Multi-omics models predict treatment response and overall survival for non-small cell lung cancer patients following chemo-radiotherapy: A multi-center study

Multi-omics models predict treatment response and overall survival for non-small cell lung cancer patients following chemo-radiotherapy: A multi-center 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 Multi-omics models predict treatment response and overall survival for non-small cell lung cancer patients following chemo-radiotherapy: A multi-center study Yuteng Pan, Liting Shi, Yuan Liu, Jyh-cheng Chen, Jianfeng Qiu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4076424/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: Varying chemoradiotherapy outcomes in individuals arose from the intricate physical conditions and tumor heterogeneity characteristic of non-small cell lung cancer patients. This study aimed to develop and validate multi-omics models based on the radiomics, pathomics, dosiomics and clinical information for illustrating the heterogeneity and predicting treatment response and overall survival of non-small cell lung cancer patients. Methods: This retrospective study including 220 non-small cell lung cancer patients treated with chemoradiotherapy from three hospitals for overall survival prediction, with 142 of these patients specifically assessed for treatment response prediction. Radiomics and dosiomcis features were obtained from the region of interest, including first-order and texture features. Pathomics features were derived from whole slide images by Resnet34 network. Lasso regression, random forest, and extreme gradient boosting were employed for treatment response prediction to identify the most predictive biomarkers, with model performance evaluated through area under the curve and box plots. Overall survival analysis also involved three different feature selection methods, and model evaluation incorporated area under the curve, concordance index, Kaplan-Meier curves, and calibration curves. The shapley values calculated the contribution of different modality features to the models. Results: Multi-omics models consistently exhibited superior discriminative ability compared to single-modality models in predicting treatment response and overall survival. For treatment response, the multi-omics model achieved area under the curve values of 0.85, 0.81, and 0.87 in the training set, internal validation set, and external validation set, respectively. In the analysis of overall survival, the area under the curve and concordance index of the all-modalities model were 0.83/0.79, 0.74/0.74, and 0.73/0.72 in the training set, internal validation set, and external validation set, respectively. Conclusion: Multi-omics prediction models demonstrated superior predictive ability with robustness and strong biological interpretability. By predicting treatment response and overall survival in non-small cell lung cancer patients, these models had the potential to assist clinician optimizing treatment plans, supporting individualized treatment strategies, further improving tumor control probability and prolonging the patients’ survival. Non-small cell lung cancer Pathomics Dosiomics Radiomics Deep learning Treatment response Overall survival Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Chemo-radiotherapy (CRT) is one of the standard treatment methods for advanced unresectable non-small cell lung cancer (NSCLC) patients ( 1 , 2 ). It is anticipated to enhance tumor control probability in NSCLC patients and significantly extend their overall survival (OS) ( 3 – 5 ). However, varying CRT outcomes in individuals arose from the intricate physical conditions and tumor heterogeneity characteristic of NSCLC patients ( 6 , 7 ). Therefore, it is particularly crucial to quantify the heterogeneity of tumors, predict the treatment outcomes and identify those who are most likely to benefit from CRT by pretreatment information, such as CT images, dose images, whole slide images etc ( 8 , 9 ). Radiomics, extraction high-throughput quantitative features from medical images to quantify tumor heterogeneity, analyzed through predetermined algorithms, served as a tool for developing models in clinical decision support ( 10 ). Previous research had explored radiomics models to predict treatment response or survival in NSCLC patients ( 11 – 13 ). Dosiomics, an extension of radiomics, incorporated three-dimensional dose distribution information from treatment plans as an image for feature extraction ( 9 ). Pathomics is extraction quantitative features based on whole slide images (WSI) to quantify heterogeneity of the tumor tissue( 14 ). Different from radiomics and dosiomics which represented a macro perspective, pathomics explored the heterogeneity of tumor microenvironment from a micro perspective, which was expected to comprehensively explain the heterogeneity of tumors and the differences in individual treatments, thereby achieving more accurate predictions ( 15 – 17 ). Multi-omics models, bridging macroscopic and microscopic perspectives, have shown promise in discovering multi-dimensional biomarkers, offering stronger biological interpretability and predictive capabilities. For instance, some studies developed a radiopathomic model for predicting pathologic complete response ( 18 – 20 ). Zhang et al. established multiomics models to predict the occurrence of radiation pneumonia by combining radiomics, dosiomics, dose volume histograms (DVH), and clinical information ( 22 ). However, Jing et al. established a multiomics model based on radiomics and dosiomics to predict the occurrence of radiation pneumonia and explore its impact on 1-year OS in NSCLC patients ( 21 ). Multi-omics models obtained better biological interpretability and predictability compared to single-modality models. The models with multidimensional information contributed to a more precise assessment of the complex conditions of patients, guiding the clinical doctors develop individualized diagnosis and treatment plans. Precision individual treatment for patients might reduce treatment-related side effects, enhance treatment safety, and improve tumors control probability, prolong the survival of NSCLC patients( 22 , 23 ). This study focuses on constructing multi-omics models based on computed tomography (CT) images, WSI, three-dimensional dose distribution images and clinical information before treatment to predict objective response and 1-year OS in NSCLC patients undergoing CRT. The anticipated outcome is the accurate stratification of patients, enabling clinicians to develop personalized treatment modality and enhance the overall treatment benefits for patients. Methods Patients and clinical information This retrospective study involving 220 patients in three hospitals from 2021 to 2023 and was approved by their ethics committees according to the Declaration of Helsinki. A total of 158 patients were enrolled from Shandong Provincial Hospital (center 1), 21 patients from the Shandong First Medical University Affiliated Hospital (center 2) and 41 patients from the Xiangya Hospital of Central South University (center 3). The informed consent was waived because this study employed a retrospective design. The criteria for included patients were as delineated: ( 1 ) Patients were over 18 years old. ( 2 ) Patients were diagnosed with primary NSCLC. ( 3 ) Patients received treatment for the first time after diagnosis. ( 4 ) Patients received intensity modulated radiotherapy (IMRT) and were administered radiotherapy at 2 Gy per day, 5 days a week, with 20–35 fractions and with a total dose of 40–70 Gy. ( 5 ) Patients had radiation planned CT images with volumetric doses and hematoxylin eosin (HE) stained sections of lung pathological biopsy. The criteria for excluded patients were as delineated: ( 1 ) Patients underwent surgical resection of lung cancer tumor. ( 2 ) Patients with poor quality CT images and poor quality HE-stained sections. In addition to the above standards, 142 out of 220 patients were retained for evaluating treatment response. The included criteria for those patients were as delineated: patients had follow-up CT scans within 3 months after the end of the same treatment course. The process of patient inclusion and exclusion was presented in Fig. 1 . Patients were followed up every three months in the first years after diagnosis, and every year thereafter. The basic clinical characteristics of all patients were collected, including gender, age, pathological types, clinical T stage (cT), clinical N stage (cN), clinical M stage (cM), clinical stage and total dose. Imaging acquisition and preprocessing The pretreatment CT scanning parameters were detailed in Additional file1: Table S1 . Two experienced physicists from each center outlined four regions of interest (ROI) for subsequent analysis, including the gross tumor volume (GTV), the planning tumor volume (PTV), total lung volume excluded GTV (LUNG-GTV), and total lung volume excluded PTV (LUNG-PTV). The ROIs of the dose images mirrored those of the CT images. The HE-stained biopsy sections parameters were detailed in Additional file1: Table S2 . Firstly, WSI were turned into binary images as RGB images by a proper threshold based on the different depths of the HE staining in the three centers. Subsequently, tiles with tissue larger than 50% were retained after non-overlapping cutting of binary images into tiles with 224×224 pixels. Color normalization was applied to all other tiles based on a well-stained template tile to enhance the staining quality of multicenter images (method = 'vahadane'). Delineation of different tissue types, including tumor cell, stroma, lymphocyte, red blood cell, and blur area, was performed by three experienced pathologists on some WSI from the three centers. The delineated areas were cut into tiles with 224×224 pixels, and the same template tile was used for color normalization (method = 'vahadane'). The normalized tiles were divided into training, validation, and testing sets in an 8:1:1 ratio. The Resnet 152 network was employed for training a tissue classification model capable of classifying all tiles. Tiles classified as tumor cell, stroma, and lymphocyte were reserved as tumor microenvironment tiles for subsequent feature extraction. Feature extraction and selection Radiomics and dosiomics features were extracted by pyradiomics. Before extracting those features, the CT images and dose image of patients were resampled and standardized to 1mm×1mm×5mm pixels. The radiomics and dosiomics features were divided into two groups: without preprocessing and after preprocessing by laplacian of gaussian (log) filters (sigma = 3mm). Besides calculating the minimum dose, average dose, and maximum dose of GTV, PTV and whole lung, we also included V5, V10, V15, V20, V25, V30, V35, V40, V45 and V50 of the whole lung in the analysis based on DVH. The features from DVH were classified into the clinical features for further analysis. The steps for feature selection were as follows: ( 1 ) Standardized the features of each patient by Z-score normalization to ensure comparability between data. ( 2 ) For predicting treatment response, a total of 113 patients (92 from center 1 and 21 from center 2) were randomly split into a training set (n = 80) and an internal validation set (n = 33) using a 7:3 ratio with 300 seed points for random partitioning. Additionally, 29 patients from center 3 were designated as an external validation set. Single modality features, encompassing radiomics, pathomics, dosiomics, and clinical features, were selected using the least absolute shrinkage and selection operator (Lasso) regression, random forest, and extreme gradient boosting based on the same training set. ( 3 ) For predicting 1-year OS, a total of 179 patients (158 from center 1 and 21 from center 2) were randomly split into a training set (n = 126) and an internal validation set (n = 53) using a 7:3 ratio with 300 seed points for random partitioning. Additionally, 41 patients from center 3 were designated as an external validation set. Lasso cox regression, random survival forest, and extreme gradient boosting were also used to select single modality features, including radiomics, pathomics, and dosiomics, based on the same training set. Clinical features were selected through univariate and multivariate cox proportional hazards regression ( p < 0.1)( 24 ). The selected features were demonstrated to qualify their contribution for the treatment response and OS prediction model by shapley values. Chemoradiotherapy response and overall survival prediction According to the Response Evaluation Criteria in Solid Tumors 1.1 (RECIST 1.1) ( 8 ), treatment response was divided into four levels. Complete response (CR) and partial response (PR) were considered as objective response (OR) groups, while patients with stable disease (SD) and progressive disease (PD) were considered as non-OR groups ( 9 , 10 ). For the prediction of treatment response, single omics model and multi-omics model were established by support vector machine (SVM) based on the selected features from different modalities. The prediction ability of treatment response was evaluated by the area under the receiver operating characteristic (ROC) curves and boxplot. OS was defined from the start of the initial antitumor treatment until death from any cause during follow-up. For the prediction of 1-year OS, single omics model and multi-omics model were established by cox proportional hazard based on the selected features from different modalities. The following comprehensive evaluation methods were employed: the area under the ROC curves and concordance index (C-index); the survival curves of the high-risk and low-risk group which evaluated by the Kaplan-Meier (KM) method (cutoff = the median value of the predicted value), the differences between the survival curves were tested by the log-rank test; Calibration curves were calculated to evaluate the consistency between the predicted results and recorded survival results. The flowchart of survival model construction was presented in Fig. 2 . Statistical analysis The feature extraction of radiomics and dosiomics, dose calculation and DVH production were implemented in 3D-Slicer (Version 4.11, https://www.slicer.org/ ). The Kruskal–Wall test for analyzing differences in multiple data sets. T-test were used to analyze the differences between two data sets (satisfying normal distribution and homogeneity of variance). A two sides p -value < 0.05 was considered statistically significant. Features selection, model’s construction and all statistical analyses were performed by R studio (Version 3.4.0, https://www.r- project. org/). Results Patients and clinical information Total 220 patients were recruited in this study from three hospitals (ceter1=158, ceter2=21, center3=48). All the recruited patients were used to develop and validate the OS prediction models. 142 out of 220 patients were retained for constructing treatment response predicting models (center1=, center2=, center3=29). Table 1 presented the basic clinical characteristics of all patients in the training set, internal validation set and external validation. As for characteristics of patients used for predicting treatment response are shown in the Additional file1: Table S3. There were no statistically significant differences were observed in three sets, except for cM and clinical stage. Dose factors in treatment response cohort and overall survival cohort, such as minimum dose, average dose and maximum dose of GTV, PTV, LUNG and V5, V10, V15, V20, V25, V30, V35, V40, V45 and V50 of the whole lung were detailed in Additional file1: Table S4-S5, respectively. Table 1. Baseline characteristics of all patients. Characteristics Training set Internal validation set External validation set p -value No. patients 126 53 41 Age(years) 65(41-87) 65(44-84) 61(40-78) 0.202 Gender 0.420 Male 101(80.2%) 41(77.3%) 36(87.8%) Female 25(19.8%) 12(22.7%) 5(12.2%) Pathological types 0.147 Squamous 61(48.4%) 31(58.5%) 25(61.0%) Adenocarcinoma 65(51.6%) 22(41.5%) 15(36.6%) Other 0(0.0%) 0(0.0%) 1(2.4%) Clinical T stage 0.194 T1 22(17.5%) 6(11.3%) 7(17.1%) T2 33(26.2%) 19(35.9%) 20(48.7%) T3 42(33.3%) 14(26.3%) 7(17.1%) T4 29(23.0%) 14(26.3%) 7(17.1%) Clinical N stage 0.150 N0 13(10.3%) 4(7.5%) 11(26.8%) N1 10(7.9%) 6(11.3%) 5(12.2%) N2 58(46.0%) 29(54.7%) 14(34.2%) N3 45(35.8%) 14(26.5%) 11(26.8%) Clinical M stage 0.030 M0 89(71.4%) 48(90.6%) 33(80.5%) M1 37(28.6%) 5(9.4%) 8(19.5%) Clinical stage 0.020 IIIA 35(27.8%) 25(47.2%) 18(43.9%) IIIB 40(31.7%) 20(37.7%) 13(31.7%) IIIC 14(11.1%) 3(5.7%) 2(4.9%) IV Total dose (Gy) 65 37(29.4%) 5(3.9%) 8(6.3%) 96(76.2%) 17(13.6%) 5(9.4%) 3(5.6%) 9(17.0%) 38(71.7%) 3(5.7%) 8(19.5%) 1(2.4%) 6(14.6%) 32(78.0%) 2(5.0%) 0.060 Overall survival(days) 848(24-2526) 951(226-2470) 913(111-2960) 0.485 Feature extraction and selection A total of 720 radiomics features were extracted from four ROIs by Pyradiomics, with 180 features extracted from each ROI, including 36 first-order features and 144 texture features. The texture features were calculated by using Gray Level Cooccurrence Matrix (GLCM), Gray Level Dependence Matrix (GLDM), Gray Level Run Length Matrix (GLRLM), Gray Level Size Zone Matrix (GLSZM) and Neighborhood Gray-tone Difference Matrix (NGTDM). As the Additional file1: Table S6 shown, it provided the detail tiles amounts about the tissue classification model’s establishment. The model achieved accuracies of 0.93, 0.91, and 0.92 in the training set, validation set, and test set, respectively. As Figure 3A shown, the classified tiles including tumor cell, stroma, and lymphocyte were reserved as tumor microenvironment tiles. The number of tumor microenvironment tiles were available in Additional file1: Table S7. A total of 512 pathological deep learning features of each tile were extracted by Resnet34 with pre-trained weights in ImageNet (https://www.image-net.org/) as the backbone network. The maximum value of all tiles for each patient were calculated as the patient-level feature. Chemoradiotherapy response and OS prediction The AUC of feature selection were shown in Additional file1: Table S8 for treatment response prediction in the training set, internal validation set and external validation set. The selected features for predicting treatment response comprised one radiomics feature, one pathomics feature, two dosiomics features, and one clinical indicator as shown in Additional file1: Table S9. Specifically, the chosen features included: a first-order radiomics feature extracted from PTV (LFM PTV ), a pathomics feature identified as deep learning feature NO.134 (DF134), dosiomics features encompassing a first-order feature extracted from GTV (OFR GTV ) and a texture feature extracted from LUNG-GTV (LGZ LUNG-GTV ), along with the clinical indicator, age. The performance of the single omics models and mulit-omics model were displayed in Table 2 and Figure 3. The performance of other models generated by combining various modality prediction factors was shown in Additional file1: Table S10. As shown in Table 2, the area under the curve (AUC) of the multi-omics model was consistently higher than that of the single omics model in the training, internal validation, and external validation sets, surpassing 0.8. The AUC of the multi-omics model in the training set was 0.85, with an internal validation set of 0.81 and an external validation set of 0.87 (Additional file2:figure S1). Through the boxplot, it can also be seen that the combined model effectively distinguished between responsive and non-responsive groups. Table 2. Performance of single omics and multiomics model in treatment response prediction. Treatment response prediction AUC (95%CI) Accuracy (95%CI) Sensitivity (95%CI) Specificity (95%CI) Training set Radiomics 0.66 0.61 0.70 0.58 Pathomics 0.61 0.66 0.62 0.75 Dosiomics 0.78 0.75 0.72 0.78 Clinical 0.64 0.61 0.58 0.67 Combined 0.85(0.75-0.93) 0.80(0.70-0.88) 0.85(0.69-0.94) 0.76(0.60-0.88) IV set Radiomics 0.60 0.55 0.57 0.58 Pathomics 0.54 0.61 0.52 0.75 Dosiomics 0.78 0.73 0.67 0.78 Clinical 0.70 0.70 0.64 0.74 Combined 0.81(0.67-0.97) 0.79(0.61-0.91) 0.79(0.54-0.94) 0.79(0.54-0.94) EV set Radiomics 0.72 0.69 0.50 0.82 Pathomics 0.67 0.62 0.58 0.76 Dosiomics 0.79 0.68 0.55 0.79 Clinical 0.58 0.55 0.62 0.77 Combined 0.87(0.74-1.00) 0.79(0.60-0.92) 0.67(0.70-0.93) 0.85(0.62-0.97) Abbreviation: IV, Internal validation; EV, External validation; CI, Confidence interval; The AUC of feature selection were shown in Additional file1: Table S11 for OS prediction in the training set, internal validation set and external validation set. The five most effective features for predicting 1-year OS were selected, including one radiomics feature, one pathomics feature, two dosiomics features, and one clinical indicator. The information of the selected features are as follows: the radiomics feature was a texture feature extracted from GTV (OGG GTV ), the pathomics feature was deep learning feature NO.166 (DF166), the dosiomics features included two texture features extracted from GTV (OGZ GTV and LGR GTV ), and the clinical indicator was gender (Additional file1: Table S12). The C-index of feature selection for 1-year OS prediction and the details of selected features for 1-year OS prediction in the training set, internal validation set, external validation set were shown in Additional file1: Table S13-14, respectively. The performance of the single-omics models, three modalities (clinical, dosiomics and pathomics) model and final multi-omics model were displayed in Table3. The performance of other models generated by combining various modality prediction factors was shown in Additional file1: Table S15-16. Ultimately, two combined models that demonstrated excellent performance were retained. The AUC and C-index of three-modalities (clinical, dosiomics and pathomics) model and finally multi-omics model were higher than the single-omics models. The AUC and C-index of the three-modalities (clinical, dosiomics and pathomics) model in the training set, internal validation set and external validation set were 0.83/0.79, 0.73/0.73 and 0.78/0.78, respectively. The AUC and C-index of the all-modalities combined model in the training set, internal validation set and external validation set were 0.83/0.79, 0.74/0.74 and 0.73/0.72, respectively. The performance of the three-modalities model in the external validation group was slightly better than that of the multi-omics model. The KM curves revealed that the two combined models had good ability to distinguish between high-risk and low-risk groups (Additional file2: Figure S2). The calibration plots demonstrated great agreement between combined models’ prediction and the actual observation for survival (Additional file2: Figure S3). Table3. Performance of single omics model and multiomics model in OS prediction. OS prediction AUC (95%CI) C-index (95%CI) Training set 0.74 0.62 0.65 0.61 0.83(0.73-0.91) 0.83(0.74-0.91) 0.67 0.65 0.65 0.62 0.73(0.64-0.92) 0.74(0.55-0.92) 0.70 0.71 0.72 0.57 0.77(0.64-0.91) 0.73(0.58-0.88) 0.72 0.61 0.64 0.60 0.79(0.70-0.87) 0.80(0.71-0.87) 0.67 0.63 0.65 0.61 0.73(0.65-0.93) 0.74(0.52-0.98) 0.70 0.70 0.70 0.56 0.78(0.62-0.92) 0.72(0.62-0.92) Radiomics Pathomics Dosiomics Clinical C+D+P Combined IV set Radiomics Pathomics Dosiomics Clinical C+D+P Combined EV set Radiomics Pathomics Dosiomics Clinical C+D+P Combined Abbreviation: IV, Internal validation; EV, External validation; CI, Confidence interval; OS, Overall survival. Discussion Precise stratification of patients through chemo-radiotherapy response and OS before treatment is a pivotal step in clinical precision therapy. The implementation of precision therapy has significantly improved the treatment outcome of NSCLC (25). Advanced and non-invasive models hold the potential to predict treatment endpoints, aiding doctors in crafting personalized treatment plans and refining patient prognoses. In this study, we developed and validated multi-omics models with superior performance and clear biological interpretability based on the CT images, whole slide images, dose images and clinical information to predict treatment response and 1-year OS, outperforming single omics models. Currently, many studies focused on identifying biomarkers for predicting treatment response or survival of lung cancer patients by single omics. Yang et al. attempted to establish a radiomics model to differentiate responsive and non-responsive NSCLC patients undergoing first-line chemotherapy and targeted therapy, but the model’s performance of AUC values was only 0.74 and was limited to the single center model(26). Chen et al. innovatively developed an OS prediction model for NSCLC patients by combining intratumoral and peritumoral radiomics features, but the AUC of the combined model in the external validation set was also less than 0.7 (27). Dosiomics as an extended concept of radiomics, contained more three-dimensional dose distribution information(28). Some researchers had investigated radiotherapy side effects and survival in lung cancer patients by integrating radiomics and dosiomics information (21, 29, 30). However, whether dosomics can predict the chemoradiotherapy response and OS of NSCLC patients remains to be studied(31, 32). In other perspective, radiomics and dosiomics utilized high-throughput data to characterize tumor heterogeneity from a macro perspective. While pathomics explored individual differences at the cellular and tissue levels within the tumor microenvironment. Some investigation had established pathomics models based on WSI to predict the prognosis of NSCLC (33, 34). Therefore, integrating multi perspective image information from both macro and micro perspectives was expected to yield models with better biological interpretability and stronger predictive capabilities. The selected one radiomics feature were extracted from PTV and the selected two dosiomics features were extracted from GTV and LUNG-GTV in treatment response prediction model construction. This indicated that not only did the dose distribution of GTV affected the treatment response of patients, but also the feature intensity and dose distribution of peritumoral and surrounding normal lung tissue affected the treatment response(35, 36). Age is also selected to predict treatment response, which is consistent with the views of Sprave et al(37). The AUC of the multi-omics in the training set was 0.85, with an internal validation set of 0.81 and an external validation set of 0.87. Notably, multi-omics model exhibited increased stability when compared to radiomics models in previous studies, showing robust predictive performance even in independent external validation sets(38). Furthermore, the statistical test results illustrated in the box plot confirmed significant differences in the predicted values of the combined model between the responsive and non-responsive groups, effectively distinguishing the two cohorts. The selected five radiomics and dosiomics features were all texture features extracted from GTV in OS prediction model construction. It is suggested that the heterogeneity of intra-tumoral texture may be an important factor affect the survival stratification of advanced NSCLC patients (39, 40). Besides, gender had also been found as potential predictive factors, as they might be affecting the 1 year overall survival of CRT in patients due to differences in radiosensitivity between individuals (41). However, pathomics features were obtained from deep features of tiles without interpretability. In the future, the correlation between pathomics features and features from other modalities will be explored for enhancing the interpretability of the models. For 1-year overall survival prediction, gender was used to construct in both two combined models with better predictive ability. The AUC and C-index of the three modalities (clinical, dosiomics, and pathomics) and the all-modalities combined model surpassed 0.7 in the training set, internal validation set, and external validation set. The KM curves indicated effective differentiation between high-risk and low-risk groups. The calibration curves showcased minimal deviation between model predictions and actual values. While relying on patient clinical information, such as clinical TNM stages, could offer rough predictions of patient prognosis, models based on selected multimodal biomarkers demonstrated superior predictive performance. In addition, compared to the research on establishing prediction models for NSCLC patients based on genetic information (42-44), the multimodal information we included for analysis was more readily obtainable in clinical settings. The prediction cost of genes was relatively high, but CT imaging and pathological biopsy were necessary examination items with lower cost in the diagnosis and treatment process of patients (45). Easier to obtain biomarkers and more stable predictive models will provide precise references for patient survival, which will help clinical doctors develop personalized diagnosis and treatment plans, thereby improving patient prognosis. Additionally, our study had other limitations: only 220 NSCLC patients from three centers were included in the study and we plan to continue collecting data to explore information in large sample data; Besides, in addition to prefusion and post-fusion methods, it was also necessary to explore more new feature fusion methods. It was worth mentioning that in our study, we retained two models for predicting 1-year OS. The performance of the three-modalities model based on clinical, dosiomics and pathomics features in the external validation set was slightly better than muliti-omics model, which indicated that sometimes multimodality information may be redundant, and prompted consideration of improved fusion methods. Finally, the clinical basic indicators of the patients we had analysis were not comprehensive enough. In future research, more clinical indicators will be collected, such as hematological examination indicators, patient smoking and alcohol consumption status, to provide a more holistic assessment of their physics condition. Here multi-omics models were constructed by biomarkers which extracted from pretreatment CT images, WSI, dose information and clinical factors, which were easily accessible before therapy. This study captured macroscopic tumor characteristics, microscopic tumor ecosystem manifestations, and considered the impact of dose distribution on short-term and long-term treatment outcomes for each patient(46). The integration of multidimensional information will assist patients in establishing more comprehensive personal profiles, enabling the identification of more suitable treatment methods within complex therapeutic regimens. Additionally, the fusion of multi-modalities information contributed to the establishment of models with more precise predictive performance (47). Conclusion In this study, multi-omics models were established based on the CT images, whole slide image and dose information for predicting radio-chemotherapy response and 1-year OS of NSCLC patients. These models have the potential to assist doctors optimize patients’ CRT plans and supporting the precise treatment for patients by predicting the short-term treatment response and long-term survival of patients. Abbreviations CRT, chemoradiotherapy; NSCLC, non-small cell lung cancer; OS, overall survival; WSI, whole slide images; DVH, dose volume histograms; IMRT, intensity modulated radiotherapy ; CT, computed tomography; HE, hematoxylin eosin ; ROI, region of interest; GTV, gross tumor volume; PTV, planning tumor volume; LUNG-GTV, total lung volume excluded GTV; LUNG-PTV, total lung volume excluded PTV; Lasso, least absolute shrinkage and selection operator; Log, laplacian of gaussian (log); CR, Complete response; PR, partial response; OR, objective response; SD, stable disease; PD, progressive disease; SVM, support vector machine; ROC, receiver operating characteristic; C-index, concordance index; KM, Kaplan-Meier; GLCM, Gray Level Cooccurrence Matrix; GLDM, Gray Level Dependence Matrix; GLRLM, Gray Level Run Length Matrix; GLSZM, Gray Level Size Zone Matrix; NGTDM, Neighborhood Gray-tone Difference Matrix; AUC, area under the curve ; LFM, Log-sigma-3-0-mm-3D-firstorder-maximum; DF, Deep learning feature; OFR, Original-firstorder-robust-mean-absolute-deviation; LGZ, Log-sigma-3-0-mm-3D-glszm-zone-entropy; OGG, Original-glrlm- gray-level-nonuniformity; OGZ, Original-glszm-zone-variance; LGR, Log-sigma-3-0-mm-3D- glrlm- run-length-nonuniformity. Declarations Ethics approval and consent to participate Declarations Ethics approval and consent to participate This study was approved by the Shandong First Medical University Ethics Committee. Written informed consent was waived off due to the retrospective study. Consent for publication Not applicable. Availability of data and materials The datasets are available from the corresponding author during the current study. Competing interests The authors declare that they have no competing interests. Funding This work was supported by China National Key Research and Development (No. 2021YFE0204600), Science and Technology funding from Jinan (Grant number: 2020GXRC018). Author Contribution Yuteng Pan and Jianfeng Qiu concepted and designed the research. Liting Shi provised this study materials or patients. Yuan Liu collected the data. Yuteng Pan interpretated and analyzed the data. Jyh-cheng Chen and Jianfeng Qiu revised the manuscript; All authors approved the manuscript. References Baldini E, Tibaldi C, Delli Paoli C. Chemo-radiotherapy integration in unresectable locally advanced non-small-cell lung cancer: a review. Clin Transl Oncol. 2020;22(10):1681-6. Lemjabbar-Alaoui H, Hassan OU, Yang YW, Buchanan P. Lung cancer: Biology and treatment options. Biochim Biophys Acta. 2015;1856(2):189-210. Auperin A, Le Pechoux C, Rolland E, Curran WJ, Furuse K, Fournel P, et al. Meta-analysis of concomitant versus sequential radiochemotherapy in locally advanced non-small-cell lung cancer. J Clin Oncol. 2010;28(13):2181-90. Kim TE, Murren JR. Therapy for stage IIIB and stage IV non-small cell lung cancer. Clin Chest Med. 2002;23(1):209-24. Vinod SK, Hau E. Radiotherapy treatment for lung cancer: Current status and future directions. Respirology. 2020;25 Suppl 2:61-71. de Sousa VML, Carvalho L. Heterogeneity in Lung Cancer. Pathobiology. 2018;85(1-2):96-107. Arroyo-Hernandez M, Maldonado F, Lozano-Ruiz F, Munoz-Montano W, Nunez-Baez M, Arrieta O. Radiation-induced lung injury: current evidence. BMC Pulm Med. 2021;21(1):9. Rocco G, Morabito A, Leone A, Muto P, Fiore F, Budillon A. Management of non-small cell lung cancer in the era of personalized medicine. The International Journal of Biochemistry & Cell Biology. 2016;78:173-9. Yang W-C, Hsu F-M, Yang P-C. Precision radiotherapy for non-small cell lung cancer. Journal of Biomedical Science. 2020;27(1). Lambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout RG, Granton P, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer. 2012;48(4):441-6. Chetan MR, Gleeson FV. Radiomics in predicting treatment response in non-small-cell lung cancer: current status, challenges and future perspectives. European Radiology. 2020;31(2):1049-58. Chen N-B, Xiong M, Zhou R, Zhou Y, Qiu B, Luo Y-F, et al. CT radiomics-based long-term survival prediction for locally advanced non-small cell lung cancer patients treated with concurrent chemoradiotherapy using features from tumor and tumor organismal environment. Radiation Oncology. 2022;17(1). Khorrami M, Prasanna P, Gupta A, Patil P, Velu PD, Thawani R, et al. Changes in CT Radiomic Features Associated with Lymphocyte Distribution Predict Overall Survival and Response to Immunotherapy in Non–Small Cell Lung Cancer. Cancer Immunology Research. 2020;8(1):108-19. Webster JD, Dunstan RW. Whole-slide imaging and automated image analysis: considerations and opportunities in the practice of pathology. Vet Pathol. 2014;51(1):211-23. Pham HHN, Futakuchi M, Bychkov A, Furukawa T, Kuroda K, Fukuoka J. Detection of Lung Cancer Lymph Node Metastases from Whole-Slide Histopathologic Images Using a Two-Step Deep Learning Approach. Am J Pathol. 2019;189(12):2428-39. Yang H, Chen L, Cheng Z, Yang M, Wang J, Lin C, et al. Deep learning-based six-type classifier for lung cancer and mimics from histopathological whole slide images: a retrospective study. BMC Med. 2021;19(1):80. Coudray N, Ocampo PS, Sakellaropoulos T, Narula N, Snuderl M, Fenyo D, et al. Classification and mutation prediction from non-small cell lung cancer histopathology images using deep learning. Nat Med. 2018;24(10):1559-67. Zhang J, Wu Q, Yin W, Yang L, Xiao B, Wang J, et al. Development and validation of a radiopathomic model for predicting pathologic complete response to neoadjuvant chemotherapy in breast cancer patients. BMC Cancer. 2023;23(1):431. Feng L, Liu Z, Li C, Li Z, Lou X, Shao L, et al. Development and validation of a radiopathomics model to predict pathological complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer: a multicentre observational study. The Lancet Digital Health. 2022;4(1):e8-e17. Shao L, Liu Z, Feng L, Lou X, Li Z, Zhang XY, et al. Multiparametric MRI and Whole Slide Image-Based Pretreatment Prediction of Pathological Response to Neoadjuvant Chemoradiotherapy in Rectal Cancer: A Multicenter Radiopathomic Study. Ann Surg Oncol. 2020;27(11):4296-306. Niu L, Chu X, Yang X, Zhao H, Chen L, Deng F, et al. A multiomics approach-based prediction of radiation pneumonia in lung cancer patients: impact on survival outcome. Journal of Cancer Research and Clinical Oncology. 2023;149(11):8923-34. Yang S-R, Schultheis AM, Yu H, Mandelker D, Ladanyi M, Büttner R. Precision medicine in non-small cell lung cancer: Current applications and future directions. Seminars in Cancer Biology. 2022;84:184-98. Wang M, Herbst RS, Boshoff C. Toward personalized treatment approaches for non-small-cell lung cancer. Nature Medicine. 2021;27(8):1345-56. Zheng X, Liu K, Li C, Zhu C, Gao Y, Li J, et al. A CT-based radiomics nomogram for predicting the progression-free survival in small cell lung cancer: a multicenter cohort study. La radiologia medica. 2023;128(11):1386-97. Chae YK, Pan AP, Davis AA, Patel SP, Carneiro BA, Kurzrock R, et al. Path toward Precision Oncology: Review of Targeted Therapy Studies and Tools to Aid in Defining “Actionability” of a Molecular Lesion and Patient Management Support. Molecular Cancer Therapeutics. 2017;16(12):2645-55. Yang F, Zhang J, Zhou L, Xia W, Zhang R, Wei H, et al. CT-based radiomics signatures can predict the tumor response of non-small cell lung cancer patients treated with first-line chemotherapy and targeted therapy. European Radiology. 2021;32(3):1538-47. Chen Q, Shao J, Xue T, Peng H, Li M, Duan S, et al. Intratumoral and peritumoral radiomics nomograms for the preoperative prediction of lymphovascular invasion and overall survival in non-small cell lung cancer. European Radiology. 2022;33(2):947-58. Rossi L, Bijman R, Schillemans W, Aluwini S, Cavedon C, Witte M, et al. Texture analysis of 3D dose distributions for predictive modelling of toxicity rates in radiotherapy. Radiother Oncol. 2018;129(3):548-53. Zheng X, Guo W, Wang Y, Zhang J, Zhang Y, Cheng C, et al. Multi-omics to predict acute radiation esophagitis in patients with lung cancer treated with intensity-modulated radiation therapy. European Journal of Medical Research. 2023;28(1). Lee SH, Han P, Hales RK, Voong KR, Noro K, Sugiyama S, et al. Multi-view radiomics and dosiomics analysis with machine learning for predicting acute-phase weight loss in lung cancer patients treated with radiotherapy. Physics in Medicine & Biology. 2020;65(19). Wang B, Liu J, Zhang X, Wang Z, Cao Z, Lu L, et al. Prognostic value of (18)F-FDG PET/CT-based radiomics combining dosiomics and dose volume histogram for head and neck cancer. EJNMMI Res. 2023;13(1):14. Zhang Z, Wang Z, Yan M, Yu J, Dekker A, Zhao L, et al. Radiomics and Dosiomics Signature From Whole Lung Predicts Radiation Pneumonitis: A Model Development Study With Prospective External Validation and Decision-curve Analysis. Int J Radiat Oncol Biol Phys. 2023;115(3):746-58. Ding H, Feng Y, Huang X, Xu J, Zhang T, Liang Y, et al. Deep learning-based classification and spatial prognosis risk score on whole-slide images of lung adenocarcinoma. Histopathology. 2023;83(2):211-28. Yu KH, Zhang C, Berry GJ, Altman RB, Re C, Rubin DL, et al. Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun. 2016;7:12474. Ma Y, Li Q. An integrated model combined intra- and peritumoral regions for predicting chemoradiation response of non small cell lung cancers based on radiomics and deep learning. Cancer/Radiothérapie. 2023;27(8):705-11. Tunali I, Hall LO, Napel S, Cherezov D, Guvenis A, Gillies RJ, et al. Stability and reproducibility of computed tomography radiomic features extracted from peritumoral regions of lung cancer lesions. Medical Physics. 2019;46(11):5075-85. Sprave T, Rühle A, Stoian R, Weber A, Zamboglou C, Nieder C, et al. Radiotherapy for nonagenarians: the value of biological versus chronological age. Radiation Oncology. 2020;15(1). van Timmeren JE, van Elmpt W, Leijenaar RTH, Reymen B, Monshouwer R, Bussink J, et al. Longitudinal radiomics of cone-beam CT images from non-small cell lung cancer patients: Evaluation of the added prognostic value for overall survival and locoregional recurrence. Radiother Oncol. 2019;136:78-85. Liu W, Sun X, Qi Y, Jia X, Huang Y, Liu N, et al. Integrated texture parameter of 18F-FDG PET may be a stratification factor for the survival of nonoperative patients with locally advanced non-small-cell lung cancer. Nuclear Medicine Communications. 2018;39(8):732-40. Andersen MB, Harders SW, Thygesen J, Ganeshan B, Torp Madsen HH, Rasmussen F. Potential impact of texture analysis in contrast enhanced CT in non-small cell lung cancer as a marker of survival: A retrospective feasibility study. Medicine. 2022;101(48). De Courcy L, Bezak E, Marcu LG. Gender-dependent radiotherapy: The next step in personalised medicine? Critical Reviews in Oncology/Hematology. 2020;147. Jing Y, Mao Z, Zhu J, Ma X, Liu H, Chen F. TRAIP serves as a potential prognostic biomarker and correlates with immune infiltrates in lung adenocarcinoma. Int Immunopharmacol. 2023;122:110605. Lu D, Chen Y, Ke L, Wu W, Yuan L, Feng S, et al. Machine learning-assisted global DNA methylation fingerprint analysis for differentiating early-stage lung cancer from benign lung diseases. Biosens Bioelectron. 2023;235:115235. Donker HC, van Es B, Tamminga M, Lunter GA, van Kempen L, Schuuring E, et al. Using genomic scars to select immunotherapy beneficiaries in advanced non-small cell lung cancer. Sci Rep. 2023;13(1):6581. Collins LG, Haines C, Perkel R, Enck RE. Lung cancer: diagnosis and management. Am Fam Physician. 2007;75(1):56-63. Chen P, Rojas FR, Hu X, Serrano A, Zhu B, Chen H, et al. Pathomic Features Reveal Immune and Molecular Evolution From Lung Preneoplasia to Invasive Adenocarcinoma. Modern Pathology. 2023;36(12). Kang W, Qiu X, Luo Y, Luo J, Liu Y, Xi J, et al. Application of radiomics-based multiomics combinations in the tumor microenvironment and cancer prognosis. Journal of Translational Medicine. 2023;21(1). Additional Declarations No competing interests reported. Supplementary Files Addaitionalfile1.docx Additionalfile2.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4076424","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":278493007,"identity":"01717d67-a3ba-469b-b637-8b7da05aacd7","order_by":0,"name":"Yuteng Pan","email":"","orcid":"","institution":"Shandong First Medical University, Shandong Academy of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Yuteng","middleName":"","lastName":"Pan","suffix":""},{"id":278493009,"identity":"a76e14ca-71d6-4651-991f-bba7417c88e5","order_by":1,"name":"Liting Shi","email":"","orcid":"","institution":"Shandong First Medical University, Shandong Academy of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Liting","middleName":"","lastName":"Shi","suffix":""},{"id":278493010,"identity":"1ae87bb2-402f-438f-87d2-97021295799a","order_by":2,"name":"Yuan Liu","email":"","orcid":"","institution":"Shandong First Medical University, Shandong Academy of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Liu","suffix":""},{"id":278493011,"identity":"82e6bf1b-d7ef-4c21-b659-289cbf3f4387","order_by":3,"name":"Jyh-cheng Chen","email":"","orcid":"","institution":"National Yang-Ming Chiao-Tung University","correspondingAuthor":false,"prefix":"","firstName":"Jyh-cheng","middleName":"","lastName":"Chen","suffix":""},{"id":278493014,"identity":"76edcfe0-1048-4658-8b98-81a176c29a63","order_by":4,"name":"Jianfeng Qiu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzUlEQVRIiWNgGAWjYBACxuYDDAcYDCR4+EG8hAJitLQlALUU2MhJNoC0GBBjDVsCkPiQZmxwAMQjRgtzG/PGwzwGhxM3n1+d+OGBAYM8v9gBQg5jKwBr2Xbj7WYJoMMMZ85OIKBlfo8BVMvZDSAtCQa3CWlp44Fo2Tzj7OYfpGgBep+/dxuxtrAVHJxjYCMncYN3m0WCgQRhvxi2MW/+8OYPMCr7z26++aPCRp5fmpCWBlhcSIBVSuBXDgLy8OjjP0BY9SgYBaNgFIxMAADxNEdfmYp69gAAAABJRU5ErkJggg==","orcid":"","institution":"Shandong First Medical University, Shandong Academy of Medical Sciences","correspondingAuthor":true,"prefix":"","firstName":"Jianfeng","middleName":"","lastName":"Qiu","suffix":""}],"badges":[],"createdAt":"2024-03-11 16:02:01","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4076424/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4076424/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":52623567,"identity":"753dac78-715a-40d5-b341-5a1e69eabd67","added_by":"auto","created_at":"2024-03-13 17:20:50","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":990746,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe process of patient inclusion and exclusion.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure1..jpg","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/e0d65f34cfad9d9f19504d12.jpg"},{"id":52623644,"identity":"7c3c141b-b9fe-4cb0-8fd9-d6d6923ffb45","added_by":"auto","created_at":"2024-03-13 17:20:57","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1100595,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe workflow of this study.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure2..jpg","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/00f87b0e17839ba4fc657d56.jpg"},{"id":52623588,"identity":"ca8008db-0bea-4752-9a9d-556ac1d0b64e","added_by":"auto","created_at":"2024-03-13 17:20:53","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1439679,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eVisualization of tiles classification and evaluation of treatment response prediction models.\u003c/strong\u003e(A)The visualization of tiles classification model which classified all tiles into five categories: tumor, stroma, lymphocytes, blurry and red cell. (B) The ROC curves of radiomics, dosiomics, pathomics, clinical and combined models in the internal validation set. (C) The ROC curves of radiomics, dosiomics, pathomics, clinical and combined models in the external validation set. \u003cem\u003eAbbreviations: \u003c/em\u003eR: responsive; NR, non-responsive; EV, external validation set; IV, internal validation set; Train, training set.\u003c/p\u003e","description":"","filename":"Figure3..jpg","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/69417e62c2d925975b293bba.jpg"},{"id":52623642,"identity":"6e24d017-33f2-41f9-b754-a7667e393ead","added_by":"auto","created_at":"2024-03-13 17:20:57","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":586335,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe shapley values of different features in treatment response and OS models.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure4..jpg","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/aa571b57ee917b73fb5ee885.jpg"},{"id":53106337,"identity":"d8430736-f8c9-4481-ae70-bc3d47e6df18","added_by":"auto","created_at":"2024-03-20 16:21:28","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":808378,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/57e33f5c-f789-40ec-a480-42904949a695.pdf"},{"id":52623570,"identity":"a0655faa-9d18-42db-b054-07824a98b2b8","added_by":"auto","created_at":"2024-03-13 17:20:51","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":46961,"visible":true,"origin":"","legend":"","description":"","filename":"Addaitionalfile1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/5ceaa7f825508c21bc2a73e0.docx"},{"id":52623569,"identity":"8b259c33-f89e-48b9-bcee-7d34ea098d75","added_by":"auto","created_at":"2024-03-13 17:20:51","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":538870,"visible":true,"origin":"","legend":"","description":"","filename":"Additionalfile2.docx","url":"https://assets-eu.researchsquare.com/files/rs-4076424/v1/d5904b41950a33ca4eadb18d.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multi-omics models predict treatment response and overall survival for non-small cell lung cancer patients following chemo-radiotherapy: A multi-center study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eChemo-radiotherapy (CRT) is one of the standard treatment methods for advanced unresectable non-small cell lung cancer (NSCLC) patients (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). It is anticipated to enhance tumor control probability in NSCLC patients and significantly extend their overall survival (OS) (\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). However, varying CRT outcomes in individuals arose from the intricate physical conditions and tumor heterogeneity characteristic of NSCLC patients (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). Therefore, it is particularly crucial to quantify the heterogeneity of tumors, predict the treatment outcomes and identify those who are most likely to benefit from CRT by pretreatment information, such as CT images, dose images, whole slide images etc (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRadiomics, extraction high-throughput quantitative features from medical images to quantify tumor heterogeneity, analyzed through predetermined algorithms, served as a tool for developing models in clinical decision support (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Previous research had explored radiomics models to predict treatment response or survival in NSCLC patients (\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Dosiomics, an extension of radiomics, incorporated three-dimensional dose distribution information from treatment plans as an image for feature extraction (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Pathomics is extraction quantitative features based on whole slide images (WSI) to quantify heterogeneity of the tumor tissue(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Different from radiomics and dosiomics which represented a macro perspective, pathomics explored the heterogeneity of tumor microenvironment from a micro perspective, which was expected to comprehensively explain the heterogeneity of tumors and the differences in individual treatments, thereby achieving more accurate predictions (\u003cspan additionalcitationids=\"CR16\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Multi-omics models, bridging macroscopic and microscopic perspectives, have shown promise in discovering multi-dimensional biomarkers, offering stronger biological interpretability and predictive capabilities. For instance, some studies developed a radiopathomic model for predicting pathologic complete response (\u003cspan additionalcitationids=\"CR19\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). Zhang et al. established multiomics models to predict the occurrence of radiation pneumonia by combining radiomics, dosiomics, dose volume histograms (DVH), and clinical information (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). However, Jing et al. established a multiomics model based on radiomics and dosiomics to predict the occurrence of radiation pneumonia and explore its impact on 1-year OS in NSCLC patients (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). Multi-omics models obtained better biological interpretability and predictability compared to single-modality models. The models with multidimensional information contributed to a more precise assessment of the complex conditions of patients, guiding the clinical doctors develop individualized diagnosis and treatment plans. Precision individual treatment for patients might reduce treatment-related side effects, enhance treatment safety, and improve tumors control probability, prolong the survival of NSCLC patients(\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study focuses on constructing multi-omics models based on computed tomography (CT) images, WSI, three-dimensional dose distribution images and clinical information before treatment to predict objective response and 1-year OS in NSCLC patients undergoing CRT. The anticipated outcome is the accurate stratification of patients, enabling clinicians to develop personalized treatment modality and enhance the overall treatment benefits for patients.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatients and clinical information\u003c/h2\u003e \u003cp\u003e This retrospective study involving 220 patients in three hospitals from 2021 to 2023 and was approved by their ethics committees according to the Declaration of Helsinki. A total of 158 patients were enrolled from Shandong Provincial Hospital (center 1), 21 patients from the Shandong First Medical University Affiliated Hospital (center 2) and 41 patients from the Xiangya Hospital of Central South University (center 3). The informed consent was waived because this study employed a retrospective design.\u003c/p\u003e \u003cp\u003eThe criteria for included patients were as delineated: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Patients were over 18 years old. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Patients were diagnosed with primary NSCLC. (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Patients received treatment for the first time after diagnosis. (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) Patients received intensity modulated radiotherapy (IMRT) and were administered radiotherapy at 2 Gy per day, 5 days a week, with 20\u0026ndash;35 fractions and with a total dose of 40\u0026ndash;70 Gy. (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) Patients had radiation planned CT images with volumetric doses and hematoxylin eosin (HE) stained sections of lung pathological biopsy. The criteria for excluded patients were as delineated: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Patients underwent surgical resection of lung cancer tumor. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Patients with poor quality CT images and poor quality HE-stained sections. In addition to the above standards, 142 out of 220 patients were retained for evaluating treatment response. The included criteria for those patients were as delineated: patients had follow-up CT scans within 3 months after the end of the same treatment course. The process of patient inclusion and exclusion was presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Patients were followed up every three months in the first years after diagnosis, and every year thereafter. The basic clinical characteristics of all patients were collected, including gender, age, pathological types, clinical T stage (cT), clinical N stage (cN), clinical M stage (cM), clinical stage and total dose.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eImaging acquisition and preprocessing\u003c/h2\u003e \u003cp\u003eThe pretreatment CT scanning parameters were detailed in Additional file1: Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. Two experienced physicists from each center outlined four regions of interest (ROI) for subsequent analysis, including the gross tumor volume (GTV), the planning tumor volume (PTV), total lung volume excluded GTV (LUNG-GTV), and total lung volume excluded PTV (LUNG-PTV). The ROIs of the dose images mirrored those of the CT images.\u003c/p\u003e \u003cp\u003eThe HE-stained biopsy sections parameters were detailed in Additional file1: Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e. Firstly, WSI were turned into binary images as RGB images by a proper threshold based on the different depths of the HE staining in the three centers. Subsequently, tiles with tissue larger than 50% were retained after non-overlapping cutting of binary images into tiles with 224\u0026times;224 pixels. Color normalization was applied to all other tiles based on a well-stained template tile to enhance the staining quality of multicenter images (method = 'vahadane').\u003c/p\u003e \u003cp\u003eDelineation of different tissue types, including tumor cell, stroma, lymphocyte, red blood cell, and blur area, was performed by three experienced pathologists on some WSI from the three centers. The delineated areas were cut into tiles with 224\u0026times;224 pixels, and the same template tile was used for color normalization (method = 'vahadane'). The normalized tiles were divided into training, validation, and testing sets in an 8:1:1 ratio. The Resnet 152 network was employed for training a tissue classification model capable of classifying all tiles. Tiles classified as tumor cell, stroma, and lymphocyte were reserved as tumor microenvironment tiles for subsequent feature extraction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eFeature extraction and selection\u003c/h2\u003e \u003cp\u003eRadiomics and dosiomics features were extracted by pyradiomics. Before extracting those features, the CT images and dose image of patients were resampled and standardized to 1mm\u0026times;1mm\u0026times;5mm pixels. The radiomics and dosiomics features were divided into two groups: without preprocessing and after preprocessing by laplacian of gaussian (log) filters (sigma\u0026thinsp;=\u0026thinsp;3mm). Besides calculating the minimum dose, average dose, and maximum dose of GTV, PTV and whole lung, we also included V5, V10, V15, V20, V25, V30, V35, V40, V45 and V50 of the whole lung in the analysis based on DVH. The features from DVH were classified into the clinical features for further analysis.\u003c/p\u003e \u003cp\u003eThe steps for feature selection were as follows: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Standardized the features of each patient by Z-score normalization to ensure comparability between data. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) For predicting treatment response, a total of 113 patients (92 from center 1 and 21 from center 2) were randomly split into a training set (n\u0026thinsp;=\u0026thinsp;80) and an internal validation set (n\u0026thinsp;=\u0026thinsp;33) using a 7:3 ratio with 300 seed points for random partitioning. Additionally, 29 patients from center 3 were designated as an external validation set. Single modality features, encompassing radiomics, pathomics, dosiomics, and clinical features, were selected using the least absolute shrinkage and selection operator (Lasso) regression, random forest, and extreme gradient boosting based on the same training set. (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) For predicting 1-year OS, a total of 179 patients (158 from center 1 and 21 from center 2) were randomly split into a training set (n\u0026thinsp;=\u0026thinsp;126) and an internal validation set (n\u0026thinsp;=\u0026thinsp;53) using a 7:3 ratio with 300 seed points for random partitioning. Additionally, 41 patients from center 3 were designated as an external validation set. Lasso cox regression, random survival forest, and extreme gradient boosting were also used to select single modality features, including radiomics, pathomics, and dosiomics, based on the same training set. Clinical features were selected through univariate and multivariate cox proportional hazards regression (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.1)(\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). The selected features were demonstrated to qualify their contribution for the treatment response and OS prediction model by shapley values.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eChemoradiotherapy response and overall survival prediction\u003c/h2\u003e \u003cp\u003eAccording to the Response Evaluation Criteria in Solid Tumors 1.1 (RECIST 1.1) (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e), treatment response was divided into four levels. Complete response (CR) and partial response (PR) were considered as objective response (OR) groups, while patients with stable disease (SD) and progressive disease (PD) were considered as non-OR groups (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). For the prediction of treatment response, single omics model and multi-omics model were established by support vector machine (SVM) based on the selected features from different modalities. The prediction ability of treatment response was evaluated by the area under the receiver operating characteristic (ROC) curves and boxplot.\u003c/p\u003e \u003cp\u003eOS was defined from the start of the initial antitumor treatment until death from any cause during follow-up. For the prediction of 1-year OS, single omics model and multi-omics model were established by cox proportional hazard based on the selected features from different modalities. The following comprehensive evaluation methods were employed: the area under the ROC curves and concordance index (C-index); the survival curves of the high-risk and low-risk group which evaluated by the Kaplan-Meier (KM) method (cutoff\u0026thinsp;=\u0026thinsp;the median value of the predicted value), the differences between the survival curves were tested by the log-rank test; Calibration curves were calculated to evaluate the consistency between the predicted results and recorded survival results. The flowchart of survival model construction was presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eThe feature extraction of radiomics and dosiomics, dose calculation and DVH production were implemented in 3D-Slicer (Version 4.11, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.slicer.org/\u003c/span\u003e\u003cspan address=\"https://www.slicer.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The Kruskal\u0026ndash;Wall test for analyzing differences in multiple data sets. T-test were used to analyze the differences between two data sets (satisfying normal distribution and homogeneity of variance). A two sides \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant. Features selection, model\u0026rsquo;s construction and all statistical analyses were performed by R studio (Version 3.4.0, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.r-\u003c/span\u003e\u003cspan address=\"https://www.r-\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e project. org/).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cem\u003ePatients and clinical information\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTotal 220 patients were recruited in this study from three hospitals (ceter1=158, ceter2=21, center3=48). All the recruited patients were used to develop and validate the OS prediction models. 142 out of 220 patients were retained for constructing treatment response predicting models (center1=, center2=, center3=29). Table 1 presented the basic clinical characteristics of all patients in the training set, internal validation set and external validation. As for characteristics of patients used for predicting treatment response are shown in the Additional file1: Table S3. There were no statistically significant differences were observed in three sets, except for cM and clinical stage. Dose factors in treatment response cohort and overall survival cohort, such as minimum dose, average dose and maximum dose of GTV, PTV, LUNG and V5, V10, V15, V20, V25, V30, V35, V40, V45 and V50 of the whole lung were detailed in Additional file1: Table S4-S5, respectively.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1. Baseline characteristics of all patients.\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"567\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eCharacteristics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003eTraining\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eset\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003eInternal\u0026nbsp;\u003c/p\u003e\n \u003cp\u003evalidation set\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003eExternal\u003c/p\u003e\n \u003cp\u003evalidation set\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eNo. patients\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e126\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eAge(years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e65(41-87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e65(44-84)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e61(40-78)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.202\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.420\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Male\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e101(80.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e41(77.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e36(87.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e25(19.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e12(22.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e5(12.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003ePathological types\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.147\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eSquamous\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e61(48.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e31(58.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e25(61.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eAdenocarcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e65(51.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e22(41.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e15(36.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e0(0.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e0(0.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e1(2.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eClinical T stage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.194\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;T1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e22(17.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e6(11.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e7(17.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eT2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e33(26.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e19(35.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e20(48.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eT3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e42(33.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e14(26.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e7(17.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eT4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e29(23.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e14(26.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e7(17.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eClinical N stage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.150\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eN0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e13(10.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e4(7.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e11(26.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eN1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e10(7.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e6(11.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e5(12.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eN2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e58(46.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e29(54.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e14(34.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eN3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e45(35.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e14(26.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e11(26.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eClinical M stage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.030\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eM0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e89(71.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e48(90.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e33(80.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eM1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e37(28.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e5(9.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e8(19.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eClinical stage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eIIIA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e35(27.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e25(47.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e18(43.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eIIIB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e40(31.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e20(37.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e13(31.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eIIIC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e14(11.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e3(5.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e2(4.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eIV\u003c/p\u003e\n \u003cp\u003eTotal dose (Gy)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u0026lt;45\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;45-55\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;55-65\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u0026gt;65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e37(29.4%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e5(3.9%)\u003c/p\u003e\n \u003cp\u003e8(6.3%)\u003c/p\u003e\n \u003cp\u003e96(76.2%)\u003c/p\u003e\n \u003cp\u003e17(13.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e5(9.4%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3(5.6%)\u003c/p\u003e\n \u003cp\u003e9(17.0%)\u003c/p\u003e\n \u003cp\u003e38(71.7%)\u003c/p\u003e\n \u003cp\u003e3(5.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e8(19.5%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e1(2.4%)\u003c/p\u003e\n \u003cp\u003e6(14.6%)\u003c/p\u003e\n \u003cp\u003e32(78.0%)\u003c/p\u003e\n \u003cp\u003e2(5.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.060\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"30.035335689045937%\" valign=\"top\"\u003e\n \u003cp\u003eOverall survival(days)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"18.374558303886925%\" valign=\"top\"\u003e\n \u003cp\u003e848(24-2526)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e951(226-2470)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.964664310954063%\" valign=\"top\"\u003e\n \u003cp\u003e913(111-2960)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"11.66077738515901%\" valign=\"top\"\u003e\n \u003cp\u003e0.485\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eFeature extraction and selection\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eA total of 720 radiomics features were extracted from four ROIs by Pyradiomics, with 180 features extracted from each ROI, including 36 first-order features and 144 texture features. The texture features were calculated by using Gray Level Cooccurrence Matrix (GLCM), Gray Level Dependence Matrix (GLDM), Gray Level Run Length Matrix (GLRLM), Gray Level Size Zone Matrix (GLSZM) and Neighborhood Gray-tone Difference Matrix (NGTDM).\u003c/p\u003e\n\u003cp\u003eAs the Additional file1: Table S6 shown, it provided the detail tiles amounts about the tissue classification model\u0026rsquo;s establishment. The model achieved accuracies of 0.93, 0.91, and 0.92 in the training set, validation set, and test set, respectively. As Figure 3A shown, the classified tiles including tumor cell, stroma, and lymphocyte were reserved as tumor microenvironment tiles. The number of tumor microenvironment tiles were available in Additional file1: Table S7. A total of 512 pathological deep learning features of each tile were extracted by Resnet34 with pre-trained weights in ImageNet (https://www.image-net.org/) as the backbone network. The maximum value of all tiles for each patient were calculated as the patient-level feature.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eChemoradiotherapy response and OS prediction\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe AUC of feature selection were shown in Additional file1: Table S8 for treatment response prediction in the training set, internal validation set and external validation set. The selected features for predicting treatment response comprised one radiomics feature, one pathomics feature, two dosiomics features, and one clinical indicator as shown in Additional file1: Table S9. Specifically, the chosen features included: a first-order radiomics feature extracted from PTV (LFM\u003csub\u003ePTV\u003c/sub\u003e), a pathomics feature identified as deep learning feature NO.134 (DF134), dosiomics features encompassing a first-order feature extracted from GTV (OFR\u003csub\u003eGTV\u003c/sub\u003e) and a texture feature extracted from LUNG-GTV (LGZ\u003csub\u003eLUNG-GTV\u003c/sub\u003e), along with the clinical indicator, age. The performance of the single omics models and mulit-omics model were displayed in Table 2 and Figure 3. The performance of other models generated by combining various modality prediction factors was shown in Additional file1: Table S10. As shown in Table 2, the area under the curve (AUC) of the multi-omics model was consistently higher than that of the single omics model in the training, internal validation, and external validation sets, surpassing 0.8. The AUC of the multi-omics model in the training set was 0.85, with an internal validation set of 0.81 and an external validation set of 0.87 (Additional file2:figure S1). Through the boxplot, it can also be seen that the combined model effectively distinguished between responsive and non-responsive groups.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2.\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ePerformance of single omics and multiomics model in treatment response prediction.\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"626\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" colspan=\"5\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTreatment response prediction\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003eAUC (95%CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003eAccuracy (95%CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003eSensitivity (95%CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003eSpecificity (95%CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTraining set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Radiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Pathomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Dosiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Clinical\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Combined\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.85(0.75-0.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.80(0.70-0.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.85(0.69-0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.76(0.60-0.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eIV set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Radiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Pathomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Dosiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Clinical\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Combined\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.81(0.67-0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.79(0.61-0.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.79(0.54-0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.79(0.54-0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eEV set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Radiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Pathomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Dosiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Clinical\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"15.151515151515152%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Combined\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.29824561403509%\" valign=\"top\"\u003e\n \u003cp\u003e0.87(0.74-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.05263157894737%\" valign=\"top\"\u003e\n \u003cp\u003e0.79(0.60-0.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.169059011164276%\" valign=\"top\"\u003e\n \u003cp\u003e0.67(0.70-0.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"22.328548644338117%\" valign=\"top\"\u003e\n \u003cp\u003e0.85(0.62-0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviation: IV, Internal validation; EV, External validation; CI, Confidence interval;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe AUC of feature selection were shown in Additional file1: Table S11 for OS prediction in the training set, internal validation set and external validation set. The five most effective features for predicting 1-year OS were selected, including one radiomics feature, one pathomics feature, two dosiomics features, and one clinical indicator. The information of the selected features are as follows: the radiomics feature was a texture feature extracted from GTV (OGG\u003csub\u003eGTV\u003c/sub\u003e), the pathomics feature was deep learning feature NO.166 (DF166), the dosiomics features included two texture features extracted from GTV (OGZ\u003csub\u003eGTV\u003c/sub\u003e and LGR\u003csub\u003eGTV\u003c/sub\u003e), and the clinical indicator was gender (Additional file1: Table S12).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThe C-index of feature selection for 1-year OS prediction and the details of selected features for 1-year OS prediction in the training set, internal validation set, external validation set were shown in Additional file1: Table S13-14, respectively. The performance of the single-omics models, three modalities (clinical, dosiomics and pathomics) model and final multi-omics model were displayed in Table3. The performance of other models generated by combining various modality prediction factors was shown in Additional file1: Table S15-16. Ultimately, two combined models that demonstrated excellent performance were retained. The AUC and C-index of three-modalities (clinical, dosiomics and pathomics) model and finally multi-omics model were higher than the single-omics models. The AUC and C-index of the three-modalities (clinical, dosiomics and pathomics) model in the training set, internal validation set and external validation set were 0.83/0.79, 0.73/0.73 and 0.78/0.78, respectively. The AUC and C-index of the all-modalities combined model in the training set, internal validation set and external validation set were 0.83/0.79, 0.74/0.74 and 0.73/0.72, respectively. The performance of the three-modalities model in the external validation group was slightly better than that of the multi-omics model. The KM curves revealed that the two combined models had good ability to distinguish between high-risk and low-risk groups (Additional file2: Figure S2). The calibration plots demonstrated great agreement between combined models\u0026rsquo; prediction and the actual observation for survival (Additional file2: Figure S3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable3. Performance of single omics model and multiomics model in OS prediction.\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"626\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" colspan=\"3\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eOS prediction\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25.76%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"37.76%\" valign=\"top\"\u003e\n \u003cp\u003eAUC (95%CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"36.48%\" valign=\"top\"\u003e\n \u003cp\u003eC-index (95%CI)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"25.76%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTraining set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"37.76%\" rowspan=\"18\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003cp\u003e0.83(0.73-0.91)\u003c/p\u003e\n \u003cp\u003e0.83(0.74-0.91)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003cp\u003e0.73(0.64-0.92)\u003c/p\u003e\n \u003cp\u003e0.74(0.55-0.92)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003cp\u003e0.77(0.64-0.91)\u003c/p\u003e\n \u003cp\u003e0.73(0.58-0.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"36.48%\" rowspan=\"18\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003cp\u003e0.79(0.70-0.87)\u003c/p\u003e\n \u003cp\u003e0.80(0.71-0.87)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003cp\u003e0.73(0.65-0.93)\u003c/p\u003e\n \u003cp\u003e0.74(0.52-0.98)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003cp\u003e0.78(0.62-0.92)\u003c/p\u003e\n \u003cp\u003e0.72(0.62-0.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Radiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Pathomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Dosiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Clinical\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;C+D+P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Combined\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eIV set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Radiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Pathomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Dosiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Clinical\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;C+D+P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Combined\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eEV set\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Radiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Pathomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Dosiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Clinical\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;C+D+P\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;Combined\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviation: IV, Internal validation; EV, External validation; CI, Confidence interval; OS, Overall survival.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003ePrecise stratification of patients through chemo-radiotherapy response and OS before treatment is a pivotal step in clinical precision therapy. The implementation of precision therapy has significantly improved the treatment outcome of NSCLC\u0026nbsp;(25). Advanced and non-invasive models hold the potential to predict treatment endpoints, aiding doctors in crafting personalized treatment plans and refining patient prognoses. In this study, we developed and validated multi-omics models with superior performance and clear biological interpretability based on the CT images, whole slide images, dose images and clinical information to predict treatment response and 1-year OS, outperforming single omics models.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCurrently, many studies focused on identifying biomarkers for predicting treatment response or survival of lung cancer patients by single omics. Yang et al. attempted to establish a radiomics model to differentiate responsive and non-responsive NSCLC patients undergoing first-line chemotherapy and targeted therapy, but the model\u0026rsquo;s performance of AUC values was only 0.74 and was limited to the single center model(26). Chen et al. innovatively developed an OS prediction model for NSCLC patients by combining intratumoral and peritumoral radiomics features, but the AUC of the combined model in the external validation set was also less than 0.7\u0026nbsp;(27). Dosiomics as an extended concept of radiomics, contained more three-dimensional dose distribution information(28). Some researchers had investigated radiotherapy side effects and survival in lung cancer patients by integrating radiomics and dosiomics information\u0026nbsp;(21, 29, 30). However, whether dosomics can predict the chemoradiotherapy response and OS of NSCLC patients remains to be studied(31, 32). In other perspective, radiomics and dosiomics utilized high-throughput data to characterize tumor heterogeneity from a macro perspective. While pathomics explored individual differences at the cellular and tissue levels within the tumor microenvironment. Some investigation had established pathomics models based on WSI to predict the prognosis of NSCLC\u0026nbsp;(33, 34). Therefore, integrating multi perspective image information from both macro and micro perspectives was expected to yield models with better biological interpretability and stronger predictive capabilities.\u003c/p\u003e\n\u003cp\u003eThe selected one radiomics feature were extracted from PTV and the selected two dosiomics features were extracted from GTV and LUNG-GTV in treatment response prediction model construction. This indicated that not only did the dose distribution of GTV affected the treatment response of patients, but also the feature intensity and dose distribution of peritumoral and surrounding normal lung tissue affected the treatment response(35, 36). Age is also selected to predict treatment response, which is consistent with the views of Sprave et al(37). The AUC of the multi-omics in the training set was 0.85, with an internal validation set of 0.81 and an external validation set of 0.87. Notably, multi-omics model exhibited increased stability when compared to radiomics models in previous studies, showing robust predictive performance even in independent external validation sets(38). Furthermore, the statistical test results illustrated in the box plot confirmed significant differences in the predicted values of the combined model between the responsive and non-responsive groups, effectively distinguishing the two cohorts. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe selected five radiomics and dosiomics features were all texture features extracted from GTV in OS prediction model construction. It is suggested that the heterogeneity of intra-tumoral texture may be an important factor affect the survival stratification of advanced NSCLC patients\u0026nbsp;(39, 40). Besides, gender had also been found as potential predictive factors, as they might be affecting the 1 year overall survival of CRT in patients due to differences in radiosensitivity between individuals\u0026nbsp;(41). However, pathomics features were obtained from deep features of tiles without interpretability. In the future, the correlation between pathomics features and features from other modalities will be explored for enhancing the interpretability of the models. For 1-year overall survival prediction, gender was used to construct in both two combined models with better predictive ability. The AUC and C-index of the three modalities (clinical, dosiomics, and pathomics) and the all-modalities combined model surpassed 0.7 in the training set, internal validation set, and external validation set. The KM curves indicated effective differentiation between high-risk and low-risk groups. The calibration curves showcased minimal deviation between model predictions and actual values.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhile relying on patient clinical information, such as clinical TNM stages, could offer rough predictions of patient prognosis, models based on selected multimodal biomarkers demonstrated superior predictive performance. In addition, compared to the research on establishing prediction models for NSCLC patients based on genetic information\u0026nbsp;(42-44), the multimodal information we included for analysis was more readily obtainable in clinical settings. The prediction cost of genes was relatively high, but CT imaging and pathological biopsy were necessary examination items with lower cost in the diagnosis and treatment process of patients\u0026nbsp;(45). Easier to obtain biomarkers and more stable predictive models will provide precise references for patient survival, which will help clinical doctors develop personalized diagnosis and treatment plans, thereby improving patient prognosis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditionally, our study had other limitations: only 220 NSCLC patients from three centers were included in the study and we plan to continue collecting data to explore information in large sample data; Besides, in addition to prefusion and post-fusion methods, it was also necessary to explore more new feature fusion methods. It was worth mentioning that in our study, we retained two models for predicting 1-year OS. The performance of the three-modalities model based on clinical, dosiomics and pathomics features in the external validation set was slightly better than muliti-omics model, which indicated that sometimes multimodality information may be redundant, and prompted consideration of improved fusion methods. Finally, the clinical basic indicators of the patients we had analysis were not comprehensive enough. In future research, more clinical indicators will be collected, such as hematological examination indicators, patient smoking and alcohol consumption status, to provide a more holistic assessment of their physics condition.\u003c/p\u003e\n\u003cp\u003eHere multi-omics models were constructed by biomarkers which extracted from pretreatment CT images, WSI, dose information and clinical factors, which were easily accessible before therapy. This study captured macroscopic tumor characteristics, microscopic tumor ecosystem manifestations, and considered the impact of dose distribution on short-term and long-term treatment outcomes for each patient(46). The integration of multidimensional information will assist patients in establishing more comprehensive personal profiles, enabling the identification of more suitable treatment methods within complex therapeutic regimens. Additionally, the fusion of multi-modalities information contributed to the establishment of models with more precise predictive performance (47).\u0026nbsp;\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this study, multi-omics models were established based on the CT images, whole slide image and dose information for predicting radio-chemotherapy response and 1-year OS of NSCLC patients. These models have the potential to assist doctors optimize patients\u0026rsquo; CRT plans and supporting the precise treatment for patients by predicting the short-term treatment response and long-term survival of patients.\u0026nbsp;\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eCRT, chemoradiotherapy; NSCLC, non-small cell lung cancer; OS, overall survival; WSI, whole slide images; DVH, dose volume histograms; IMRT, intensity modulated radiotherapy ; CT, computed tomography; HE, hematoxylin eosin ; ROI, region of interest; GTV, gross tumor volume; PTV, planning tumor volume; LUNG-GTV, total lung volume excluded GTV; LUNG-PTV, total lung volume excluded PTV; Lasso, least absolute shrinkage and selection operator; Log, laplacian of gaussian (log); CR, Complete response; PR, partial response; OR, objective response; SD, stable disease; PD, progressive disease; SVM, support vector machine; ROC, receiver operating characteristic; C-index, concordance index; KM, Kaplan-Meier; GLCM, Gray Level Cooccurrence Matrix; GLDM, Gray Level Dependence Matrix; GLRLM, Gray Level Run Length Matrix; GLSZM, Gray Level Size Zone Matrix; NGTDM, Neighborhood Gray-tone Difference Matrix; AUC, area under the curve ; LFM, Log-sigma-3-0-mm-3D-firstorder-maximum; DF, Deep learning feature; OFR, Original-firstorder-robust-mean-absolute-deviation; LGZ, Log-sigma-3-0-mm-3D-glszm-zone-entropy; OGG, Original-glrlm- gray-level-nonuniformity; OGZ, Original-glszm-zone-variance; LGR, Log-sigma-3-0-mm-3D- glrlm- run-length-nonuniformity.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDeclarations Ethics approval and consent to participate This study was approved by the Shandong First Medical University Ethics Committee. Written informed consent was waived off due to the retrospective study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets are available from the corresponding author during the current study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by China National Key Research and Development (No. 2021YFE0204600), Science and Technology funding from Jinan (Grant number: 2020GXRC018).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eYuteng Pan and Jianfeng Qiu concepted and designed the research. Liting Shi provised this study materials or patients. Yuan Liu collected the data. Yuteng Pan interpretated and analyzed the data. Jyh-cheng Chen and Jianfeng Qiu revised the manuscript; All authors approved the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBaldini E, Tibaldi C, Delli Paoli C. Chemo-radiotherapy integration in unresectable locally advanced non-small-cell lung cancer: a review. Clin Transl Oncol. 2020;22(10):1681-6.\u003c/li\u003e\n\u003cli\u003eLemjabbar-Alaoui H, Hassan OU, Yang YW, Buchanan P. Lung cancer: Biology and treatment options. Biochim Biophys Acta. 2015;1856(2):189-210.\u003c/li\u003e\n\u003cli\u003eAuperin A, Le Pechoux C, Rolland E, Curran WJ, Furuse K, Fournel P, et al. Meta-analysis of concomitant versus sequential radiochemotherapy in locally advanced non-small-cell lung cancer. J Clin Oncol. 2010;28(13):2181-90.\u003c/li\u003e\n\u003cli\u003eKim TE, Murren JR. Therapy for stage IIIB and stage IV non-small cell lung cancer. Clin Chest Med. 2002;23(1):209-24.\u003c/li\u003e\n\u003cli\u003eVinod SK, Hau E. Radiotherapy treatment for lung cancer: Current status and future directions. Respirology. 2020;25 Suppl 2:61-71.\u003c/li\u003e\n\u003cli\u003ede Sousa VML, Carvalho L. Heterogeneity in Lung Cancer. Pathobiology. 2018;85(1-2):96-107.\u003c/li\u003e\n\u003cli\u003eArroyo-Hernandez M, Maldonado F, Lozano-Ruiz F, Munoz-Montano W, Nunez-Baez M, Arrieta O. Radiation-induced lung injury: current evidence. BMC Pulm Med. 2021;21(1):9.\u003c/li\u003e\n\u003cli\u003eRocco G, Morabito A, Leone A, Muto P, Fiore F, Budillon A. Management of non-small cell lung cancer in the era of personalized medicine. The International Journal of Biochemistry \u0026amp; Cell Biology. 2016;78:173-9.\u003c/li\u003e\n\u003cli\u003eYang W-C, Hsu F-M, Yang P-C. Precision radiotherapy for non-small cell lung cancer. Journal of Biomedical Science. 2020;27(1).\u003c/li\u003e\n\u003cli\u003eLambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout RG, Granton P, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer. 2012;48(4):441-6.\u003c/li\u003e\n\u003cli\u003eChetan MR, Gleeson FV. Radiomics in predicting treatment response in non-small-cell lung cancer: current status, challenges and future perspectives. European Radiology. 2020;31(2):1049-58.\u003c/li\u003e\n\u003cli\u003eChen N-B, Xiong M, Zhou R, Zhou Y, Qiu B, Luo Y-F, et al. CT radiomics-based long-term survival prediction for locally advanced non-small cell lung cancer patients treated with concurrent chemoradiotherapy using features from tumor and tumor organismal environment. Radiation Oncology. 2022;17(1).\u003c/li\u003e\n\u003cli\u003eKhorrami M, Prasanna P, Gupta A, Patil P, Velu PD, Thawani R, et al. Changes in CT Radiomic Features Associated with Lymphocyte Distribution Predict Overall Survival and Response to Immunotherapy in Non\u0026ndash;Small Cell Lung Cancer. Cancer Immunology Research. 2020;8(1):108-19.\u003c/li\u003e\n\u003cli\u003eWebster JD, Dunstan RW. Whole-slide imaging and automated image analysis: considerations and opportunities in the practice of pathology. Vet Pathol. 2014;51(1):211-23.\u003c/li\u003e\n\u003cli\u003ePham HHN, Futakuchi M, Bychkov A, Furukawa T, Kuroda K, Fukuoka J. Detection of Lung Cancer Lymph Node Metastases from Whole-Slide Histopathologic Images Using a Two-Step Deep Learning Approach. Am J Pathol. 2019;189(12):2428-39.\u003c/li\u003e\n\u003cli\u003eYang H, Chen L, Cheng Z, Yang M, Wang J, Lin C, et al. Deep learning-based six-type classifier for lung cancer and mimics from histopathological whole slide images: a retrospective study. BMC Med. 2021;19(1):80.\u003c/li\u003e\n\u003cli\u003eCoudray N, Ocampo PS, Sakellaropoulos T, Narula N, Snuderl M, Fenyo D, et al. Classification and mutation prediction from non-small cell lung cancer histopathology images using deep learning. Nat Med. 2018;24(10):1559-67.\u003c/li\u003e\n\u003cli\u003eZhang J, Wu Q, Yin W, Yang L, Xiao B, Wang J, et al. Development and validation of a radiopathomic model for predicting pathologic complete response to neoadjuvant chemotherapy in breast cancer patients. BMC Cancer. 2023;23(1):431.\u003c/li\u003e\n\u003cli\u003eFeng L, Liu Z, Li C, Li Z, Lou X, Shao L, et al. Development and validation of a radiopathomics model to predict pathological complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer: a multicentre observational study. The Lancet Digital Health. 2022;4(1):e8-e17.\u003c/li\u003e\n\u003cli\u003eShao L, Liu Z, Feng L, Lou X, Li Z, Zhang XY, et al. Multiparametric MRI and Whole Slide Image-Based Pretreatment Prediction of Pathological Response to Neoadjuvant Chemoradiotherapy in Rectal Cancer: A Multicenter Radiopathomic Study. Ann Surg Oncol. 2020;27(11):4296-306.\u003c/li\u003e\n\u003cli\u003eNiu L, Chu X, Yang X, Zhao H, Chen L, Deng F, et al. A multiomics approach-based prediction of radiation pneumonia in lung cancer patients: impact on survival outcome. Journal of Cancer Research and Clinical Oncology. 2023;149(11):8923-34.\u003c/li\u003e\n\u003cli\u003eYang S-R, Schultheis AM, Yu H, Mandelker D, Ladanyi M, B\u0026uuml;ttner R. Precision medicine in non-small cell lung cancer: Current applications and future directions. Seminars in Cancer Biology. 2022;84:184-98.\u003c/li\u003e\n\u003cli\u003eWang M, Herbst RS, Boshoff C. Toward personalized treatment approaches for non-small-cell lung cancer. Nature Medicine. 2021;27(8):1345-56.\u003c/li\u003e\n\u003cli\u003eZheng X, Liu K, Li C, Zhu C, Gao Y, Li J, et al. A CT-based radiomics nomogram for predicting the progression-free survival in small cell lung cancer: a multicenter cohort study. La radiologia medica. 2023;128(11):1386-97.\u003c/li\u003e\n\u003cli\u003eChae YK, Pan AP, Davis AA, Patel SP, Carneiro BA, Kurzrock R, et al. Path toward Precision Oncology: Review of Targeted Therapy Studies and Tools to Aid in Defining \u0026ldquo;Actionability\u0026rdquo; of a Molecular Lesion and Patient Management Support. Molecular Cancer Therapeutics. 2017;16(12):2645-55.\u003c/li\u003e\n\u003cli\u003eYang F, Zhang J, Zhou L, Xia W, Zhang R, Wei H, et al. CT-based radiomics signatures can predict the tumor response of non-small cell lung cancer patients treated with first-line chemotherapy and targeted therapy. European Radiology. 2021;32(3):1538-47.\u003c/li\u003e\n\u003cli\u003eChen Q, Shao J, Xue T, Peng H, Li M, Duan S, et al. Intratumoral and peritumoral radiomics nomograms for the preoperative prediction of lymphovascular invasion and overall survival in non-small cell lung cancer. European Radiology. 2022;33(2):947-58.\u003c/li\u003e\n\u003cli\u003eRossi L, Bijman R, Schillemans W, Aluwini S, Cavedon C, Witte M, et al. Texture analysis of 3D dose distributions for predictive modelling of toxicity rates in radiotherapy. Radiother Oncol. 2018;129(3):548-53.\u003c/li\u003e\n\u003cli\u003eZheng X, Guo W, Wang Y, Zhang J, Zhang Y, Cheng C, et al. Multi-omics to predict acute radiation esophagitis in patients with lung cancer treated with intensity-modulated radiation therapy. European Journal of Medical Research. 2023;28(1).\u003c/li\u003e\n\u003cli\u003eLee SH, Han P, Hales RK, Voong KR, Noro K, Sugiyama S, et al. Multi-view radiomics and dosiomics analysis with machine learning for predicting acute-phase weight loss in lung cancer patients treated with radiotherapy. Physics in Medicine \u0026amp; Biology. 2020;65(19).\u003c/li\u003e\n\u003cli\u003eWang B, Liu J, Zhang X, Wang Z, Cao Z, Lu L, et al. Prognostic value of (18)F-FDG PET/CT-based radiomics combining dosiomics and dose volume histogram for head and neck cancer. EJNMMI Res. 2023;13(1):14.\u003c/li\u003e\n\u003cli\u003eZhang Z, Wang Z, Yan M, Yu J, Dekker A, Zhao L, et al. Radiomics and Dosiomics Signature From Whole Lung Predicts Radiation Pneumonitis: A Model Development Study With Prospective External Validation and Decision-curve Analysis. Int J Radiat Oncol Biol Phys. 2023;115(3):746-58.\u003c/li\u003e\n\u003cli\u003eDing H, Feng Y, Huang X, Xu J, Zhang T, Liang Y, et al. Deep learning-based classification and spatial prognosis risk score on whole-slide images of lung adenocarcinoma. Histopathology. 2023;83(2):211-28.\u003c/li\u003e\n\u003cli\u003eYu KH, Zhang C, Berry GJ, Altman RB, Re C, Rubin DL, et al. Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun. 2016;7:12474.\u003c/li\u003e\n\u003cli\u003eMa Y, Li Q. An integrated model combined intra- and peritumoral regions for predicting chemoradiation response of non small cell lung cancers based on radiomics and deep learning. Cancer/Radioth\u0026eacute;rapie. 2023;27(8):705-11.\u003c/li\u003e\n\u003cli\u003eTunali I, Hall LO, Napel S, Cherezov D, Guvenis A, Gillies RJ, et al. Stability and reproducibility of computed tomography radiomic features extracted from peritumoral regions of lung cancer lesions. Medical Physics. 2019;46(11):5075-85.\u003c/li\u003e\n\u003cli\u003eSprave T, R\u0026uuml;hle A, Stoian R, Weber A, Zamboglou C, Nieder C, et al. Radiotherapy for nonagenarians: the value of biological versus chronological age. Radiation Oncology. 2020;15(1).\u003c/li\u003e\n\u003cli\u003evan Timmeren JE, van Elmpt W, Leijenaar RTH, Reymen B, Monshouwer R, Bussink J, et al. Longitudinal radiomics of cone-beam CT images from non-small cell lung cancer patients: Evaluation of the added prognostic value for overall survival and locoregional recurrence. Radiother Oncol. 2019;136:78-85.\u003c/li\u003e\n\u003cli\u003eLiu W, Sun X, Qi Y, Jia X, Huang Y, Liu N, et al. Integrated texture parameter of 18F-FDG PET may be a stratification factor for the survival of nonoperative patients with locally advanced non-small-cell lung cancer. Nuclear Medicine Communications. 2018;39(8):732-40.\u003c/li\u003e\n\u003cli\u003eAndersen MB, Harders SW, Thygesen J, Ganeshan B, Torp Madsen HH, Rasmussen F. Potential impact of texture analysis in contrast enhanced CT in non-small cell lung cancer as a marker of survival: A retrospective feasibility study. Medicine. 2022;101(48).\u003c/li\u003e\n\u003cli\u003eDe Courcy L, Bezak E, Marcu LG. Gender-dependent radiotherapy: The next step in personalised medicine? Critical Reviews in Oncology/Hematology. 2020;147.\u003c/li\u003e\n\u003cli\u003eJing Y, Mao Z, Zhu J, Ma X, Liu H, Chen F. TRAIP serves as a potential prognostic biomarker and correlates with immune infiltrates in lung adenocarcinoma. Int Immunopharmacol. 2023;122:110605.\u003c/li\u003e\n\u003cli\u003eLu D, Chen Y, Ke L, Wu W, Yuan L, Feng S, et al. Machine learning-assisted global DNA methylation fingerprint analysis for differentiating early-stage lung cancer from benign lung diseases. Biosens Bioelectron. 2023;235:115235.\u003c/li\u003e\n\u003cli\u003eDonker HC, van Es B, Tamminga M, Lunter GA, van Kempen L, Schuuring E, et al. Using genomic scars to select immunotherapy beneficiaries in advanced non-small cell lung cancer. Sci Rep. 2023;13(1):6581.\u003c/li\u003e\n\u003cli\u003eCollins LG, Haines C, Perkel R, Enck RE. Lung cancer: diagnosis and management. Am Fam Physician. 2007;75(1):56-63.\u003c/li\u003e\n\u003cli\u003eChen P, Rojas FR, Hu X, Serrano A, Zhu B, Chen H, et al. Pathomic Features Reveal Immune and Molecular Evolution From Lung Preneoplasia to Invasive Adenocarcinoma. Modern Pathology. 2023;36(12).\u003c/li\u003e\n\u003cli\u003eKang W, Qiu X, Luo Y, Luo J, Liu Y, Xi J, et al. Application of radiomics-based multiomics combinations in the tumor microenvironment and cancer prognosis. Journal of Translational Medicine. 2023;21(1).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Non-small cell lung cancer, Pathomics, Dosiomics, Radiomics, Deep learning, Treatment response, Overall survival","lastPublishedDoi":"10.21203/rs.3.rs-4076424/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4076424/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground:\u003c/h2\u003e \u003cp\u003eVarying chemoradiotherapy outcomes in individuals arose from the intricate physical conditions and tumor heterogeneity characteristic of non-small cell lung cancer patients. This study aimed to develop and validate multi-omics models based on the radiomics, pathomics, dosiomics and clinical information for illustrating the heterogeneity and predicting treatment response and overall survival of non-small cell lung cancer patients.\u003c/p\u003e\u003ch2\u003eMethods:\u003c/h2\u003e \u003cp\u003eThis retrospective study including 220 non-small cell lung cancer patients treated with chemoradiotherapy from three hospitals for overall survival prediction, with 142 of these patients specifically assessed for treatment response prediction. Radiomics and dosiomcis features were obtained from the region of interest, including first-order and texture features. Pathomics features were derived from whole slide images by Resnet34 network. Lasso regression, random forest, and extreme gradient boosting were employed for treatment response prediction to identify the most predictive biomarkers, with model performance evaluated through area under the curve and box plots. Overall survival analysis also involved three different feature selection methods, and model evaluation incorporated area under the curve, concordance index, Kaplan-Meier curves, and calibration curves. The shapley values calculated the contribution of different modality features to the models.\u003c/p\u003e\u003ch2\u003eResults:\u003c/h2\u003e \u003cp\u003eMulti-omics models consistently exhibited superior discriminative ability compared to single-modality models in predicting treatment response and overall survival. For treatment response, the multi-omics model achieved area under the curve values of 0.85, 0.81, and 0.87 in the training set, internal validation set, and external validation set, respectively. In the analysis of overall survival, the area under the curve and concordance index of the all-modalities model were 0.83/0.79, 0.74/0.74, and 0.73/0.72 in the training set, internal validation set, and external validation set, respectively.\u003c/p\u003e\u003ch2\u003eConclusion:\u003c/h2\u003e \u003cp\u003eMulti-omics prediction models demonstrated superior predictive ability with robustness and strong biological interpretability. By predicting treatment response and overall survival in non-small cell lung cancer patients, these models had the potential to assist clinician optimizing treatment plans, supporting individualized treatment strategies, further improving tumor control probability and prolonging the patients\u0026rsquo; survival.\u003c/p\u003e","manuscriptTitle":"Multi-omics models predict treatment response and overall survival for non-small cell lung cancer patients following chemo-radiotherapy: A multi-center study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-13 17:20:28","doi":"10.21203/rs.3.rs-4076424/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"891fb924-8bb7-44fb-bded-95a1bc368c71","owner":[],"postedDate":"March 13th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-03-20T16:13:21+00:00","versionOfRecord":[],"versionCreatedAt":"2024-03-13 17:20:28","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4076424","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4076424","identity":"rs-4076424","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}