and molecular factors to boost predictive accuracy. In conclusion, this research highlights the potential of utilizing super-resolution CT technology for the effective classification of PD-L1 status in lung adenocarcinoma patients. The findings suggest that intratumoral and peritumoral radiomics analysis can provide valuable insights into tumor immunobiology and aid in the development of personalized treatment strategies. Abbreviations ANOVA Analysis of variance CT Computed tomography GANs Generative adversarial networks GLCM Gray-level co-occurrence matrix GLDM Gray-level dependence matrix GLRLM Gray-level run-length matrix GLSZM Gray-level size zone matrix IAC Invasive adenocarcinoma LASSO Least absolute shrinkage and selection operator NGTDM Neighborhood gray-tone difference matrix MLP Multi-layer perceptron PD-L1 Programmed death-ligand 1 ROC-AUC Receiver operating characteristic curve-area under the curve SMOTE Synthetic minority oversampling technique SVM Support vector machine VOI Volume of interest Declarations Ethics approval and consent to participate: This retrospective study was approved by the Medical Ethics Committee of Zhejiang Provincial People’s Hospital (No. 2020QT108). All procedures were performed in accordance with the 1975 Declaration of Helsinki and its later amendments. The informed consent was waived for this retrospective study by the Medical Ethics Committee of hospitals. Consent for publication: Not applicable. Competing interests: The authors declare no competing interests. Human Ethics and Consent to Participate declarations Not applicable. Funding: This study was supported by the Medical and Health Research Projects of Health Commission of Zhejiang Province (2022KY040/2023KY472), and Zhejiang Provincial Natural Science Foundation of China (LTGY24H180017). Author Contribution Conceptualization: Yanqing Ma. Data curation and investigation: all authors. Project administration and methodology: Yanqing Ma and Zheng Guan. Supervision: Yi Lin and Wenjie Liang. Writing-original draft: Yanqing Ma and Huizhi Ni. Writing-review & editing: all authors. Acknowledgement: Not applicable. Data Availability The datasets generated and/or analysed during the current study are not publicly available due Data security, but are available from the corresponding author on reasonable request. References Safi S, Benner A, Beckhove P, et al. Low tumour interleukin-1β expression predicts a limited effect of adjuvant platinum-based chemotherapy for patients with completely resected lung adenocarcinoma: An identification and validation study. Authors´reply Pulmonol. 2025;31(1):2447637. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229–63. Rahnea-Nita G, Corobcean N, Constantin GB et al. Unexpected Long-Term Survival and Downstaging in Oligometastatic Non-Small Cell Lung Cancer Treated with Multimodal Therapy. J Clin Med 2025; 14(10). Hektoen HH, Tsuruda KM, Brustugun OT, et al. Real-world comparison of pembrolizumab alone and combined with chemotherapy in metastatic lung adenocarcinoma patients with PD-L1 expression ≥ 50. ESMO open. 2025;10(5):105073. Chen ML, Shi AH, Li XT, et al. Is there any correlation between spectral CT imaging parameters and PD-L1 expression of lung adenocarcinoma? Thorac cancer. 2020;11(2):362–8. Nishimori M, Iwasa H, Miyatake K, et al. Correlation between PD-L1 expression and FDG-PET/CT visual assessments in non-small cell lung cancer resected specimens. Nucl Med Commun. 2025;46(7):636–42. Caruso D, Polici M, Zerunian M, et al. Radiomics in Oncology, Part 1: Technical Principles and Gastrointestinal Application in CT and MRI. Cancers (Basel). 2021;13(11):2522. Yuan H, Xu X, Tu S, et al. The CT-based intratumoral and peritumoral machine learning radiomics analysis in predicting lymph node metastasis in rectal carcinoma. BMC Gastroenterol. 2022;22(1):463. Wu Y, Shieh A, Cen S et al. Prediction of PD-L1 and CD68 in Clear Cell Renal Cell Carcinoma with Green Learning. J Imaging 2025; 11(6). Zhang Z, Yu S, Qin W, et al. Self-supervised CT super-resolution with hybrid model. Comput Biol Med. 2021;138:104775. Wang L, Jiang W, Zhan C, et al. Lymph node metastasis in clinical stage IA peripheral lung cancer. Lung Cancer. 2015;90(1):41–6. Ma Z, Wang Z, Li Y, et al. Detection and treatment of lung adenocarcinoma at pre-/minimally invasive stage: is it lead-time bias? J Cancer Res Clin Oncol. 2022;148(10):2717–22. Ma Y, Li J, Xu X, et al. The CT delta-radiomics based machine learning approach in evaluating multiple primary lung adenocarcinoma. BMC Cancer. 2022;22(1):949. Irandoust K, Alipour V, Arabloo J, et al. Economic burden of five common cancers in Iran: a systematic review of cost-of-illness with a focus on healthcare resource utilization. BMC Health Serv Res. 2025;25(1):800. Kawachi H, Yamada T, Tamiya M, et al. Clinical impact of cancer cachexia on the outcome of patients with non-small cell lung cancer with PD-L1 tumor proportion scores of ≥ 50% receiving pembrolizumab monotherapy versus immune checkpoint inhibitor with chemotherapy. Oncoimmunology. 2025;14(1):2442116. Salimi M, Vadipour P, Khosravi A, et al. CT-Based Radiomics for Predicting PD-L1 Expression in Non-small Cell Lung Cancer: A Systematic Review and Meta-analysis. Acad Radiol; 2025. Xu J, Liu L, Ji Y, et al. Enhanced CT-Based Intratumoral and Peritumoral Radiomics Nomograms Predict High-Grade Patterns of Invasive Lung Adenocarcinoma. Acad Radiol. 2025;32(1):482–92. Bassanelli M, Sioletic S, Martini M, et al. Heterogeneity of PD-L1 Expression and Relationship with Biology of NSCLC. Anticancer Res. 2018;38(7):3789–96. Itoh S, Yoshizumi T, Yugawa K, et al. Impact of Immune Response on Outcomes in Hepatocellular Carcinoma: Association With Vascular Formation. Hepatology. 2020;72(6):1987–99. Koivu A, Sairanen M, Airola A, et al. Synthetic minority oversampling of vital statistics data with generative adversarial networks. J Am Med Inf Assoc. 2020;27(11):1667–74. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 14 Jan, 2026 Reviewers agreed at journal 14 Jan, 2026 Reviewers invited by journal 07 Jan, 2026 Editor invited by journal 17 Dec, 2025 Editor assigned by journal 02 Oct, 2025 Submission checks completed at journal 26 Sep, 2025 First submitted to journal 24 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-7629216","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":571193751,"identity":"75882a7d-1949-4ccc-be4d-508408da382a","order_by":0,"name":"Yanqing Ma","email":"","orcid":"","institution":"The First Affiliated Hospital, College of Medicine, Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Yanqing","middleName":"","lastName":"Ma","suffix":""},{"id":571193752,"identity":"dc2e9c85-f643-433c-a7e1-3209aeb19cb2","order_by":1,"name":"Huizhi Ni","email":"","orcid":"","institution":"Changsha Medical University","correspondingAuthor":false,"prefix":"","firstName":"Huizhi","middleName":"","lastName":"Ni","suffix":""},{"id":571193753,"identity":"fb9bd262-47f3-4875-a161-cd245195b8ed","order_by":2,"name":"Zheng Guan","email":"","orcid":"","institution":"Zhejiang Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zheng","middleName":"","lastName":"Guan","suffix":""},{"id":571193754,"identity":"9eb203ad-5c12-4333-9545-9608fb8e63de","order_by":3,"name":"Yi Lin","email":"","orcid":"","institution":"Zhejiang Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yi","middleName":"","lastName":"Lin","suffix":""},{"id":571193755,"identity":"6e971d8f-2050-46a8-afce-6c1f9639506a","order_by":4,"name":"Wenjie Liang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIie3RMQrCMBSA4VcCnZ52fS6tR4h0zU1c0qWb4OjQISLo6Oo9BOfAA6eoV+gR2guI6eaWjIL5t0C+kJcApFI/mYNs2BEWQnAfTQQ5VS5OeSsjCQLMjm0tX7ikKFDtHzxQzs2VESR0ah0kmXlquUVubjyzPdzbjQkRAU5qoonMtcwMh0nuiUXpL3ZASVEEwa0Maj++iCUErhZkVUnsH1nHzFJdXD2Ob/+VZ+Z+6FSYANmvhQ5unyrCp6ZSqdS/9wGhlz8v+JvJjAAAAABJRU5ErkJggg==","orcid":"","institution":"The First Affiliated Hospital, College of Medicine, Zhejiang University","correspondingAuthor":true,"prefix":"","firstName":"Wenjie","middleName":"","lastName":"Liang","suffix":""}],"badges":[],"createdAt":"2025-09-16 10:23:29","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7629216/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7629216/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100361151,"identity":"25cd4bbb-27fd-485d-9229-825f9e105640","added_by":"auto","created_at":"2026-01-16 07:44:30","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1114451,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/877305602645e8e8319a51f9.docx"},{"id":100010157,"identity":"96f631bc-8080-4dc1-8389-91ab12986f4f","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6939,"visible":true,"origin":"","legend":"","description":"","filename":"20ae5b429da34dd7a4cfbc20576ef2e8.json","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/414d3b6e670abf9c1e9b5a97.json"},{"id":100010161,"identity":"808567e5-332a-4aaf-926a-0510bc3bfb13","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69933,"visible":true,"origin":"","legend":"","description":"","filename":"20ae5b429da34dd7a4cfbc20576ef2e81enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/76b07afed8a16d4df6dc8f43.xml"},{"id":100010168,"identity":"4cdcd6f8-c942-4b16-a7ca-84b712354f4e","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27626,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/0c3d7b2a384b9947791e7392.png"},{"id":100361117,"identity":"02c28127-0513-45bc-b2b1-823b87a6464e","added_by":"auto","created_at":"2026-01-16 07:44:28","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75350,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/6ad8c3af95e946199b2f4cfb.png"},{"id":100010164,"identity":"07f4518f-b1ac-48a2-892a-d43125117ce2","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75470,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/15b53f54c29a269a405e7d9a.png"},{"id":100361919,"identity":"69c6808e-189e-4601-8d27-c4057d04172c","added_by":"auto","created_at":"2026-01-16 07:45:56","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54079,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/6c8966c65ca32aae89e0af57.png"},{"id":100010169,"identity":"d02e6595-bd26-4750-9ba5-cb961845b0ea","added_by":"auto","created_at":"2026-01-12 06:04:47","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68139,"visible":true,"origin":"","legend":"","description":"","filename":"20ae5b429da34dd7a4cfbc20576ef2e81structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/ab4d1cbd76ac7796cebdc638.xml"},{"id":100010163,"identity":"2f431d67-2b73-4b20-a3fa-bbb5d92442b9","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"html","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75844,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/6036f128170e45e0c5ef4849.html"},{"id":100010159,"identity":"91a96475-e326-47b8-b2d8-039f120db21a","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":211523,"visible":true,"origin":"","legend":"\u003cp\u003eThe patient enrollment process.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/760c27c16715ffece3112b8e.jpeg"},{"id":100010162,"identity":"10aa19e4-a825-4812-8f6f-48ccf9945981","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":303383,"visible":true,"origin":"","legend":"\u003cp\u003eThe comprehensive workflow of the super-reconstruction.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/58886d83830d9e4098938e74.jpeg"},{"id":100010166,"identity":"8bc9fd82-5730-4d65-b8f4-9079d6708970","added_by":"auto","created_at":"2026-01-12 06:04:46","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":442383,"visible":true,"origin":"","legend":"\u003cp\u003eThe framework of study.\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/13f4a5a52a22a22e73392e19.jpeg"},{"id":100361678,"identity":"886f8e1e-b304-47eb-aba5-0eea1e8a24e5","added_by":"auto","created_at":"2026-01-16 07:45:30","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":90900,"visible":true,"origin":"","legend":"\u003cp\u003eThe ROC curves of Intra/Peri-Radio model based on MLP algorithm.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/64341a085ab73a6bf45f1138.jpeg"},{"id":100380824,"identity":"08c3e3cc-a19a-40af-a31e-4fd48597ce72","added_by":"auto","created_at":"2026-01-16 10:34:51","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1706716,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7629216/v1/3aa58851-19ec-400f-a14a-86bc065abd00.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Super-Resolution CT-Based Intratumoral and Peritumoral Radiomics Analysis for Trinary Classification of PD-L1 Status in Lung Adenocarcinoma ≥ 1cm","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLung adenocarcinoma is acknowledged as one of the most common malignant neoplasms, contributing significantly to the global healthcare burden due to its elevated mortality rates\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. The Global Cancer Statistics reveal that lung cancer remains the leading cause of cancer-related deaths, emphasizing the urgent requirement for effective diagnostic and therapeutic measures\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. Despite advancements in targeted therapies and immunotherapy have improved patient survival rates\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e, precise evaluation of programmed death-ligand 1(PD-L1) status remains challenging\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. Conventional imaging techniques frequently fail to differentiate the various biological traits of tumors\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e, highlighting the necessity for innovative diagnostic approaches that can improve the accuracy of PD-L1 status evaluation\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn recent years, radiomics has emerged as a promising discipline, utilizing advanced imaging technologies to extract quantitative features from medical images\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e, thereby facilitating a deeper understanding of tumor heterogeneity and its micro-environment\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. This study aims to employ super-resolution computed tomography (CT) to perform a radiomics analysis of both the intratumoral and peritumoral tissues in patients with lung adenocarcinoma. Previous studies have suggested that radiomics features can uncover vital insights into the tumor micro-environment and their correlation with PD-L1 expression levels\u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. This creates an opportunity to enhance patient stratification and guide personalized treatment plans based on imaging-derived biomarkers. The methodological approach of this research integrates super-resolution CT imaging to achieve superior image resolution and enhanced analytical precision\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. By utilizing machine learning models alongside radiomics feature extraction, the study aspires to categorize PD-L1 status into three distinct classes. This strategy not only addresses the shortcomings associated with traditional imaging modalities but also offers a robust framework for managing intricate datasets.\u003c/p\u003e \u003cp\u003eLung adenocarcinoma with a diameter of \u0026ge;\u0026thinsp;1cm typically exhibit higher invasiveness and metastatic potential\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e, and accurate evaluation of PD-L1 status can assist physicians in devising more tailored treatment plans prior to surgery\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. The primary objective of the research is to harness super-resolution CT and machine learning to provide a comprehensive analysis of PD-L1 status through radiomics profiling, thereby establishing a foundation for customized therapeutic interventions.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e This retrospective study received approval from the Medical Ethics Committee of Zhejiang Provincial People\u0026rsquo;s Hospital (Approval No.2020QT108), and requirement for informed consent was waived. All procedures adhered to the principles outlined in the Declaration of Helsinki and complied with relevant regulatory guidelines.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatient population\u003c/h2\u003e \u003cp\u003eBetween January 2016 and December 2024, a cohort of 949 patients with pathologically confirmed lung adenocarcinoma was enrolled. The inclusion criteria mandated that CT scans be performed within two weeks prior to surgery, CT images be of high quality and free from motion or respiratory artifacts, and patients have pathological confirmation of invasive adenocarcinoma (IAC) with a diameter exceeding 10 mm, as well as undergo PD-L1 immunohistochemical detection. Exclusion criteria encompassed individuals who underwent preoperative radiotherapy or chemotherapy, those with a previous diagnosis of primary cancers posing a considerable risk for metastasis, and patients with active infections or a documented history of particular infections. Ultimately, 949 patients were categorized into three distinct groups based on the TPS derived from PD-L1 immunohistochemical analysis. The classifications were as follows: TPS\u0026thinsp;\u0026lt;\u0026thinsp;1% (negative/-, with less than 1% of tumor cells demonstrating positive PD-L1 staining), 1% \u0026le; TPS\u0026thinsp;\u0026lt;\u0026thinsp;50% (low-positive/+, with 1% to 49% of tumor cells positive for PD-L1), and TPS\u0026thinsp;\u0026ge;\u0026thinsp;50% (high-positive/++, with 50% or more of tumor cells exhibiting positive PD-L1 staining). The patient enrollment process is depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eCT imaging and super-resolution reconstruction\u003c/h3\u003e\n\u003cp\u003eThe CT imaging data were obtained using 64 or 128-row detector CT scanners, specifically the Somatom Definition AS/AS\u0026thinsp;+\u0026thinsp;and Philips Incisive CT models. The chest image, captured during the patient's breath-hold status after a deep inhalation, is a volumetric scan extending from the lung apices to the adrenal glands, with the patient positioned supine and arms elevated to minimize motion and beam-hardening artifacts. The scanning parameters are as follows: 120 kVp tube voltage, automatic current regulation, standard algorithm reconstruction, and slice thickness/interval of 1\u0026ndash;2 mm. Super-resolution reconstruction entails resampling isotropic voxels to achieve a resolution of 1\u0026times;1\u0026times;1 mm\u003csup\u003e3\u003c/sup\u003e, alongside the implementation of optimized lung window settings, specifically a width of 1500 HU and a level of -600 HU, to standardize CT images. Meanwhile, a hybrid architecture integrating generative adversarial networks (GANs) with deep transfer learning algorithms is employed to enhance the in-plane resolution fourfold into 1\u0026times;1\u0026times;0.25 mm\u003csup\u003e3\u003c/sup\u003e. In this process, GANs operate on a competitive learning paradigm where the generator network synthesizes realistic images approximating true data distributions. The comprehensive workflow of the super-reconstruction analysis is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eRadiomics feature extraction and selection\u003c/h3\u003e\n\u003cp\u003eRadiomics features are categorized into geometric properties depicting the 3D shape and structure of tumors, first-order statistics measuring voxel intensity distributions within lesions to reflect internal uniformity and variability, and texture characteristics analyzing the spatial arrangement of intensity values using methods like gray-level co-occurrence matrix (GLCM), gray-level dependence matrix (GLDM), gray-level run-length matrix (GLRLM), gray-level size zone matrix (GLSZM) ,and neighborhood gray-tone difference matrix (NGTDM) to capture complex voxel relationships. Following a consensus reached through discussion, two radiologists (10 and 15 years of experience) delineated the intratumoral volume of interest (VOI) layer by layer, and subsequently used Python software to automatically expand the VOI outward by 3 mm to generate peritumoral VOI.\u003c/p\u003e \u003cp\u003eFinally, 1561 intratumoral and peritumoral radiomics features including 306 first-order, 14 shape, 374 GLCM, 272 GLRLM, 238 GLDM, 272 GLSZM, and 85 NGTDM features, were extracted using in-house software based on PyRadiomics (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://PyRadiomics.readthedocs.io\u003c/span\u003e\u003cspan address=\"http://PyRadiomics.readthedocs.io\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), respectively. After calculating radiomics features and normalizing them via Z-score standardization, synthetic minority oversampling technique (SMOTE), univariate analysis including ANOVA (analysis of variance), correlation analysis (to remove features with a correlation coefficient exceeding 0.9), the mRMR (maximal redundancy minimal relevance) algorithm (for filtering unnecessary/redundant features), and LASSO (least absolute shrinkage and selection operator) regression (to identify the most relevant features by shrinking coefficients of less important ones to zero) is conducted, with the selected features further validated through 5-fold cross-validation.\u003c/p\u003e\n\u003ch3\u003eModel construction and evaluation\u003c/h3\u003e\n\u003cp\u003eThe 949 cases were randomly divided into training and testing groups at a 7:3 ratio. Various machine learning methods, including support vector machine(SVM), RandomForest, Logistic Regression, LightGBM, ExtraTrees, XGBoost, and multi-layer perceptron (MLP), were used to construct Trinary Classification models for predicting PD-L1 status based on the selected intratumoral and peritumoral radiomics features. Ultimately, three radiomics classification models were developed: the intratumoral radiomics (Intral-Radio) model, the peritumoral radiomics (Peri-Radio) model, and the combined intratumoral-peritumoral radiomics (Intra/Peri-Radio) model. The performance of these three-class machine learning models was rigorously evaluated across training and testing cohorts using metrics including accuracy, precision, sensitivity, specificity, PPV, NPV, recall, F1 score, and ROC-AUC (receiver operating characteristic curve-area under the curve). Calibration curves assessed model consistency, while decision curve analysis evaluated net clinical benefit. The study\u0026rsquo;s framework is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eData analysis was performed with IBM SPSS (Version 24.0.0), R (Version 3.4.1), and Python (Version 3.7.12). Quantitative data are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD and categorical data shown as counts (percentages). Clinical baseline data of patients were compared using one-way ANOVA if they met normality and homogeneity assumptions; otherwise, the Kruskal-Wallis test was applied. Categorical variables were assessed via the chi-square test. Statistical significance was set at \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 (two-tailed).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eClinical baseline of patients\u003c/h2\u003e \u003cp\u003eA cohort of 949 patients were divided into three groups, with 519 in the TPS- group, 324 in the TPS\u0026thinsp;+\u0026thinsp;group and 106 in the TPS\u0026thinsp;+\u0026thinsp;+\u0026thinsp;group. There was statistical differences in the clinical data regarding gender (χ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;18.966, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) and smoke (χ\u003csup\u003e2\u003c/sup\u003e=117.018, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.008), but no significant statistical differences were observed in age (H\u0026thinsp;=\u0026thinsp;6.710, \u0026lt;\u0026thinsp;0.05), tumor history (χ\u003csup\u003e2\u003c/sup\u003e=0.357, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.837), drink (χ\u003csup\u003e2\u003c/sup\u003e=2.683, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.261), hypertension (χ\u003csup\u003e2\u003c/sup\u003e=0.208, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.901), and diameter (H\u0026thinsp;=\u0026thinsp;81.563, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) between three groups. The cohort was divided into the training group (TPS-, TPS+, and TPS++) and testing group (TPS-, TPS+, and TPS++) with a random proportion of 7:3 for model reconstruction and validation. The specific clinical data was shown in 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\u003eThe clinical information of patients.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTPS-\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTPS+\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTPS++\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCohort\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender(female/male)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e263(27.7%)/\u003c/p\u003e \u003cp\u003e256(27.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e170(17.9%)/\u003c/p\u003e \u003cp\u003e154(16.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e78(8.2%)/\u003c/p\u003e \u003cp\u003e28(3.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e511(53.8%)/\u003c/p\u003e \u003cp\u003e438(46.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge(mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63.561\u0026thinsp;\u0026plusmn;\u0026thinsp;0.443\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e64.595\u0026thinsp;\u0026plusmn;\u0026thinsp;0.522\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65.962\u0026thinsp;\u0026plusmn;\u0026thinsp;0.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e64.182\u0026thinsp;\u0026plusmn;\u0026thinsp;9.735\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiameter(mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.722\u0026thinsp;\u0026plusmn;\u0026thinsp;0.034\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.966\u0026thinsp;\u0026plusmn;\u0026thinsp;0.054\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.879\u0026thinsp;\u0026plusmn;\u0026thinsp;0.144\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.934\u0026thinsp;\u0026plusmn;\u0026thinsp;1.010\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmoke(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e35(3.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e84(8.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e49(5.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e168(17.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDrink(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e58(6.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e43(4.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8(0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e109(11.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor history(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e72(7.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e49(5.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e14(1.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e135(14.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypertension(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e102(10.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e61(6.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22(2.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e185(19.5%)\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\n\u003ch3\u003eThe comparison of models\u003c/h3\u003e\n\u003cp\u003eFollowing the feature selection, 19 intratumoral and peritumoral radiomics features were retained to develop trinary classification models using SVM, RandomForest, Logistic Regression, LightGBM, ExtraTrees, XGBoost, and MLP algorithms. The accuracy of MLP algorithms (0.651 in the training group and 0.512 in the testing group) was the highest compared with other methods. The Intra/Peri-Radio model using the MLP algorithm had micro AUC values of 0.826 (95%CI: 0.808 to 0.845) in the training group and 0.713 (95%CI: 0.678 to 0.748) in the testing group, and macro AUC values of 0.818 (95%CI: 0.781 to 0.853) in the training group and 0.660 (95%CI: 0.587 to 0.728) in the testing group. The AUC values for this model\u0026rsquo;s standalone classification for category TPS- were 0.787 and 0.680, for category TPS\u0026thinsp;+\u0026thinsp;were 0.769 and 0.498, for category TPS\u0026thinsp;+\u0026thinsp;+\u0026thinsp;were 0.895 and 0.794, in the training and testing groups respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhile, the Intra-Radio and Peri-Radio model based on MLP machine learning method show similar micro and macro AUC values in the training group (micro AUC: 0.751 vs. 0.738 ,macro AUC: 0.738 vs. 0.734) and testing group (micro AUC: 0.654 vs. 0.596,macro AUC: 0.633 vs. 0.520). The detail results of three models were illustrated in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\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\u003eThe detail performance of three models.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMicro-AUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMacro-AUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTPS- AUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTPS\u0026thinsp;+\u0026thinsp;AUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTPS\u0026thinsp;+\u0026thinsp;+\u0026thinsp;AUC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntra-Radio model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraining group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.751\u003c/p\u003e \u003cp\u003e(0.730\u0026ndash;0.773)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.738\u003c/p\u003e \u003cp\u003e(0.696\u0026ndash;0.778)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.723\u003c/p\u003e \u003cp\u003e(0.684\u0026ndash;0.761)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.632\u003c/p\u003e \u003cp\u003e(0.587\u0026ndash;0.677)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.856\u003c/p\u003e \u003cp\u003e(0.817\u0026ndash;0.895)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTesting group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.654\u003c/p\u003e \u003cp\u003e(0.618\u0026ndash;0.691)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.633\u003c/p\u003e \u003cp\u003e(0.555\u0026ndash;0.705)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.625\u003c/p\u003e \u003cp\u003e(0.561\u0026ndash;0.690)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.476\u003c/p\u003e \u003cp\u003e(0.408\u0026ndash;0.544)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003cp\u003e(0.698\u0026ndash;0.881)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePeri-Radio model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraining group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.738\u003c/p\u003e \u003cp\u003e(0.715\u0026ndash;0.760)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.734\u003c/p\u003e \u003cp\u003e(0.689\u0026ndash;0.777)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.708\u003c/p\u003e \u003cp\u003e(0.669\u0026ndash;0.747)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.697\u003c/p\u003e \u003cp\u003e(0.655\u0026ndash;0.738)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.794\u003c/p\u003e \u003cp\u003e(0.743\u0026ndash;0.846)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTesting group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.596\u003c/p\u003e \u003cp\u003e(0.557\u0026ndash;0.635)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.520\u003c/p\u003e \u003cp\u003e(0.438\u0026ndash;0.597)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.562\u003c/p\u003e \u003cp\u003e(0.495\u0026ndash;0.630)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.491\u003c/p\u003e \u003cp\u003e(0.419\u0026ndash;0.562)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.500\u003c/p\u003e \u003cp\u003e(0.401\u0026ndash;0.598)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntra/Peri-Radio model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraining group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.826 (0.808\u0026ndash;0.845)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.818 (0.781\u0026ndash;0.853)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.787 (0.753\u0026ndash;0.821)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.769\u003c/p\u003e \u003cp\u003e(0.732\u0026ndash;0.807)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.895\u003c/p\u003e \u003cp\u003e(0.859\u0026ndash;0.930)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTesting group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.713\u003c/p\u003e \u003cp\u003e(0.678\u0026ndash;0.748)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.660\u003c/p\u003e \u003cp\u003e(0.587\u0026ndash;0.728)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.680\u003c/p\u003e \u003cp\u003e(0.616\u0026ndash;0.744)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.498\u003c/p\u003e \u003cp\u003e(0.423\u0026ndash;0.573)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.794\u003c/p\u003e \u003cp\u003e(0.721\u0026ndash;0.868)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe increasing incidence of lung adenocarcinoma poses a significant public health challenge\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e, as it not only impacts patients' survival rates but also imposes substantial economic burdens on healthcare systems\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. Despite advancements in targeted therapies and immunotherapies, accurate assessment of PD-L1 expression status, an important biomarker for predicting treatment efficacy, remains a challenge\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. Given the limitations of traditional imaging techniques in discriminating the diverse biological characteristics of tumors\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e, this study explores the potential of super-resolution CT to analyze radiomics features of lung adenocarcinoma and classify PD-L1 status into three distinct categories, aiming to enhance decision-making in personalized medicine. The Intra-Radio and Peri-Radio models based on the MLP machine learning method exhibited comparable micro (trainging group: 0.751 vs. 0.738, testing group: 0.654 vs. 0.596) and macro AUC (trainging group: 0.738 vs. 0.734, testing group: 0.633 vs. 0.520) values in both the training and testing groups. The comparable performance underscores the significant value of peritumoral tissue in predicting PD-L1 status and highlights the necessity of incorporating peritumoral information into future diagnostic strategies to enhance prediction accuracy\u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBy leveraging advanced imaging technology and machine learning methodologies, we investigate the relationship between intratumoral radiomics characteristics of tumors and their peritumoral micro-environments and PD-L1 expression levels. The methodology utilized in this research modestly contributes to the field by offering a three-class classification of PD-L1 expression, an area that has not been extensively investigated in prior literature. While previous studies have focused primarily on the correlation between PD-L1 expression and clinical outcomes or treatment responses\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e, it has been noted that PD-L1 expression is more commonly found in male smokers with adenocarcinoma histology who had EGFR mutation and EML4-ALK fusion protein\u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. Our research shows that radiomics features from super-resolution CT images can effectively differentiate PD-L1 status in lung adenocarcinoma patients, echoing Wang et al. (2021), who stressed the importance of integrating imaging biomarkers into clinical decision-making to improve cancer immunotherapy efficacy\u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. For the Intra/Peri-Radio model in classifying status of PD-L1, the micro and macro AUC values in both training and testing groups exceed 0.5, demonstrating that this model\u0026rsquo;s ability to rank the expression levels of PD-L1 is significantly better than random chance. However, the fact that the macro AUCs (0.818 in training group and 0.660 in testing group) were lower than their micro AUCs (0.826 in training and 0.713 in testing group), reflecting an imbalance sample size and suggesting that some classes have weaker distinguishing capabilities, which in turn lower the average performance. To address the issue of imbalanced sample sizes, we utilized the SMOTE technique to create additional synthetic samples from the existing minority class instances, thereby increasing their quantity and facilitating the model's ability to effectively learn characteristics\u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. Further verification of Intra/Peri-Radio model's ability to distinguish different TPS statuses in the testing group revealed that category TPS\u0026thinsp;+\u0026thinsp;+\u0026thinsp;had an AUC of 0.794 (95%CI, 0.721\u0026ndash;0.868), indicating robust differentiation from Categories TPS- and TPS+. And category TPS-, with an AUC of 0.680 (95%CI, 0.616\u0026ndash;0.744), exhibited a moderate level of classification. In contrast, category TPS+'s AUC of 0.498 (95%CI, 0.423\u0026ndash;0.573) revealed weaker identification and insufficient distinction from the other two categories. This results implied that during the assessing of TPS\u0026thinsp;+\u0026thinsp;status, it is crucial to heed the subtle indicators of potential limitations or the underlying complexities in category differentiation.\u003c/p\u003e \u003cp\u003eThe limitations of this study include the use of single-center data, which may impact the statistical power and limit the generalizability of the findings. Additionally, the lack of a longitudinal study design limits our capacity to evaluate the long-term predictive value of the radiomics features for patient outcomes. Future research should include extended follow-up and investigate combining these imaging biomarkers with other clinical and molecular factors to boost predictive accuracy.\u003c/p\u003e \u003cp\u003eIn conclusion, this research highlights the potential of utilizing super-resolution CT technology for the effective classification of PD-L1 status in lung adenocarcinoma patients. The findings suggest that intratumoral and peritumoral radiomics analysis can provide valuable insights into tumor immunobiology and aid in the development of personalized treatment strategies.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eANOVA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAnalysis of variance\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eComputed tomography\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGANs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGenerative adversarial networks\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLCM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray-level co-occurrence matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLDM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray-level dependence matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLRLM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray-level run-length matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLSZM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray-level size zone matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIAC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eInvasive adenocarcinoma\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLASSO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLeast absolute shrinkage and selection operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNGTDM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNeighborhood gray-tone difference matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMLP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMulti-layer perceptron\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePD-L1\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eProgrammed death-ligand 1\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC-AUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eReceiver operating characteristic curve-area under the curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSMOTE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSynthetic minority oversampling technique\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSupport vector machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eVOI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eVolume of interest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eEthics approval and consent to participate:\u003c/h2\u003e \u003cp\u003e This retrospective study was approved by the Medical Ethics Committee of Zhejiang Provincial People\u0026rsquo;s Hospital (No. 2020QT108). All procedures were performed in accordance with the 1975 Declaration of Helsinki and its later amendments. The informed consent was waived for this retrospective study by the Medical Ethics Committee of hospitals.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for publication:\u003c/strong\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCompeting interests:\u003c/strong\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eHuman Ethics and Consent to Participate declarations\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eThis study was supported by the Medical and Health Research Projects of Health Commission of Zhejiang Province (2022KY040/2023KY472), and Zhejiang Provincial Natural Science Foundation of China (LTGY24H180017).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eConceptualization: Yanqing Ma. Data curation and investigation: all authors. Project administration and methodology: Yanqing Ma and Zheng Guan. Supervision: Yi Lin and Wenjie Liang. Writing-original draft: Yanqing Ma and Huizhi Ni. Writing-review \u0026amp; editing: all authors.\u003c/p\u003e\u003ch2\u003eAcknowledgement:\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated and/or analysed during the current study are not publicly available due Data security, but are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSafi S, Benner A, Beckhove P, et al. Low tumour interleukin-1β expression predicts a limited effect of adjuvant platinum-based chemotherapy for patients with completely resected lung adenocarcinoma: An identification and validation study. Authors\u0026acute;reply Pulmonol. 2025;31(1):2447637.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahnea-Nita G, Corobcean N, Constantin GB et al. Unexpected Long-Term Survival and Downstaging in Oligometastatic Non-Small Cell Lung Cancer Treated with Multimodal Therapy. J Clin Med 2025; 14(10).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHektoen HH, Tsuruda KM, Brustugun OT, et al. Real-world comparison of pembrolizumab alone and combined with chemotherapy in metastatic lung adenocarcinoma patients with PD-L1 expression\u0026thinsp;\u0026ge;\u0026thinsp;50. ESMO open. 2025;10(5):105073.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen ML, Shi AH, Li XT, et al. Is there any correlation between spectral CT imaging parameters and PD-L1 expression of lung adenocarcinoma? Thorac cancer. 2020;11(2):362\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNishimori M, Iwasa H, Miyatake K, et al. Correlation between PD-L1 expression and FDG-PET/CT visual assessments in non-small cell lung cancer resected specimens. Nucl Med Commun. 2025;46(7):636\u0026ndash;42.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCaruso D, Polici M, Zerunian M, et al. Radiomics in Oncology, Part 1: Technical Principles and Gastrointestinal Application in CT and MRI. Cancers (Basel). 2021;13(11):2522.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan H, Xu X, Tu S, et al. The CT-based intratumoral and peritumoral machine learning radiomics analysis in predicting lymph node metastasis in rectal carcinoma. BMC Gastroenterol. 2022;22(1):463.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu Y, Shieh A, Cen S et al. Prediction of PD-L1 and CD68 in Clear Cell Renal Cell Carcinoma with Green Learning. J Imaging 2025; 11(6).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Z, Yu S, Qin W, et al. Self-supervised CT super-resolution with hybrid model. Comput Biol Med. 2021;138:104775.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang L, Jiang W, Zhan C, et al. Lymph node metastasis in clinical stage IA peripheral lung cancer. Lung Cancer. 2015;90(1):41\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa Z, Wang Z, Li Y, et al. Detection and treatment of lung adenocarcinoma at pre-/minimally invasive stage: is it lead-time bias? J Cancer Res Clin Oncol. 2022;148(10):2717\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa Y, Li J, Xu X, et al. The CT delta-radiomics based machine learning approach in evaluating multiple primary lung adenocarcinoma. BMC Cancer. 2022;22(1):949.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIrandoust K, Alipour V, Arabloo J, et al. Economic burden of five common cancers in Iran: a systematic review of cost-of-illness with a focus on healthcare resource utilization. BMC Health Serv Res. 2025;25(1):800.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKawachi H, Yamada T, Tamiya M, et al. Clinical impact of cancer cachexia on the outcome of patients with non-small cell lung cancer with PD-L1 tumor proportion scores of \u0026ge;\u0026thinsp;50% receiving pembrolizumab monotherapy versus immune checkpoint inhibitor with chemotherapy. Oncoimmunology. 2025;14(1):2442116.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalimi M, Vadipour P, Khosravi A, et al. CT-Based Radiomics for Predicting PD-L1 Expression in Non-small Cell Lung Cancer: A Systematic Review and Meta-analysis. Acad Radiol; 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu J, Liu L, Ji Y, et al. Enhanced CT-Based Intratumoral and Peritumoral Radiomics Nomograms Predict High-Grade Patterns of Invasive Lung Adenocarcinoma. Acad Radiol. 2025;32(1):482\u0026ndash;92.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBassanelli M, Sioletic S, Martini M, et al. Heterogeneity of PD-L1 Expression and Relationship with Biology of NSCLC. Anticancer Res. 2018;38(7):3789\u0026ndash;96.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eItoh S, Yoshizumi T, Yugawa K, et al. Impact of Immune Response on Outcomes in Hepatocellular Carcinoma: Association With Vascular Formation. Hepatology. 2020;72(6):1987\u0026ndash;99.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoivu A, Sairanen M, Airola A, et al. Synthetic minority oversampling of vital statistics data with generative adversarial networks. J Am Med Inf Assoc. 2020;27(11):1667\u0026ndash;74.\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":false,"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":"adenocarcinoma, lung, PD-L1, radiomics, computed tomography","lastPublishedDoi":"10.21203/rs.3.rs-7629216/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7629216/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e \u003cp\u003eTo explore the potential of super-resolution CT-based intratumoral and peritumoral radiomics analysis for the trinary classification of PD-L1 status in lung adenocarcinoma with a diameter of \u0026ge;\u0026thinsp;1cm, aiming to enhance the accuracy of PD-L1 status determination and support the development of personalized treatment strategies.\u003c/p\u003e\u003ch2\u003eMaterials and methods\u003c/h2\u003e \u003cp\u003eBetween 2016 and 2024, 949 lung adenocarcinoma patients was divided into three PD-L1 status groups, including TPS- (n\u0026thinsp;=\u0026thinsp;519), TPS+ (n\u0026thinsp;=\u0026thinsp;324), and TPS++ (n\u0026thinsp;=\u0026thinsp;106). Clinical data were collected, and radiomics features were extracted from intratumoral and peritumoral regions using super-resolution CT images. After feature selection, Intra-Radio model, Peri-Radio model, and Intra/Peri-Radio model was developed using machine learning algorithms.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eSignificant differences were found in gender (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), smoke (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.008), age (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.035), and diameter (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The multi-layer perceptron (MLP) algorithm showed the highest accuracy (0.651 and 0.512 in the training and testing groups). The Intra/Peri-Radio model using MLP achieved micro AUC values of 0.826 (95%CI: 0.808 to 0.845) in the training group and 0.713 (95%CI: 0.678 to 0.748) in the testing group, and macro AUC values of 0.818 (95%CI: 0.781 to 0.853) in the training group and 0.660 (95%CI: 0.587 to 0.728) in the testing group. The Intra-Radio and Peri-Radio models based on MLP showed similar micro and macro AUC values, which were lower than those of Intra/Peri-Radio model.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eSuper-resolution CT-based intratumoral and peritumoral radiomics analysis effectively classified PD-L1 status in lung adenocarcinoma\u0026thinsp;\u0026ge;\u0026thinsp;1cm. MLP-based Intra/Peri-Radio model demonstrated strong performance, highlighting the potential of radiomics and machine learning for personalized treatments.\u003c/p\u003e","manuscriptTitle":"Super-Resolution CT-Based Intratumoral and Peritumoral Radiomics Analysis for Trinary Classification of PD-L1 Status in Lung Adenocarcinoma ≥ 1cm","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-12 06:04:41","doi":"10.21203/rs.3.rs-7629216/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-01-14T09:11:02+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"164745095412858890418119755699215466523","date":"2026-01-14T07:49:08+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-07T07:55:30+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-12-17T06:38:29+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-02T18:30:03+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-26T18:36:12+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-09-24T11:55:45+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":"f6a9aa28-8cce-4d70-8035-11d0c082057d","owner":[],"postedDate":"January 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-01-12T06:04:42+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-12 06:04:41","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7629216","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7629216","identity":"rs-7629216","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}