Tumor Radiomics Features on Pretreatment CT to Predict Response to First-Line Chemotherapy in Thymic Carcinoma: A Multicenter Study

Tumor Radiomics Features on Pretreatment CT to Predict Response to First-Line Chemotherapy in Thymic Carcinoma: A Multicenter Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Tumor Radiomics Features on Pretreatment CT to Predict Response to First-Line Chemotherapy in Thymic Carcinoma: A Multicenter Study Si Yu Wu, Ying Wang, Ping Fan, Tianqi Xu, Yong Huang, Pengxi Han, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3988200/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Objectives Platinum-based chemotherapy as first-line treatment have shown promising results against tumors in thymic carcinoma (TC). However, clinical benefit has been limited to a small proportion of patients. We developed and validated a combined radiomics model to predict progressive disease (PD) in patients suffering from TC following first-line chemotherapy. Methods Patients receiving platinum-based chemotherapy as first-line treatment from four centers in Shandong, China, were retrospectively included (n = 134); 93 and 41 were entered into the training and validation sets. Radiomics features were extracted from pretreatment enhanced CT. After feature selection, radiomics score (RS) was developed by Linear Discriminant Analysis (LDA) and TNM, clinicopathological and clinicopathological-radiomics models (combined radiomics model) were developed by using logistic regression algorithm. Models were assessed for performance, incremental predictive value of radiomics features versus clinicopathological features, and the relationship of RS and clinical factors to survival. Results The clinicopathological model was a modest predictor of PD, with area under curve (AUC) of 0.879 (95% CI: 0.709–0.890) and 0.799 (95% CI: 0.778–0.980) in the training and validation sets. The AUC of the combined radiomics model was 0.937 (95% CI: 0.891–0.984) and 0.899 (95% CI: 0.800-0.997), which is of good calibration and clinical application. The incremental predictive value of radiomics features for clinicopathological features was 27% ( p < 0.001) and 0.4% ( p = 0.108) in the training and validation sets. In addition, RS, CD5, and distant metastasis were associated with progression-free survival in both the training and validation sets. Conclusion Radiomics features extracted from pretreatment enhanced CT allow prediction of individualized objective responses to platinum-based chemotherapy as first-line treatment in TC, providing incremental predictive value for clinicopathological features, and are associated with progression-free survival after initiation of this combination regimen. Machine learning Thymic carcinoma Radiomics CT Multicenter study Linear Discriminant Analysis Prediction model Progression-free survival Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Thymic carcinoma (TC) is a rare malignant tumor occurring in the pre-vascular mediastinum, with an incidence between 0.07 and 0.38/100,000/year [ 1 ]. Most of these tumors are squamous cell carcinomas that infiltrate the thymic epithelial peritoneum invading mediastinal structures and spreading to the pleura. Thus, thymic carcinoma is usually diagnosed as locally advanced or metastatic and require aggressive multidisciplinary treatment [ 2 ]. Platinum-based chemotherapy remains the standard of first-line treatment for patients with advanced TC who are not candidates for locoregional therapy [ 3 ]. A retrospective study of the efficacy of first-line treatment (platinum-based chemotherapy) in patients with advanced thymic carcinoma showed remission rates of 12%-67% and no significant differences among chemotherapy combinations. However, in clinical practice, there are significant differences in prognosis even in patients with the same staging [ 4 ]. Therefore, in order to improve the prognosis of TC treatment, a new biomarker is needed to assess the treatment response after first-line chemotherapy in TC. Enhanced CT plays an important role in assessing the response of tumors to treatment. However, conventional diagnostic imaging relies heavily on the personal experience of physicians, which is subjective. Moreover, it is difficult for the human eye vision to fully capture all the information in images. Comparatively speaking, radiomics, which combines artificial intelligence with diagnostic imaging, can fully mine image information to objectively and efficiently achieve differential diagnosis of diseases, prediction of treatment response, and judgement of prognosis [ 5 , 6 ]. Despite the great potential of radiomics, the use of different CT scanners and imaging parameters in different centers places greater demands on the reproducibility and generalizability of the models. In this study, we collected TC data from different centers and sought to develop and validate a non-invasive radiomics biomarker for predicting response to first-line chemotherapy in TC by integrating clinical, pathological and imaging factors. Materials and Methods Patients To predict response after first-line chemotherapy for TC, TC patients from four hospitals in Shandong province from April 2014 to September 2023 were retrospectively included in this study (Supplementary Material). All patients received platinum-based first-line chemotherapy [ 7 ]. Inclusion criteria: 1) enhanced CT scan at the hospital within 1 month prior to treatment; 2) patients with a pathological diagnosis of thymic carcinoma; 3) well-developed post-chemotherapy imaging, able to judge efficacy according to RECIST 1.1 criteria; 4) complete records of patient demographic and clinical characteristics. Exclusion criteria: 1) poor image quality due to human factors or other reasons; 2) previous history of treatment for thymic carcinoma (including chemotherapy, radiotherapy and volume-reducing surgery); 3) history of other malignant tumors. In the end, we included a total of 134 patients. The patients from the four centers were randomly divided into training and validation sets in a 7:3 ratio. Clinical data and Assessments The following preoperative clinical and laboratory data were collected: basic demographics, immunohistochemical information, TNM tumor staging (8th edition), lactate dehydrogenase (LDH), carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), neutrophils (NE), lymphocytes (Lym). LDH, CEA, NSE, and NE thresholds were 250 U/ ml, 10, 16.3 ng/ml, and 6.3 × 10 9 /L, respectively, and Lym range was 1.1–3.2 × 10 9 /L. Radiographic Data and Image Segmentation Pretreatment enhanced CT (non-enhanced and arterial enhanced phases) images were collected. See supplementary material for CT acquisition and parameter details. Intratumoral and mediastinal ROIs were outlined with the Deepwise Multimodal Research Platform ( https://keyan.deepwise.com,V2.0 ) (Beijing Deepwise & League of PHD Technology Co., Ltd., Beijing, China). Mediastinal ROIs were outlined avoiding bones, heart, great vessels, esophagus and trachea. If there are metastatic tumors from other sites, they are all outlined. Images from 15 patients from the training set were then randomly selected and the same segmentation procedure was repeated 2 weeks later by the same two radiologists to assess inter-observer and intra-observer agreement. Feature Extraction Radiomics features were extracted in non-enhanced and enhanced intratumoral and mediastinal ROIs, respectively. The voxels of all images were normalized, resampled at a size of 1 × 1 × 1 cubic millimeter and features were extracted from within each dimension in a 2D manner. Detailed parameters are given in the supplementary material. Ultimately 1781, 1781, 1781 and 1781 features were extracted from the tumor and mediastinal regions in each case, respectively. Classify features into 7 categories: 1) First Order Features; 2) Shape Features; 3) GLCM Features; 4) GLSZM Features; 5) GLRLM Features; 6) GLDM Features; 7) NGTDM Features. Feature Selection Firstly, radiomics features with high inter-observer and intra-observer stability (ICC ≥ 0.75) were selected. Then, feature correlation analysis was used to eliminate features with high similarity (> 0.7). Finally, modelling features were identified by F test. Clinicopathological features were selected for subsequent model construction by univariate and multivariate logistic regression. Radiomics Imaging Biomarker Development The selected features were analyzed using Linear Discriminant Analysis (LDA) algorithm to construct four models: non-enhanced intratumoral model, non-enhanced mediastinal model, enhanced intratumoral model and enhanced mediastinal model. The discrimination was quantified by the area under the curve (AUC). The better two models are selected based on the AUC results, their features are merged, and then the feature selection step is repeated to develop the radiomics score (RS) using LDA algorithm. Finally, the combined RS and clinicopathological features obtained from screening were used together to construct the combined radiomics model. Model Validation In this study, the performance of each model was evaluated using a confusion matrix, and the prediction accuracy of the combined model was evaluated using calibration curves and the Hosmer-Lemeshow test [ 8 ]. Practical application of combined radiomics model assessed using decision curve analysis (DCA). Incremental differences in evaluating combinatorial models using integrated discriminatory improvement (IDI). Patients in each set were stratified according to risk predicted by RS and combined radiomics model and Kaplan-Meier analysis. Univariate and multivariate Cox regression analysis of RS and other clinicopathological variables were performed to select independent predictors of survival. In addition, we tried several different machine learning classifiers and found that LDA performed the best (Table S1 ). Follow-up Patients were followed up after treatment at each center as recommended by the guidelines. CT scans were performed every 6 months for the first 3 years after receiving first-line chemotherapy [ 9 ]. Progression-free survival (PFS) was defined as the time between the start of chemotherapy and the onset of tumor progression or (for any reason) death. Statistical Analysis Statistical analysis were performed using SPSS (version 26), R software (version 3.6.2) and Python (version 3.8). For skewed distribution data, they were described as median [interquartile range, IQR]. Comparison of different ROC curves using the Delong test. Radiomics features selection and model construction were carried out using the Deepwise Multimodal Research Platform. In this study, we used the "Matplotlib" library in python to draw the confusion matrix. Moreover, we used univariate and multivariate Cox regression analysis to screen the independent predictors of PFS. A two-sided p < 0.05 was considered statistically significant. Results Patient Characteristics This study included 134 TC patients from four independent centers. The overall design of the study is shown in Fig. 1 . The clinicopathological characteristics of the training set (n = 93) and validation set (n = 41) are listed in Table S2. There were 87 (64.9%) males and 47 (35.1%) females, with a median age of 57 (49.25-64) years, and the majority of patients (n = 122, 91%) were in stage III or stage IV. The progression rates (PD: Progressive disease) were 56.1% and 58.1% for the training and validation sets. Feature Selection and Establishment of Radiomics Models Firstly, features with low concordance (ICC < 0.75) were excluded from this study, and 1611, 1360, 1546 and 1418 radiomics features were retained in the non-enhanced intratumor, non-enhanced mediastinum, enhanced intratumor and enhanced mediastinum, respectively. Then, after feature correlation analysis and F-test, the final 6, 9, 10 and 6 radiomics features were generated in this study, respectively. The LDA algorithm was used to build the radiomics model, with AUC values ranging from 0.725–0.839 in the training set and from 0.672–0.730 in the validation set. After comparing AUC, this study derived RS based on unenhanced intratumoural and enhanced intratumoural features using LDA algorithm. Detailed formulae for RS are given in the supplementary results. In terms of prediction of progression after chemotherapy, the AUC of RS was 0.911 and 0.785 in the training and validation sets. Combined Radiomics Model Development Univariate and multivariate analysis showed that CD5 and Distant metastasis were independently associated with PD (Fig. S1 ). The combined radiomics model was built by combining the above two clinical features with RS. After adding the radiomics features, the IDI of the combined radiomics model in the training set was significantly improved compared to the clinical model, however, no significant improvement was found in the validation set (Table S3). Internal Validation of the Combined Radiomics Model As can be seen from the ROC curves (Fig. 2 A and 2 B), the combined radiomics model has the highest AUC values in the training and validation sets, which are 0.937 (95% CI 0.891–0.984) and 0.899 (95% CI 0.800-0.997), respectively. Both the training and validation sets showed good agreement on the calibration curves (Fig. 3 A and 3 B), and the Hosmer-Lemeshow test showed that there was no significant difference between the predicted probabilities and the reproduced probabilities of the actual observations ( p = 0.152, 0.899), which supports the accuracy of the model. The confusion matrix for each model is shown in Fig. S2. Then, we compared each model by Delong test, and the differences between the combined model and the other models were all statistically significant ( p < 0.05) in the training set, and no significant differences were found between the models in the validation set (Table S4). In terms of practical application, as shown in Fig. 4 , it helps clinicians to identify patients with progression pre-chemotherapy in order to develop better treatment and follow-up plans, such as early adjustment of chemotherapy regimen, combination immunotherapy, or shortening of follow-up intervals. It helps to reduce the risk of PD and improve patient prognosis. Additionally, the DCA plot (Fig. 5 ) indicates that the combined radiomics model can predict the treatment response of patients within a probability threshold range of 0.25 to 0.70. Relationship of RS and Combined Radiomics Model to PFS after First-line Chemotherapy Until January 14, 2024, the median follow-up was 32.5 months (IQR: 18.2–57.1) for the training set and 35.1 months (IQR: 21.7–66.2) for the validation set; the median treatment duration was 33.9 months (IQR: 18.1–56.9) for the training set and 34.6 months for the validation set (IQR. 20.6–65.9). Progression occurred in 54 (58.1%) of 93 patients in the training set and in 23 (56.1%) of 41 patients in the validation set. The median PFS was 11.5 months and 16.1 months in the training and validation sets, respectively. Univariate and multivariate COX regression analysis showed that high RS, CD5 (-) and the presence of distant metastasis were independently associated with poorer PFS (Table 1 ). In addition, Kaplan-Meier analysis demonstrated the ability of the RS and combined radiomics model to stratify patients in each cohort pre-chemotherapy in both the training and validation sets, with log-rank tests suggesting good discriminatory properties of the models (Fig. 6 ). Furthermore, as can be seen from the heatmap in Fig. 7 , patients with high RS, CD5 (-) and M (+) tend to develop PD after first-line chemotherapy. Table 1 Multivariate Cox regression analysis for progression-free survival in patients with thymic carcinoma in the training and validation sets Variables Progression-free survival RR (95% CI) p Training set RS 1.929–16.037 0.001 CD5 (positive vs. negative) 0.226–0.780 0.006 Distant metastasis (positive vs. negative) 3.629–16.851 <0.001 Validation set RS 1.295–23.600 0.021 CD5 (positive vs. negative) 0.170–0.879 0.023 Distant metastasis (positive vs. negative) 2.063–16.923 0.001 Discussion In this multicenter study, we established and evaluated the effectiveness of combined radiomics model to predict progression after first-line chemotherapy for TC. The highest model prediction performance was achieved when radiomics features were combined with clinicopathological variables, with AUCs of 0.937 (95% CI 0.891–0.984) in the training set and 0.899 (95% CI 0.800-0.997) in the validation set. Additionally, the calibration curves and Hosmer-Lemeshow test indicate that the model has good predictive accuracy. DCA demonstrated that the combined radiomics model in this study resulted in better net clinical gains. These results suggest that radiomics features and clinicopathological factors can provide support for predicting disease progression after first-line chemotherapy. More and more studies are focusing on the tumor microenvironment to expect more information on disease progression and treatment response [ 10 , 11 ]. However, in this study, the performance of the mediastinum-based model was lower compared to the performance of the intratumor-based model. This is similar to past studies on other tumors [ 12 , 13 ]. The following reasons may explain these results. On the one hand, past studies have shown that it is possible to distinguish thymoma from thymic squamous cell carcinoma by the collagen fiber pattern on T2-weighted MR, with linear or reticular fibrous intervals often present within thymomas; however, in thymic carcinoma, the collagen fibers are poorly delineated and are distributed in eccentric patches of greater size [ 14 , 15 ]. Collagen fibers are associated with increased stiffness of the extracellular matrix (ECM). Past studies have shown that ECM contributes to immune rejection [ 16 ]. Moreover, in breast cancer, increased ECM hardness promotes epithelial mesenchymal transition (EMT), cell invasion and metastasis [ 17 ]. Thus, the collagen fiber pattern may reflect increased invasiveness of TC, however, in the context of fat in the mediastinum, it is difficult to differentiate between the tumor collagen fiber region and other regions of the tumor. Therefore, in our manual segmentation, collagen fiber regions that actually have information about invasiveness may be included in the intratumor ROI. On the other hand, using mediastinum as the ROI may incorporate more redundant features, which will inevitably lead to an impact on the model performance despite the fact that this study screened the features by ICC analysis, feature correlation analysis, and F-test. Although the mediastinum-based model has no additional value in predicting response to TC treatment, it is a novel attempt to explore the peritumoral region and could provide a reference for subsequent studies. It has been shown that radiomics performs well in risk stratification and identification of thymic epithelial tumors (TETs) (AUC range, 0.76–0.978) [ 18 – 21 ]. Moreover, models that combine radiomics features and clinicopathological factors may reflect different characteristics of tumors associated with treatment response from different perspectives [ 22 ]. Although, there is a known association between radiomics features and treatment response, no studies have been reported on predicting treatment response in TC [ 23 , 24 ]. To our knowledge, this study is the first to predict treatment response after first-line chemotherapy for TC by establishing a combined radiomics model. The variables of the combined radiomics model in this study included 21 radiomics features and 2 clinicopathological features. Notably, 81% (17/21) of the radiomics features were 3D and wavelet features, which is in line with previous findings, this suggests that tumor heterogeneity can be reflected as differences in the spatial distribution of voxel intensities in images [ 25 , 26 ]. After unifactorial and multifactorial analysis, CD5 and distant metastasis were finally selected as clinicopathological variables for the establishment of the combined radiomics model. They have been confirmed in previous studies and are considered to be associated with tumor prognosis. In a study evaluating the prognosis of patients with advanced thymic carcinoma receiving first-line chemotherapy, researchers found that CD5 positivity may be associated with better PFS [ 27 ]. In the present study, we found that CD5 positivity was associated with stable disease and better PFS after first-line chemotherapy, similar to past studies. Studies have shown that immune cells such as neutrophils and lymphocytes can regulate tumor cell activity by secreting large amounts of cytokines [ 28 ]. In addition, high levels of neutrophil-to-lymphocyte ratio (NLR) are thought to be strongly associated with poor tumor outcomes [ 29 , 30 ]. Therefore, neutrophils, lymphocytes and NLR in peripheral blood were included in this study and analyzed for their association with progression after first-line chemotherapy for TC as well as with PFS. The results showed that they were not independent predictors of progression and PFS after first-line chemotherapy for TC. This may be related to the decrease in neutrophil levels due to glucocorticoid use in some TC patients [ 31 ]. There are some limitations in this study. Firstly, in the validation set, Delong test showed that there was no significant difference between the models as well as no significant improvement was found in the combined identification improvement of the combined model over the clinical model. This may be related to the smaller number of cases in the validation set. Future studies could include a larger sample for external validation of the models. Secondly, in this study, we selected non-enhanced and arterial phase images for analysis, which theoretically may have reduced predictive performance compared to the full-phase CT image model. In the future, we also plan to perform full-phase CT-based feature extraction and analysis. Thirdly, considering the accuracy of image segmentation, we use manual segmentation of ROIs, but this will take more time. With the development of AI technology, automatic segmentation techniques may further improve the efficiency of image segmentation. Conclusion This study demonstrates that the combined radiomics model can be used to predict objective response to first-line chemotherapy in TC patients with an improvement over clinicopathological models. It was also correlated with PFS after the initiation of first-line chemotherapy. Abbreviations AUC Area under the curve CEA carcinoembryonic antigen DCA Decision curve analysis ECM Extracellular matrix EMT Epithelial mesenchymal transition IDI Integrated discriminatory improvement IQR Interquartile range LDA Linear discriminant analysis LDH Lactate dehydrogenase Lym Lymphocytes NE Neutrophils NLR Neutrophil-to-lymphocyte ratio NSE Neuron-specific enolase PD Progressive disease PFS Progression-free survival RS radiomics score SD stable disease TC Thymic carcinoma TETs Thymic epithelial tumors Declarations Acknowledgment The authors would like to thank all the participants. Authors ’ contributions Siyu Wu: Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing. Ying Wang: Data curation. Ping Fan: Data curation. Tianqi Xu: Data curation. Yong Huang: Data curation. Pengxi Han: Data curation. Yan Deng: Data curation. Ximing Wang : Conceptualization, Data curation, Project administration, Supervision, Writing – review & editing. All authors reviewed the manuscript. Funding The present study was supported by the National Natural Science Foundation of China (Grant Nos. 8187354, 81571672), and Academic promotion program of Shandong First Medical University (Grant No. 2019QL023). Availability of data and materials The data that support the fundings of this study are available from the corresponding author upon reasonable request. Ethics approval and consent to participate This retrospective study was approved by the Ethics Committee of the Shandong Provincial Hospital and written informed consent was discarded because of the retrospective nature of this study. We confirmed that the study was carried out in accordance with relevant guidelines and regulations of Helsinki Declaration. Consent for publication Not applicable. Competing interests All the authors declare that they have no conflict of interest. References Roden AC, Ahmad U, Cardillo G, Girard N, Jain D, Marom EM, Marx A, Moreira AL, Nicholson AG, Rajan A, et al. Thymic Carcinomas-A Concise Multidisciplinary Update on Recent Developments From the Thymic Carcinoma Working Group of the International Thymic Malignancy Interest Group. J Thorac Oncol. 2022;17(5):637–50. https://www.ncbi.nlm.nih.gov/pubmed/35227908 . Chun SG, Rimner A, Amini A, Chang JY, Donington J, Edelman MJ, Geng Y, Gubens MA, Higgins KA, Iyengar P, et al. American Radium Society Appropriate Use Criteria for Radiation Therapy in the Multidisciplinary Management of Thymic Carcinoma. JAMA Oncol. 2023;9(7):971–80. https://www.ncbi.nlm.nih.gov/pubmed/37186595 . Remon J, Villacampa G, Facchinetti F, Di Maio M, Marcuse F, Tiseo M, Hochstenbag M, Hendriks LEL, Besse B. Immune checkpoint blockers in patients with unresectable or metastatic thymic epithelial tumours: A meta-analysis. Eur J Cancer. 2023;180:117–24. https://www.ncbi.nlm.nih.gov/pubmed/36592506 . Ko R, Shukuya T, Okuma Y, Tateishi K, Imai H, Iwasawa S, Miyauchi E, Fujiwara A, Sugiyama T, Azuma K, et al. Prognostic Factors and Efficacy of First-Line Chemotherapy in Patients with Advanced Thymic Carcinoma: A Retrospective Analysis of 286 Patients from NEJ023 Study. Oncologist. 2018;23(10):1210–7. https://www.ncbi.nlm.nih.gov/pubmed/29567820 . Chen Q, Zhang L, Liu S, You J, Chen L, Jin Z, Zhang S, Zhang B. Radiomics in precision medicine for gastric cancer: opportunities and challenges. Eur Radiol. 2022;32(9):5852–68. https://www.ncbi.nlm.nih.gov/pubmed/35316364 . Guiot J, Vaidyanathan A, Deprez L, Zerka F, Danthine D, Frix AN, Lambin P, Bottari F, Tsoutzidis N, Miraglio B, et al. A review in radiomics: Making personalized medicine a reality via routine imaging. Med Res Rev. 2022;42(1):426–40. https://www.ncbi.nlm.nih.gov/pubmed/34309893 . Berghmans T, Durieux V, Holbrechts S, Jungels C, Lafitte JJ, Meert AP, Moretti L, Ocak S, Roelandts M, Girard N. Systemic treatments for thymoma and thymic carcinoma: A systematic review. Lung Cancer. 2018;126:25–31. https://www.ncbi.nlm.nih.gov/pubmed/30527189 . Kramer AA, Zimmerman JE. Assessing the calibration of mortality benchmarks in critical care: The Hosmer-Lemeshow test revisited. Crit Care Med. 2007;35(9):2052–6. https://www.ncbi.nlm.nih.gov/pubmed/17568333 . Scorsetti M, Leo F, Trama A, D'Angelillo R, Serpico D, Macerelli M, Zucali P, Gatta G, Garassino MC. Thymoma and thymic carcinomas. Crit Rev Oncol Hematol. 2016;99:332–50. https://www.ncbi.nlm.nih.gov/pubmed/26818050 . Huang W, Jiang Y, Xiong W, Sun Z, Chen C, Yuan Q, Zhou K, Han Z, Feng H, Chen H, et al. Noninvasive imaging of the tumor immune microenvironment correlates with response to immunotherapy in gastric cancer. Nat Commun. 2022;13(1):5095. https://www.ncbi.nlm.nih.gov/pubmed/36042205 . Wang T, She Y, Yang Y, Liu X, Chen S, Zhong Y, Deng J, Zhao M, Sun X, Xie D, et al. Radiomics for Survival Risk Stratification of Clinical and Pathologic Stage IA Pure-Solid Non–Small Cell Lung Cancer. Radiology. 2022;302(2):425–34. https://www.ncbi.nlm.nih.gov/pubmed/34726531 . Song W, Yu X, Guo D, Liu H, Tang Z, Liu X, Zhou J, Zhang H, Liu Y, Liu X. MRI-Based Radiomics: Associations With the Recurrence-Free Survival of Patients With Hepatocellular Carcinoma Treated With Conventional Transcatheter Arterial Chemoembolization. J Magn Reson Imaging. 2020;52(2):461–73. https://www.ncbi.nlm.nih.gov/pubmed/31675174 . Lin Z, Wang T, Li Q, Bi Q, Wang Y, Luo Y, Feng F, Xiao M, Gu Y, Qiang J, et al. Development and validation of MRI-based radiomics model to predict recurrence risk in patients with endometrial cancer: a multicenter study. Eur Radiol. 2023;33(8):5814–24. https://www.ncbi.nlm.nih.gov/pubmed/37171486 . Hu YC, Yan WQ, Yan LF, Xiao G, Han Y, Liu CX, Wang SZ, Li GF, Wang SM, Yang G, et al. Differentiating thymoma, thymic carcinoma and lymphoma based on collagen fibre patterns with T2- and diffusion-weighted magnetic resonance imaging. Eur Radiol. 2022;32(1):194–204. https://www.ncbi.nlm.nih.gov/pubmed/34215941 . Takahashi K, Al-Janabi NJ. Computed tomography and magnetic resonance imaging of mediastinal tumors. J Magn Reson Imaging. 2010;32(6):1325–39. https://www.ncbi.nlm.nih.gov/pubmed/21105138 . Sun X, Wu B, Chiang H-C, Deng H, Zhang X, Xiong W, Liu J, Rozeboom AM, Harris BT, Blommaert E, et al. Tumour DDR1 promotes collagen fibre alignment to instigate immune exclusion. Nature. 2021;599(7886):673–8. https://www.ncbi.nlm.nih.gov/pubmed/34732895 . Fattet L, Jung HY, Matsumoto MW, Aubol BE, Kumar A, Adams JA, Chen AC, Sah RL, Engler AJ, Pasquale EB, et al. Matrix Rigidity Controls Epithelial-Mesenchymal Plasticity and Tumor Metastasis via a Mechanoresponsive EPHA2/LYN Complex. Dev Cell. 2020;54(3):302–e316307. .https://www.ncbi.nlm.nih.gov/pubmed/32574556 . Ohira R, Yanagawa M, Suzuki Y, Hata A, Miyata T, Kikuchi N, Yoshida Y, Yamagata K, Doi S, Ninomiya K, et al. CT-based radiomics analysis for differentiation between thymoma and thymic carcinoma. J Thorac Dis. 2022;14(5):1342–52. https://www.ncbi.nlm.nih.gov/pubmed/35693628 . Mahmoudi S, Gruenewald LD, Eichler K, Althoff FC, Martin SS, Bernatz S, Booz C, Yel I, Kinzler MN, Ziegengeist NS, et al. Multiparametric Evaluation of Radiomics Features and Dual-Energy CT Iodine Maps for Discrimination and Outcome Prediction of Thymic Masses. Acad Radiol. 2023;30(12):3010–21. https://www.ncbi.nlm.nih.gov/pubmed/37105804 . Yu C, Li T, Yang X, Zhang R, Xin L, Zhao Z, Cui J. Contrast-enhanced CT-based radiomics model for differentiating risk subgroups of thymic epithelial tumors. BMC Med Imaging. 2022;22(1):37. https://www.ncbi.nlm.nih.gov/pubmed/35249531 . Zhao W, Ozawa Y, Hara M, Okuda K, Hiwatashi A. Computed tomography radiomic feature analysis of thymic epithelial tumors: Differentiation of thymic epithelial tumors from thymic cysts and prediction of histological subtypes. Jpn J Radiol 2023. https://www.ncbi.nlm.nih.gov/pubmed/38010596 . Xu B, Dong SY, Bai XL, Song TQ, Zhang BH, Zhou LD, Chen YJ, Zeng ZM, Wang K, Zhao HT, et al. Tumor Radiomic Features on Pretreatment MRI to Predict Response to Lenvatinib plus an Anti-PD-1 Antibody in Advanced Hepatocellular Carcinoma: A Multicenter Study. Liver Cancer. 2023;12(3):262–76. https://www.ncbi.nlm.nih.gov/pubmed/37601982 . Zhao J, He Y, Yang X, Tian P, Zeng L, Huang K, Zhao J, Zhou J, Zhu Y, Wang Q et al. Assessing treatment outcomes of chemoimmunotherapy in extensive-stage small cell lung cancer: an integrated clinical and radiomics approach. J Immunother Cancer 2023, 11(9). https://www.ncbi.nlm.nih.gov/pubmed/37730276 . Wang R, Dai W, Gong J, Huang M, Hu T, Li H, Lin K, Tan C, Hu H, Tong T, et al. Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients. J Hematol Oncol. 2022;15(1):11. https://www.ncbi.nlm.nih.gov/pubmed/35073937 . Song Y, Zhou G, Zhou Y, Xu Y, Zhang J, Zhang K, He P, Chen M, Liu Y, Sun J, et al. Artificial intelligence CT radiomics to predict early recurrence of intrahepatic cholangiocarcinoma: a multicenter study. Hepatol Int. 2023;17(4):1016–27. https://www.ncbi.nlm.nih.gov/pubmed/36821045 . Xu L, Yang P, Liang W, Liu W, Wang W, Luo C, Wang J, Peng Z, Xing L, Huang M, et al. A radiomics approach based on support vector machine using MR images for preoperative lymph node status evaluation in intrahepatic cholangiocarcinoma. Theranostics. 2019;9(18):5374–85. https://www.ncbi.nlm.nih.gov/pubmed/31410221 . Ho IW, Pan YL, Lai JI, Liu CY. Characteristics and outcome of systemic treatment for metastatic or unresectable thymic carcinoma: A single institution experience. Thorac Cancer 2023, n/a( n/a).https://www.ncbi.nlm.nih.gov/pubmed/38149471 . Coffelt SB, Wellenstein MD, de Visser KE. Neutrophils in cancer: neutral no more. Nat Rev Cancer. 2016;16(7):431–46. https://www.ncbi.nlm.nih.gov/pubmed/27282249 . Ouyang H, Xiao B, Huang Y, Wang Z. Baseline and early changes in the neutrophil-lymphocyte ratio (NLR) predict survival outcomes in advanced colorectal cancer patients treated with immunotherapy. Int Immunopharmacol. 2023;123:110703. https://www.ncbi.nlm.nih.gov/pubmed/37536184 . Hu Z, Zhou J, Li Y, Luan Y, Li H, Jia B, Xie Z, Cheng B, Wu T. Peripheral immune signature resembles tumor microenvironment and predicts clinical outcomes in head and neck squamous cell carcinoma. Front Immunol 2022, 13915207 https://www.ncbi.nlm.nih.gov/pubmed/36148222 . Yang AP, Liu JP, Tao WQ, Li HM. The diagnostic and predictive role of NLR, d-NLR and PLR in COVID-19 patients. Int Immunopharmacol. 2020;84:106504. https://www.ncbi.nlm.nih.gov/pubmed/32304994 . Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 17 Mar, 2024 Editor invited by journal 11 Mar, 2024 Editor assigned by journal 11 Mar, 2024 Submission checks completed at journal 11 Mar, 2024 First submitted to journal 25 Feb, 2024 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-3988200","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":278464583,"identity":"103c8591-0449-4197-9657-151bb50e0024","order_by":0,"name":"Si Yu Wu","email":"","orcid":"","institution":"Shandong University","correspondingAuthor":false,"prefix":"","firstName":"Si","middleName":"Yu","lastName":"Wu","suffix":""},{"id":278464584,"identity":"e8277571-ee7e-4250-8498-9edd156288a1","order_by":1,"name":"Ying Wang","email":"","orcid":"","institution":"Shandong Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Wang","suffix":""},{"id":278464586,"identity":"e41a96a3-ec06-48fe-ab6e-da48b8422450","order_by":2,"name":"Ping Fan","email":"","orcid":"","institution":"Weifang Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ping","middleName":"","lastName":"Fan","suffix":""},{"id":278464588,"identity":"e24a8022-b55e-4371-93ae-8cfb3783c9b5","order_by":3,"name":"Tianqi Xu","email":"","orcid":"","institution":"Shandong Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Tianqi","middleName":"","lastName":"Xu","suffix":""},{"id":278464590,"identity":"99e75125-edeb-4aa1-ae60-ded5bfc4edf4","order_by":4,"name":"Yong Huang","email":"","orcid":"","institution":"Shandong Tumor Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yong","middleName":"","lastName":"Huang","suffix":""},{"id":278464591,"identity":"6922e1a6-188b-4d17-8a82-1bb9d1c8cf6a","order_by":5,"name":"Pengxi Han","email":"","orcid":"","institution":"Shandong Provincial QianFoShan Hospital","correspondingAuthor":false,"prefix":"","firstName":"Pengxi","middleName":"","lastName":"Han","suffix":""},{"id":278464592,"identity":"281f4f4f-68ba-4875-98f2-e2543b509f09","order_by":6,"name":"Yan Deng","email":"","orcid":"","institution":"Qilu Hospital of Shandong University","correspondingAuthor":false,"prefix":"","firstName":"Yan","middleName":"","lastName":"Deng","suffix":""},{"id":278464593,"identity":"77036678-7bc2-401f-84ea-1587d2e59b03","order_by":7,"name":"Ximing Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsElEQVRIiWNgGAWjYHAC5gcfKmx4+PkbiNfCZjjjTJqM5IwDJFgjzdl22MagIYFI5QY3shOMGdvO8xgwHGD88DGHGC1nzm54XHDuNo85cwOz5MxtRGgxO967wXhG2W0ey4YDbMy8RGk5zLtBmoftHI/BgQRitQBtkeZpO0CCFvszZ7cBAzmZR3LGwWbi/CI5I3czMCrt7Pn5mw9++EiMFiTA2ECa+lEwCkbBKBgFuAEArhE6E950tRAAAAAASUVORK5CYII=","orcid":"","institution":"Shandong University","correspondingAuthor":true,"prefix":"","firstName":"Ximing","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2024-02-25 14:31:02","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3988200/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3988200/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":52623021,"identity":"48a34753-9419-478d-b453-5a4105093ae0","added_by":"auto","created_at":"2024-03-13 17:17:38","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":250699,"visible":true,"origin":"","legend":"\u003cp\u003eFramework and flowchart for this study.\u003cstrong\u003e A \u003c/strong\u003eComparing the performance of models built on different ROIs\u003cstrong\u003e B \u003c/strong\u003eBuilding the combined radiomics model using logistic regression algorithm\u003cstrong\u003e C \u003c/strong\u003eEvaluating the performance of combined radiomics model for predicting progression after TC chemotherapy and for stratifying populations at risk\u003c/p\u003e","description":"","filename":"Fig.1.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/9380c825ba19b1f9116eb2ed.png"},{"id":52623019,"identity":"886ffb79-2b50-45ae-8535-ddeb976e35c5","added_by":"auto","created_at":"2024-03-13 17:17:37","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":93906,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of ROC curves for radiomics model, TNM stage model, clinical model and combined radiomics model in training set (\u003cstrong\u003eA\u003c/strong\u003e) and validation set (\u003cstrong\u003eB\u003c/strong\u003e)\u003c/p\u003e","description":"","filename":"Fig.2.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/2a79c3ce568382902d13735e.png"},{"id":52623018,"identity":"9fc32378-e547-42e7-9ef8-8061a9c23557","added_by":"auto","created_at":"2024-03-13 17:17:37","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":29117,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration curves for the training set (\u003cstrong\u003eA\u003c/strong\u003e) and validation set (\u003cstrong\u003eB\u003c/strong\u003e). The dashed line on the middle diagonal is the ideal perfect case, the other dashed line indicates the predictive performance of the current model, and the solid line indicates the predictive performance of the calibrated model. The closer the two dashed lines are, the better the model's goodness-of-fit is\u003c/p\u003e","description":"","filename":"Fig.3.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/cab3dd5a645f679434352b7e.png"},{"id":52623024,"identity":"b78b3ea9-1894-4cc0-abbd-ab7c794b1791","added_by":"auto","created_at":"2024-03-13 17:17:38","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":492337,"visible":true,"origin":"","legend":"\u003cp\u003eEnhanced CT images of two representative cases of thymic carcinoma:\u003cstrong\u003e(A)\u003c/strong\u003e Anterior mediastinal mass on enhanced CT. \u003cstrong\u003e(B)\u003c/strong\u003e Enlarged right hilar lymph node (arrow) in the same patient as \u003cstrong\u003eA\u003c/strong\u003e. No distant metastasis occurred. Postoperative pathology shows CD5 (+) and TNM stage IV. Using the combined radiomics model, this patient was classified as stable disease (SD), with progression-free survival until last follow-up (118.1 months); \u003cstrong\u003e(C) A\u003c/strong\u003enterior mediastinal mass, with no vascular invasion of the lesion, and no lymph node metastases on enhanced CT.\u003cstrong\u003e (D)\u003c/strong\u003e Liver metastases (arrow) occurred in the same patient as \u003cstrong\u003eC\u003c/strong\u003e. Postoperative pathology showed CD5 (-) and TNM stage IV. Using the combined radiomics model, the patient was classified as PD, with tumor progression 9.5 months after chemotherapy\u003c/p\u003e","description":"","filename":"Fig.4.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/22c0d77adb850dbcead0caa5.png"},{"id":52623025,"identity":"f2bf3a11-f593-4c60-a464-2b2ae07c0437","added_by":"auto","created_at":"2024-03-13 17:17:38","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":620636,"visible":true,"origin":"","legend":"\u003cp\u003eThe decision curve analysis of the combined radiomics model in the training set and validation set\u003c/p\u003e","description":"","filename":"Fig.5.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/fb4e99c5c97f0b4225992e9f.png"},{"id":52623022,"identity":"1ab99157-8df0-42af-883c-76e7a468042b","added_by":"auto","created_at":"2024-03-13 17:17:38","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":103332,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier plots for progression-free survival (PFS) in low- and high-risk groups predicted by radiomics model and the combined radiomics model in training set (\u003cstrong\u003eA\u003c/strong\u003e and \u003cstrong\u003eB\u003c/strong\u003e) and validation set (\u003cstrong\u003eC\u003c/strong\u003e and \u003cstrong\u003eD\u003c/strong\u003e)\u003c/p\u003e","description":"","filename":"Fig.6.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/01cc04bf1cc00f041101c933.png"},{"id":52623026,"identity":"626bee01-4e7c-450b-825b-982bec1b612d","added_by":"auto","created_at":"2024-03-13 17:17:38","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1034740,"visible":true,"origin":"","legend":"\u003cp\u003eHeatmap showing treatment response after first-line chemotherapy for thymic carcinoma in relation to clinicopathological variables (CD5 and M) and RS in all datasets. Patients with CD5 (-), M (+), and high RS tended to develop PD after first-line chemotherapy. Color bars indicate the status of CD5 (1: positive), M (1: positive), and response to treatment (1: PD), as well as the RS value.\u003c/p\u003e","description":"","filename":"Fig.7.png","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/ffc13ff58353a955342338c2.png"},{"id":52623971,"identity":"6e748b96-20c2-4daa-bc46-d2f328153343","added_by":"auto","created_at":"2024-03-13 17:25:38","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1891602,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/8ec3f28c-f5b0-4435-9770-de5533917a7e.pdf"},{"id":52623020,"identity":"66b40569-e386-408a-94c6-d35d176acc2c","added_by":"auto","created_at":"2024-03-13 17:17:37","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":234758,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-3988200/v1/8a506539a1ea84a09a8f9b06.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Tumor Radiomics Features on Pretreatment CT to Predict Response to First-Line Chemotherapy in Thymic Carcinoma: A Multicenter Study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThymic carcinoma (TC) is a rare malignant tumor occurring in the pre-vascular mediastinum, with an incidence between 0.07 and 0.38/100,000/year [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Most of these tumors are squamous cell carcinomas that infiltrate the thymic epithelial peritoneum invading mediastinal structures and spreading to the pleura. Thus, thymic carcinoma is usually diagnosed as locally advanced or metastatic and require aggressive multidisciplinary treatment [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePlatinum-based chemotherapy remains the standard of first-line treatment for patients with advanced TC who are not candidates for locoregional therapy [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. A retrospective study of the efficacy of first-line treatment (platinum-based chemotherapy) in patients with advanced thymic carcinoma showed remission rates of 12%-67% and no significant differences among chemotherapy combinations. However, in clinical practice, there are significant differences in prognosis even in patients with the same staging [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Therefore, in order to improve the prognosis of TC treatment, a new biomarker is needed to assess the treatment response after first-line chemotherapy in TC.\u003c/p\u003e \u003cp\u003eEnhanced CT plays an important role in assessing the response of tumors to treatment. However, conventional diagnostic imaging relies heavily on the personal experience of physicians, which is subjective. Moreover, it is difficult for the human eye vision to fully capture all the information in images. Comparatively speaking, radiomics, which combines artificial intelligence with diagnostic imaging, can fully mine image information to objectively and efficiently achieve differential diagnosis of diseases, prediction of treatment response, and judgement of prognosis [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite the great potential of radiomics, the use of different CT scanners and imaging parameters in different centers places greater demands on the reproducibility and generalizability of the models. In this study, we collected TC data from different centers and sought to develop and validate a non-invasive radiomics biomarker for predicting response to first-line chemotherapy in TC by integrating clinical, pathological and imaging factors.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e \u003cb\u003ePatients\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTo predict response after first-line chemotherapy for TC, TC patients from four hospitals in Shandong province from April 2014 to September 2023 were retrospectively included in this study (Supplementary Material). All patients received platinum-based first-line chemotherapy [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Inclusion criteria: 1) enhanced CT scan at the hospital within 1 month prior to treatment; 2) patients with a pathological diagnosis of thymic carcinoma; 3) well-developed post-chemotherapy imaging, able to judge efficacy according to RECIST 1.1 criteria; 4) complete records of patient demographic and clinical characteristics. Exclusion criteria: 1) poor image quality due to human factors or other reasons; 2) previous history of treatment for thymic carcinoma (including chemotherapy, radiotherapy and volume-reducing surgery); 3) history of other malignant tumors. In the end, we included a total of 134 patients. The patients from the four centers were randomly divided into training and validation sets in a 7:3 ratio.\u003c/p\u003e\n\u003ch3\u003eClinical data and Assessments\u003c/h3\u003e\n\u003cp\u003eThe following preoperative clinical and laboratory data were collected: basic demographics, immunohistochemical information, TNM tumor staging (8th edition), lactate dehydrogenase (LDH), carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), neutrophils (NE), lymphocytes (Lym). LDH, CEA, NSE, and NE thresholds were 250 U/ ml, 10, 16.3 ng/ml, and 6.3 \u0026times; 10\u003csup\u003e9\u003c/sup\u003e/L, respectively, and Lym range was 1.1\u0026ndash;3.2 \u0026times; 10\u003csup\u003e9\u003c/sup\u003e/L.\u003c/p\u003e\n\u003ch3\u003eRadiographic Data and Image Segmentation\u003c/h3\u003e\n\u003cp\u003ePretreatment enhanced CT (non-enhanced and arterial enhanced phases) images were collected. See supplementary material for CT acquisition and parameter details. Intratumoral and mediastinal ROIs were outlined with the Deepwise Multimodal Research Platform (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://keyan.deepwise.com,V2.0\u003c/span\u003e\u003cspan address=\"https://keyan.deepwise.com,V2.0\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) (Beijing Deepwise \u0026amp; League of PHD Technology Co., Ltd., Beijing, China). Mediastinal ROIs were outlined avoiding bones, heart, great vessels, esophagus and trachea. If there are metastatic tumors from other sites, they are all outlined. Images from 15 patients from the training set were then randomly selected and the same segmentation procedure was repeated 2 weeks later by the same two radiologists to assess inter-observer and intra-observer agreement.\u003c/p\u003e\n\u003ch3\u003eFeature Extraction\u003c/h3\u003e\n\u003cp\u003eRadiomics features were extracted in non-enhanced and enhanced intratumoral and mediastinal ROIs, respectively. The voxels of all images were normalized, resampled at a size of 1 \u0026times; 1 \u0026times; 1 cubic millimeter and features were extracted from within each dimension in a 2D manner. Detailed parameters are given in the supplementary material. Ultimately 1781, 1781, 1781 and 1781 features were extracted from the tumor and mediastinal regions in each case, respectively. Classify features into 7 categories: 1) First Order Features; 2) Shape Features; 3) GLCM Features; 4) GLSZM Features; 5) GLRLM Features; 6) GLDM Features; 7) NGTDM Features.\u003c/p\u003e\n\u003ch3\u003eFeature Selection\u003c/h3\u003e\n\u003cp\u003eFirstly, radiomics features with high inter-observer and intra-observer stability (ICC\u0026thinsp;\u0026ge;\u0026thinsp;0.75) were selected. Then, feature correlation analysis was used to eliminate features with high similarity (\u0026gt;\u0026thinsp;0.7). Finally, modelling features were identified by F test. Clinicopathological features were selected for subsequent model construction by univariate and multivariate logistic regression.\u003c/p\u003e\n\u003ch3\u003eRadiomics Imaging Biomarker Development\u003c/h3\u003e\n\u003cp\u003eThe selected features were analyzed using Linear Discriminant Analysis (LDA) algorithm to construct four models: non-enhanced intratumoral model, non-enhanced mediastinal model, enhanced intratumoral model and enhanced mediastinal model. The discrimination was quantified by the area under the curve (AUC). The better two models are selected based on the AUC results, their features are merged, and then the feature selection step is repeated to develop the radiomics score (RS) using LDA algorithm. Finally, the combined RS and clinicopathological features obtained from screening were used together to construct the combined radiomics model.\u003c/p\u003e\n\u003ch3\u003eModel Validation\u003c/h3\u003e\n\u003cp\u003eIn this study, the performance of each model was evaluated using a confusion matrix, and the prediction accuracy of the combined model was evaluated using calibration curves and the Hosmer-Lemeshow test [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Practical application of combined radiomics model assessed using decision curve analysis (DCA). Incremental differences in evaluating combinatorial models using integrated discriminatory improvement (IDI). Patients in each set were stratified according to risk predicted by RS and combined radiomics model and Kaplan-Meier analysis. Univariate and multivariate Cox regression analysis of RS and other clinicopathological variables were performed to select independent predictors of survival. In addition, we tried several different machine learning classifiers and found that LDA performed the best (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eFollow-up\u003c/h3\u003e\n\u003cp\u003e Patients were followed up after treatment at each center as recommended by the guidelines. CT scans were performed every 6 months for the first 3 years after receiving first-line chemotherapy [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Progression-free survival (PFS) was defined as the time between the start of chemotherapy and the onset of tumor progression or (for any reason) death.\u003c/p\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eStatistical analysis were performed using SPSS (version 26), R software (version 3.6.2) and Python (version 3.8). For skewed distribution data, they were described as median [interquartile range, IQR]. Comparison of different ROC curves using the Delong test. Radiomics features selection and model construction were carried out using the Deepwise Multimodal Research Platform. In this study, we used the \"Matplotlib\" library in python to draw the confusion matrix. Moreover, we used univariate and multivariate Cox regression analysis to screen the independent predictors of PFS. A two-sided \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e \u003cb\u003ePatient Characteristics\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThis study included 134 TC patients from four independent centers. The overall design of the study is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The clinicopathological characteristics of the training set (n\u0026thinsp;=\u0026thinsp;93) and validation set (n\u0026thinsp;=\u0026thinsp;41) are listed in Table S2. There were 87 (64.9%) males and 47 (35.1%) females, with a median age of 57 (49.25-64) years, and the majority of patients (n\u0026thinsp;=\u0026thinsp;122, 91%) were in stage III or stage IV. The progression rates (PD: Progressive disease) were 56.1% and 58.1% for the training and validation sets.\u003c/p\u003e\n\u003ch3\u003eFeature Selection and Establishment of Radiomics Models\u003c/h3\u003e\n\u003cp\u003eFirstly, features with low concordance (ICC\u0026thinsp;\u0026lt;\u0026thinsp;0.75) were excluded from this study, and 1611, 1360, 1546 and 1418 radiomics features were retained in the non-enhanced intratumor, non-enhanced mediastinum, enhanced intratumor and enhanced mediastinum, respectively. Then, after feature correlation analysis and F-test, the final 6, 9, 10 and 6 radiomics features were generated in this study, respectively. The LDA algorithm was used to build the radiomics model, with AUC values ranging from 0.725\u0026ndash;0.839 in the training set and from 0.672\u0026ndash;0.730 in the validation set. After comparing AUC, this study derived RS based on unenhanced intratumoural and enhanced intratumoural features using LDA algorithm. Detailed formulae for RS are given in the supplementary results. In terms of prediction of progression after chemotherapy, the AUC of RS was 0.911 and 0.785 in the training and validation sets.\u003c/p\u003e\n\u003ch3\u003eCombined Radiomics Model Development\u003c/h3\u003e\n\u003cp\u003eUnivariate and multivariate analysis showed that CD5 and Distant metastasis were independently associated with PD (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). The combined radiomics model was built by combining the above two clinical features with RS. After adding the radiomics features, the IDI of the combined radiomics model in the training set was significantly improved compared to the clinical model, however, no significant improvement was found in the validation set (Table S3).\u003c/p\u003e\n\u003ch3\u003eInternal Validation of the Combined Radiomics Model\u003c/h3\u003e\n\u003cp\u003eAs can be seen from the ROC curves (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB), the combined radiomics model has the highest AUC values in the training and validation sets, which are 0.937 (95% CI 0.891\u0026ndash;0.984) and 0.899 (95% CI 0.800-0.997), respectively. Both the training and validation sets showed good agreement on the calibration curves (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB), and the Hosmer-Lemeshow test showed that there was no significant difference between the predicted probabilities and the reproduced probabilities of the actual observations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.152, 0.899), which supports the accuracy of the model. The confusion matrix for each model is shown in Fig. S2.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThen, we compared each model by Delong test, and the differences between the combined model and the other models were all statistically significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the training set, and no significant differences were found between the models in the validation set (Table S4).\u003c/p\u003e \u003cp\u003eIn terms of practical application, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, it helps clinicians to identify patients with progression pre-chemotherapy in order to develop better treatment and follow-up plans, such as early adjustment of chemotherapy regimen, combination immunotherapy, or shortening of follow-up intervals. It helps to reduce the risk of PD and improve patient prognosis. Additionally, the DCA plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) indicates that the combined radiomics model can predict the treatment response of patients within a probability threshold range of 0.25 to 0.70.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eRelationship of RS and Combined Radiomics Model to PFS after First-line Chemotherapy\u003c/h3\u003e\n\u003cp\u003eUntil January 14, 2024, the median follow-up was 32.5 months (IQR: 18.2\u0026ndash;57.1) for the training set and 35.1 months (IQR: 21.7\u0026ndash;66.2) for the validation set; the median treatment duration was 33.9 months (IQR: 18.1\u0026ndash;56.9) for the training set and 34.6 months for the validation set (IQR. 20.6\u0026ndash;65.9). Progression occurred in 54 (58.1%) of 93 patients in the training set and in 23 (56.1%) of 41 patients in the validation set. The median PFS was 11.5 months and 16.1 months in the training and validation sets, respectively.\u003c/p\u003e \u003cp\u003eUnivariate and multivariate COX regression analysis showed that high RS, CD5 (-) and the presence of distant metastasis were independently associated with poorer PFS (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). In addition, Kaplan-Meier analysis demonstrated the ability of the RS and combined radiomics model to stratify patients in each cohort pre-chemotherapy in both the training and validation sets, with log-rank tests suggesting good discriminatory properties of the models (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Furthermore, as can be seen from the heatmap in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, patients with high RS, CD5 (-) and M (+) tend to develop PD after first-line chemotherapy.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMultivariate Cox regression analysis for progression-free survival in patients with thymic carcinoma in the training and validation sets\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eProgression-free survival\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraining set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.929\u0026ndash;16.037\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCD5 (positive vs. negative)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.226\u0026ndash;0.780\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003e0.006\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistant metastasis (positive vs. negative)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.629\u0026ndash;16.851\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eValidation set\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e1.295\u0026ndash;23.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e0.021\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCD5 (positive vs. negative)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e0.170\u0026ndash;0.879\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistant metastasis (positive vs. negative)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e2.063\u0026ndash;16.923\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this multicenter study, we established and evaluated the effectiveness of combined radiomics model to predict progression after first-line chemotherapy for TC. The highest model prediction performance was achieved when radiomics features were combined with clinicopathological variables, with AUCs of 0.937 (95% CI 0.891\u0026ndash;0.984) in the training set and 0.899 (95% CI 0.800-0.997) in the validation set. Additionally, the calibration curves and Hosmer-Lemeshow test indicate that the model has good predictive accuracy. DCA demonstrated that the combined radiomics model in this study resulted in better net clinical gains. These results suggest that radiomics features and clinicopathological factors can provide support for predicting disease progression after first-line chemotherapy.\u003c/p\u003e \u003cp\u003eMore and more studies are focusing on the tumor microenvironment to expect more information on disease progression and treatment response [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. However, in this study, the performance of the mediastinum-based model was lower compared to the performance of the intratumor-based model. This is similar to past studies on other tumors [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. The following reasons may explain these results. On the one hand, past studies have shown that it is possible to distinguish thymoma from thymic squamous cell carcinoma by the collagen fiber pattern on T2-weighted MR, with linear or reticular fibrous intervals often present within thymomas; however, in thymic carcinoma, the collagen fibers are poorly delineated and are distributed in eccentric patches of greater size [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Collagen fibers are associated with increased stiffness of the extracellular matrix (ECM). Past studies have shown that ECM contributes to immune rejection [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Moreover, in breast cancer, increased ECM hardness promotes epithelial mesenchymal transition (EMT), cell invasion and metastasis [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Thus, the collagen fiber pattern may reflect increased invasiveness of TC, however, in the context of fat in the mediastinum, it is difficult to differentiate between the tumor collagen fiber region and other regions of the tumor. Therefore, in our manual segmentation, collagen fiber regions that actually have information about invasiveness may be included in the intratumor ROI. On the other hand, using mediastinum as the ROI may incorporate more redundant features, which will inevitably lead to an impact on the model performance despite the fact that this study screened the features by ICC analysis, feature correlation analysis, and F-test. Although the mediastinum-based model has no additional value in predicting response to TC treatment, it is a novel attempt to explore the peritumoral region and could provide a reference for subsequent studies.\u003c/p\u003e \u003cp\u003eIt has been shown that radiomics performs well in risk stratification and identification of thymic epithelial tumors (TETs) (AUC range, 0.76\u0026ndash;0.978) [\u003cspan additionalcitationids=\"CR19 CR20\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Moreover, models that combine radiomics features and clinicopathological factors may reflect different characteristics of tumors associated with treatment response from different perspectives [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Although, there is a known association between radiomics features and treatment response, no studies have been reported on predicting treatment response in TC [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. To our knowledge, this study is the first to predict treatment response after first-line chemotherapy for TC by establishing a combined radiomics model.\u003c/p\u003e \u003cp\u003eThe variables of the combined radiomics model in this study included 21 radiomics features and 2 clinicopathological features. Notably, 81% (17/21) of the radiomics features were 3D and wavelet features, which is in line with previous findings, this suggests that tumor heterogeneity can be reflected as differences in the spatial distribution of voxel intensities in images [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. After unifactorial and multifactorial analysis, CD5 and distant metastasis were finally selected as clinicopathological variables for the establishment of the combined radiomics model. They have been confirmed in previous studies and are considered to be associated with tumor prognosis. In a study evaluating the prognosis of patients with advanced thymic carcinoma receiving first-line chemotherapy, researchers found that CD5 positivity may be associated with better PFS [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. In the present study, we found that CD5 positivity was associated with stable disease and better PFS after first-line chemotherapy, similar to past studies. Studies have shown that immune cells such as neutrophils and lymphocytes can regulate tumor cell activity by secreting large amounts of cytokines [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. In addition, high levels of neutrophil-to-lymphocyte ratio (NLR) are thought to be strongly associated with poor tumor outcomes [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Therefore, neutrophils, lymphocytes and NLR in peripheral blood were included in this study and analyzed for their association with progression after first-line chemotherapy for TC as well as with PFS. The results showed that they were not independent predictors of progression and PFS after first-line chemotherapy for TC. This may be related to the decrease in neutrophil levels due to glucocorticoid use in some TC patients [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThere are some limitations in this study. Firstly, in the validation set, Delong test showed that there was no significant difference between the models as well as no significant improvement was found in the combined identification improvement of the combined model over the clinical model. This may be related to the smaller number of cases in the validation set. Future studies could include a larger sample for external validation of the models. Secondly, in this study, we selected non-enhanced and arterial phase images for analysis, which theoretically may have reduced predictive performance compared to the full-phase CT image model. In the future, we also plan to perform full-phase CT-based feature extraction and analysis. Thirdly, considering the accuracy of image segmentation, we use manual segmentation of ROIs, but this will take more time. With the development of AI technology, automatic segmentation techniques may further improve the efficiency of image segmentation.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates that the combined radiomics model can be used to predict objective response to first-line chemotherapy in TC patients with an improvement over clinicopathological models. It was also correlated with PFS after the initiation of first-line chemotherapy.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAUC Area under the curve\u003c/p\u003e\u003cp\u003eCEA carcinoembryonic antigen\u003c/p\u003e\u003cp\u003eDCA Decision curve analysis\u003c/p\u003e\u003cp\u003eECM Extracellular matrix\u003c/p\u003e\u003cp\u003eEMT Epithelial mesenchymal transition\u003c/p\u003e\u003cp\u003eIDI Integrated discriminatory improvement\u003c/p\u003e\u003cp\u003eIQR Interquartile range\u003c/p\u003e\u003cp\u003eLDA Linear discriminant analysis\u003c/p\u003e\u003cp\u003eLDH Lactate dehydrogenase\u003c/p\u003e\u003cp\u003eLym Lymphocytes\u003c/p\u003e\u003cp\u003eNE Neutrophils\u003c/p\u003e\u003cp\u003eNLR Neutrophil-to-lymphocyte ratio\u003c/p\u003e\u003cp\u003eNSE Neuron-specific enolase\u003c/p\u003e\u003cp\u003ePD Progressive disease\u003c/p\u003e\u003cp\u003ePFS Progression-free survival\u003c/p\u003e\u003cp\u003eRS radiomics score\u003c/p\u003e\u003cp\u003eSD stable disease\u003c/p\u003e\u003cp\u003eTC Thymic carcinoma\u003c/p\u003e\u003cp\u003eTETs Thymic epithelial tumors\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgment\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank all the participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u003c/strong\u003e\u003cstrong\u003e’\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSiyu Wu: Formal analysis, Methodology, Visualization, Writing\u0026nbsp;–\u0026nbsp;original draft, Writing\u0026nbsp;–\u0026nbsp;review \u0026amp; editing. Ying Wang: Data curation. Ping Fan: Data curation. Tianqi Xu: Data curation. Yong Huang: Data curation. Pengxi Han: Data curation. Yan Deng: Data curation. Ximing Wang : Conceptualization, Data curation, Project administration, Supervision, Writing\u0026nbsp;–\u0026nbsp;review \u0026amp; editing. All authors reviewed the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe present study was supported by the National Natural Science Foundation of China (Grant Nos. 8187354, 81571672), and Academic promotion program of Shandong First Medical University (Grant No. 2019QL023).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the fundings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study was approved by the Ethics Committee of the Shandong Provincial Hospital and written informed consent was discarded because of the retrospective nature of this study. We confirmed that the study was carried out in accordance with relevant guidelines and regulations of Helsinki Declaration.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll the authors declare that they have no conflict of interest.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRoden AC, Ahmad U, Cardillo G, Girard N, Jain D, Marom EM, Marx A, Moreira AL, Nicholson AG, Rajan A, et al. Thymic Carcinomas-A Concise Multidisciplinary Update on Recent Developments From the Thymic Carcinoma Working Group of the International Thymic Malignancy Interest Group. J Thorac Oncol. 2022;17(5):637\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/35227908\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/35227908\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChun SG, Rimner A, Amini A, Chang JY, Donington J, Edelman MJ, Geng Y, Gubens MA, Higgins KA, Iyengar P, et al. American Radium Society Appropriate Use Criteria for Radiation Therapy in the Multidisciplinary Management of Thymic Carcinoma. JAMA Oncol. 2023;9(7):971\u0026ndash;80. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/37186595\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/37186595\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRemon J, Villacampa G, Facchinetti F, Di Maio M, Marcuse F, Tiseo M, Hochstenbag M, Hendriks LEL, Besse B. Immune checkpoint blockers in patients with unresectable or metastatic thymic epithelial tumours: A meta-analysis. Eur J Cancer. 2023;180:117\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36592506\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36592506\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo R, Shukuya T, Okuma Y, Tateishi K, Imai H, Iwasawa S, Miyauchi E, Fujiwara A, Sugiyama T, Azuma K, et al. Prognostic Factors and Efficacy of First-Line Chemotherapy in Patients with Advanced Thymic Carcinoma: A Retrospective Analysis of 286 Patients from NEJ023 Study. Oncologist. 2018;23(10):1210\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/29567820\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/29567820\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Q, Zhang L, Liu S, You J, Chen L, Jin Z, Zhang S, Zhang B. Radiomics in precision medicine for gastric cancer: opportunities and challenges. Eur Radiol. 2022;32(9):5852\u0026ndash;68. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/35316364\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/35316364\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuiot J, Vaidyanathan A, Deprez L, Zerka F, Danthine D, Frix AN, Lambin P, Bottari F, Tsoutzidis N, Miraglio B, et al. A review in radiomics: Making personalized medicine a reality via routine imaging. Med Res Rev. 2022;42(1):426\u0026ndash;40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34309893\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34309893\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerghmans T, Durieux V, Holbrechts S, Jungels C, Lafitte JJ, Meert AP, Moretti L, Ocak S, Roelandts M, Girard N. Systemic treatments for thymoma and thymic carcinoma: A systematic review. Lung Cancer. 2018;126:25\u0026ndash;31. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/30527189\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/30527189\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKramer AA, Zimmerman JE. Assessing the calibration of mortality benchmarks in critical care: The Hosmer-Lemeshow test revisited. Crit Care Med. 2007;35(9):2052\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/17568333\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/17568333\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScorsetti M, Leo F, Trama A, D'Angelillo R, Serpico D, Macerelli M, Zucali P, Gatta G, Garassino MC. Thymoma and thymic carcinomas. Crit Rev Oncol Hematol. 2016;99:332\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/26818050\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/26818050\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang W, Jiang Y, Xiong W, Sun Z, Chen C, Yuan Q, Zhou K, Han Z, Feng H, Chen H, et al. Noninvasive imaging of the tumor immune microenvironment correlates with response to immunotherapy in gastric cancer. Nat Commun. 2022;13(1):5095. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36042205\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36042205\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang T, She Y, Yang Y, Liu X, Chen S, Zhong Y, Deng J, Zhao M, Sun X, Xie D, et al. Radiomics for Survival Risk Stratification of Clinical and Pathologic Stage IA Pure-Solid Non\u0026ndash;Small Cell Lung Cancer. Radiology. 2022;302(2):425\u0026ndash;34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34726531\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34726531\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSong W, Yu X, Guo D, Liu H, Tang Z, Liu X, Zhou J, Zhang H, Liu Y, Liu X. MRI-Based Radiomics: Associations With the Recurrence-Free Survival of Patients With Hepatocellular Carcinoma Treated With Conventional Transcatheter Arterial Chemoembolization. J Magn Reson Imaging. 2020;52(2):461\u0026ndash;73. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/31675174\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/31675174\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin Z, Wang T, Li Q, Bi Q, Wang Y, Luo Y, Feng F, Xiao M, Gu Y, Qiang J, et al. Development and validation of MRI-based radiomics model to predict recurrence risk in patients with endometrial cancer: a multicenter study. Eur Radiol. 2023;33(8):5814\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/37171486\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/37171486\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu YC, Yan WQ, Yan LF, Xiao G, Han Y, Liu CX, Wang SZ, Li GF, Wang SM, Yang G, et al. Differentiating thymoma, thymic carcinoma and lymphoma based on collagen fibre patterns with T2- and diffusion-weighted magnetic resonance imaging. Eur Radiol. 2022;32(1):194\u0026ndash;204. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34215941\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34215941\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakahashi K, Al-Janabi NJ. Computed tomography and magnetic resonance imaging of mediastinal tumors. J Magn Reson Imaging. 2010;32(6):1325\u0026ndash;39. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/21105138\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/21105138\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun X, Wu B, Chiang H-C, Deng H, Zhang X, Xiong W, Liu J, Rozeboom AM, Harris BT, Blommaert E, et al. Tumour DDR1 promotes collagen fibre alignment to instigate immune exclusion. Nature. 2021;599(7886):673\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34732895\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34732895\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFattet L, Jung HY, Matsumoto MW, Aubol BE, Kumar A, Adams JA, Chen AC, Sah RL, Engler AJ, Pasquale EB, et al. Matrix Rigidity Controls Epithelial-Mesenchymal Plasticity and Tumor Metastasis via a Mechanoresponsive EPHA2/LYN Complex. Dev Cell. 2020;54(3):302\u0026ndash;e316307. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e.https://www.ncbi.nlm.nih.gov/pubmed/32574556\u003c/span\u003e\u003cspan address=\"http://.https://www.ncbi.nlm.nih.gov/pubmed/32574556\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOhira R, Yanagawa M, Suzuki Y, Hata A, Miyata T, Kikuchi N, Yoshida Y, Yamagata K, Doi S, Ninomiya K, et al. CT-based radiomics analysis for differentiation between thymoma and thymic carcinoma. J Thorac Dis. 2022;14(5):1342\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/35693628\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/35693628\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahmoudi S, Gruenewald LD, Eichler K, Althoff FC, Martin SS, Bernatz S, Booz C, Yel I, Kinzler MN, Ziegengeist NS, et al. Multiparametric Evaluation of Radiomics Features and Dual-Energy CT Iodine Maps for Discrimination and Outcome Prediction of Thymic Masses. Acad Radiol. 2023;30(12):3010\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/37105804\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/37105804\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu C, Li T, Yang X, Zhang R, Xin L, Zhao Z, Cui J. Contrast-enhanced CT-based radiomics model for differentiating risk subgroups of thymic epithelial tumors. BMC Med Imaging. 2022;22(1):37. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/35249531\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/35249531\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao W, Ozawa Y, Hara M, Okuda K, Hiwatashi A. Computed tomography radiomic feature analysis of thymic epithelial tumors: Differentiation of thymic epithelial tumors from thymic cysts and prediction of histological subtypes. Jpn J Radiol 2023.\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/38010596\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/38010596\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu B, Dong SY, Bai XL, Song TQ, Zhang BH, Zhou LD, Chen YJ, Zeng ZM, Wang K, Zhao HT, et al. Tumor Radiomic Features on Pretreatment MRI to Predict Response to Lenvatinib plus an Anti-PD-1 Antibody in Advanced Hepatocellular Carcinoma: A Multicenter Study. Liver Cancer. 2023;12(3):262\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/37601982\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/37601982\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao J, He Y, Yang X, Tian P, Zeng L, Huang K, Zhao J, Zhou J, Zhu Y, Wang Q et al. Assessing treatment outcomes of chemoimmunotherapy in extensive-stage small cell lung cancer: an integrated clinical and radiomics approach. J Immunother Cancer 2023, 11(9).\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/37730276\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/37730276\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang R, Dai W, Gong J, Huang M, Hu T, Li H, Lin K, Tan C, Hu H, Tong T, et al. Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients. J Hematol Oncol. 2022;15(1):11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/35073937\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/35073937\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSong Y, Zhou G, Zhou Y, Xu Y, Zhang J, Zhang K, He P, Chen M, Liu Y, Sun J, et al. Artificial intelligence CT radiomics to predict early recurrence of intrahepatic cholangiocarcinoma: a multicenter study. Hepatol Int. 2023;17(4):1016\u0026ndash;27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36821045\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36821045\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu L, Yang P, Liang W, Liu W, Wang W, Luo C, Wang J, Peng Z, Xing L, Huang M, et al. A radiomics approach based on support vector machine using MR images for preoperative lymph node status evaluation in intrahepatic cholangiocarcinoma. Theranostics. 2019;9(18):5374\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/31410221\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/31410221\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHo IW, Pan YL, Lai JI, Liu CY. Characteristics and outcome of systemic treatment for metastatic or unresectable thymic carcinoma: A single institution experience. Thorac Cancer 2023, n/a(\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003en/a).https://www.ncbi.nlm.nih.gov/pubmed/38149471\u003c/span\u003e\u003cspan address=\"http://n/a).https://www.ncbi.nlm.nih.gov/pubmed/38149471\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoffelt SB, Wellenstein MD, de Visser KE. Neutrophils in cancer: neutral no more. Nat Rev Cancer. 2016;16(7):431\u0026ndash;46. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/27282249\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/27282249\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOuyang H, Xiao B, Huang Y, Wang Z. Baseline and early changes in the neutrophil-lymphocyte ratio (NLR) predict survival outcomes in advanced colorectal cancer patients treated with immunotherapy. Int Immunopharmacol. 2023;123:110703. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/37536184\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/37536184\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu Z, Zhou J, Li Y, Luan Y, Li H, Jia B, Xie Z, Cheng B, Wu T. Peripheral immune signature resembles tumor microenvironment and predicts clinical outcomes in head and neck squamous cell carcinoma. Front Immunol 2022, 13915207\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36148222\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36148222\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang AP, Liu JP, Tao WQ, Li HM. The diagnostic and predictive role of NLR, d-NLR and PLR in COVID-19 patients. Int Immunopharmacol. 2020;84:106504. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/32304994\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/32304994\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Machine learning, Thymic carcinoma, Radiomics, CT, Multicenter study, Linear Discriminant Analysis, Prediction model, Progression-free survival","lastPublishedDoi":"10.21203/rs.3.rs-3988200/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3988200/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eObjectives\u003c/b\u003e\u003c/p\u003e \u003cp\u003ePlatinum-based chemotherapy as first-line treatment have shown promising results against tumors in thymic carcinoma (TC). However, clinical benefit has been limited to a small proportion of patients. We developed and validated a combined radiomics model to predict progressive disease (PD) in patients suffering from TC following first-line chemotherapy.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e \u003cp\u003ePatients receiving platinum-based chemotherapy as first-line treatment from four centers in Shandong, China, were retrospectively included (n\u0026thinsp;=\u0026thinsp;134); 93 and 41 were entered into the training and validation sets. Radiomics features were extracted from pretreatment enhanced CT. After feature selection, radiomics score (RS) was developed by Linear Discriminant Analysis (LDA) and TNM, clinicopathological and clinicopathological-radiomics models (combined radiomics model) were developed by using logistic regression algorithm. Models were assessed for performance, incremental predictive value of radiomics features versus clinicopathological features, and the relationship of RS and clinical factors to survival.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe clinicopathological model was a modest predictor of PD, with area under curve (AUC) of 0.879 (95% CI: 0.709\u0026ndash;0.890) and 0.799 (95% CI: 0.778\u0026ndash;0.980) in the training and validation sets. The AUC of the combined radiomics model was 0.937 (95% CI: 0.891\u0026ndash;0.984) and 0.899 (95% CI: 0.800-0.997), which is of good calibration and clinical application. The incremental predictive value of radiomics features for clinicopathological features was 27% (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and 0.4% (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.108) in the training and validation sets. In addition, RS, CD5, and distant metastasis were associated with progression-free survival in both the training and validation sets.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusion\u003c/b\u003e\u003c/p\u003e \u003cp\u003eRadiomics features extracted from pretreatment enhanced CT allow prediction of individualized objective responses to platinum-based chemotherapy as first-line treatment in TC, providing incremental predictive value for clinicopathological features, and are associated with progression-free survival after initiation of this combination regimen.\u003c/p\u003e","manuscriptTitle":"Tumor Radiomics Features on Pretreatment CT to Predict Response to First-Line Chemotherapy in Thymic Carcinoma: A Multicenter Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-13 17:17:32","doi":"10.21203/rs.3.rs-3988200/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2024-03-17T10:36:44+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-03-11T11:03:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-03-11T10:54:53+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-03-11T10:54:52+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2024-02-25T14:24:06+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"caa51212-667d-4386-b4ce-c63d5f8033b9","owner":[],"postedDate":"March 13th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2024-03-13T17:17:33+00:00","versionOfRecord":[],"versionCreatedAt":"2024-03-13 17:17:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3988200","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3988200","identity":"rs-3988200","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}