Multi-omics-based prognostic prediction for locally advanced hypopharyngeal cancer treated with postoperative chemoradiotherapy: a dual-center study

Multi-omics-based prognostic prediction for locally advanced hypopharyngeal cancer treated with postoperative chemoradiotherapy: a dual-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-based prognostic prediction for locally advanced hypopharyngeal cancer treated with postoperative chemoradiotherapy: a dual-center study Sixue Dong, Zian Yao, Zhiyuan Zhang, Jiazhou Wang, Guo Ying, Lei Tao, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5861722/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 Purpose This study aimed to predict the progression-free survival (PFS) of the patients who were diagnosed with hypopharyngeal cancer and received postoperative chemoradiotherapy by using multi-omics method which integrating clinical factors, dosimetric and radiomic features. Materials and methods This study retrospectively collected the pretreatment T1-weighted MR imaging data of 88 hypopharyngeal cancer patients with postoperative chemoradiotherapy, including 56 cases from one center (training and testing cohorts) and 32 cases from another center (external validation cohort), and the gross tumor volumes (GTV) were countered for all cases. A Python-based library, pyradiomics was used to extract the radiomics features from each GTV. Least absolute shrinkage and selection operator (LASSO) regression was used to identify the most important features for classifier establishment. On the other hand, complete radiotherapy data are retained for 48 patients among them, and the planning tumor volumes (PTV) were countered for radiotherapy planning. The dose distribution features extracted by using pyradiomics and the dosimetric parameters were combined with the radiomics features to establish the classifiers. The probabilities of positive sample calculated from the best classifier, the radiomics and multi-omics signatures were obtained for establish the Cox proportional hazards models. Results The ensemble learning (EL) model was selected as the superior model with the higher area under the receiver operating characteristic curve (AUC) values than other classifier during the radiomics-only analysis, and the EL model with stacking technique showed the best performance, yielding AUC values of 0.93, 0.79, and 0.78 for the training, testing, and external validation cohorts, respectively. Furthermore, the multi-omics analysis integrating radiomics and dosiomics improved the effectiveness of the EL model with AUC values of 0.98 and 0.88 for the training and testing cohorts, respectively. Furthermore, the C-index of the Cox proportional hazards models resulted in a 0.099 improvement in the testing cohort when employing the multi-omics signature versus the radiomics signature. Conclusion Regarding the patients with hypopharyngeal cancer receiving postoperative chemoradiotherapy, the multi-omics-based prognostic prediction could achieve a more robust predictive capability than the radiomics-only study. This approach warrants further validation through prospective studies. Hypopharyngeal cancer postoperative chemoradiotherapy Multi-omics Machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction The incidence of hypopharyngeal cancer constitutes approximately 3–5% of head and neck cancers, with higher rates observed in China and Eastern Europe. According to data from the Taiwan Cancer Registry, the crude incidence rate of hypopharyngeal cancer is 5.15 per 100,000 individuals[1, 2]. Hypopharyngeal cancer represents a malignancy characterized by its heightened aggressiveness, which (70–85%) are diagnosed at an advanced stage and are characterized by indistinct symptoms[3, 4]. The National Comprehensive Cancer Network (NCCN) recommends the implementation of a comprehensive multidisciplinary approach, incorporating chemotherapy, surgery, and radiotherapy, to optimize functional preservation and improve patient outcomes[5]. Postoperative chemoradiotherapy was recommended as one of the treatment strategies in the guidelines and correlated with the survival benefit, however, the prognosis was suboptimal with a 5-year overall survival (OS) rate of approximately 40% and the discrepancies were in patient outcome[6–8]. Consequently, the acquisition of follow-up data poses challenges, and the limited sample size also presents a significant challenge to the performance of the prognostic prediction model. In recent years, the integration of omics analysis into clinical models has significantly advanced cancer research. In a one-year PFS prediction study of hypopharyngeal cancer, the RF classifier based solely on clinical features without incorporating omics data was only able to improve the area under the receiver operating characteristic (ROC) curve (AUC) to a maximum of 0.739 in the training cohort[9]. In contrast, radiomics-based prognostic models have gained significant attention in cancer research. Radiomics entails the analysis and mining of numerous features derived from medical imaging data[10–12]. By analyzing imaging data of postoperative chemoradiotherapy patients, radiomics signatures related to tumor characteristics, treatment response, and prognosis can be extracted. These signatures, along with clinical data such as age, gender, and tumor stage, can be utilized to construct predictive model, aims to assess the prognostic risk of patients and identify high-risk individuals[13–15]. Although previous studies have predominantly relied on computed tomography (CT) signatures, and demonstrated robustness and predictive utility for various clinical endpoints, their efficacy is limited by suboptimal soft tissue contrast[1]. Magnetic resonance imaging (MRI) provides enhanced visualization of soft tissues and superior anatomical resolution, typically used for the staging of head and neck cancers[16–19]. Previous studies utilized MRI-based radiomics to predict progression-free survival (PFS) in nasopharyngeal carcinoma (NPC) with different treatment regimens by collecting 1872 samples from four centers, and achieved a reliable nomogram with the C-index range from 0.784 to 0.921[20]. Additionally, multi-omics holds immense promise in advancing the prognosis and treatment management of individuals afflicted with cancer[21, 22]. Related studies showed that the introduction of dosiomics is beneficial for effectively predicting the prognosis of tumor recurrence and radioactive diseases (such as radiation esophagitis and radiation pneumonitis)[23–25]. Considering the fewer studies of prognostic prediction of postoperative chemoradiotherapy for hypopharyngeal cancer, and the bottleneck in radiomics-based model performance, a more effective method was required to be developed for facilitating the clinical decision. This study endeavors to use multi-omics method which integrate dosimetric, radiomic, and clinical factors to predict PFS of the patients who were diagnosed with hypopharyngeal cancer and received postoperative chemoradiotherapy. Materials and methods Patients and data collection This is a retrospective study of patients with locally advanced hypopharyngeal cancer treated by surgery and postoperative chemoradiotherapy. Inclusion criteria: patients diagnosed between January 2015 and December 2021, pathologically confirmed, resectable locally advanced hypopharyngeal squamous cell carcinoma (T2-4a, N0-resectable N3, M0), preoperatively examined by contrast enhanced MRI with complete imaging data, and treated with surgery and postoperative chemoradiotherapy. Exclusion criteria: stage I/II patients, T4b patients, suffered from malignant tumors except cervical carcinoma in situ, papillary thyroid carcinoma, or skin cancer (non- melanoma) within five years, distant metastasis before treatment, and incomplete treatment or missing clinical data. The enrollment diagram was shown in Fig. 1 a. In the radiomics-only analysis of this study, 56 patients with hypopharyngeal cancer treated between 2015 and 2021 were selected from Fudan University Shanghai Cancer Center, and were divided into a training cohort (n = 39) and a testing cohort (n = 17) at a 7:3 ratio. Additionally, an external validation cohort of 32 hypopharyngeal cancer patients treated between 2015 and 2021 was selected from Eye and ENT Hospital of Fudan University. In the multi-omics analysis integrating radiomics and dosiomics of this study, 8 patients without complete radiotherapy data were excluded. All patients underwent a follow-up period of at least three years, with an interval of six months. Imaging follow-up, either with CT and/or MRI, was performed alternatively at intervals of 3 months during the first year after treatment and 6 months during the second year. OS (defined as the time between the date of initial pathologic diagnosis and the date of death or the last follow-up) and PFS (defined as the duration between the date of initial pathologic diagnosis and the date of the first observed sign of disease progression, death, or the last follow-up) were considered the most important indicators in the present follow-up assessments. This retrospective study received approval from the Ethics Committees of Fudan University Shanghai Cancer Center (A) and Eye & ENT Hospital of Fudan University The subsequent workflow of this study was illustrated in Fig. 1 b). The radiomics and dosiomics data was prepared for feature extraction and selection. Omics-based signatures and the positive sample probabilities calculated from the classifiers were obtained to do the survival analysis. Feature extraction and selection MRI images of the patients before the surgery were obtained using a 3.0T Magnetom Skyra (Siemens, Germany). The gross tumor volume (GTV, including the primary tumor and metastatic lymph nodes) was countered as the region of interest (ROI) on MIM Software (v7.3.6). by a head and neck radiologist with the title of associate chief physician and 15 years of clinical practice. The planning target volume (PTV) was countered on the CT images and intensity modulated radiation therapy (IMRT) and/or volumetric modulated arc therapy (VMAT) technique was utilized during treatment planning with the Pinnacle 3 (v9.10) treatment planning system (TPS). The radiotherapy process was conducted by Varian Vital Beam linear accelerator (6 MV, 600 MU/min). An in-house Python-based script were developed to extract the radiomics and dose distribution features by using Pyradiomics (v3.1.0) package. In total of 1688 features were extracted for the two modalities, respectively including first order features (Energy, standard deviation, uniformity, etc.), shape features (2D and 3D), texture features (Gray Level Co-occurrence Matrix, GLCM. Gray Level Run Length Matrix, GLRLM. Gray Level Size Zone Matrix, GLSZM. Neighborhood Gray Tone Difference Matrix, NGTDM. Gray Level Dependence Matrix, GLDM)[26]. During the preprocessing of images, sitkBSpline interpolator of Pyradiomics package was used to resample all images to a uniform image spacing of 1×1×1 mm 3 . Besides the dose distribution features, dosimetric parameters based on dose-volume histogram (DVH) were also collected along with the PTV in the dosiomics analysis, including D 90 -D 100 (i.e., the minimum absolute dose covering 90%, 95%, 100% of PTV volume, respectively), V 50 -V 70 (i.e., the maximum relative PTV volume covered by 50Gy, 55Gy, 60Gy, 65Gy, 70Gy), homogeneity index (HI), standard deviation (SD) of dose distribution, maximum absolute dose (Dmax), minimum absolute dose (Dmin), and mean absolute dose (Dmean). Feature standardization by the StandardScaler command of the scikit-learn (v1.4.1) package was proceeded prior to feature selection. Least absolute shrinkage and selection operator (LASSO) regression with ten-fold cross-validation was used to identify the most important features and prevent overfitting, thereby enhancing model generalization and interpretability. Furthermore, the radiomics signature and the multi-omics signature were obtained separately. Model establishment and statistical analysis In the radiomics-only analysis, four single-classifiers including support vector machine (SVM) model, deep learning (DL) model, logistic regression (LR) model, and random forest (RF) model were trained. Furthermore, ensemble learning (EL) models integrated by the four single-classifiers with four techniques including boosting, stacking, bagging, and voting were also trained. On the other hand, the four EL models were trained in the multi-omics analysis. For all of the models, the hyperparameters were tuned to achieve the best F1 score by using Bayesian optimization. The parameter "random_state" of each classifier was randomly modified 100 times to achieve various AUC values, and the inter-model differences were evaluated by Student’s t-test. The parameters "random_state" which yielded the highest AUC values were employed to establish various classifiers. P-values (A classification result with P < 0.05 was considered to be statistically significant) of the Kaplan–Meier (K-M) survival analysis were evaluated in the training, testing, and external validation cohorts, respectively. The probabilities of positive sample were calculated from the best classifier. Single-factor Cox regression was used to identify the most important clinical factors, and two Cox proportional hazards models were established by incorporating the selected clinical factors with the signatures and positive sample probabilities of radiomics-only and multi-omics analysis separately. The C-indexes were evaluated, and the multi-omics-based nomogram were established. Results Clinical characteristics The study samples consisted of 88 male patients with locally advanced hypopharyngeal carcinoma. The clinical characteristics of training, testing, and external validation cohorts with (radiomics-only) were shown in Table 1 a, the corresponding median ages (age ranges) were 63.3 (48-77) years, 62.5 (46-76) years, and 61.84 (46-76) years, respectively. The median follow-up duration of the whole cohort was 36 months (range, 3–92 months). Tumor regions from both cohorts were mostly located in the pyriform sinus, respectively with a proportion of 87.2% and 82.4%. Overall, 30 (34.1%) patients were classified as stage III (AJCC 8th), while 51 (58.0%) were stage IVA and 7 (8.0%) were stage IVB. Upon the last follow-up, 7 (17.9%) in the training cohort and 4 (23.5%) patients in the testing cohort experienced a confirmed disease progression. Particularly, there were no significant differences (P > 0.05) in all of the terms. Table 1 b illustrates the dosimetric characteristics of the population. Specifically, the median EQD2 dose across all patients was 60 Gy (inter quartile range [IQR], 60–66 Gy). Moreover, other dosimetric parameters including D90-D10, V50-V70, homogeneity index (HI), standard deviation (SD) of dose distribution, maximum absolute dose (Dmax), minimum absolute dose (Dmin), and mean absolute dose (Dmean) of training and testing cohorts were shown. Notably, no significant differences (P > 0.05) were observed between the two cohorts, with exception of SD of dose distribution (p = 0.024) and V55 (p = 0.018). Classifiers comparison of radiomics-only and multi-omics The features selected by using LASSO regression were shown in Table 2 with their coefficients. The ROCs comparison derived from various classifiers of radiomics and multi-omics were shown in Figure 2 and Figure 3, respectively. The performance of each classifier was recorded in Table 3, Figure 2, and Figure 3. The radiomics-based classifiers were evaluated in three cohorts: training, testing, and external validation cohorts. The multi-omics-based classifiers were evaluated in two cohorts: training and testing cohorts. Initially, four single-classifiers including SVM, DL, LF, RF models were tested in the radiomics-only analysis, the AUC values for various single-classifiers ranged from 0.85 to 0.93, 0.58 to 0.75, and 0.57 to 0.65 in the training, testing, and external validation cohorts, respectively. Subsequently, four EL models with various techniques were added for further comparison against the single-classifiers. In general, the EL model performances were superior to each single-classifiers, and the AUC values ranged from 0.93 to 0.99 in the training cohort. While in the testing and external validation cohorts, the AUC values range from 0.72 to 0.79 and from 0.69 to 0.78, respectively. The highest AUC value of 0.99 in the training cohort was achieved by using the voting technique, while using the stacking technique could yield the optimal AUC in testing (0.79) and external validation cohorts (0.78). The K-M survival curves derived from the EL model with the stacking technique with were displayed to identify the high- and low-risk groups in Figure 2 b), and the P-values < 0.05 in all the cohorts. In the multi-omics analysis, the AUC values for various EL models ranged from 0.98 to 0.99 in the training cohort. While in the testing cohort, the AUC values range from 0.77 to 0.88. The integration of dosiomics could enhance the average AUC values by 0.038±0.022 and 0.080±0.019 in the training and testing cohorts, respectively. The EL model with the stacking technique still exhibited superior performance with the AUCs of 0.98 and 0.88, and the corresponding K-M survival curves were also displayed in Figure 3 b), the P-values in the training and testing cohorts were both lower than those in the radiomics-only analysis. By performing different "random_state", the P-value between the AUC values of various classifiers were lower than 0.05 except in the cases of the DL model versus LR and RF models. Evaluation of Cox proportional hazards models The valuable clinical factors including overall stage and age were identified by using single-factor Cox regression. According to the performance of classifiers, EL model with stacking technique was chosen to calculate the positive sample probabilities, in addition, incorporating the signatures derived from the LASSO regression and the valuable clinical factors, two Cox proportional hazards models were established for radiomics-only and multi-omics analysis. In the radiomics-only analysis, the C-indexes were 0.777, 0.685, and 0.663 in the training, testing, and external validation cohorts, respectively. On the other hand, in the multi-omics analysis, the C-indexes were 0.783 and 0.784 in the training and testing cohorts, respectively. The multi-omics-based nomogram (Figure 4) were established to predict the 1-year, 2-year, 3-year, and 4-year PFSs. The prediction of a given sample were also shown in the nomogram, the 1-year, 2-year, 3-year, and 4-year PFSs were 69.7%, 52.9%, 43.2%, and 34.0%, respectively. Discussion Clinical factors played a crucial role in the conventional prognostic investigation of hypopharyngeal cancer. Previous studies demonstrated the treatment modalities were correlated with OS and cancer-specific survival (CSS), while postoperative chemoradiotherapy stood out as the most beneficial treatment modality for enhancing the OS and CSS of patients[27–29]. However, despite the generally favorable prognosis for patients, the high recurrence rate remained one of the significant factors affecting survival and quality of life, particularly for advanced hypopharyngeal cancer, and the analysis of PFS became necessary. Thus far, most of the studies on prognostic predictions for hypopharyngeal cancer have primarily focused on surgery and/or adjuvant chemotherapy, with the method largely relying on radiomics-only analysis[2, 30–32]. In the present study, a prediction for the PFS of hypopharyngeal cancer with postoperative chemoradiotherapy based on the multi-omics method was developed. The results showed the analysis of multi-omics including radiomics and dosiomics exhibited superior performance than the radiomics-only analysis, whether in the classifier model (with an average AUC of 0.828 ± 0.040 in the testing cohort) or the Cox proportional hazards model (with a C-index of 0.784 in the testing cohort). Targeting patients who have received radiotherapy, the recurrence of tumors correlates intricately with the dose distribution within the PTV, and the integration of dosiomics could emerge as the significant factor for predicting PFS [33–35]. Our present study focused on the method the multi-omics, compared the performance of omics predictive models with and without the inclusion of dosimetric omics, and successfully demonstrated the robust correlation between dosimetric parameters and PFS. It should be noted that the shape of the target volume and the radiotherapy planner varies among individual patients, thus substantial variations exist in dosimetric parameters such as HI, Dmax, and Dmin, however, none of the DVH-based dosimetric features were identified. Based on these findings, there is justification to suggest that there is no significant correlation between the other DVH parameters and PFS when the target area receives adequate coverage at the prescribed dose(D 95% >100% or V 95% >100%). Moreover, the category of selected features with LASSO regression were almost “texture”, representing the heterogeneity of the tumor, and higher heterogeneity with irregular margins were typically associated with an elevated risk of tumor recurrence and reduced the PFS. During the radiotherapy, the technique of VMAT could yield a more homogeneous dose distribution at the target area with a smoother edge compared to IMRT, while two texture features of dosiomics were selected with negative correlation coefficients. In our perspective, VMAT might could potentially be a better option in terms of mitigating the risk of recurrence. By incorporating a greater number of samples from diverse centers and additional factors related to radiotherapy, further research is being conducted to substantiate our perspective and enhance the generalization capability of the model. During the process of classifier modeling, three machine learning techniques (SVM, LR, RF) and one DL technique were employed initially, while the generalization abilities in both the testing and external validation cohorts were not satisfactory. Therefore, four EL methods were introduced to modeling the classifiers, it could reduce the prediction generalization error and make more accurate predictions than a single-classifier. Nevertheless, this enhancement came at the expense of sacrificing the model interpretability[36–38]. Although the AUC value of the EL model with stacking technique in the training cohort was not the highest (0.93 in the radiomics-only analysis and 0.98 in the multi-omics analysis, even the lowest among EL models), it achieves the highest AUC in both the testing and external validation cohorts, that indicated the stacking technique demonstrated relatively strong generalization ability while maintaining modeling accuracy, in other words, it exhibited more balance performance than other techniques. Therefore, in this study, it was regarded as the best classifier for estimating the probabilities of the positive sample. Besides, in the figures of K-M survival curve, the classifier could also successfully divide high- and low-risk groups in each cohort (P-value < 0.05). Benefited to the incorporation of dosiomics, there was a marginal improvement in the AUC value in the training cohort (0.038 ± 0.022), whereas a more substantial improvement in the testing cohort (0.080 ± 0.019). A more satisfactory outcome in the external validation cohort was anticipated and necessitated to be confirmed in subsequent research. Most previous studies on survival analysis have incorporated omics signatures and/or risk levels to enhance the predictive performance of the models[39–42], however, the categorization of risk levels (e.g., high-, medium-, and low-risk) might be somewhat imprecise due to its binary or ternary nature. This study integrated the positive sample probabilities predicted by the EL classifier into a Cox proportional hazards model as a continuous indicator, aiming to combine machine learning with statistical analysis to obtain more refined predictive results. Age and overall stage are considered the two most highly correlated clinical factors with PFS. Elderly patients with advanced-stage hypopharyngeal cancer exhibited higher risk scores and were associated with lower probabilities of PFS. Previous studies showed that clinical factors-based Cox proportional hazards model for predicting OS achieved a c-index of approximately 0.72[43, 44], when upon the incorporated of radiomics, the C-index increased to 0.78[18]. Our presented study focused on predicting the PFS, and the introduction of radiomics achieves a comparable performance to predicting the OS (C-index = 0.777 in the training cohort), while its capacity for generalization is somewhat constrained (C-index = 0.685 in the testing cohort). With the further research, the introduction of multi-omics signatures had addressed the issue of poor generalization, and achieved an increase in the C-index to 0.784 in the training cohort. Multi-omics methodology was successfully employed in this study to establish predictive models for PFS in patients with hypopharyngeal cancer and received postoperative chemoradiotherapy, moreover, the comparative advantages with radiomics-only methodology were analyzed. Accordingly, for surgical patients diagnosed with locally advanced hypopharyngeal carcinoma, the multi-omics prognostic model has provided essential biomarkers to select patients treated with surgery and postoperative chemoradiotherapy to address hypopharyngeal cancer who likely suffered early tumor progression, thus quantitatively evaluating the improvement of clinicians' decision-making and patients' outcome. Nevertheless, there were still some deficiencies and limitations: 1) As a dual-center study, this study had not yet obtained complete radiotherapy data in the external validation cohort, leading to a missing link in the external validation of multi-omics analysis. 2) Meanwhile, due to the relatively small number of patients with hypopharyngeal cancer and the other treatment modalities were excluded apart from postoperative chemoradiotherapy, there were only 39 and 33 samples for training the radiomics-based and multi-omics-based predictive model, respectively. The limited sample size might potentially influence the accuracy and generalization ability of the predictive model. 3) All patients included in the study were males aged 46 years and above, despite the lower incidence of hypopharyngeal cancer among younger populations and females, it still needed to be considered to incorporate them into the predictive model. 4) Insufficient prospective studies for validation. Concerning these limitations, it was necessary to expand the sample size and integrate more clinical factors such as gender (male or female), radiotherapy techniques (IMRT, VMAT, helical tomotherapy, or 3D-conformal radiotherapy), age (≤ 45), etc. It was also required to Optimize the model performance and perform validations by using multicenter data to translate the multi-omics methodology into clinical practice in our further studies. Conclusion In this retrospective study, a multi-omics methodology, integrating clinical parameters, dosimetric, and radiomic features, was harnessed to establish predictive models to predict the PFS of patients with hypopharyngeal cancer who received postoperative chemoradiotherapy. The EL model with the stacking technique demonstrated superior performance in classifying high- and low-risk groups. Furthermore, multi-omics significantly enhanced the predictive performance and generalization ability of classifiers and Cox proportional hazards models. The presented models could facilitate the clinical decision by assessing the requirement for postoperative chemotherapy based on varied probabilities of time-dependent PFS and risk levels. Despite limitations such as limited sample size, our research effectively demonstrated the potential clinical utility of multi-omics analysis in the prognostic prediction of hypopharyngeal cancer. Future prospective, multicenter studies were needed to refine and validate this clinical translational approach. Abbreviations PFS progression-free survival GTV gross tumor volume LASSO least absolute shrinkage and selection operator PTV planning tumor volume EL ensemble learning ROC receiver operating characteristic AUC area under the receiver operating characteristic curve NCCN National Comprehensive Cancer Network OS overall survival CT computed tomography MRI magnetic resonance imaging NPC nasopharyngeal carcinoma IMRT intensity modulated radiation therapy VMAT volumetric modulated arc therapy TPS treatment planning system Dmax maximum absolute dose Dmin minimum absolute dose Dmean mean absolute dose SVM support vector machine DL deep learning LR logistic regression RF random forest K-M Kaplan–Meier IQR inter quartile range CSS cancer-specific survival Declarations Ethics approval and consent to participate: All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Fudan University Shanghai Cancer Center. Informed consent or substitute for it was obtained from all patients for being included in the study. Consent for publication: The Author confirms: that the work described has not been published before; that it is not under consideration for publication elsewhere; that its publication has been approved by all co-authors; that its publication has been approved by the responsible authorities at the institution where the work is carried out. Availability of data and material: Due to the nature of this research, the participants did not agree to share their data publicly, so supporting data is not available. Competing interests: There are no conflicts of interest or financial ties to disclose from any author. Funding: Chinese Society of Clinical Oncology Foundation (Y-Young2021–0127, to Xiaomin Ou). Authors' contributions: Author initials: Sixue Dong (Sx. D), Zian Yao (Za. Y), Zhiyuan Zhang (Zy. Z), Jiazhou Wang (Jz. W), Guo Ying (G. Y), Lei Tao (L. T), Xiaomin Ou (Xm. O), Weigang Hu (Wg. H), Chaosu Hu (Cs. H) Sx. D, Methodology, Experimental procedures, Data Analysis, Writing, Editing, Za. Y, Data Collection, Review and Editing Zy. Z, Methodology, Review Jz. W, Methodology, Experimental procedures, Review G. Y, Methodology, Review L. T, Data Collection Xm. O, Data Collection, funding acquisition, Review Wg. H, Conceptualization, Supervision Cs. H, Conceptualization, Supervision Acknowledgements: N/A. Authors' information: N/A. References Siow, T.Y., et al., MRI Radiomics for Predicting Survival in Patients with Locally Advanced Hypopharyngeal Cancer Treated with Concurrent Chemoradiotherapy. Cancers (Basel), 2022. 14 (24). Chiesa-Estomba, C.M., et al. Radiomics in Hypopharyngeal Cancer Management: A State-of-the-Art Review . Biomedicines, 2023. 11 , DOI: 10.3390/biomedicines11030805. Buckley, J.G. and K. MacLennan, Cervical node metastases in laryngeal and hypopharyngeal cancer: a prospective analysis of prevalence and distribution. Head Neck, 2000. 22 (4): p. 380-5. Kuo, P., et al., Hypopharyngeal cancer incidence, treatment, and survival: temporal trends in the United States. Laryngoscope, 2014. 124 (9): p. 2064-9. Kwon, D.I., et al., Hypopharyngeal carcinoma: Do you know your guidelines? Head & Neck, 2019. 41 (3): p. 569-576. Newman, J.R., et al., Survival trends in hypopharyngeal cancer: a population-based review. Laryngoscope, 2015. 125 (3): p. 624-9. Janoray, G., et al., Long-term Results of a Multicenter Randomized Phase III Trial of Induction Chemotherapy With Cisplatin, 5-fluorouracil, ± Docetaxel for Larynx Preservation. J Natl Cancer Inst, 2016. 108 (4). Burbure, N., et al., Prognostic significance of human papillomavirus status and treatment modality in hypopharyngeal cancer. Head Neck, 2021. 43 (10): p. 3042-3052. Zhang, Y.F., et al., Predicting survival of advanced laryngeal squamous cell carcinoma: comparison of machine learning models and Cox regression models. Sci Rep, 2023. 13 (1): p. 18498. Qi, M., et al., An MRI-based radiomics nomogram for detecting cervical esophagus invasion in hypopharyngeal squamous cell carcinoma. Cancer Imaging, 2023. 23 (1): p. 120. Wu, T.-C., et al., Radiomics analysis for the prediction of locoregional recurrence of locally advanced oropharyngeal cancer and hypopharyngeal cancer. European Archives of Oto-Rhino-Laryngology, 2024. 281 (3): p. 1473-1481. Rizzo, S., et al., Radiomics: the facts and the challenges of image analysis. Eur Radiol Exp, 2018. 2 (1): p. 36. Mo, X., et al., Prognostic value of the radiomics-based model in progression-free survival of hypopharyngeal cancer treated with chemoradiation. Eur Radiol, 2020. 30 (2): p. 833-843. Hsu, C.Y., et al., Magnetic resonance imaging-derived radiomic signature predicts locoregional failure after organ preservation therapy in patients with hypopharyngeal squamous cell carcinoma. Clin Transl Radiat Oncol, 2020. 25 : p. 1-9. Guo, R., et al., CT-based radiomics features in the prediction of thyroid cartilage invasion from laryngeal and hypopharyngeal squamous cell carcinoma. Cancer Imaging, 2020. 20 (1): p. 81. Romeo, V., et al., Prediction of Tumor Grade and Nodal Status in Oropharyngeal and Oral Cavity Squamous-cell Carcinoma Using a Radiomic Approach. Anticancer Res, 2020. 40 (1): p. 271-280. Mukherjee, P., et al., CT-based Radiomic Signatures for Predicting Histopathologic Features in Head and Neck Squamous Cell Carcinoma. Radiol Imaging Cancer, 2020. 2 (3): p. e190039. Chen, J., et al., An MRI-based radiomics-clinical nomogram for the overall survival prediction in patients with hypopharyngeal squamous cell carcinoma: a multi-cohort study. Eur Radiol, 2022. 32 (3): p. 1548-1557. Katsoulakis, E., et al., Radiomic analysis identifies tumor subtypes associated with distinct molecular and microenvironmental factors in head and neck squamous cell carcinoma. Oral Oncol, 2020. 110 : p. 104877. Zhong, L., et al., A deep learning-based radiomic nomogram for prognosis and treatment decision in advanced nasopharyngeal carcinoma: A multicentre study. EBioMedicine, 2021. 70 : p. 103522. Boehm, K.M., et al., Harnessing multimodal data integration to advance precision oncology. Nat Rev Cancer, 2022. 22 (2): p. 114-126. Liu, Z., et al., Integrated multi-omics profiling yields a clinically relevant molecular classification for esophageal squamous cell carcinoma. Cancer Cell, 2023. 41 (1): p. 181-195.e9. Wu, A., et al., Dosiomics improves prediction of locoregional recurrence for intensity modulated radiotherapy treated head and neck cancer cases. Oral Oncology, 2020. 104 : p. 104625. Zheng, X., et al., Multi-omics to predict acute radiation esophagitis in patients with lung cancer treated with intensity-modulated radiation therapy. Eur J Med Res, 2023. 28 (1): p. 126. Nie, T., et al., Integration of dosimetric parameters, clinical factors, and radiomics to predict symptomatic radiation pneumonitis in lung cancer patients undergoing combined immunotherapy and radiotherapy. Radiother Oncol, 2024. 190 : p. 110047. van Griethuysen, J.J.M., et al., Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Research, 2017. 77 (21): p. e104-e107. Zheng, L., et al., Optimal treatment strategy and prognostic analysis for hypopharyngeal squamous-cell carcinoma patients with T3-T4 or node-positive: A population-based study. Eur J Surg Oncol, 2023. 49 (7): p. 1162-1170. Tang, X., et al., A novel prognostic model predicting the long-term cancer-specific survival for patients with hypopharyngeal squamous cell carcinoma. BMC Cancer, 2020. 20 (1): p. 1095. Forastiere, A.A., et al., Long-term results of RTOG 91-11: a comparison of three nonsurgical treatment strategies to preserve the larynx in patients with locally advanced larynx cancer. J Clin Oncol, 2013. 31 (7): p. 845-52. Grasl, S., et al., A new nomogram to predict oncological outcome in laryngeal and hypopharyngeal carcinoma patients after laryngopharyngectomy. Eur Arch Otorhinolaryngol, 2023. 280 (3): p. 1381-1390. Liu, X., et al., CT-based radiomics signature analysis for evaluation of response to induction chemotherapy and progression-free survival in locally advanced hypopharyngeal carcinoma. European Radiology, 2022. 32 (11): p. 7755-7766. Li, F., et al., A Nomogram to Predict Nodal Response after Induction Chemotherapy for Hypopharyngeal Carcinoma. Laryngoscope, 2023. 133 (4): p. 849-855. Cavalieri, S., et al., Development of a multiomics database for personalized prognostic forecasting in head and neck cancer: The Big Data to Decide EU Project. Head Neck, 2021. 43 (2): p. 601-612. Huang, Y., et al., Radiation pneumonitis prediction after stereotactic body radiation therapy based on 3D dose distribution: dosiomics and/or deep learning-based radiomics features. Radiation Oncology, 2022. 17 (1): p. 188. Yang, S.S., et al., Dosiomics Risk Model for Predicting Radiation Induced Temporal Lobe Injury and Guiding Individual Intensity-Modulated Radiation Therapy. Int J Radiat Oncol Biol Phys, 2023. 115 (5): p. 1291-1300. Shorewala, V., Early detection of coronary heart disease using ensemble techniques. Informatics in Medicine Unlocked, 2021. 26 : p. 100655. Chandra Joshi, R., et al., Ensemble based machine learning approach for prediction of glioma and multi-grade classification. Computers in Biology and Medicine, 2021. 137 : p. 104829. Zhao, S., et al., Stacking Ensemble Learning-Based [(18)F]FDG PET Radiomics for Outcome Prediction in Diffuse Large B-Cell Lymphoma. J Nucl Med, 2023. 64 (10): p. 1603-1609. Wang, R., et al., Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients. Journal of Hematology & Oncology, 2022. 15 (1): p. 11. Shen, L.L., et al., Delta computed tomography radiomics features-based nomogram predicts long-term efficacy after neoadjuvant chemotherapy in advanced gastric cancer. Radiol Med, 2023. 128 (4): p. 402-414. Kong, C., et al., Prediction of tumor response via a pretreatment MRI radiomics-based nomogram in HCC treated with TACE. Eur Radiol, 2021. 31 (10): p. 7500-7511. Ma, X., et al., Radiomics nomogram based on optimal VOI of multi-sequence MRI for predicting microvascular invasion in intrahepatic cholangiocarcinoma. Radiol Med, 2023. 128 (11): p. 1296-1309. Tian, S., et al., Development and Validation of a Prognostic Nomogram for Hypopharyngeal Carcinoma. Frontiers in Oncology, 2021. 11 . Zhang, D., et al., Prognostic Nomogram for Postoperative Hypopharyngeal Squamous Cell Carcinoma to Assist Decision Making for Adjuvant Chemotherapy. Journal of Clinical Medicine, 2022. 11 (19): p. 5801. Tables Table 1 Clinical characteristics of the patients a) Clinical characteristics of patients in the training, testing and external validation cohorts Clinical Characteristics All Patients ( n = 56) Training Cohort ( n = 39) Testing Cohort ( n = 17) p External Validation Cohort ( n = 32) p Age 0.742 0.432 Median (range) 63.1 (46-77) 63.3 (48-77) 62.5 (46-76) 61.84 (46-76) Pathology (%) 0.961 0.254 High grade 8 (14.3) 5 (12.8%) 3 (17.6) 3 (9.4) Intermediate grade 34 (60.7) 24 (61.5%) 10 (58.8) 26 (81.2) Low grade 10 (17.9) 7 (17.9%) 3 (17.6) 3 (9.4) Unknown 4 (7.1) 3 (7.7) 1 (5.9) 0 (0.0) cT stage (AJCC 8th) (%) 0.411 0.320 T1 6 (10.7) 3 (7.7) 3 (17.6) 3 (9.4) T2 26 (46.4) 19 (48.7) 7 (41.2) 10 (31.2) T3 15 (26.8) 12 (30.8) 3 (17.6) 10 (31.2) T4 9 (16.1) 5 (12.8) 4 (23.5) 9 (28.1) cN stage (AJCC 8th) (%) 0.328 0.222 N0 11 (19.6) 9 (23.1) 2 (11.8) 3 (9.4) N1 13 (23.2) 7 (17.9) 6 (35.3) 11 (34.4) N2 28 (50.0) 21 (53.8) 7 (41.2) 15 (46.9) N3 4 (7.1) 2 (5.1) 2 (11.8) 3 (9.4) cStage (AJCC 8th) (%) 0.736 0.607 III 21 (37.5) 15 (38.5) 6 (35.3) 9 (28.1) IVA 31 (55.4) 22 (56.4) 9 (52.9) 20 (62.5) IVB 4 (7.1) 2 (5.1) 2 (11.8) 3 (9.4) Smoking history (%) 0.541 1.000 Yes 39 (69.6) 26 (66.7) 13 (76.5) 22 (68.8) No 17 (30.4) 13 (33.3) 4 (23.5) 10 (31.3) Alcohol history (%) 0.395 0.240 Yes 31 (55.4) 20 (51.3) 11 (64.7) 21 (65.6) No 25 (44.6) 19 (48.7) 6 (35.3) 11 (34.4) Location (%) 0.833 0.629 Pyriform sinus 48 (85.7) 34 (87.2) 14 (82.4) 25 (78.1) Posterior pharyngeal wall 5 (8.9) 3 (7.7) 2 (11.8) 5 (15.6) Post cricoid 3 (5.4) 2 (5.1) 1 (5.9) 2 (6.2) Concurrent Chemotherapy 0.513 1.000 Yes 41 (73.2) 27 (69.2) 14 (82.4) 24 (72.7) No 15 (26.8) 12 (30.8) 3 (17.6) 9 (27.3) Radiation technology 0.519 0.585 IMRT 54 (96.4) 38 (97.4) 16 (94.1) 31 (93.9) 2-D RT /3-D CRT 2 (3.6) 1 (2.6) 1 (5.9) 2 (6.1) b) Dosimetric characteristics of patients in the training and testing cohorts for multi-omics study Dosimetric Characteristics All Patients ( n = 48) Training Cohort ( n = 33) Testing Cohort ( n = 15) p Radiation EQD2 (Gy), Median (IQR) 60.00 (60.00-66.00) 60.00 (60.00-66.00) 60.00 (60.00-66.00) 0.785 Dmax (Gy), Median (IQR) 72.91 (70.35-74.30) 73.22 (70.64-74.48) 72.11 (68.33-74.21) 0.392 Dmin (Gy), Median (IQR) 16.29 (2.68-38.33) 15.26 (1.87-36.72) 19.08 (3.54-43.57) 0.648 Dmean (Gy), Median (IQR) 62.40 (61.71-62.77) 62.23 (61.57-62.84) 62.43 (62.22-62.69) 0.484 Dose distribution SD, Median (IQR) 2.04 (1.58-2.95) 2.26 (1.72-2.98) 1.59 (1.39-2.17) 0.024 Homogeneity Index, Median (IQR) 0.75 (0.42-0.96) 0.76 (0.44-0.97) 0.74 (0.35-0.95) 0.681 D90 (Gy), Median (IQR) 60.59 (59.25-60.89) 60.47 (56.94-60.94) 60.64 (60.49-60.83) 0.417 D95 (Gy), Median (IQR) 59.85 (58.31-60.25) 59.69 (55.77-60.28) 59.93 (59.70-60.23) 0.186 D100 (Gy), Median (IQR) 16.29 (2.68-38.33) 15.26 (1.87-36.72) 19.08 (3.54-43.57) 0.648 V50 (%), Median (IQR) 99.87 (99.66-99.97) 99.84 (99.63-99.94) 99.94 (99.87-99.98) 0.053 V55 (%), Median (IQR) 99.54 (98.47-99.80) 99.42 (97.98-99.69) 99.66 (99.51-99.94) 0.018 V60 (%), Median (IQR) 94.06 (85.14-95.95) 93.77 (80.04-95.81) 94.69 (93.43-96.16) 0.217 V65 (%), Median (IQR) 3.93 (1.60-8.49) 4.26 (1.61-15.09) 2.85 (1.51-6.18) 0.368 V70 (%), Median (IQR) 1.00 (0.09-1.03) 1.00 (0.08-1.06) 1.00 (0.10-1.02) 0.569 Table 2 The results of feature selection from LASSO regression Features Category Coefficient Radiomics rad_squareroot_glcm_Imc1 texture -0.172948 rad_lbp-3D-k_glszm_ZoneEntropy texture -0.065080 rad_gradient_firstorder_Skewness first order 0.002844 rad_wavelet-LHH_glcm_InverseVariance texture -0.018525 rad_squareroot_glszm_GrayLevelNonUniformity texture -0.011088 rad_lbp-3D-k_gldm_SmallDependenceEmphasis texture 0.016451 Multi-omics dos_logarithm_glcm_Imc1 texture -0.008965 dos_logarithm_ngtdm_Coarseness texture -0.056607 rad_gradient_glcm_Imc2 texture 0.006655 rad_lbp-3D-k_gldm_SmallDependenceEmphasis texture 0.001976 rad_lbp-3D-k_glszm_ZoneEntropy texture -0.019096 rad_squareroot_glcm_Imc1 texture -0.021856 rad_squareroot_glszm_GrayLevelNonUniformity texture -0.003041 rad_wavelet-LLH_gldm_DependenceEntropy texture -0.010745 Table 3 The performance of each classifier in the training, testing, and external validation cohorts. Classifiers Training (Average) Testing External Validation Radiomics SVM model 0.790 0.72 0.57 DL model 0.868 0.58 0.62 LR model 0.871 0.67 0.61 RF model 0.862 0.75 0.65 EL_Boosting 0.881 0.76 0.72 EL_Stacking 0.907 0.79 0.78 EL_Bagging 0.896 0.76 0.69 EL_Voting 0.929 0.72 0.70 Multi-omics EL_Boosting 0.960 0.86 N/A EL_Stacking 0.963 0.88 EL_Bagging 0.966 0.84 EL_Voting 0.947 0.77 Additional Declarations No competing interests reported. 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-5861722","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":406715569,"identity":"1132b7a0-e594-48e5-ab33-2b1653dc1695","order_by":0,"name":"Sixue Dong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIiWNgGAWjYBAC+WYw9QeImQ8ACQsGBh4CWgwOg6ljQMyWwMCQIEGEFggF0shjQKQWdh4ziQ8Mh+XM+dd8fMz7Q0KOn+cA44ePOfj8wpYmOYPhsLHljLebjXkSJIwlexuYJWduw2PNYeZj0jwMhxM33Di7TRqoJXHDeQY2Zl68WhjbpP+AtZx5RqwWoC0MIC3ne9ggWs424NdicJgt2bLH4JixwQ02Y8M5aUC/9BxsxusX+f4zhjd+VPyRMzh/+OGDNzY2wBBLPvjhIz6HQewCYokEGI+xgZB6KOA/QKTCUTAKRsEoGHEAALh4Tj9bG6H/AAAAAElFTkSuQmCC","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":true,"prefix":"","firstName":"Sixue","middleName":"","lastName":"Dong","suffix":""},{"id":406715570,"identity":"97610053-8835-4088-953c-fd870e3a1def","order_by":1,"name":"Zian Yao","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Zian","middleName":"","lastName":"Yao","suffix":""},{"id":406715571,"identity":"ad5e7767-6808-4c08-9505-7f82dff79aa4","order_by":2,"name":"Zhiyuan Zhang","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Zhiyuan","middleName":"","lastName":"Zhang","suffix":""},{"id":406715572,"identity":"a8f89c84-31c4-4565-8d83-954344fc1f9c","order_by":3,"name":"Jiazhou Wang","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Jiazhou","middleName":"","lastName":"Wang","suffix":""},{"id":406715573,"identity":"17683e43-3703-4372-b498-1faa16da359d","order_by":4,"name":"Guo Ying","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Guo","middleName":"","lastName":"Ying","suffix":""},{"id":406715574,"identity":"b26ab9e3-7a8a-4572-b615-8b26ef30d475","order_by":5,"name":"Lei Tao","email":"","orcid":"","institution":"Eye \u0026 ENT Hospital of Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Tao","suffix":""},{"id":406715575,"identity":"d6309b88-8da7-4377-a1d1-6a27f5a0ad70","order_by":6,"name":"Xiaomin Ou","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Xiaomin","middleName":"","lastName":"Ou","suffix":""},{"id":406715576,"identity":"b24e34eb-5126-4051-ad0d-421563894496","order_by":7,"name":"Weigang Hu","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Weigang","middleName":"","lastName":"Hu","suffix":""},{"id":406715577,"identity":"d166c348-e49d-4d11-9953-23e45c1bc9b4","order_by":8,"name":"Chaosu Hu","email":"","orcid":"","institution":"Fudan University Shanghai Cancer Center","correspondingAuthor":false,"prefix":"","firstName":"Chaosu","middleName":"","lastName":"Hu","suffix":""}],"badges":[],"createdAt":"2025-01-20 01:23:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5861722/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5861722/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":75311974,"identity":"86609dd1-52c5-4d27-8f4d-828ec49ff18b","added_by":"auto","created_at":"2025-02-03 09:10:44","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":89255,"visible":true,"origin":"","legend":"\u003cp\u003eThe workflow of the study.\u003c/p\u003e\n\u003cp\u003ea) The enrollment diagram for radiomics-only and multi-omics analysis. b) The workflow chart of the present study.\u003c/p\u003e","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5861722/v1/846f9d470ecd3fdf278421fc.png"},{"id":75310474,"identity":"61d55439-c9ba-4aff-affb-8e5ea0acff7a","added_by":"auto","created_at":"2025-02-03 09:02:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":106239,"visible":true,"origin":"","legend":"\u003cp\u003eThe performance of different classifiers in the radiomics-only analysis.\u003c/p\u003e\n\u003cp\u003ea) The ROCs of the 8 different classifiers in the training (maximum values), testing, and external validation cohorts, respectively, and the AUCs were shown in the legend. b) The performance of the EL model with the stacking technique was superior to other models and selected to plot the K-M survival curves in the three cohorts, and the corresponding P-values were shown. c) The p-values between the AUC values of different classifiers in radiomics.\u003c/p\u003e","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5861722/v1/0f0b8f39138610ca8de82d70.png"},{"id":75310471,"identity":"842bc2d4-aa0d-4411-b02a-63d7a8be6e4b","added_by":"auto","created_at":"2025-02-03 09:02:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":44854,"visible":true,"origin":"","legend":"\u003cp\u003eThe performance of different classifiers in the multi-omics analysis.\u003c/p\u003e\n\u003cp\u003ea) The ROCs of the 4 EL models with different techniques in the training (maximum values), and testing cohorts, respectively, and the AUCs were shown in the legend. b) The performance of the EL model with the stacking technique was superior to other models and selected to plot the K-M survival curves in the two cohorts, and the corresponding P-values were shown. c) The p-values between the AUC values of different classifiers in multi-omics\u003c/p\u003e","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5861722/v1/fe299562dd57686df67e5821.png"},{"id":75310478,"identity":"a3d54a98-c87b-4f15-b921-aad2eb9e3a98","added_by":"auto","created_at":"2025-02-03 09:02:44","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":24436,"visible":true,"origin":"","legend":"\u003cp\u003eA multi-omics-based nomogram was developed to predict the 1-year, 2-year, 3-year, and 4-year PFSs of the hypopharyngeal cancer with postoperative chemoradiotherapy. The density or box plot illustrated the distribution of age, overall stage, multi-omics signature, positive probability, and total points. The selected sample for presentation was 60 years old with an overall stage of Ⅲ, exhibiting a prediction of 69.7%, 52.9%, 43.2%, and 34.0% for 1-year (\u0026gt;12), 2-year (\u0026gt;24), 3-year (\u0026gt;36), and 4-year (\u0026gt;48) PFSs, respectively.\u003c/p\u003e","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5861722/v1/4f5bab45779eef50a46fe964.png"},{"id":77205832,"identity":"b86fc837-58fb-4a76-8324-0506d9dc1bec","added_by":"auto","created_at":"2025-02-26 08:17:13","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1361028,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5861722/v1/c21bdbf3-e2ac-4483-8d36-e73bea1dc131.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multi-omics-based prognostic prediction for locally advanced hypopharyngeal cancer treated with postoperative chemoradiotherapy: a dual-center study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe incidence of hypopharyngeal cancer constitutes approximately 3\u0026ndash;5% of head and neck cancers, with higher rates observed in China and Eastern Europe. According to data from the Taiwan Cancer Registry, the crude incidence rate of hypopharyngeal cancer is 5.15 per 100,000 individuals[1, 2].\u003c/p\u003e \u003cp\u003eHypopharyngeal cancer represents a malignancy characterized by its heightened aggressiveness, which (70\u0026ndash;85%) are diagnosed at an advanced stage and are characterized by indistinct symptoms[3, 4]. The National Comprehensive Cancer Network (NCCN) recommends the implementation of a comprehensive multidisciplinary approach, incorporating chemotherapy, surgery, and radiotherapy, to optimize functional preservation and improve patient outcomes[5]. Postoperative chemoradiotherapy was recommended as one of the treatment strategies in the guidelines and correlated with the survival benefit, however, the prognosis was suboptimal with a 5-year overall survival (OS) rate of approximately 40% and the discrepancies were in patient outcome[6\u0026ndash;8]. Consequently, the acquisition of follow-up data poses challenges, and the limited sample size also presents a significant challenge to the performance of the prognostic prediction model.\u003c/p\u003e \u003cp\u003eIn recent years, the integration of omics analysis into clinical models has significantly advanced cancer research. In a one-year PFS prediction study of hypopharyngeal cancer, the RF classifier based solely on clinical features without incorporating omics data was only able to improve the area under the receiver operating characteristic (ROC) curve (AUC) to a maximum of 0.739 in the training cohort[9]. In contrast, radiomics-based prognostic models have gained significant attention in cancer research. Radiomics entails the analysis and mining of numerous features derived from medical imaging data[10\u0026ndash;12]. By analyzing imaging data of postoperative chemoradiotherapy patients, radiomics signatures related to tumor characteristics, treatment response, and prognosis can be extracted. These signatures, along with clinical data such as age, gender, and tumor stage, can be utilized to construct predictive model, aims to assess the prognostic risk of patients and identify high-risk individuals[13\u0026ndash;15]. Although previous studies have predominantly relied on computed tomography (CT) signatures, and demonstrated robustness and predictive utility for various clinical endpoints, their efficacy is limited by suboptimal soft tissue contrast[1]. Magnetic resonance imaging (MRI) provides enhanced visualization of soft tissues and superior anatomical resolution, typically used for the staging of head and neck cancers[16\u0026ndash;19]. Previous studies utilized MRI-based radiomics to predict progression-free survival (PFS) in nasopharyngeal carcinoma (NPC) with different treatment regimens by collecting 1872 samples from four centers, and achieved a reliable nomogram with the C-index range from 0.784 to 0.921[20]. Additionally, multi-omics holds immense promise in advancing the prognosis and treatment management of individuals afflicted with cancer[21, 22]. Related studies showed that the introduction of dosiomics is beneficial for effectively predicting the prognosis of tumor recurrence and radioactive diseases (such as radiation esophagitis and radiation pneumonitis)[23\u0026ndash;25].\u003c/p\u003e \u003cp\u003eConsidering the fewer studies of prognostic prediction of postoperative chemoradiotherapy for hypopharyngeal cancer, and the bottleneck in radiomics-based model performance, a more effective method was required to be developed for facilitating the clinical decision. This study endeavors to use multi-omics method which integrate dosimetric, radiomic, and clinical factors to predict PFS of the patients who were diagnosed with hypopharyngeal cancer and received postoperative chemoradiotherapy.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatients and data collection\u003c/h2\u003e \u003cp\u003eThis is a retrospective study of patients with locally advanced hypopharyngeal cancer treated by surgery and postoperative chemoradiotherapy. Inclusion criteria: patients diagnosed between January 2015 and December 2021, pathologically confirmed, resectable locally advanced hypopharyngeal squamous cell carcinoma (T2-4a, N0-resectable N3, M0), preoperatively examined by contrast enhanced MRI with complete imaging data, and treated with surgery and postoperative chemoradiotherapy. Exclusion criteria: stage I/II patients, T4b patients, suffered from malignant tumors except cervical carcinoma in situ, papillary thyroid carcinoma, or skin cancer (non- melanoma) within five years, distant metastasis before treatment, and incomplete treatment or missing clinical data. The enrollment diagram was shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn the radiomics-only analysis of this study, 56 patients with hypopharyngeal cancer treated between 2015 and 2021 were selected from Fudan University Shanghai Cancer Center, and were divided into a training cohort (n\u0026thinsp;=\u0026thinsp;39) and a testing cohort (n\u0026thinsp;=\u0026thinsp;17) at a 7:3 ratio. Additionally, an external validation cohort of 32 hypopharyngeal cancer patients treated between 2015 and 2021 was selected from Eye and ENT Hospital of Fudan University. In the multi-omics analysis integrating radiomics and dosiomics of this study, 8 patients without complete radiotherapy data were excluded.\u003c/p\u003e \u003cp\u003eAll patients underwent a follow-up period of at least three years, with an interval of six months. Imaging follow-up, either with CT and/or MRI, was performed alternatively at intervals of 3 months during the first year after treatment and 6 months during the second year. OS (defined as the time between the date of initial pathologic diagnosis and the date of death or the last follow-up) and PFS (defined as the duration between the date of initial pathologic diagnosis and the date of the first observed sign of disease progression, death, or the last follow-up) were considered the most important indicators in the present follow-up assessments.\u003c/p\u003e \u003cp\u003eThis retrospective study received approval from the Ethics Committees of Fudan University Shanghai Cancer Center (A) and Eye \u0026amp; ENT Hospital of Fudan University\u003c/p\u003e \u003cp\u003eThe subsequent workflow of this study was illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb). The radiomics and dosiomics data was prepared for feature extraction and selection. Omics-based signatures and the positive sample probabilities calculated from the classifiers were obtained to do the survival analysis.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eFeature extraction and selection\u003c/h3\u003e\n\u003cp\u003eMRI images of the patients before the surgery were obtained using a 3.0T Magnetom Skyra (Siemens, Germany). The gross tumor volume (GTV, including the primary tumor and metastatic lymph nodes) was countered as the region of interest (ROI) on MIM Software (v7.3.6). by a head and neck radiologist with the title of associate chief physician and 15 years of clinical practice. The planning target volume (PTV) was countered on the CT images and intensity modulated radiation therapy (IMRT) and/or volumetric modulated arc therapy (VMAT) technique was utilized during treatment planning with the Pinnacle\u003csup\u003e3\u003c/sup\u003e (v9.10) treatment planning system (TPS). The radiotherapy process was conducted by Varian Vital Beam linear accelerator (6 MV, 600 MU/min).\u003c/p\u003e \u003cp\u003eAn in-house Python-based script were developed to extract the radiomics and dose distribution features by using Pyradiomics (v3.1.0) package. In total of 1688 features were extracted for the two modalities, respectively including first order features (Energy, standard deviation, uniformity, etc.), shape features (2D and 3D), texture features (Gray Level Co-occurrence Matrix, GLCM. Gray Level Run Length Matrix, GLRLM. Gray Level Size Zone Matrix, GLSZM. Neighborhood Gray Tone Difference Matrix, NGTDM. Gray Level Dependence Matrix, GLDM)[26]. During the preprocessing of images, sitkBSpline interpolator of Pyradiomics package was used to resample all images to a uniform image spacing of 1\u0026times;1\u0026times;1 mm\u003csup\u003e3\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBesides the dose distribution features, dosimetric parameters based on dose-volume histogram (DVH) were also collected along with the PTV in the dosiomics analysis, including D\u003csub\u003e90\u003c/sub\u003e-D\u003csub\u003e100\u003c/sub\u003e (i.e., the minimum absolute dose covering 90%, 95%, 100% of PTV volume, respectively), V\u003csub\u003e50\u003c/sub\u003e-V\u003csub\u003e70\u003c/sub\u003e (i.e., the maximum relative PTV volume covered by 50Gy, 55Gy, 60Gy, 65Gy, 70Gy), homogeneity index (HI), standard deviation (SD) of dose distribution, maximum absolute dose (Dmax), minimum absolute dose (Dmin), and mean absolute dose (Dmean).\u003c/p\u003e \u003cp\u003eFeature standardization by the StandardScaler command of the scikit-learn (v1.4.1) package was proceeded prior to feature selection. Least absolute shrinkage and selection operator (LASSO) regression with ten-fold cross-validation was used to identify the most important features and prevent overfitting, thereby enhancing model generalization and interpretability. Furthermore, the radiomics signature and the multi-omics signature were obtained separately.\u003c/p\u003e\n\u003ch3\u003eModel establishment and statistical analysis\u003c/h3\u003e\n\u003cp\u003eIn the radiomics-only analysis, four single-classifiers including support vector machine (SVM) model, deep learning (DL) model, logistic regression (LR) model, and random forest (RF) model were trained. Furthermore, ensemble learning (EL) models integrated by the four single-classifiers with four techniques including boosting, stacking, bagging, and voting were also trained. On the other hand, the four EL models were trained in the multi-omics analysis. For all of the models, the hyperparameters were tuned to achieve the best F1 score by using Bayesian optimization. The parameter \"random_state\" of each classifier was randomly modified 100 times to achieve various AUC values, and the inter-model differences were evaluated by Student\u0026rsquo;s t-test. The parameters \"random_state\" which yielded the highest AUC values were employed to establish various classifiers. P-values (A classification result with P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered to be statistically significant) of the Kaplan\u0026ndash;Meier (K-M) survival analysis were evaluated in the training, testing, and external validation cohorts, respectively.\u003c/p\u003e \u003cp\u003eThe probabilities of positive sample were calculated from the best classifier. Single-factor Cox regression was used to identify the most important clinical factors, and two Cox proportional hazards models were established by incorporating the selected clinical factors with the signatures and positive sample probabilities of radiomics-only and multi-omics analysis separately. The C-indexes were evaluated, and the multi-omics-based nomogram were established.\u003c/p\u003e"},{"header":"Results","content":"\u003ch2\u003eClinical characteristics\u003c/h2\u003e\n\u003cp\u003eThe study samples consisted of 88 male patients with locally advanced hypopharyngeal carcinoma. The clinical characteristics of training, testing, and external validation cohorts with (radiomics-only) were shown in Table 1 a, the corresponding median ages (age ranges) were 63.3 (48-77) years, 62.5 (46-76) years, and 61.84 (46-76) years, respectively. The median follow-up duration of the whole cohort was 36 months (range, 3\u0026ndash;92 months). Tumor regions from both cohorts were mostly located in the pyriform sinus, respectively with a proportion of 87.2% and 82.4%. Overall, 30 (34.1%) patients were classified as stage III (AJCC 8th), while 51 (58.0%) were stage IVA and 7 (8.0%) were stage IVB. Upon the last follow-up, 7 (17.9%) in the training cohort and 4 (23.5%) patients in the testing cohort experienced a confirmed disease progression. Particularly, there were no significant differences (P \u0026gt; 0.05) in all of the terms.\u003c/p\u003e\n\u003cp\u003eTable 1 b illustrates the dosimetric characteristics of the population. Specifically, the median EQD2 dose across all patients was 60 Gy (inter quartile range [IQR], 60\u0026ndash;66 Gy). Moreover, other dosimetric parameters including D90-D10, V50-V70, homogeneity index (HI), standard deviation (SD) of dose distribution, maximum absolute dose (Dmax), minimum absolute dose (Dmin), and mean absolute dose (Dmean) of training and testing cohorts were shown. Notably, no significant differences (P \u0026gt; 0.05) were observed between the two cohorts, with exception of SD of dose distribution (p = 0.024) and V55 (p = 0.018).\u003c/p\u003e\n\u003ch2\u003eClassifiers comparison of radiomics-only and multi-omics\u003c/h2\u003e\n\u003cp\u003eThe features selected by using LASSO regression were shown in Table 2 with their coefficients. The ROCs comparison derived from various classifiers of radiomics and multi-omics were shown in Figure 2 and Figure 3, respectively.\u003c/p\u003e\n\u003cp\u003eThe performance of each classifier was recorded in Table 3, Figure 2, and Figure 3. The radiomics-based classifiers were evaluated in three cohorts: training, testing, and external validation cohorts. The multi-omics-based classifiers were evaluated in two cohorts: training and testing cohorts.\u003c/p\u003e\n\u003cp\u003eInitially, four single-classifiers including SVM, DL, LF, RF models were tested in the radiomics-only analysis, the AUC values for various single-classifiers ranged from 0.85 to 0.93, 0.58 to 0.75, and 0.57 to 0.65 in the training, testing, and external validation cohorts, respectively. Subsequently, four EL models with various techniques were added for further comparison against the single-classifiers. In general, the EL model performances were superior to each single-classifiers, and the AUC values ranged from 0.93 to 0.99 in the training cohort. While in the testing and external validation cohorts, the AUC values range from 0.72 to 0.79 and from 0.69 to 0.78, respectively. The highest AUC value of 0.99 in the training cohort was achieved by using the voting technique, while using the stacking technique could yield the optimal AUC in testing (0.79) and external validation cohorts (0.78). The K-M survival curves derived from the EL model with the stacking technique with were displayed to identify the high- and low-risk groups in Figure 2 b), and the P-values \u0026lt; 0.05 in all the cohorts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the multi-omics analysis, the AUC values for various EL models ranged from 0.98 to 0.99 in the training cohort. While in the testing cohort, the AUC values range from 0.77 to 0.88. The integration of dosiomics could enhance the average AUC values by 0.038\u0026plusmn;0.022 and 0.080\u0026plusmn;0.019 in the training and testing cohorts, respectively. The EL model with the stacking technique still exhibited superior performance with the AUCs of 0.98 and 0.88, and the corresponding K-M survival curves were also displayed in Figure 3 b), the P-values in the training and testing cohorts were both lower than those in the radiomics-only analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBy performing different \u0026quot;random_state\u0026quot;, the P-value between the AUC values of various classifiers were lower than 0.05 except in the cases of the DL model versus LR and RF models.\u003c/p\u003e\n\u003ch2\u003eEvaluation of Cox proportional hazards models\u003c/h2\u003e\n\u003cp\u003eThe valuable clinical factors including overall stage and age were identified by using single-factor Cox regression. According to the performance of classifiers, EL model with stacking technique was chosen to calculate the positive sample probabilities, in addition, incorporating the signatures derived from the LASSO regression and the valuable clinical factors, two Cox proportional hazards models were established for radiomics-only and multi-omics analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the radiomics-only analysis, the C-indexes were 0.777, 0.685, and 0.663 in the training, testing, and external validation cohorts, respectively. On the other hand, in the multi-omics analysis, the C-indexes were 0.783 and 0.784 in the training and testing cohorts, respectively. The multi-omics-based nomogram (Figure 4) were established to predict the 1-year, 2-year, 3-year, and 4-year PFSs. The prediction of a given sample were also shown in the nomogram, the 1-year, 2-year, 3-year, and 4-year PFSs were 69.7%, 52.9%, 43.2%, and 34.0%, respectively.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eClinical factors played a crucial role in the conventional prognostic investigation of hypopharyngeal cancer. Previous studies demonstrated the treatment modalities were correlated with OS and cancer-specific survival (CSS), while postoperative chemoradiotherapy stood out as the most beneficial treatment modality for enhancing the OS and CSS of patients[27\u0026ndash;29]. However, despite the generally favorable prognosis for patients, the high recurrence rate remained one of the significant factors affecting survival and quality of life, particularly for advanced hypopharyngeal cancer, and the analysis of PFS became necessary. Thus far, most of the studies on prognostic predictions for hypopharyngeal cancer have primarily focused on surgery and/or adjuvant chemotherapy, with the method largely relying on radiomics-only analysis[2, 30\u0026ndash;32]. In the present study, a prediction for the PFS of hypopharyngeal cancer with postoperative chemoradiotherapy based on the multi-omics method was developed. The results showed the analysis of multi-omics including radiomics and dosiomics exhibited superior performance than the radiomics-only analysis, whether in the classifier model (with an average AUC of 0.828\u0026thinsp;\u0026plusmn;\u0026thinsp;0.040 in the testing cohort) or the Cox proportional hazards model (with a C-index of 0.784 in the testing cohort).\u003c/p\u003e \u003cp\u003eTargeting patients who have received radiotherapy, the recurrence of tumors correlates intricately with the dose distribution within the PTV, and the integration of dosiomics could emerge as the significant factor for predicting PFS [33\u0026ndash;35]. Our present study focused on the method the multi-omics, compared the performance of omics predictive models with and without the inclusion of dosimetric omics, and successfully demonstrated the robust correlation between dosimetric parameters and PFS. It should be noted that the shape of the target volume and the radiotherapy planner varies among individual patients, thus substantial variations exist in dosimetric parameters such as HI, Dmax, and Dmin, however, none of the DVH-based dosimetric features were identified. Based on these findings, there is justification to suggest that there is no significant correlation between the other DVH parameters and PFS when the target area receives adequate coverage at the prescribed dose(D\u003csub\u003e95%\u003c/sub\u003e\u0026gt;100% or V\u003csub\u003e95%\u003c/sub\u003e\u0026gt;100%). Moreover, the category of selected features with LASSO regression were almost \u0026ldquo;texture\u0026rdquo;, representing the heterogeneity of the tumor, and higher heterogeneity with irregular margins were typically associated with an elevated risk of tumor recurrence and reduced the PFS. During the radiotherapy, the technique of VMAT could yield a more homogeneous dose distribution at the target area with a smoother edge compared to IMRT, while two texture features of dosiomics were selected with negative correlation coefficients. In our perspective, VMAT might could potentially be a better option in terms of mitigating the risk of recurrence. By incorporating a greater number of samples from diverse centers and additional factors related to radiotherapy, further research is being conducted to substantiate our perspective and enhance the generalization capability of the model.\u003c/p\u003e \u003cp\u003eDuring the process of classifier modeling, three machine learning techniques (SVM, LR, RF) and one DL technique were employed initially, while the generalization abilities in both the testing and external validation cohorts were not satisfactory. Therefore, four EL methods were introduced to modeling the classifiers, it could reduce the prediction generalization error and make more accurate predictions than a single-classifier. Nevertheless, this enhancement came at the expense of sacrificing the model interpretability[36\u0026ndash;38]. Although the AUC value of the EL model with stacking technique in the training cohort was not the highest (0.93 in the radiomics-only analysis and 0.98 in the multi-omics analysis, even the lowest among EL models), it achieves the highest AUC in both the testing and external validation cohorts, that indicated the stacking technique demonstrated relatively strong generalization ability while maintaining modeling accuracy, in other words, it exhibited more balance performance than other techniques. Therefore, in this study, it was regarded as the best classifier for estimating the probabilities of the positive sample. Besides, in the figures of K-M survival curve, the classifier could also successfully divide high- and low-risk groups in each cohort (P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Benefited to the incorporation of dosiomics, there was a marginal improvement in the AUC value in the training cohort (0.038\u0026thinsp;\u0026plusmn;\u0026thinsp;0.022), whereas a more substantial improvement in the testing cohort (0.080\u0026thinsp;\u0026plusmn;\u0026thinsp;0.019). A more satisfactory outcome in the external validation cohort was anticipated and necessitated to be confirmed in subsequent research.\u003c/p\u003e \u003cp\u003eMost previous studies on survival analysis have incorporated omics signatures and/or risk levels to enhance the predictive performance of the models[39\u0026ndash;42], however, the categorization of risk levels (e.g., high-, medium-, and low-risk) might be somewhat imprecise due to its binary or ternary nature. This study integrated the positive sample probabilities predicted by the EL classifier into a Cox proportional hazards model as a continuous indicator, aiming to combine machine learning with statistical analysis to obtain more refined predictive results. Age and overall stage are considered the two most highly correlated clinical factors with PFS. Elderly patients with advanced-stage hypopharyngeal cancer exhibited higher risk scores and were associated with lower probabilities of PFS. Previous studies showed that clinical factors-based Cox proportional hazards model for predicting OS achieved a c-index of approximately 0.72[43, 44], when upon the incorporated of radiomics, the C-index increased to 0.78[18]. Our presented study focused on predicting the PFS, and the introduction of radiomics achieves a comparable performance to predicting the OS (C-index\u0026thinsp;=\u0026thinsp;0.777 in the training cohort), while its capacity for generalization is somewhat constrained (C-index\u0026thinsp;=\u0026thinsp;0.685 in the testing cohort). With the further research, the introduction of multi-omics signatures had addressed the issue of poor generalization, and achieved an increase in the C-index to 0.784 in the training cohort.\u003c/p\u003e \u003cp\u003eMulti-omics methodology was successfully employed in this study to establish predictive models for PFS in patients with hypopharyngeal cancer and received postoperative chemoradiotherapy, moreover, the comparative advantages with radiomics-only methodology were analyzed. Accordingly, for surgical patients diagnosed with locally advanced hypopharyngeal carcinoma, the multi-omics prognostic model has provided essential biomarkers to select patients treated with surgery and postoperative chemoradiotherapy to address hypopharyngeal cancer who likely suffered early tumor progression, thus quantitatively evaluating the improvement of clinicians' decision-making and patients' outcome. Nevertheless, there were still some deficiencies and limitations: 1) As a dual-center study, this study had not yet obtained complete radiotherapy data in the external validation cohort, leading to a missing link in the external validation of multi-omics analysis. 2) Meanwhile, due to the relatively small number of patients with hypopharyngeal cancer and the other treatment modalities were excluded apart from postoperative chemoradiotherapy, there were only 39 and 33 samples for training the radiomics-based and multi-omics-based predictive model, respectively. The limited sample size might potentially influence the accuracy and generalization ability of the predictive model. 3) All patients included in the study were males aged 46 years and above, despite the lower incidence of hypopharyngeal cancer among younger populations and females, it still needed to be considered to incorporate them into the predictive model. 4) Insufficient prospective studies for validation. Concerning these limitations, it was necessary to expand the sample size and integrate more clinical factors such as gender (male or female), radiotherapy techniques (IMRT, VMAT, helical tomotherapy, or 3D-conformal radiotherapy), age (\u0026le;\u0026thinsp;45), etc. It was also required to Optimize the model performance and perform validations by using multicenter data to translate the multi-omics methodology into clinical practice in our further studies.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this retrospective study, a multi-omics methodology, integrating clinical parameters, dosimetric, and radiomic features, was harnessed to establish predictive models to predict the PFS of patients with hypopharyngeal cancer who received postoperative chemoradiotherapy. The EL model with the stacking technique demonstrated superior performance in classifying high- and low-risk groups. Furthermore, multi-omics significantly enhanced the predictive performance and generalization ability of classifiers and Cox proportional hazards models. The presented models could facilitate the clinical decision by assessing the requirement for postoperative chemotherapy based on varied probabilities of time-dependent PFS and risk levels. Despite limitations such as limited sample size, our research effectively demonstrated the potential clinical utility of multi-omics analysis in the prognostic prediction of hypopharyngeal cancer. Future prospective, multicenter studies were needed to refine and validate this clinical translational approach.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePFS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eprogression-free survival\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGTV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003egross tumor volume\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLASSO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eleast absolute shrinkage and selection operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePTV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eplanning tumor volume\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eEL\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eensemble learning\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ereceiver operating characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003earea under the receiver operating characteristic curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNCCN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNational Comprehensive Cancer Network\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eoverall survival\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ecomputed tomography\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMRI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emagnetic resonance imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNPC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003enasopharyngeal carcinoma\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIMRT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eintensity modulated radiation therapy\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eVMAT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003evolumetric modulated arc therapy\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTPS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etreatment planning system\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDmax\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emaximum absolute dose\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDmin\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eminimum absolute dose\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDmean\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emean absolute dose\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003esupport vector machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDL\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003edeep learning\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elogistic regression\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003erandom forest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eK-M\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eKaplan\u0026ndash;Meier\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIQR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003einter quartile range\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCSS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ecancer-specific survival\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate:\u0026nbsp;\u003c/strong\u003eAll procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Fudan University Shanghai Cancer Center. Informed consent or substitute for it was obtained from all patients for being included in the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication:\u0026nbsp;\u003c/strong\u003eThe Author confirms: that the work described has not been published before; that it is not under consideration for publication elsewhere; that its publication has been approved by all co-authors; that its publication has been approved by the responsible authorities at the institution where the work is carried out.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and material:\u0026nbsp;\u003c/strong\u003eDue to the nature of this research, the participants did not agree to share their data publicly, so supporting data is not available.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u0026nbsp;\u003c/strong\u003eThere are no conflicts of interest or financial ties to disclose from any author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eChinese Society of Clinical Oncology Foundation (Y-Young2021\u0026ndash;0127, to Xiaomin Ou).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthor initials: Sixue Dong (Sx. D), Zian Yao (Za. Y), Zhiyuan Zhang (Zy. Z), Jiazhou Wang (Jz. W), Guo Ying (G. Y), Lei Tao (L. T), Xiaomin Ou (Xm. O), Weigang Hu (Wg. H), Chaosu Hu (Cs. H)\u003c/p\u003e\n\u003cp\u003eSx. D, Methodology, Experimental procedures, Data Analysis, Writing, Editing,\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eZa. Y, Data Collection, Review and Editing\u003c/p\u003e\n\u003cp\u003eZy. Z, Methodology, Review\u003c/p\u003e\n\u003cp\u003eJz. W, Methodology, Experimental procedures, Review\u003c/p\u003e\n\u003cp\u003eG. Y, Methodology, Review\u003c/p\u003e\n\u003cp\u003eL. T, Data Collection\u003c/p\u003e\n\u003cp\u003eXm. O, Data Collection, funding acquisition, Review\u003c/p\u003e\n\u003cp\u003eWg. H, Conceptualization, Supervision\u003c/p\u003e\n\u003cp\u003eCs. H, Conceptualization, Supervision\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u0026nbsp;\u003c/strong\u003eN/A.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; information:\u0026nbsp;\u003c/strong\u003eN/A.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSiow, T.Y., et al., \u003cem\u003eMRI Radiomics for Predicting Survival in Patients with Locally Advanced Hypopharyngeal Cancer Treated with Concurrent Chemoradiotherapy.\u003c/em\u003e Cancers (Basel), 2022. \u003cstrong\u003e14\u003c/strong\u003e(24).\u003c/li\u003e\n\u003cli\u003eChiesa-Estomba, C.M., et al. \u003cem\u003eRadiomics in Hypopharyngeal Cancer Management: A State-of-the-Art Review\u003c/em\u003e. Biomedicines, 2023. \u003cstrong\u003e11\u003c/strong\u003e, DOI: 10.3390/biomedicines11030805.\u003c/li\u003e\n\u003cli\u003eBuckley, J.G. and K. MacLennan, \u003cem\u003eCervical node metastases in laryngeal and hypopharyngeal cancer: a prospective analysis of prevalence and distribution.\u003c/em\u003e Head Neck, 2000. \u003cstrong\u003e22\u003c/strong\u003e(4): p. 380-5.\u003c/li\u003e\n\u003cli\u003eKuo, P., et al., \u003cem\u003eHypopharyngeal cancer incidence, treatment, and survival: temporal trends in the United States.\u003c/em\u003e Laryngoscope, 2014. \u003cstrong\u003e124\u003c/strong\u003e(9): p. 2064-9.\u003c/li\u003e\n\u003cli\u003eKwon, D.I., et al., \u003cem\u003eHypopharyngeal carcinoma: Do you know your guidelines?\u003c/em\u003e Head \u0026amp; Neck, 2019. \u003cstrong\u003e41\u003c/strong\u003e(3): p. 569-576.\u003c/li\u003e\n\u003cli\u003eNewman, J.R., et al., \u003cem\u003eSurvival trends in hypopharyngeal cancer: a population-based review.\u003c/em\u003e Laryngoscope, 2015. \u003cstrong\u003e125\u003c/strong\u003e(3): p. 624-9.\u003c/li\u003e\n\u003cli\u003eJanoray, G., et al., \u003cem\u003eLong-term Results of a Multicenter Randomized Phase III Trial of Induction Chemotherapy With Cisplatin, 5-fluorouracil, \u0026plusmn; Docetaxel for Larynx Preservation.\u003c/em\u003e J Natl Cancer Inst, 2016. \u003cstrong\u003e108\u003c/strong\u003e(4).\u003c/li\u003e\n\u003cli\u003eBurbure, N., et al., \u003cem\u003ePrognostic significance of human papillomavirus status and treatment modality in hypopharyngeal cancer.\u003c/em\u003e Head Neck, 2021. \u003cstrong\u003e43\u003c/strong\u003e(10): p. 3042-3052.\u003c/li\u003e\n\u003cli\u003eZhang, Y.F., et al., \u003cem\u003ePredicting survival of advanced laryngeal squamous cell carcinoma: comparison of machine learning models and Cox regression models.\u003c/em\u003e Sci Rep, 2023. \u003cstrong\u003e13\u003c/strong\u003e(1): p. 18498.\u003c/li\u003e\n\u003cli\u003eQi, M., et al., \u003cem\u003eAn MRI-based radiomics nomogram for detecting cervical esophagus invasion in hypopharyngeal squamous cell carcinoma.\u003c/em\u003e Cancer Imaging, 2023. \u003cstrong\u003e23\u003c/strong\u003e(1): p. 120.\u003c/li\u003e\n\u003cli\u003eWu, T.-C., et al., \u003cem\u003eRadiomics analysis for the prediction of locoregional recurrence of locally advanced oropharyngeal cancer and hypopharyngeal cancer.\u003c/em\u003e European Archives of Oto-Rhino-Laryngology, 2024. \u003cstrong\u003e281\u003c/strong\u003e(3): p. 1473-1481.\u003c/li\u003e\n\u003cli\u003eRizzo, S., et al., \u003cem\u003eRadiomics: the facts and the challenges of image analysis.\u003c/em\u003e Eur Radiol Exp, 2018. \u003cstrong\u003e2\u003c/strong\u003e(1): p. 36.\u003c/li\u003e\n\u003cli\u003eMo, X., et al., \u003cem\u003ePrognostic value of the radiomics-based model in progression-free survival of hypopharyngeal cancer treated with chemoradiation.\u003c/em\u003e Eur Radiol, 2020. \u003cstrong\u003e30\u003c/strong\u003e(2): p. 833-843.\u003c/li\u003e\n\u003cli\u003eHsu, C.Y., et al., \u003cem\u003eMagnetic resonance imaging-derived radiomic signature predicts locoregional failure after organ preservation therapy in patients with hypopharyngeal squamous cell carcinoma.\u003c/em\u003e Clin Transl Radiat Oncol, 2020. \u003cstrong\u003e25\u003c/strong\u003e: p. 1-9.\u003c/li\u003e\n\u003cli\u003eGuo, R., et al., \u003cem\u003eCT-based radiomics features in the prediction of thyroid cartilage invasion from laryngeal and hypopharyngeal squamous cell carcinoma.\u003c/em\u003e Cancer Imaging, 2020. \u003cstrong\u003e20\u003c/strong\u003e(1): p. 81.\u003c/li\u003e\n\u003cli\u003eRomeo, V., et al., \u003cem\u003ePrediction of Tumor Grade and Nodal Status in Oropharyngeal and Oral Cavity Squamous-cell Carcinoma Using a Radiomic Approach.\u003c/em\u003e Anticancer Res, 2020. \u003cstrong\u003e40\u003c/strong\u003e(1): p. 271-280.\u003c/li\u003e\n\u003cli\u003eMukherjee, P., et al., \u003cem\u003eCT-based Radiomic Signatures for Predicting Histopathologic Features in Head and Neck Squamous Cell Carcinoma.\u003c/em\u003e Radiol Imaging Cancer, 2020. \u003cstrong\u003e2\u003c/strong\u003e(3): p. e190039.\u003c/li\u003e\n\u003cli\u003eChen, J., et al., \u003cem\u003eAn MRI-based radiomics-clinical nomogram for the overall survival prediction in patients with hypopharyngeal squamous cell carcinoma: a multi-cohort study.\u003c/em\u003e Eur Radiol, 2022. \u003cstrong\u003e32\u003c/strong\u003e(3): p. 1548-1557.\u003c/li\u003e\n\u003cli\u003eKatsoulakis, E., et al., \u003cem\u003eRadiomic analysis identifies tumor subtypes associated with distinct molecular and microenvironmental factors in head and neck squamous cell carcinoma.\u003c/em\u003e Oral Oncol, 2020. \u003cstrong\u003e110\u003c/strong\u003e: p. 104877.\u003c/li\u003e\n\u003cli\u003eZhong, L., et al., \u003cem\u003eA deep learning-based radiomic nomogram for prognosis and treatment decision in advanced nasopharyngeal carcinoma: A multicentre study.\u003c/em\u003e EBioMedicine, 2021. \u003cstrong\u003e70\u003c/strong\u003e: p. 103522.\u003c/li\u003e\n\u003cli\u003eBoehm, K.M., et al., \u003cem\u003eHarnessing multimodal data integration to advance precision oncology.\u003c/em\u003e Nat Rev Cancer, 2022. \u003cstrong\u003e22\u003c/strong\u003e(2): p. 114-126.\u003c/li\u003e\n\u003cli\u003eLiu, Z., et al., \u003cem\u003eIntegrated multi-omics profiling yields a clinically relevant molecular classification for esophageal squamous cell carcinoma.\u003c/em\u003e Cancer Cell, 2023. \u003cstrong\u003e41\u003c/strong\u003e(1): p. 181-195.e9.\u003c/li\u003e\n\u003cli\u003eWu, A., et al., \u003cem\u003eDosiomics improves prediction of locoregional recurrence for intensity modulated radiotherapy treated head and neck cancer cases.\u003c/em\u003e Oral Oncology, 2020. \u003cstrong\u003e104\u003c/strong\u003e: p. 104625.\u003c/li\u003e\n\u003cli\u003eZheng, X., et al., \u003cem\u003eMulti-omics to predict acute radiation esophagitis in patients with lung cancer treated with intensity-modulated radiation therapy.\u003c/em\u003e Eur J Med Res, 2023. \u003cstrong\u003e28\u003c/strong\u003e(1): p. 126.\u003c/li\u003e\n\u003cli\u003eNie, T., et al., \u003cem\u003eIntegration of dosimetric parameters, clinical factors, and radiomics to predict symptomatic radiation pneumonitis in lung cancer patients undergoing combined immunotherapy and radiotherapy.\u003c/em\u003e Radiother Oncol, 2024. \u003cstrong\u003e190\u003c/strong\u003e: p. 110047.\u003c/li\u003e\n\u003cli\u003evan Griethuysen, J.J.M., et al., \u003cem\u003eComputational Radiomics System to Decode the Radiographic Phenotype.\u003c/em\u003e Cancer Research, 2017. \u003cstrong\u003e77\u003c/strong\u003e(21): p. e104-e107.\u003c/li\u003e\n\u003cli\u003eZheng, L., et al., \u003cem\u003eOptimal treatment strategy and prognostic analysis for hypopharyngeal squamous-cell carcinoma patients with T3-T4 or node-positive: A population-based study.\u003c/em\u003e Eur J Surg Oncol, 2023. \u003cstrong\u003e49\u003c/strong\u003e(7): p. 1162-1170.\u003c/li\u003e\n\u003cli\u003eTang, X., et al., \u003cem\u003eA novel prognostic model predicting the long-term cancer-specific survival for patients with hypopharyngeal squamous cell carcinoma.\u003c/em\u003e BMC Cancer, 2020. \u003cstrong\u003e20\u003c/strong\u003e(1): p. 1095.\u003c/li\u003e\n\u003cli\u003eForastiere, A.A., et al., \u003cem\u003eLong-term results of RTOG 91-11: a comparison of three nonsurgical treatment strategies to preserve the larynx in patients with locally advanced larynx cancer.\u003c/em\u003e J Clin Oncol, 2013. \u003cstrong\u003e31\u003c/strong\u003e(7): p. 845-52.\u003c/li\u003e\n\u003cli\u003eGrasl, S., et al., \u003cem\u003eA new nomogram to predict oncological outcome in laryngeal and hypopharyngeal carcinoma patients after laryngopharyngectomy.\u003c/em\u003e Eur Arch Otorhinolaryngol, 2023. \u003cstrong\u003e280\u003c/strong\u003e(3): p. 1381-1390.\u003c/li\u003e\n\u003cli\u003eLiu, X., et al., \u003cem\u003eCT-based radiomics signature analysis for evaluation of response to induction chemotherapy and progression-free survival in locally advanced hypopharyngeal carcinoma.\u003c/em\u003e European Radiology, 2022. \u003cstrong\u003e32\u003c/strong\u003e(11): p. 7755-7766.\u003c/li\u003e\n\u003cli\u003eLi, F., et al., \u003cem\u003eA Nomogram to Predict Nodal Response after Induction Chemotherapy for Hypopharyngeal Carcinoma.\u003c/em\u003e Laryngoscope, 2023. \u003cstrong\u003e133\u003c/strong\u003e(4): p. 849-855.\u003c/li\u003e\n\u003cli\u003eCavalieri, S., et al., \u003cem\u003eDevelopment of a multiomics database for personalized prognostic forecasting in head and neck cancer: The Big Data to Decide EU Project.\u003c/em\u003e Head Neck, 2021. \u003cstrong\u003e43\u003c/strong\u003e(2): p. 601-612.\u003c/li\u003e\n\u003cli\u003eHuang, Y., et al., \u003cem\u003eRadiation pneumonitis prediction after stereotactic body radiation therapy based on 3D dose distribution: dosiomics and/or deep learning-based radiomics features.\u003c/em\u003e Radiation Oncology, 2022. \u003cstrong\u003e17\u003c/strong\u003e(1): p. 188.\u003c/li\u003e\n\u003cli\u003eYang, S.S., et al., \u003cem\u003eDosiomics Risk Model for Predicting Radiation Induced Temporal Lobe Injury and Guiding Individual Intensity-Modulated Radiation Therapy.\u003c/em\u003e Int J Radiat Oncol Biol Phys, 2023. \u003cstrong\u003e115\u003c/strong\u003e(5): p. 1291-1300.\u003c/li\u003e\n\u003cli\u003eShorewala, V., \u003cem\u003eEarly detection of coronary heart disease using ensemble techniques.\u003c/em\u003e Informatics in Medicine Unlocked, 2021. \u003cstrong\u003e26\u003c/strong\u003e: p. 100655.\u003c/li\u003e\n\u003cli\u003eChandra Joshi, R., et al., \u003cem\u003eEnsemble based machine learning approach for prediction of glioma and multi-grade classification.\u003c/em\u003e Computers in Biology and Medicine, 2021. \u003cstrong\u003e137\u003c/strong\u003e: p. 104829.\u003c/li\u003e\n\u003cli\u003eZhao, S., et al., \u003cem\u003eStacking Ensemble Learning-Based [(18)F]FDG PET Radiomics for Outcome Prediction in Diffuse Large B-Cell Lymphoma.\u003c/em\u003e J Nucl Med, 2023. \u003cstrong\u003e64\u003c/strong\u003e(10): p. 1603-1609.\u003c/li\u003e\n\u003cli\u003eWang, R., et al., \u003cem\u003eDevelopment of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients.\u003c/em\u003e Journal of Hematology \u0026amp; Oncology, 2022. \u003cstrong\u003e15\u003c/strong\u003e(1): p. 11.\u003c/li\u003e\n\u003cli\u003eShen, L.L., et al., \u003cem\u003eDelta computed tomography radiomics features-based nomogram predicts long-term efficacy after neoadjuvant chemotherapy in advanced gastric cancer.\u003c/em\u003e Radiol Med, 2023. \u003cstrong\u003e128\u003c/strong\u003e(4): p. 402-414.\u003c/li\u003e\n\u003cli\u003eKong, C., et al., \u003cem\u003ePrediction of tumor response via a pretreatment MRI radiomics-based nomogram in HCC treated with TACE.\u003c/em\u003e Eur Radiol, 2021. \u003cstrong\u003e31\u003c/strong\u003e(10): p. 7500-7511.\u003c/li\u003e\n\u003cli\u003eMa, X., et al., \u003cem\u003eRadiomics nomogram based on optimal VOI of multi-sequence MRI for predicting microvascular invasion in intrahepatic cholangiocarcinoma.\u003c/em\u003e Radiol Med, 2023. \u003cstrong\u003e128\u003c/strong\u003e(11): p. 1296-1309.\u003c/li\u003e\n\u003cli\u003eTian, S., et al., \u003cem\u003eDevelopment and Validation of a Prognostic Nomogram for Hypopharyngeal Carcinoma.\u003c/em\u003e Frontiers in Oncology, 2021. \u003cstrong\u003e11\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003eZhang, D., et al., \u003cem\u003ePrognostic Nomogram for Postoperative Hypopharyngeal Squamous Cell Carcinoma to Assist Decision Making for Adjuvant Chemotherapy.\u003c/em\u003e Journal of Clinical Medicine, 2022. \u003cstrong\u003e11\u003c/strong\u003e(19): p. 5801.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1\u003c/strong\u003e Clinical characteristics of the patients\u003c/p\u003e\n\u003cp\u003ea) Clinical characteristics of patients in the training, testing and external validation cohorts\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"651\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eClinical Characteristics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003eAll\u003c/p\u003e\n \u003cp\u003ePatients\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u0026nbsp;\u003c/em\u003e= 56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003cp\u003eCohort\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u0026nbsp;\u003c/em\u003e= 39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003eTesting\u003c/p\u003e\n \u003cp\u003eCohort\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u003c/em\u003e = 17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003eExternal\u003c/p\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003cp\u003eCohort\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u003c/em\u003e = 32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.742\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.432\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eMedian\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;(range)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e63.1 (46-77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e63.3 (48-77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e62.5 (46-76)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e61.84 (46-76)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ePathology (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.961\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.254\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eHigh grade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e8 (14.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e5 (12.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (17.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e3 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eIntermediate\u003c/p\u003e\n \u003cp\u003egrade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e34 (60.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e24 (61.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e10 (58.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e26 (81.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eLow grade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e10 (17.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e7 (17.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (17.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e3 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eUnknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e4 (7.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (7.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e1 (5.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e0 (0.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ecT stage\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(AJCC 8th) (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.411\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.320\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eT1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e6 (10.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (7.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (17.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e3 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eT2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e26 (46.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e19 (48.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e7 (41.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e10 (31.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eT3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e15 (26.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e12 (30.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (17.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e10 (31.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eT4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e9 (16.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e5 (12.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e4 (23.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e9 (28.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ecN stage\u003c/p\u003e\n \u003cp\u003e(AJCC 8th) (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.328\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.222\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eN0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e11 (19.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e9 (23.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (11.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e3 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eN1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e13 (23.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e7 (17.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e6 (35.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e11 (34.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eN2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e28 (50.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e21 (53.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e7 (41.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e15 (46.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eN3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e4 (7.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (5.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (11.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e3 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ecStage\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(AJCC 8th) (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.736\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.607\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eIII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e21 (37.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e15 (38.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e6 (35.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e9 (28.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eIVA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e31 (55.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e22 (56.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e9 (52.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e20 (62.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eIVB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e4 (7.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (5.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (11.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e3 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eSmoking history (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.541\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e39 (69.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e26 (66.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e13 (76.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e22 (68.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e17 (30.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e13 (33.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e4 (23.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e10 (31.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eAlcohol history (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.395\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.240\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e31 (55.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e20 (51.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e11 (64.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e21 (65.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e25 (44.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e19 (48.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e6 (35.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e11 (34.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eLocation (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.833\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.629\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ePyriform\u003c/p\u003e\n \u003cp\u003esinus\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e48 (85.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e34 (87.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e14 (82.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e25 (78.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ePosterior\u003c/p\u003e\n \u003cp\u003epharyngeal\u003c/p\u003e\n \u003cp\u003ewall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e5 (8.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (7.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (11.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e5 (15.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003ePost\u003c/p\u003e\n \u003cp\u003ecricoid\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (5.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (5.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e1 (5.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e2 (6.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eConcurrent\u003c/p\u003e\n \u003cp\u003eChemotherapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.513\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e41 (73.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e27 (69.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e14 (82.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e24 (72.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e15 (26.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e12 (30.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e3 (17.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e9 (27.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eRadiation\u003c/p\u003e\n \u003cp\u003etechnology\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e0.519\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e0.585\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003eIMRT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e54 (96.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e38 (97.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e16 (94.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e31 (93.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 21.6923%;\"\u003e\n \u003cp\u003e2-D RT\u003c/p\u003e\n \u003cp\u003e/3-D CRT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e2 (3.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e1 (2.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.6154%;\"\u003e\n \u003cp\u003e1 (5.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 10.1538%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 14.7692%;\"\u003e\n \u003cp\u003e2 (6.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.53846%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eb) Dosimetric characteristics of patients in the training and testing cohorts for multi-omics study\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"673\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eDosimetric Characteristics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22.4368%;\"\u003e\n \u003cp\u003eAll\u003c/p\u003e\n \u003cp\u003ePatients\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u0026nbsp;\u003c/em\u003e= 48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 22.4368%;\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003cp\u003eCohort\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u0026nbsp;\u003c/em\u003e= 33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21.2481%;\"\u003e\n \u003cp\u003eTesting\u003c/p\u003e\n \u003cp\u003eCohort\u003c/p\u003e\n \u003cp\u003e(\u003cem\u003en\u003c/em\u003e = 15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eRadiation EQD2 (Gy), Median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e60.00 (60.00-66.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e60.00 (60.00-66.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e60.00 (60.00-66.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.785\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eDmax (Gy),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e72.91 (70.35-74.30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e73.22 (70.64-74.48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e72.11 (68.33-74.21)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.392\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eDmin (Gy),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e16.29 (2.68-38.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e15.26 (1.87-36.72)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e19.08 (3.54-43.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.648\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eDmean (Gy),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e62.40 (61.71-62.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e62.23 (61.57-62.84)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e62.43 (62.22-62.69)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.484\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eDose distribution SD,\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e2.04 (1.58-2.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e2.26 (1.72-2.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e1.59 (1.39-2.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.024\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eHomogeneity Index, Median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e0.75 (0.42-0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e0.76 (0.44-0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e0.74 (0.35-0.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.681\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eD90 (Gy),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e60.59 (59.25-60.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e60.47 (56.94-60.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e60.64 (60.49-60.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.417\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eD95 (Gy),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e59.85 (58.31-60.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e59.69 (55.77-60.28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e59.93 (59.70-60.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.186\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eD100 (Gy),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e16.29 (2.68-38.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e15.26 (1.87-36.72)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e19.08 (3.54-43.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.648\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eV50 (%),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e99.87 (99.66-99.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e99.84 (99.63-99.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e99.94 (99.87-99.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.053\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eV55 (%),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e99.54 (98.47-99.80)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e99.42 (97.98-99.69)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e99.66 (99.51-99.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.018\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eV60 (%),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e94.06 (85.14-95.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e93.77 (80.04-95.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e94.69 (93.43-96.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.217\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eV65 (%),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e3.93 (1.60-8.49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e4.26 (1.61-15.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e2.85 (1.51-6.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.368\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 24.9629%;\"\u003e\n \u003cp\u003eV70 (%),\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e1.00 (0.09-1.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22.4368%;\"\u003e\n \u003cp\u003e1.00 (0.08-1.06)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21.2481%;\"\u003e\n \u003cp\u003e1.00 (0.10-1.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8.9153%;\"\u003e\n \u003cp\u003e0.569\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cbr\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2\u003c/strong\u003e The results of feature selection from LASSO regression\u003c/p\u003e\n \u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"612\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eFeatures\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003eCategory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003eCoefficient\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"6\" style=\"width: 96px;\"\u003e\n \u003cp\u003eRadiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_squareroot_glcm_Imc1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.172948\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_lbp-3D-k_glszm_ZoneEntropy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.065080\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_gradient_firstorder_Skewness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003efirst order\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e0.002844\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_wavelet-LHH_glcm_InverseVariance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.018525\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_squareroot_glszm_GrayLevelNonUniformity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.011088\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_lbp-3D-k_gldm_SmallDependenceEmphasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e0.016451\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"8\" style=\"width: 96px;\"\u003e\n \u003cp\u003eMulti-omics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003edos_logarithm_glcm_Imc1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.008965\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003edos_logarithm_ngtdm_Coarseness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.056607\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_gradient_glcm_Imc2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e0.006655\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_lbp-3D-k_gldm_SmallDependenceEmphasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e0.001976\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_lbp-3D-k_glszm_ZoneEntropy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.019096\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_squareroot_glcm_Imc1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.021856\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_squareroot_glszm_GrayLevelNonUniformity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.003041\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003erad_wavelet-LLH_gldm_DependenceEntropy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 90px;\"\u003e\n \u003cp\u003etexture\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 89px;\"\u003e\n \u003cp\u003e-0.010745\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cbr\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3\u003c/strong\u003e The performance of each classifier in the training, testing, and external validation cohorts.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"454\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 199px;\"\u003e\n \u003cp\u003eClassifiers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003cp\u003e(Average)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003eTesting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003eExternal\u003c/p\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"8\" style=\"width: 99px;\"\u003e\n \u003cp\u003eRadiomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eSVM model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.790\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eDL model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.868\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eLR model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.871\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eRF model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.862\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Boosting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.881\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Stacking\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.907\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Bagging\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.896\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Voting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.929\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 99px;\"\u003e\n \u003cp\u003eMulti-omics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Boosting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.960\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" style=\"width: 92px;\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Stacking\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.963\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Bagging\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.966\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 100px;\"\u003e\n \u003cp\u003eEL_Voting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003e0.947\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 73px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\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":"Hypopharyngeal cancer, postoperative chemoradiotherapy, Multi-omics, Machine learning","lastPublishedDoi":"10.21203/rs.3.rs-5861722/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5861722/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eThis study aimed to predict the progression-free survival (PFS) of the patients who were diagnosed with hypopharyngeal cancer and received postoperative chemoradiotherapy by using multi-omics method which integrating clinical factors, dosimetric and radiomic features.\u003c/p\u003e\u003ch2\u003eMaterials and methods\u003c/h2\u003e \u003cp\u003eThis study retrospectively collected the pretreatment T1-weighted MR imaging data of 88 hypopharyngeal cancer patients with postoperative chemoradiotherapy, including 56 cases from one center (training and testing cohorts) and 32 cases from another center (external validation cohort), and the gross tumor volumes (GTV) were countered for all cases. A Python-based library, pyradiomics was used to extract the radiomics features from each GTV. Least absolute shrinkage and selection operator (LASSO) regression was used to identify the most important features for classifier establishment. On the other hand, complete radiotherapy data are retained for 48 patients among them, and the planning tumor volumes (PTV) were countered for radiotherapy planning. The dose distribution features extracted by using pyradiomics and the dosimetric parameters were combined with the radiomics features to establish the classifiers. The probabilities of positive sample calculated from the best classifier, the radiomics and multi-omics signatures were obtained for establish the Cox proportional hazards models.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe ensemble learning (EL) model was selected as the superior model with the higher area under the receiver operating characteristic curve (AUC) values than other classifier during the radiomics-only analysis, and the EL model with stacking technique showed the best performance, yielding AUC values of 0.93, 0.79, and 0.78 for the training, testing, and external validation cohorts, respectively. Furthermore, the multi-omics analysis integrating radiomics and dosiomics improved the effectiveness of the EL model with AUC values of 0.98 and 0.88 for the training and testing cohorts, respectively. Furthermore, the C-index of the Cox proportional hazards models resulted in a 0.099 improvement in the testing cohort when employing the multi-omics signature versus the radiomics signature.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eRegarding the patients with hypopharyngeal cancer receiving postoperative chemoradiotherapy, the multi-omics-based prognostic prediction could achieve a more robust predictive capability than the radiomics-only study. This approach warrants further validation through prospective studies.\u003c/p\u003e","manuscriptTitle":"Multi-omics-based prognostic prediction for locally advanced hypopharyngeal cancer treated with postoperative chemoradiotherapy: a dual-center study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-03 09:02:39","doi":"10.21203/rs.3.rs-5861722/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":"dae0f9ea-9cd3-4d4b-94b9-a90ea3b07a5b","owner":[],"postedDate":"February 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-02-26T08:08:58+00:00","versionOfRecord":[],"versionCreatedAt":"2025-02-03 09:02:39","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5861722","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5861722","identity":"rs-5861722","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}