Investigating the correlation between PD-L1 expression and radiomics predictions in non-small cell lung cancer using PET/CT imaging analysis

Investigating the correlation between PD-L1 expression and radiomics predictions in non-small cell lung cancer using PET/CT imaging analysis | 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 Investigating the correlation between PD-L1 expression and radiomics predictions in non-small cell lung cancer using PET/CT imaging analysis Ruxi chang, Cong Shen, Liang Luo, xiang Liu, Yan Li, Xiaoyi Duan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4207471/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background This study aims to build a clinical factor model by incorporating clinical factors and metabolic parameters, as well as lesion imaging features from PET/CT images. Additionally, radiomics models are established based on PET-CT images to assess its capability in predicting PD-L1 expression in patients with non-small cell lung cancer. Methods After retrospective data collection, based on the clinical factor logistic regression results, a clinical factor model was constructed. The regions of interest (ROIs) for PET in radiomics were delineated using a semi-automatic method, while those for diagnosis CT were manually delineated. After extracting radiomic features, feature selection was performed using variance analysis, correlation analysis, and Gradient Boosting Decision Tree (GBDT). PET, diagnosis CT, and combined models were constructed. Predictive power was evaluated through ROC analysis comparing different models. Result In all 104 cases(mean age, 63.90years+/-8.99, 62males) were evaluated. The SUVmax in the PD-L1 positive group was higher than that in the negative group (P = 0.04), but both metabolic parameters and imaging features showed no correlation with PD-L1 expression. The radiomics models outperformed the clinical factor model (AUC = 0.712), yet the clinical factor model exhibited higher specificity than all radiomics models (Specificity = 0.765). The predictive performance of the PET model surpassed that of the diagnosis CT model (AUC: 0.838 vs 0.723). The combined model demonstrated enhanced predictive performance (AUC = 0.874). Conclusion The radiomics models perform better in predicting PD-L1 expression than the clinical factor model. The radiomics model combining PET and diagnosis CT exhibits the best predictive performance. Positron emission tomography/computed tomography Radiomics Lung cancer PD-L1 Textural analysis Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Lung cancer is a severe disease impacting human life. The 2023 cancer statistics reveal that among newly diagnosed cancer patients who died in 2023, lung cancer ranked first, accounting for 21% [ 1 ] . Non-small cell lung cancer (NSCLC) constitutes over 85% of lung cancer cases. The NCCN guidelines state that immune checkpoint inhibitors (ICIs) can improve the prognosis of NSCLC patients, providing longer disease-free survival. Moreover, the group with high PD-L1 expression exhibits better survival benefits [ 2 ] . The current gold standard for PD-L1 detection is immunohistochemical (IHC) evaluation based on PD-L1 expression in cell/tissue tumor specimens. However, IHC typically requires biopsied tissue for biopsy, causing trauma to patients. Furthermore, biopsy results may vary at different times, and the heterogeneity of tumors may impact the results [ 3 ] . Therefore, a non-invasive method to measure PD-L1 status is needed to support clinical decisions. Due to the advantages of easy accessibility and repeatability, imaging examinations can be used to explore their predictive effect on PD-L1 status. Previous studies have predicted PD-L1 based on quantitative analysis results of 18 F-FDG PET/CT examinations, finding that metabolic parameters can predict the expression of PD-L1 in lung adenocarcinoma patients [ 4 ] . Radiomics or deep learning methods can complement the predictive ability of PD-L1 by imaging features in CT images [ 5 , 6 ] . This study focuses on non-small cell lung cancer (NSCLC) patients, exploring the relationship between patients' clinical data, PET/CT imaging data, and radiomics, and PD-L1 expression, providing a basis for immunotherapy in NSCLC patients. Materials and Methods Patients This retrospective study collected data from patients who underwent 18 F-FDG PET/CT examinations at the PET/CT departmen from January 2020 to December 2023. Inclusion criteria were as follows: ① Confirmation of pulmonary nodules through CT imaging, ② Confirmed through biopsy or surgical pathology as non-small cell lung cancer, with complete PD-L1 testing results., ③ Interval between biopsy or surgical procedure and 18 F-FDG PET/CT examination not exceeding one month, along with additional diagnosis chest CT and 1mm thin-layer reconstruction, ④ No anti-tumor, anti-inflammatory, or related treatment before biopsy or surgery. Exclusion criteria included: ① Nodules with extensive adhesions to the pleura or mediastinum, affecting the accuracy of ROI delineation, ② SPN diameter < 0.8cm, ③ Other lung lesions affecting the accuracy of ROI depiction. Figure 1 illustrates the patient inclusion process. Image acquisition The Philips Gemini TF PET/CT scanner was utilized. Intravenous injection of 3. 7MBq/kg of 18 F-FDG was administered based on patient weight, and imaging was conducted 60 minutes later. The scanning range extended from the base of the eye to the upper femur. diagnosis CT parameters consisted of : 120V voltage, automated tube current modulation at 200mAs, 1mm slice thickness, and a matrix selection of 512×512. PET parameters included: 5mm slice thickness, matrix selection of 128×128, 8 bed positions per patient, and 1.5 minutes per bed position. Before the PET examination, a CT scan of the same position was performed for each bed position, serving as a reference anatomical position for each layer of the PET scan. Image analysis Two nuclear medicine physicians conducted an analysis of PET/CT images. Quantitative data included patient age, Tumor size in CT images: in the maximum level of the tumor in the axial plane, (long diameter + short diameter)/2 of the lesion, Maximum standardized uptake value (SUVmax), and Metabolic tumor volume (MTV) in PET images. The quantitative results were measured twice by one nuclear medicine physician. If the results were inconsistent, another physician would perform the measurements. Qualitative data included patient gender and age and whether the following radiological signs appeared in CT images of lung nodules: lobulation sign, spiculation sign, vacuole sign, pleural retraction, tracheal cutoff, and bronchovascular bundle. For qualitative results, two physicians independently evaluated them and discussed any discrepancies before making a decision. Lung lesion segmentation and radiomics feature extraction on image The DICOM 3.0 format PET/CT images of patients included in the study, as well as diagnosis CT images, were imported into the 3D_slicer software(version 5.2.2, www.slicer.com ). Two physicians conducted the delineation of lung lesions. The PET image delineation method utilized the PETtumor automatic delineation approach. Subsequently, the ROI delineated on the PET image served as a reference for the meticulous manual delineation of lesions layer by layer within the diagnosis CT images. Figure 2 portrays the results of the delineation process facilitated by the 3D Slicer software. For feature extraction, the Pyradiomics V3.1.0 software was chosen. The extracted feature categories encompass First Order Statistics, Shape-based (2D and 3D), Gray Level Co-occurrence Matrix (GLCM), Gray Level Run Length Matrix (GLRLM), Gray Level Size Zone Matrix (GLSZM), Gray Level Dependence Matrix (GLDM), Neighboring Gray Tone Difference Matrix (NGTDM), as well as other types. Feature selection and model development Patients were divided into training and testing sets with a ratio of 7:3. Missing values were imputed using the median, and Z-Score normalization was applied. During feature selection, the Variance method was initially employed to select features. Features with a variance greater than 0.3 were retained. Subsequently, the Correlation method was used to analyze the correlation between these selected features, and features with a correlation coefficient lower than 0.7 were retained. To further narrow down the CT feature set and select the most impactful ones, the Gradient Boosting Decision Tree (GBDT) method was introduced to enhance model effectiveness and accuracy. Moreover, features extracted from PET images were combined with features extracted from diagnosis CT images to generate bimodal image fusion features. To evaluate the impact of extracted features on predicting PD-L1 expression, Tree and Forest methods were selected to build models for each modality, choosing the model with better performance. The comparison of different modeling methods is available in the appendix, resulting in the PET model, diagnosis CT model, and combined model. The aforementioned procedures were executed using scikit-learn 1.2.0. Statistical Analysis Patient data were presented using Python 3.9 and R software 4.2 in the form of mean ± standard deviation or percentage (if applicable). Based on the normality test results for continuous data, independent two-sample t-tests or Mann–Whitney U tests were selected for different data. Chi-square test was used for analyzing categorical data, evaluating clinical data differences between PD-L1 positive (TPS ≥ 1%) and negative (TPS < 1%) patients. Logistic regression was employed for the analysis of clinical and imaging data to obtain predictive probabilities and construct Receiver Operating Characteristic (ROC) curves. In the radiomics analysis, separate ROC curves were generated for different modalities, and the area under the curve (AUC) was calculated to determine the corresponding sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). These models were first assessed within the training cohort for predicting PD-L1 in non-small cell lung cancer patients and then evaluated in the testing cohort. A significance level of P < 0.05 was considered statistically significant. In accordance with the Declaration of Helsinki, this study has been approved by the Institutional Review Board (IRB) with a number provided and the informed consent of the participants has been waived. Results Patient Characteristics and Imaging Features In our study, a total of 104 patients were included. Among them, 70 patients were PD-L1 positive, and 34 patients were PD-L1 negative. Table 1 presents their clinical data and imaging features. Pathologically confirmed, there were 89 cases diagnosed with adenocarcinoma, 14 cases with squamous cell carcinoma, and 1 case pathologically confirmed as non-small cell lung cancer without further classification. The average age of the positive group was 65.12 ± 10.28, while the negative group was 61.51 ± 8.28. The difference between the two groups was statistically significant (P = 0.046). Additionally, there was a statistically significant difference between the two groups in terms of SUVmax (P = 0.040). There were no statistically significant differences between the malignant and benign groups in terms of patient gender, SPN imaging features (lobulated, spiculated, Vacuole sign, Pleural stretch sign, Tracheal cutoff sign, Bronchovascular Bundle), Tumor size, and MTV. Table.1 Characteristics of patients Positive(64) Negitive(39) t Z χ2 P Age, years (mean ± SD) 64.70 ± 9.43 61.92 ± 8.87 -1.484 - - 0.141 Gender - - 2.346 0.126 Male 41(64.06%) 19(48.72%) Female 23(35.94%) 20(51.28%) SUVmax (median[25%,75%]) 7.20[5.28,10.38] 6.55[2.80,9.55] - -1.397 - 0.162 MTV 6.35[3.98,11.65] 5.80[3.65,14.39] - -0.105 - 0.916 TLU 23.48[14.46,66.30] 21.12[5.81,94.14] - -0.639 - 0.523 COV 28.36[26.44,30.36] 26.58[22.49,28.50] - -2.529 - 0.011* SD, standard deviation, MTV, metabolic tumor volume, TLU, total lesion uptake, COV, coefficient of variation.*,P < 0.05. Imaging Analysis The results of logistic regression analysis predicting PD-L1 expression based on clinical and PET/CT imaging data are shown in Table 2. Univariate analysis indicates that age and gender in demographic data are associated with PD-L1 expression, with P-values both less than 0.1. Other imaging data are not correlated with PD-L1 expression (P > 0.1). Further multivariate analysis shows that the factors included in the study are not independent predictors of PD-L1 expression. Table.2 Evaluation of the efficacy of metabolic parameters in predicting PD-L1 expression Parameter AUC(25%, 75%) Se Sp P SUVmax 0.582 (0.481, 0.679) 0.875 0.359 0.2197 MTV 0.506 (0.406, 0.606) 0.297 0.590 0.0441* TLG 0.538 (0.437, 0.636) 0.922 0.308 0.0827 COV 0.649 (0.549, 0.741) 0.469 0.821 - Se, sensitivity, Sp, specificity, MTV, metabolic tumor volume, TLU, total lesion uptake, COV, coefficient of variation. P represents the comparison between the ROC curve constructed by other metabolic parameters and COV, *, P < 0.05. Feature Selection 106 features were extracted from PET images, and after feature selection, 8 features were finally chosen, including 1 shape feature, 2 First Order Statistics features, 3 gray-level dependence matrix features, and 2 gray-level size zone matrix features. From diagnosis CT images, 107 features were extracted, and after feature selection, 6 features were chosen, including 1 shape feature, 2 First Order Statistics features, 2 gray-level dependence matrix features, and 1 Neighboring Gray Tone Difference Matrix feature.. The extracted features are shown in Fig. 3 . Model Comparison Based on the logistic regression results, a clinical factors model was constructed using the predicted probabilities. The results of comparing the predictive abilities of the clinical factor model and radiomics model for PD-L1 expression are shown in Table 3 and Fig. 4 . The AUC of the clinical factor model is 0.712, with a sensitivity of 0.571 and specificity of 0.765, indicating a relatively high rate of false negatives. The AUC for the diagnosis CT model is 0.723, while the PET model is 0.838. The PET model has higher sensitivity than the diagnosis CT model and comparable specificity, suggesting that in the single-modality radiological model, PET outperforms diagnosis CT. By combining the features of the two models, a joint PET-CT model was created with a sensitivity of 0.857 and specificity of 0.636, yielding an AUC of 0.857. Table.3 Univariate and multivariate logistic regression analysis results Univariate analysis Multivariate analysis Parameter OR 95%CI P OR 95%CI P Age 1.034 0.989–1.081 0.144 Gender 1.876 0.835–4.215 0.127 SUVmax(< 3.7 vs.≥3.7) 3.500 1.293–9.476 0.014* 1.007 0.197–5.136 0.994 MTV(< 9.4cm3 vs. ≥9.4cm3) 0.653 0.285–1.496 0.314 TLU(< 6.36 vs. ≥6.36) 4.636 1.470-14.623 0.009* 3.387 0.579–19.809 0.176 COV(< 28.9 vs. ≥28.9) 3.419 1.364–8.574 0.009* 2.707 1.005–7.287 0.049* MTV, metabolic tumor volume, TLU, total lesion uptake, COV, coefficient of variation. *, P < 0.05. Discussion Immune checkpoint inhibitors have been widely used in the treatment of advanced non-small cell lung cancer patients, and they can provide sustained relief, with PD-L1 positivity being associated with significantly higher objective response rates [ 7 ] . To reduce the trauma caused by patients undergoing invasive procedures, alternative non-invasive methods to measure PD-L1 status are of significant importance for clinical decision support [ 8 , 9 ] . This study focused on patients with non-small cell lung cancer and explored the value of clinical models constructed from clinical factors, PET, diagnostic CT, and PET-diagnostic CT models in predicting PD-L1 expression. The expression of PD-L1 is somewhat correlated with the demographic factors of patients, and there is currently no unified conclusion in the research. It is commonly believed to be associated with oxidative stress induced by smoking. WU et al.'s study suggests that PD-L1 expression is higher in male patients, which is related to the higher smoking rate among male patients [ 10 ] . Moreover, with increasing age, the likelihood of cellular mutations is greater. Therefore, in elderly patients, the tumor burden is heavier, resulting in higher PD-L1 expression, and better efficacy of ICIs for this patient group [ 11 ] . Previous studies have predicted PD-L1 expression in lung adenocarcinoma based on radiological features from CT images, indicating a higher PD-L1 positivity rate in patients with tumors larger than 2cm in diameter. In our study, we chose to calculate tumor size using (long diameter + short diameter)/2, which, compared to tumor transverse diameter, demonstrated better effectiveness in reflecting 3D tumor morphology [ 12 , 13 ] . The solid-to-tumor ratio (CTR) is significantly correlated with PD-L1 expression (P = 0.003). Various lesion morphologies, such as pleural indentation (P = 0.007), spiculation (P < 0.01), and ground-glass opacity (P < 0.01), are associated with PD-L1 positivity [ 14 , 15 ] . This contrasts with our research findings, and we speculate that it may be due to the inclusion of a population confirmed as having a solitary lesion, comprising both squamous cell carcinoma and adenocarcinoma. Additionally, the difference may be related to our study's set threshold for PD-L1 positivity at TPS = 1%. The model constructed based on CT image features has an area under the curve (AUC) of 0.783, sensitivity of 81.1%, and specificity of 64.1%. The AUC is similar to our research results, but our findings exhibit higher specificity and lower sensitivity. 18 F-FDG PET/CT serves as a non-invasive method, providing both anatomical and metabolic information of lesions, and is an important imaging tool for the diagnosis and staging of lung cancer. Takada et al. found that the SUVmax of non-small cell lung cancer (NSCLC) patients expressing PD-L1 protein (TPS > 5%) was significantly higher than that of PD-L1 non-expressing NSCLC patients (P < 0.0001). Multivariate analysis indicated that high SUVmax is an independent predictor of PD-L1 positivity, with a threshold of 4.2 providing the best predictive ability for PD-L1 expression [ 16 ] . In research targeting PD-L1 expression in lung adenocarcinoma patients, tumor PD-L1 expression is positively correlated with the maximum standardized uptake value (SUVmax) and total lesion glycolysis (TLG) (both P < 0.0001). SUVmax serves as an independent predictor of tumor PD-L1 expression, with an optimal threshold of 9.5 [ 17 ] . Some studies propose that when using TPS = 1% as the threshold for defining PD-L1 positivity, there are intergroup differences in SUVmax, TLG, and MTV. However, SUVmax is not correlated with PD-L1 status. This is similar to our research conclusion, where SUVmax shows intergroup differences (P = 0.040), but logistic regression analysis indicates no correlation between SUVmax and PD-L1 expression (P = 0.12) [ 18 ] . Currently, there is a considerable amount of research on the relationship between radiomics and PD-L1 expression. Sun et al. found that combining radiomic features from CT images with clinical-pathological factors resulted in the highest accuracy in predicting PD-L1 expression levels in non-small cell lung cancer patients. The AUC in the test set was 0.848, with a sensitivity of 83.3%, providing valuable assistance in accurately detecting positive patients [ 19 ] . Yoon et al. similarly found that adding radiomic features to a model based on clinical factors improves the predictive ability of the model (c-statistic = 0.646 vs. 0.550, P = 0.0299) [ 20 ] . Similarly, research has explored the predictive capability of PET/CT radiomics for PD-L1 expression. It was found that combining radiomic features and clinical-pathological features in the prediction model can achieve good results, with an AUC of 0.762 for expression levels greater than 1% and an AUC of 0.814 for levels greater than 50% [ 21 ] . Jiang compared the predictive performance of PET, CT, and PET/CT models individually and found that the CT model had the best performance (AUC = 0.86). Interestingly, combining PET features actually decreased the predictive ability of the model (AUC = 0.85) [ 22 ] . This differs from our research results, and we attribute it to the data source. Firstly, in their study, low-dose CT was used, which may obscure some fine texture features due to its lower resolution. Additionally, their method of delineating ROI on CT images directly corresponding to PET images may lead to incorrect matching and affect the results. In our study, compared to the clinical factors model, the radiomics model shows better overall performance, but the clinical model is more accurate in predicting positive cases. Utilizing radiomics from PET images alone yields favorable results, indicating that metabolic imaging is more effective than anatomical imaging in predicting PD-L1. The combination of PET and diagnosis CT models only increases the model's AUC, with sensitivity at the cut-off point not as high as the standalone PET model. Among the features extracted from PET images, Sphericity represents the tumor's spherical ratio, indicating that PD-L1 expression is related to the tumor's surrounding morphology. However, this is primarily reflected in metabolic images rather than anatomical images. Consistent with our observations of tumor signs in CT images (lobulation, spiculation), different signs cannot predict PD-L1 expression. Zhang et al.'s study suggests that sphericity is closely related to EGFR mutations, and EGFR is an important factor in the PD-L1 regulation pathway [ 23 , 24 ] . The Coarseness feature in CT images represents the roughness of the image texture, reflecting the variation in grayscale levels. This is consistent with the results of Suda et al.'s study, where they found that lesions with ground-glass opacity had a lower incidence of PD-L1 expression positivity compared to solid lesions (4% vs. 25%, P<0.01) [ 25 ] . Among all the features extracted from PET and CT images, including morphological features, first-order histogram features, and texture features, a total of 10 features are correlated with lesion heterogeneity. This indicates that the metabolic distribution within the tumor can reflect PD-L1 expression, and further exploration will be conducted in future work. Our study has some limitations: Firstly, it is a single-center study with a limited number of cases, which may restrict the generalizability of the results. Secondly, the experiment is retrospective, and patients who undergo 18 F-FDG-PET/CT examinations are usually those with suspicious lesions detected through CT scans or hematological examinations, introducing inherent selection bias. To further validate these findings, larger-scale prospective multicenter studies are needed. Thirdly, the described segmentation method may introduce measurement errors. Although automatic segmentation is used for lesion depiction in PET images, manual delineation is employed for CT images based on PET-defined ROIs. This manual delineation may have lower repeatability. In future studies, we will explore automatic segmentation methods to address this issue and improve accuracy. Our study found that the clinical factor model can achieve good results in predicting PD-L1 expression. However, it has a lower sensitivity, leading to a significant number of false negatives. Models constructed based on radiomics outperform the clinical factor model. The PET model can do a great job in predicting PD-L1 expression, while the combination of PET and diagnosis CT in a joint radiomics model yields the best results, ensuring the correct detection of PD-L1-positive cases. Abbreviations NSCLC non-small cell lung cancer ICIs Immune checkpoint inhibitors PD-L1 programmed cell death ligand TLU total lesion uptake Declarations Ethics approval and consent to participate In accordance with the Declaration of Helsinki, this study has been approved by the Institutional Review Board (IRB) with a number provided and the informed consent of the participants has been waived. Consent for publication Not applicable Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Competing interests The authors declare that they have no competing interests. Funding This study has received funding by the National Natural Science Foundation of China (No. 82001772), the Natural Science Foundation of Shaanxi Province, China (2021SF-062, 2020JZ38) and the New Medical and Technology of the First Affiliated Hospital of Xi’an Jiaotong University (XJYFY-2019J1). Authors' contributions Ruxi Chang, design of the work; have drafted the work. Cong Shen, analysis of data Liang Luo, acquisition of data Xiang Liu, acquisition of data Yan Li, interpretation of data Xiaoyi Duan, conception of the work, revised the paper. Acknowledgements Not applicable References Siegel RL, Miller KD, Wagle NS, Jemal A (2023) Cancer statistics, 2023. CA Cancer J Clin DOI:10.3322/caac.21763 Ettinger DS, Wood DE, Aisner DL, et al (2022) Non-Small Cell Lung Cancer, Version 3.2022, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw DOI:10.6004/jnccn.2022.0025 Ilie M, Long-Mira E, Bence C, et al (2016) Comparative study of the PD-L1 status between surgically resected specimens and matched biopsies of NSCLC patients reveal major discordances: a potential issue for anti-PD-L1 therapeutic strategies. Ann Oncol DOI:10.1093/annonc/mdv489 Hu B, Xiao J, Xiu Y, Fu Z, Shi H, Cheng D (2020) Correlation of PD-L1 expression on tumor cell and tumor infiltrating immune cell with 18F-fluorodeoxyglucose uptake on PET/computed tomography in surgically resected pulmonary adenocarcinoma. Nucl Med Commun DOI:10.1097/MNM.0000000000001136 Cook GJR, Goh V (2020) A Role for FDG PET Radiomics in Personalized Medicine?. Semin Nucl Med DOI:10.1053/j.semnuclmed.2020.05.002 Nakajo M, Jinguji M, Aoki M, Tani A, Sato M, Yoshiura T (2020) The clinical value of texture analysis of dual-time-point 18 F-FDG-PET/CT imaging to differentiate between 18 F-FDG-avid benign and malignant pulmonary lesions. Eur Radiol DOI:10.1007/s00330-019-06463-7 Rosner S, Reuss JE, Forde PM (2019) PD-1 Blockade in Early-Stage Lung Cancer. Annu Rev Med DOI:10.1146/annurev-med-050217-025205 Yu H, Boyle TA, Zhou C, Rimm DL, Hirsch FR (2016) PD-L1 Expression in Lung Cancer. J Thorac Oncol DOI:10.1016/j.jtho.2016.04.014 Rossi S, Toschi L, Castello A, Grizzi F, Mansi L, Lopci E (2017) Clinical characteristics of patient selection and imaging predictors of outcome in solid tumors treated with checkpoint-inhibitors. Eur J Nucl Med Mol Imaging DOI:10.1007/s00259-017-3802-5 Wu X, Huang Y, Zhao Q, et al (2020) PD-L1 expression correlation with metabolic parameters of FDG PET/CT and clinicopathological characteristics in non-small cell lung cancer. EJNMMI Re DOI:10.1186/s13550-020-00639-9 Nosrati A, Tsai KK, Goldinger SM, et al (2017) Evaluation of clinicopathological factors in PD-1 response: derivation and validation of a prediction scale for response to PD-1 monotherapy. Br J Cancer DOI:10.1038/bjc.2017.70 Kovalchik SA, Tammemagi M, Berg CD, et al (2013) Targeting of low-dose CT screening according to the risk of lung-cancer death. N Engl J Med DOI:10.1056/NEJMoa1301851 MacMahon H, Austin JH, Gamsu G, et al (2005) Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society. Radiology DOI:10.1148/radiol.2372041887 Wu T, Zhou F, Soodeen-Lalloo AK, et al (2019) The Association Between Imaging Features of TSCT and the Expression of PD-L1 in Patients With Surgical Resection of Lung Adenocarcinoma. Clin Lung Cancer DOI:10.1016/j.cllc.2018.10.012 Toyokawa G, Takada K, Okamoto T, et al (2017) Computed Tomography Features of Lung Adenocarcinomas With Programmed Death Ligand 1 Expression. Clin Lung Cancer DOI:10.1016/j.cllc.2017.03.008 Takada K, Toyokawa G, Okamoto T, et al (2017) Metabolic characteristics of programmed cell death-ligand 1-expressing lung cancer on 18 F-fluorodeoxyglucose positron emission tomography/computed tomography. Cancer Med DOI:10.1002/cam4.1215 Cui Y, Li X, Du B, Diao Y, Li Y (2020) PD-L1 in Lung Adenocarcinoma: Insights into the Role of 18 F-FDG PET/CT. Cancer Manag Res DOI:10.2147/CMAR.S256871 Polverari G, Ceci F, Bertaglia V, et al (2020) 18 F-FDG Pet Parameters and Radiomics Features Analysis in Advanced Nsclc Treated with Immunotherapy as Predictors of Therapy Response and Survival. Cancers (Basel) DOI:10.3390/cancers12051163 Sun Z, Hu S, Ge Y, et al (2020) Radiomics study for predicting the expression of PD-L1 in non-small cell lung cancer based on CT images and clinicopathologic features. J Xray Sci Technol DOI:10.3233/XST-200642 Yoon J, Suh YJ, Han K, et al (2020) Utility of CT radiomics for prediction of PD-L1 expression in advanced lung adenocarcinomas. Thorac Cancer DOI:10.1111/1759-7714.13352 Li J, Ge S, Sang S, Hu C, Deng S (2021) Evaluation of PD-L1 Expression Level in Patients With Non-Small Cell Lung Cancer by 18 F-FDG PET/CT Radiomics and Clinicopathological Characteristics. Front Oncol DOI:10.3389/fonc.2021.789014 Jiang M, Sun D, Guo Y, et al (2020) Assessing PD-L1 Expression Level by Radiomic Features From PET/CT in Nonsmall Cell Lung Cancer Patients: An Initial Result. Acad Radio DOI:10.1016/j.acra.2019.04.016 Zhang J, Zhao X, Zhao Y, et al (2020) Value of pre-therapy 18 F-FDG PET/CT radiomics in predicting EGFR mutation status in patients with non-small cell lung cancer. Eur J Nucl Med Mol Imaging DOI:10.1007/s00259-019-04592-1 Shi Y (2018) Regulatory mechanisms of PD-L1 expression in cancer cells. Cancer Immunol Immunother DOI:10.1007/s00262-018-2226-9 Suda K, Shimoji M, Shimizu S, et al (2019) Comparison of PD-L1 Expression Status between Pure-Solid Versus Part-Solid Lung Adenocarcinomas. Biomolecules DOI:10.3390/biom9090456 Additional Declarations No competing interests reported. Supplementary Files supplyment.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4207471","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":288460439,"identity":"5e2c1f16-6cfa-4baa-9b5b-199c9f8e6068","order_by":0,"name":"Ruxi chang","email":"","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Ruxi","middleName":"","lastName":"chang","suffix":""},{"id":288460440,"identity":"a97f9eed-4102-473a-9523-7142e6c3dd93","order_by":1,"name":"Cong Shen","email":"","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Cong","middleName":"","lastName":"Shen","suffix":""},{"id":288460441,"identity":"ec70080b-a917-4369-a4c1-2ca8c3e82c5e","order_by":2,"name":"Liang Luo","email":"","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Liang","middleName":"","lastName":"Luo","suffix":""},{"id":288460442,"identity":"6c274050-c896-48d2-8070-b1184af289c7","order_by":3,"name":"xiang Liu","email":"","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"xiang","middleName":"","lastName":"Liu","suffix":""},{"id":288460443,"identity":"9561dc91-f0ee-4ff0-8707-a61ca997e4ed","order_by":4,"name":"Yan Li","email":"","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Yan","middleName":"","lastName":"Li","suffix":""},{"id":288460444,"identity":"957142e8-3dd5-4c21-881f-94dbb52e6d25","order_by":5,"name":"Xiaoyi Duan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwklEQVRIiWNgGAWjYDCCA2BSQg5IsJGmxRimRYJYLQyJDURr4buR/Owxb5tF+objzc8eMO6wqSOoRfJGmrnhzDaJ3A1njpkbMJ5JI2yLwY0EM4mPIC0gBmPbYWK0pH+TSGyTSDe4//wbUMt/YrTkgG1JMLjBA7LlAGEtkmfelEnOOCdhOPNMThnQumTJBkJa+I6nb5PmKauT5zt+fBvQOjt+graAASMwRhQOABkJxKkHgT8MDPIEHTQKRsEoGAUjFgAApyU9gWJZOVQAAAAASUVORK5CYII=","orcid":"","institution":"First Affiliated Hospital of Xi'an Jiaotong University","correspondingAuthor":true,"prefix":"","firstName":"Xiaoyi","middleName":"","lastName":"Duan","suffix":""}],"badges":[],"createdAt":"2024-04-02 14:42:30","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4207471/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4207471/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":54313909,"identity":"b5cc7fc5-b6b0-4e9f-bbee-83b66f01b177","added_by":"auto","created_at":"2024-04-08 17:30:24","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":358047,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the SPN patients set.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4207471/v1/4ba277da537b54cbe2119881.png"},{"id":54313905,"identity":"8f4d67f2-d2fc-4f6f-a5e8-bdf3d90bbeae","added_by":"auto","created_at":"2024-04-08 17:30:23","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":925633,"visible":true,"origin":"","legend":"\u003cp\u003eOne patient with a lung malignancy with the SUVmax of 1.7.(a-c)ROI delineated in axial, coronal, and sagittal views in diagnosis CT images, (d-f)ROI delineated in axial, coronal, and sagittal views in PET images.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4207471/v1/f0789f742de3f84a550685fb.png"},{"id":54314656,"identity":"04f253bb-f55c-4e89-8818-5310dd1ee066","added_by":"auto","created_at":"2024-04-08 17:38:23","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1697483,"visible":true,"origin":"","legend":"\u003cp\u003eHeat map of feature extraction. (a)Features extracted from PET images in training set, (b)Features extracted from PET images in testing set, (c)Features extracted from diagnosis CT images in training set, (d)Features extracted from diagnosis CT images in testing set.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4207471/v1/2f9964b36df82363a67dcf11.png"},{"id":54313910,"identity":"6a6b7236-348e-4e65-82f0-062c464cae91","added_by":"auto","created_at":"2024-04-08 17:30:26","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":131467,"visible":true,"origin":"","legend":"\u003cp\u003eThe ROC curves of the clinical factors, diagnosis CT, PET, and PET-diagnosis CT models constructed on the testing set.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-4207471/v1/6c4ffc69ed02e5705f4b634a.png"},{"id":57045779,"identity":"ffa42161-6029-40eb-9620-1849e064b793","added_by":"auto","created_at":"2024-05-24 00:48:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4247903,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4207471/v1/90ba81c7-a7f6-476f-9358-fdcf1551c63d.pdf"},{"id":54313907,"identity":"4275ed5e-85a7-4c93-b2f6-828853ccf8a6","added_by":"auto","created_at":"2024-04-08 17:30:23","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":13106,"visible":true,"origin":"","legend":"","description":"","filename":"supplyment.docx","url":"https://assets-eu.researchsquare.com/files/rs-4207471/v1/d220c5208b634a895b4193e8.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Investigating the correlation between PD-L1 expression and radiomics predictions in non-small cell lung cancer using PET/CT imaging analysis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLung cancer is a severe disease impacting human life. The 2023 cancer statistics reveal that among newly diagnosed cancer patients who died in 2023, lung cancer ranked first, accounting for 21%\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Non-small cell lung cancer (NSCLC) constitutes over 85% of lung cancer cases. The NCCN guidelines state that immune checkpoint inhibitors (ICIs) can improve the prognosis of NSCLC patients, providing longer disease-free survival. Moreover, the group with high PD-L1 expression exhibits better survival benefits\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. The current gold standard for PD-L1 detection is immunohistochemical (IHC) evaluation based on PD-L1 expression in cell/tissue tumor specimens. However, IHC typically requires biopsied tissue for biopsy, causing trauma to patients. Furthermore, biopsy results may vary at different times, and the heterogeneity of tumors may impact the results\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. Therefore, a non-invasive method to measure PD-L1 status is needed to support clinical decisions.\u003c/p\u003e \u003cp\u003eDue to the advantages of easy accessibility and repeatability, imaging examinations can be used to explore their predictive effect on PD-L1 status. Previous studies have predicted PD-L1 based on quantitative analysis results of \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT examinations, finding that metabolic parameters can predict the expression of PD-L1 in lung adenocarcinoma patients\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. Radiomics or deep learning methods can complement the predictive ability of PD-L1 by imaging features in CT images\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. This study focuses on non-small cell lung cancer (NSCLC) patients, exploring the relationship between patients' clinical data, PET/CT imaging data, and radiomics, and PD-L1 expression, providing a basis for immunotherapy in NSCLC patients.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003ePatients\u003c/h2\u003e\n \u003cp\u003eThis retrospective study collected data from patients who underwent \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT examinations at the PET/CT departmen from January 2020 to December 2023. Inclusion criteria were as follows: ① Confirmation of pulmonary nodules through CT imaging, ② Confirmed through biopsy or surgical pathology as non-small cell lung cancer, with complete PD-L1 testing results., ③ Interval between biopsy or surgical procedure and \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT examination not exceeding one month, along with additional diagnosis chest CT and 1mm thin-layer reconstruction, ④ No anti-tumor, anti-inflammatory, or related treatment before biopsy or surgery. Exclusion criteria included: ① Nodules with extensive adhesions to the pleura or mediastinum, affecting the accuracy of ROI delineation, ② SPN diameter\u0026thinsp;\u0026lt;\u0026thinsp;0.8cm, ③ Other lung lesions affecting the accuracy of ROI depiction. Figure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the patient inclusion process.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003eImage acquisition\u003c/h2\u003e\n \u003cp\u003eThe Philips Gemini TF PET/CT scanner was utilized. Intravenous injection of 3. 7MBq/kg of \u003csup\u003e18\u003c/sup\u003eF-FDG was administered based on patient weight, and imaging was conducted 60 minutes later. The scanning range extended from the base of the eye to the upper femur. diagnosis CT parameters consisted of : 120V voltage, automated tube current modulation at 200mAs, 1mm slice thickness, and a matrix selection of 512\u0026times;512. PET parameters included: 5mm slice thickness, matrix selection of 128\u0026times;128, 8 bed positions per patient, and 1.5 minutes per bed position. Before the PET examination, a CT scan of the same position was performed for each bed position, serving as a reference anatomical position for each layer of the PET scan.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003eImage analysis\u003c/h2\u003e\n \u003cp\u003eTwo nuclear medicine physicians conducted an analysis of PET/CT images. Quantitative data included patient age, Tumor size in CT images: in the maximum level of the tumor in the axial plane, (long diameter\u0026thinsp;+\u0026thinsp;short diameter)/2 of the lesion, Maximum standardized uptake value (SUVmax), and Metabolic tumor volume (MTV) in PET images. The quantitative results were measured twice by one nuclear medicine physician. If the results were inconsistent, another physician would perform the measurements. Qualitative data included patient gender and age and whether the following radiological signs appeared in CT images of lung nodules: lobulation sign, spiculation sign, vacuole sign, pleural retraction, tracheal cutoff, and bronchovascular bundle. For qualitative results, two physicians independently evaluated them and discussed any discrepancies before making a decision.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003eLung lesion segmentation and radiomics feature extraction on image\u003c/h2\u003e\n \u003cp\u003eThe DICOM 3.0 format PET/CT images of patients included in the study, as well as diagnosis CT images, were imported into the 3D_slicer software(version 5.2.2, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.slicer.com\u003c/span\u003e\u003c/span\u003e). Two physicians conducted the delineation of lung lesions. The PET image delineation method utilized the PETtumor automatic delineation approach. Subsequently, the ROI delineated on the PET image served as a reference for the meticulous manual delineation of lesions layer by layer within the diagnosis CT images. Figure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e portrays the results of the delineation process facilitated by the 3D Slicer software. For feature extraction, the Pyradiomics V3.1.0 software was chosen. The extracted feature categories encompass First Order Statistics, Shape-based (2D and 3D), Gray Level Co-occurrence Matrix (GLCM), Gray Level Run Length Matrix (GLRLM), Gray Level Size Zone Matrix (GLSZM), Gray Level Dependence Matrix (GLDM), Neighboring Gray Tone Difference Matrix (NGTDM), as well as other types.\u003c/p\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003eFeature selection and model development\u003c/h2\u003e\n \u003cp\u003ePatients were divided into training and testing sets with a ratio of 7:3. Missing values were imputed using the median, and Z-Score normalization was applied. During feature selection, the Variance method was initially employed to select features. Features with a variance greater than 0.3 were retained. Subsequently, the Correlation method was used to analyze the correlation between these selected features, and features with a correlation coefficient lower than 0.7 were retained. To further narrow down the CT feature set and select the most impactful ones, the Gradient Boosting Decision Tree (GBDT) method was introduced to enhance model effectiveness and accuracy. Moreover, features extracted from PET images were combined with features extracted from diagnosis CT images to generate bimodal image fusion features. To evaluate the impact of extracted features on predicting PD-L1 expression, Tree and Forest methods were selected to build models for each modality, choosing the model with better performance. The comparison of different modeling methods is available in the appendix, resulting in the PET model, diagnosis CT model, and combined model. The aforementioned procedures were executed using scikit-learn 1.2.0.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eStatistical Analysis\u003c/h2\u003e\n \u003cp\u003ePatient data were presented using Python 3.9 and R software 4.2 in the form of mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation or percentage (if applicable). Based on the normality test results for continuous data, independent two-sample t-tests or Mann\u0026ndash;Whitney U tests were selected for different data. Chi-square test was used for analyzing categorical data, evaluating clinical data differences between PD-L1 positive (TPS\u0026thinsp;\u0026ge;\u0026thinsp;1%) and negative (TPS\u0026thinsp;\u0026lt;\u0026thinsp;1%) patients. Logistic regression was employed for the analysis of clinical and imaging data to obtain predictive probabilities and construct Receiver Operating Characteristic (ROC) curves. In the radiomics analysis, separate ROC curves were generated for different modalities, and the area under the curve (AUC) was calculated to determine the corresponding sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). These models were first assessed within the training cohort for predicting PD-L1 in non-small cell lung cancer patients and then evaluated in the testing cohort. A significance level of P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e\n \u003cp\u003eIn accordance with the Declaration of Helsinki, this study has been approved by the Institutional Review Board (IRB) with a number provided and the informed consent of the participants has been waived.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003ePatient Characteristics and Imaging Features\u003c/h2\u003e \u003cp\u003eIn our study, a total of 104 patients were included. Among them, 70 patients were PD-L1 positive, and 34 patients were PD-L1 negative. Table\u0026nbsp;1 presents their clinical data and imaging features. Pathologically confirmed, there were 89 cases diagnosed with adenocarcinoma, 14 cases with squamous cell carcinoma, and 1 case pathologically confirmed as non-small cell lung cancer without further classification. The average age of the positive group was 65.12\u0026thinsp;\u0026plusmn;\u0026thinsp;10.28, while the negative group was 61.51\u0026thinsp;\u0026plusmn;\u0026thinsp;8.28. The difference between the two groups was statistically significant (P\u0026thinsp;=\u0026thinsp;0.046). Additionally, there was a statistically significant difference between the two groups in terms of SUVmax (P\u0026thinsp;=\u0026thinsp;0.040). There were no statistically significant differences between the malignant and benign groups in terms of patient gender, SPN imaging features (lobulated, spiculated, Vacuole sign, Pleural stretch sign, Tracheal cutoff sign, Bronchovascular Bundle), Tumor size, and MTV.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003eTable.1 Characteristics of patients\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePositive(64)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNegitive(39)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003et\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eZ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eχ2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge, years\u003c/p\u003e \u003cp\u003e(mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e64.70\u0026thinsp;\u0026plusmn;\u0026thinsp;9.43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e61.92\u0026thinsp;\u0026plusmn;\u0026thinsp;8.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.484\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.141\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\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 \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2.346\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.126\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e41(64.06%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19(48.72%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23(35.94%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20(51.28%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmax\u003c/p\u003e \u003cp\u003e(median[25%,75%])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.20[5.28,10.38]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.55[2.80,9.55]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-1.397\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.162\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.35[3.98,11.65]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5.80[3.65,14.39]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.916\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTLU\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23.48[14.46,66.30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.12[5.81,94.14]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.639\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.523\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCOV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28.36[26.44,30.36]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26.58[22.49,28.50]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-2.529\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.011*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003eSD, standard deviation, MTV, metabolic tumor volume, TLU, total lesion uptake, COV, coefficient of variation.*,P\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eImaging Analysis\u003c/h2\u003e \u003cp\u003eThe results of logistic regression analysis predicting PD-L1 expression based on clinical and PET/CT imaging data are shown in Table\u0026nbsp;2. Univariate analysis indicates that age and gender in demographic data are associated with PD-L1 expression, with P-values both less than 0.1. Other imaging data are not correlated with PD-L1 expression (P\u0026thinsp;\u0026gt;\u0026thinsp;0.1). Further multivariate analysis shows that the factors included in the study are not independent predictors of PD-L1 expression.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabb\" border=\"1\"\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\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eTable.2 Evaluation of the efficacy of metabolic parameters in predicting PD-L1 expression\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAUC(25%, 75%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSe\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSp\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmax\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.582 (0.481, 0.679)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.875\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.359\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.2197\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.506 (0.406, 0.606)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.297\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.590\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0441*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTLG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.538 (0.437, 0.636)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.922\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.308\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0827\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCOV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.649 (0.549, 0.741)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.469\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eSe, sensitivity, Sp, specificity, MTV, metabolic tumor volume, TLU, total lesion uptake, COV, coefficient of variation. P represents the comparison between the ROC curve constructed by other metabolic parameters and COV, *, P\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eFeature Selection\u003c/h2\u003e \u003cp\u003e106 features were extracted from PET images, and after feature selection, 8 features were finally chosen, including 1 shape feature, 2 First Order Statistics features, 3 gray-level dependence matrix features, and 2 gray-level size zone matrix features. From diagnosis CT images, 107 features were extracted, and after feature selection, 6 features were chosen, including 1 shape feature, 2 First Order Statistics features, 2 gray-level dependence matrix features, and 1 Neighboring Gray Tone Difference Matrix feature.. The extracted features are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eModel Comparison\u003c/h2\u003e \u003cp\u003eBased on the logistic regression results, a clinical factors model was constructed using the predicted probabilities. The results of comparing the predictive abilities of the clinical factor model and radiomics model for PD-L1 expression are shown in Table\u0026nbsp;3 and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The AUC of the clinical factor model is 0.712, with a sensitivity of 0.571 and specificity of 0.765, indicating a relatively high rate of false negatives. The AUC for the diagnosis CT model is 0.723, while the PET model is 0.838. The PET model has higher sensitivity than the diagnosis CT model and comparable specificity, suggesting that in the single-modality radiological model, PET outperforms diagnosis CT. By combining the features of the two models, a joint PET-CT model was created with a sensitivity of 0.857 and specificity of 0.636, yielding an AUC of 0.857.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabc\" border=\"1\"\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003eTable.3 Univariate and multivariate logistic regression analysis results\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eUnivariate analysis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e \u003cp\u003eMultivariate analysis\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95%CI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e95%CI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.034\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.989\u0026ndash;1.081\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.144\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.876\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.835\u0026ndash;4.215\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.127\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSUVmax(\u0026lt;\u0026thinsp;3.7 vs.\u0026ge;3.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.293\u0026ndash;9.476\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.014*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.007\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.197\u0026ndash;5.136\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.994\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMTV(\u0026lt;\u0026thinsp;9.4cm3 vs. \u0026ge;9.4cm3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.653\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.285\u0026ndash;1.496\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.314\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTLU(\u0026lt;\u0026thinsp;6.36 vs. \u0026ge;6.36)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.636\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.470-14.623\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.009*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.387\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.579\u0026ndash;19.809\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.176\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCOV(\u0026lt;\u0026thinsp;28.9 vs. \u0026ge;28.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.419\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.364\u0026ndash;8.574\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.009*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.707\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.005\u0026ndash;7.287\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.049*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003eMTV, metabolic tumor volume, TLU, total lesion uptake, COV, coefficient of variation. *, P\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eImmune checkpoint inhibitors have been widely used in the treatment of advanced non-small cell lung cancer patients, and they can provide sustained relief, with PD-L1 positivity being associated with significantly higher objective response rates\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. To reduce the trauma caused by patients undergoing invasive procedures, alternative non-invasive methods to measure PD-L1 status are of significant importance for clinical decision support\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. This study focused on patients with non-small cell lung cancer and explored the value of clinical models constructed from clinical factors, PET, diagnostic CT, and PET-diagnostic CT models in predicting PD-L1 expression.\u003c/p\u003e \u003cp\u003eThe expression of PD-L1 is somewhat correlated with the demographic factors of patients, and there is currently no unified conclusion in the research. It is commonly believed to be associated with oxidative stress induced by smoking. WU et al.'s study suggests that PD-L1 expression is higher in male patients, which is related to the higher smoking rate among male patients\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. Moreover, with increasing age, the likelihood of cellular mutations is greater. Therefore, in elderly patients, the tumor burden is heavier, resulting in higher PD-L1 expression, and better efficacy of ICIs for this patient group\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. Previous studies have predicted PD-L1 expression in lung adenocarcinoma based on radiological features from CT images, indicating a higher PD-L1 positivity rate in patients with tumors larger than 2cm in diameter. In our study, we chose to calculate tumor size using (long diameter\u0026thinsp;+\u0026thinsp;short diameter)/2, which, compared to tumor transverse diameter, demonstrated better effectiveness in reflecting 3D tumor morphology\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. The solid-to-tumor ratio (CTR) is significantly correlated with PD-L1 expression (P\u0026thinsp;=\u0026thinsp;0.003). Various lesion morphologies, such as pleural indentation (P\u0026thinsp;=\u0026thinsp;0.007), spiculation (P\u0026thinsp;\u0026lt;\u0026thinsp;0.01), and ground-glass opacity (P\u0026thinsp;\u0026lt;\u0026thinsp;0.01), are associated with PD-L1 positivity \u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. This contrasts with our research findings, and we speculate that it may be due to the inclusion of a population confirmed as having a solitary lesion, comprising both squamous cell carcinoma and adenocarcinoma. Additionally, the difference may be related to our study's set threshold for PD-L1 positivity at TPS\u0026thinsp;=\u0026thinsp;1%. The model constructed based on CT image features has an area under the curve (AUC) of 0.783, sensitivity of 81.1%, and specificity of 64.1%. The AUC is similar to our research results, but our findings exhibit higher specificity and lower sensitivity. \u003csup\u003e18\u003c/sup\u003eF-FDG PET/CT serves as a non-invasive method, providing both anatomical and metabolic information of lesions, and is an important imaging tool for the diagnosis and staging of lung cancer. Takada et al. found that the SUVmax of non-small cell lung cancer (NSCLC) patients expressing PD-L1 protein (TPS\u0026thinsp;\u0026gt;\u0026thinsp;5%) was significantly higher than that of PD-L1 non-expressing NSCLC patients (P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). Multivariate analysis indicated that high SUVmax is an independent predictor of PD-L1 positivity, with a threshold of 4.2 providing the best predictive ability for PD-L1 expression\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. In research targeting PD-L1 expression in lung adenocarcinoma patients, tumor PD-L1 expression is positively correlated with the maximum standardized uptake value (SUVmax) and total lesion glycolysis (TLG) (both P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). SUVmax serves as an independent predictor of tumor PD-L1 expression, with an optimal threshold of 9.5 \u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. Some studies propose that when using TPS\u0026thinsp;=\u0026thinsp;1% as the threshold for defining PD-L1 positivity, there are intergroup differences in SUVmax, TLG, and MTV. However, SUVmax is not correlated with PD-L1 status. This is similar to our research conclusion, where SUVmax shows intergroup differences (P\u0026thinsp;=\u0026thinsp;0.040), but logistic regression analysis indicates no correlation between SUVmax and PD-L1 expression (P\u0026thinsp;=\u0026thinsp;0.12) \u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. Currently, there is a considerable amount of research on the relationship between radiomics and PD-L1 expression. Sun et al. found that combining radiomic features from CT images with clinical-pathological factors resulted in the highest accuracy in predicting PD-L1 expression levels in non-small cell lung cancer patients. The AUC in the test set was 0.848, with a sensitivity of 83.3%, providing valuable assistance in accurately detecting positive patients \u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. Yoon et al. similarly found that adding radiomic features to a model based on clinical factors improves the predictive ability of the model (c-statistic\u0026thinsp;=\u0026thinsp;0.646 vs. 0.550, P\u0026thinsp;=\u0026thinsp;0.0299) \u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. Similarly, research has explored the predictive capability of PET/CT radiomics for PD-L1 expression. It was found that combining radiomic features and clinical-pathological features in the prediction model can achieve good results, with an AUC of 0.762 for expression levels greater than 1% and an AUC of 0.814 for levels greater than 50% \u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. Jiang compared the predictive performance of PET, CT, and PET/CT models individually and found that the CT model had the best performance (AUC\u0026thinsp;=\u0026thinsp;0.86). Interestingly, combining PET features actually decreased the predictive ability of the model (AUC\u0026thinsp;=\u0026thinsp;0.85) \u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. This differs from our research results, and we attribute it to the data source. Firstly, in their study, low-dose CT was used, which may obscure some fine texture features due to its lower resolution. Additionally, their method of delineating ROI on CT images directly corresponding to PET images may lead to incorrect matching and affect the results. In our study, compared to the clinical factors model, the radiomics model shows better overall performance, but the clinical model is more accurate in predicting positive cases. Utilizing radiomics from PET images alone yields favorable results, indicating that metabolic imaging is more effective than anatomical imaging in predicting PD-L1. The combination of PET and diagnosis CT models only increases the model's AUC, with sensitivity at the cut-off point not as high as the standalone PET model. Among the features extracted from PET images, Sphericity represents the tumor's spherical ratio, indicating that PD-L1 expression is related to the tumor's surrounding morphology. However, this is primarily reflected in metabolic images rather than anatomical images. Consistent with our observations of tumor signs in CT images (lobulation, spiculation), different signs cannot predict PD-L1 expression. Zhang et al.'s study suggests that sphericity is closely related to EGFR mutations, and EGFR is an important factor in the PD-L1 regulation pathway \u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. The Coarseness feature in CT images represents the roughness of the image texture, reflecting the variation in grayscale levels. This is consistent with the results of Suda et al.'s study, where they found that lesions with ground-glass opacity had a lower incidence of PD-L1 expression positivity compared to solid lesions (4% vs. 25%, P\u0026lt;0.01) \u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e. Among all the features extracted from PET and CT images, including morphological features, first-order histogram features, and texture features, a total of 10 features are correlated with lesion heterogeneity. This indicates that the metabolic distribution within the tumor can reflect PD-L1 expression, and further exploration will be conducted in future work.\u003c/p\u003e \u003cp\u003eOur study has some limitations: Firstly, it is a single-center study with a limited number of cases, which may restrict the generalizability of the results. Secondly, the experiment is retrospective, and patients who undergo \u003csup\u003e18\u003c/sup\u003eF-FDG-PET/CT examinations are usually those with suspicious lesions detected through CT scans or hematological examinations, introducing inherent selection bias. To further validate these findings, larger-scale prospective multicenter studies are needed. Thirdly, the described segmentation method may introduce measurement errors. Although automatic segmentation is used for lesion depiction in PET images, manual delineation is employed for CT images based on PET-defined ROIs. This manual delineation may have lower repeatability. In future studies, we will explore automatic segmentation methods to address this issue and improve accuracy.\u003c/p\u003e \u003cp\u003eOur study found that the clinical factor model can achieve good results in predicting PD-L1 expression. However, it has a lower sensitivity, leading to a significant number of false negatives. Models constructed based on radiomics outperform the clinical factor model. The PET model can do a great job in predicting PD-L1 expression, while the combination of PET and diagnosis CT in a joint radiomics model yields the best results, ensuring the correct detection of PD-L1-positive cases.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003eNSCLC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003enon-small cell lung cancer\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003eICIs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003eImmune checkpoint inhibitors\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003ePD-L1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003eprogrammed cell death ligand\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003eTLU\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"50%\" valign=\"top\"\u003e\n \u003cp\u003etotal lesion uptake\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn accordance with the Declaration of Helsinki, this study has been approved by the Institutional Review Board (IRB) with a number provided and the informed consent of the participants has been waived.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study has received funding by the National Natural Science Foundation of China (No. 82001772), the Natural Science Foundation of Shaanxi Province, China (2021SF-062, 2020JZ38) and the New Medical and Technology of the First Affiliated Hospital of Xi\u0026rsquo;an \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Jiaotong University (XJYFY-2019J1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRuxi Chang, design of the work; have drafted the work.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCong Shen, analysis of data\u003c/p\u003e\n\u003cp\u003eLiang Luo, acquisition of data\u003c/p\u003e\n\u003cp\u003eXiang Liu, acquisition of data\u003c/p\u003e\n\u003cp\u003eYan Li, interpretation of data\u003c/p\u003e\n\u003cp\u003eXiaoyi Duan, conception of the work, revised the paper.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003e\u003cem\u003eSiegel RL, Miller KD, Wagle NS, Jemal A (2023) Cancer statistics, 2023. CA Cancer J Clin DOI:10.3322/caac.21763\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eEttinger DS, Wood DE, Aisner DL, et al \u003c/em\u003e\u003cem\u003e(2022)\u003c/em\u003e\u003cem\u003e Non-Small Cell Lung Cancer, Version 3.2022, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw DOI:10.6004/jnccn.2022.0025\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eIlie M, Long-Mira E, Bence C, et al \u003c/em\u003e\u003cem\u003e(2016) Comparative study of the PD-L1 status between surgically resected specimens and matched biopsies of NSCLC patients reveal major discordances: a potential issue for anti-PD-L1 therapeutic strategies. Ann Oncol DOI:10.1093/annonc/mdv489\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eHu B, Xiao J, Xiu Y, Fu Z, Shi H, Cheng D \u003c/em\u003e\u003cem\u003e(2020) Correlation of PD-L1 expression on tumor cell and tumor infiltrating immune cell with 18F-fluorodeoxyglucose uptake on PET/computed tomography in surgically resected pulmonary adenocarcinoma. Nucl Med Commun DOI:10.1097/MNM.0000000000001136\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eCook GJR, Goh V \u003c/em\u003e\u003cem\u003e(2020) A Role for FDG PET Radiomics in Personalized Medicine?. Semin Nucl Med DOI:10.1053/j.semnuclmed.2020.05.002\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eNakajo M, Jinguji M, Aoki M, Tani A, Sato M, Yoshiura T \u003c/em\u003e\u003cem\u003e(2020) The clinical value of texture analysis of dual-time-point \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003eF-FDG-PET/CT imaging to differentiate between \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003eF-FDG-avid benign and malignant pulmonary lesions. Eur Radiol DOI:10.1007/s00330-019-06463-7\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eRosner S, Reuss JE, Forde PM \u003c/em\u003e\u003cem\u003e(2019) PD-1 Blockade in Early-Stage Lung Cancer. Annu Rev Med DOI:10.1146/annurev-med-050217-025205\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eYu H, Boyle TA, Zhou C, Rimm DL, Hirsch FR \u003c/em\u003e\u003cem\u003e(2016) PD-L1 Expression in Lung Cancer. J Thorac Oncol DOI:10.1016/j.jtho.2016.04.014\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eRossi S, Toschi L, Castello A, Grizzi F, Mansi L, Lopci E \u003c/em\u003e\u003cem\u003e(2017) Clinical characteristics of patient selection and imaging predictors of outcome in solid tumors treated with checkpoint-inhibitors. Eur J Nucl Med Mol Imaging DOI:10.1007/s00259-017-3802-5\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eWu X, Huang Y, Zhao Q, et al \u003c/em\u003e\u003cem\u003e(2020) PD-L1 expression correlation with metabolic parameters of FDG PET/CT and clinicopathological characteristics in non-small cell lung cancer. EJNMMI Re DOI:10.1186/s13550-020-00639-9\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eNosrati A, Tsai KK, Goldinger SM, et al \u003c/em\u003e\u003cem\u003e(2017) Evaluation of clinicopathological factors in PD-1 response: derivation and validation of a prediction scale for response to PD-1 monotherapy. Br J Cancer DOI:10.1038/bjc.2017.70\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eKovalchik SA, Tammemagi M, Berg CD, et al (2013) Targeting of low-dose CT screening according to the risk of lung-cancer death. N Engl J Med DOI:10.1056/NEJMoa1301851\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eMacMahon H, Austin JH, Gamsu G, et al \u003c/em\u003e\u003cem\u003e(2005) Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society. Radiology DOI:10.1148/radiol.2372041887\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eWu T, Zhou F, Soodeen-Lalloo AK, et al \u003c/em\u003e\u003cem\u003e(2019) The Association Between Imaging Features of TSCT and the Expression of PD-L1 in Patients With Surgical Resection of Lung Adenocarcinoma. Clin Lung Cancer DOI:10.1016/j.cllc.2018.10.012\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eToyokawa G, Takada K, Okamoto T, et al \u003c/em\u003e\u003cem\u003e(2017) Computed Tomography Features of Lung Adenocarcinomas With Programmed Death Ligand 1 Expression. Clin Lung Cancer DOI:10.1016/j.cllc.2017.03.008\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eTakada K, Toyokawa G, Okamoto T, et al \u003c/em\u003e\u003cem\u003e(2017) Metabolic characteristics of programmed cell death-ligand 1-expressing lung cancer on \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003e F-fluorodeoxyglucose positron emission tomography/computed tomography. Cancer Med DOI:10.1002/cam4.1215\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eCui Y, Li X, Du B, Diao Y, Li Y \u003c/em\u003e\u003cem\u003e(2020) PD-L1 in Lung Adenocarcinoma: Insights into the Role of \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003eF-FDG PET/CT. Cancer Manag Res DOI:10.2147/CMAR.S256871\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003ePolverari G, Ceci F, Bertaglia V, et al \u003c/em\u003e\u003cem\u003e(2020) \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003eF-FDG Pet Parameters and Radiomics Features Analysis in Advanced Nsclc Treated with Immunotherapy as Predictors of Therapy Response and Survival. Cancers (Basel) DOI:10.3390/cancers12051163\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eSun Z, Hu S, Ge Y, et al \u003c/em\u003e\u003cem\u003e(2020) Radiomics study for predicting the expression of PD-L1 in non-small cell lung cancer based on CT images and clinicopathologic features. J Xray Sci Technol DOI:10.3233/XST-200642\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eYoon J, Suh YJ, Han K, et al \u003c/em\u003e\u003cem\u003e(2020) Utility of CT radiomics for prediction of PD-L1 expression in advanced lung adenocarcinomas. Thorac Cancer DOI:10.1111/1759-7714.13352\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eLi J, Ge S, Sang S, Hu C, Deng S \u003c/em\u003e\u003cem\u003e(2021) Evaluation of PD-L1 Expression Level in Patients With Non-Small Cell Lung Cancer by \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003eF-FDG PET/CT Radiomics and Clinicopathological Characteristics. Front Oncol DOI:10.3389/fonc.2021.789014\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eJiang M, Sun D, Guo Y, et al \u003c/em\u003e\u003cem\u003e(2020) Assessing PD-L1 Expression Level by Radiomic Features From PET/CT in Nonsmall Cell Lung Cancer Patients: An Initial Result. Acad Radio DOI:10.1016/j.acra.2019.04.016\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eZhang J, Zhao X, Zhao Y, et al \u003c/em\u003e\u003cem\u003e(2020) Value of pre-therapy \u003c/em\u003e\u003cem\u003e18\u003c/em\u003e\u003cem\u003eF-FDG PET/CT radiomics in predicting EGFR mutation status in patients with non-small cell lung cancer. Eur J Nucl Med Mol Imaging DOI:10.1007/s00259-019-04592-1\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eShi Y \u003c/em\u003e\u003cem\u003e(2018) Regulatory mechanisms of PD-L1 expression in cancer cells. Cancer Immunol Immunother DOI:10.1007/s00262-018-2226-9\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eSuda K, Shimoji M, Shimizu S, et al \u003c/em\u003e\u003cem\u003e(2019) Comparison of PD-L1 Expression Status between Pure-Solid Versus Part-Solid Lung Adenocarcinomas. Biomolecules DOI:10.3390/biom9090456\u003c/em\u003e\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Positron emission tomography/computed tomography, Radiomics, Lung cancer, PD-L1, Textural analysis","lastPublishedDoi":"10.21203/rs.3.rs-4207471/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4207471/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eThis study aims to build a clinical factor model by incorporating clinical factors and metabolic parameters, as well as lesion imaging features from PET/CT images. Additionally, radiomics models are established based on PET-CT images to assess its capability in predicting PD-L1 expression in patients with non-small cell lung cancer.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eAfter retrospective data collection, based on the clinical factor logistic regression results, a clinical factor model was constructed. The regions of interest (ROIs) for PET in radiomics were delineated using a semi-automatic method, while those for diagnosis CT were manually delineated. After extracting radiomic features, feature selection was performed using variance analysis, correlation analysis, and Gradient Boosting Decision Tree (GBDT). PET, diagnosis CT, and combined models were constructed. Predictive power was evaluated through ROC analysis comparing different models.\u003c/p\u003e\u003ch2\u003eResult\u003c/h2\u003e \u003cp\u003eIn all 104 cases(mean age, 63.90years+/-8.99, 62males) were evaluated. The SUVmax in the PD-L1 positive group was higher than that in the negative group (P\u0026thinsp;=\u0026thinsp;0.04), but both metabolic parameters and imaging features showed no correlation with PD-L1 expression. The radiomics models outperformed the clinical factor model (AUC\u0026thinsp;=\u0026thinsp;0.712), yet the clinical factor model exhibited higher specificity than all radiomics models (Specificity\u0026thinsp;=\u0026thinsp;0.765). The predictive performance of the PET model surpassed that of the diagnosis CT model (AUC: 0.838 vs 0.723). The combined model demonstrated enhanced predictive performance (AUC\u0026thinsp;=\u0026thinsp;0.874).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe radiomics models perform better in predicting PD-L1 expression than the clinical factor model. The radiomics model combining PET and diagnosis CT exhibits the best predictive performance.\u003c/p\u003e","manuscriptTitle":"Investigating the correlation between PD-L1 expression and radiomics predictions in non-small cell lung cancer using PET/CT imaging analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-04-08 17:30:10","doi":"10.21203/rs.3.rs-4207471/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9ea43822-a56b-45e0-a601-24abc5eb95c7","owner":[],"postedDate":"April 8th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-05-24T00:39:56+00:00","versionOfRecord":[],"versionCreatedAt":"2024-04-08 17:30:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4207471","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4207471","identity":"rs-4207471","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}