A nomogram combining clinical features, O-RADS US, and radiomics based on ultrasound imaging for diagnosing ovarian cancer

A nomogram combining clinical features, O-RADS US, and radiomics based on ultrasound imaging for diagnosing ovarian cancer | 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 Article A nomogram combining clinical features, O-RADS US, and radiomics based on ultrasound imaging for diagnosing ovarian cancer Wenting Xie, Yaoqin Wang, Zhongshi Du, Yijie Chen, Xiaohui Ke, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5468347/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 02 Jun, 2025 Read the published version in Scientific Reports → Version 1 posted 6 You are reading this latest preprint version Abstract Background We aimed to develop and validate a nomogram for diagnosing ovarian cancer from ovarian masses based on clinical information, O-RADS US, and radiomics. Methods A total of 981 patients with ovarian masses from two centers were randomly divided into the training cohort (n = 686) and the validation cohort (n = 295). We defined the region of interest (ROI) of the tumor by manually drawing the tumor contour on the ultrasound image of the lesion. The radiomics features were extracted from ultrasound images, and the radiomics score was then calculated. O-RADS US characteristics, radiomics score, and clinical features selected using the LASSO algorithm were used to develop O-RADS US + Radscore + Clinical, Radscore + Clinical, and O-RADS US + Clinical models, respectively. Receiver operating characteristic (ROC), decision curve analysis, and calibration curve were used to evaluate the performance of the nomogram models. Results Age, CA125, O-RADS US, and radiomics score were related to ovarian malignancy through univariate and multivariate logistic regression analyses. In the training and validation datasets, the areas under the ROC curve (AUC) of O-RADS US + Clinical model were 0.830 and 0.815, respectively, and those for the Radscore + Clinical model were 0.876 and 0.867, respectively. The O-RADS US + Radscore + Clinical nomogram model presented improved AUC values of 0.967 in the training group and 0.951 in the validation group, significantly higher than that of Radscore + Clinical and O-RADS US + Clinical models. The calibration curve and the clinical decision curve analysis demonstrated that the nomogram models had high clinical benefits. The O-RADS US + Radscore + Clinical model had the highest net return. Conclusions Combination nomogram model that integrates clinical features, O-RADS US, and radiomics based on ultrasound image analysis could predict ovarian malignancy with high diagnostic accuracy, indicating that this model might have a role in preoperative diagnosis for differentiating benign and malignant ovarian tumors. Biological sciences/Cancer Biological sciences/Cancer/Cancer imaging Biological sciences/Cancer/Gynaecological cancer ovarian cancer Ultrasound Ovarian-Adnexal Reporting and Data System Ultrasound radiomics Nomogram Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Ovarian cancer (OC) is a major malignant tumor worldwide and the fifth leading cause of death among women in the United States, with over 28,6100 incident cases and 17,6000 deaths in 2017[ 1 , 2 ]. Given the lack of early accurate screening tools, rarity of symptoms in early stages, and the rapid dissemination of the disease, the 5-year survival rate of patients with ovarian cancer ranges from 30–50%[ 3 ]. As the treatment planning for benign and malignant ovarian masses is quite different, preoperative and accurate discrimination of ovarian lesions is crucial for tailoring therapeutic procedures and preserving patients’ fertility[ 4 ]. Ultrasound (US) is the primary first-line imaging modality for detecting and diagnosing ovarian masses. It has been widely used to differentiate benign from malignant ovarian tumors. As ovarian tumors are heterogeneous, different OC pathologies can exhibit similar US characteristics, and similar pathological types can exhibit different US features. Many models, including International Ovarian Tumour Analysis Simple Rules (IOTA) and the Gynecologic Imaging Report and data system (GI-RADS), have been proposed to improve the discrimination of adnexal masses[ 5 ]. Despite the high diagnostic accuracy of these models[ 6 ], their application in clinical practice has been limited internationally because of their rejection in certain nations. In 2018, the American College of Radiology (ACR) released Ovarian-Adnexal Reporting and Data System Ultrasound (O-RADS US) to improve the standardization of describing adnexal masses ultrasound features and reports[ 7 ]. The system included six risk classification categories (O-RADS 0–5). O-RADS category 0 is considered an incomplete evaluation; O-RADS category 1 is the normal ovary with 0% risk of malignancy; O-RADS category 2 is almost certainly benign lesions (<1% chance of malignancy); O-RADS category 3 referring to low risk (1% to<10% chance of malignancy); O-RADS category 4 is intermediate risk (10% to<50% chance of malignancy) and O-RADS category 5 is high risk (≥ 50% chance of malignancy)[ 8 ]. Our previous study evaluated the diagnostic performance of O-RADS US to differentiate benign from malignant tumors[ 9 ]. Although O-RADS US showed a higher ability to differentiate benign from malignant ovarian masses, the diversity and complexity of ovarian tumors still pose a formidable challenge for radiologists[ 10 ]. An efficient and reproducible method for accurate diagnosis of ovarian masses is urgently needed to assist inexperienced radiologists or medical resource-lacking regions in diagnosing the malignancy of ovarian masses. Recently, radiomics has garnered attention for its ability to accurately diagnose OC through computer algorithms to extract thousands of quantitative features from radiology images[ 11 ]. Studies have shown that radiomics nomogram analysis can help diagnose and predict the outcome of OC[ 12 – 14 ]. Radiomics can be avoided inconsistent inter-observer and intra-observer circumstances for improving diagnostic accuracy in distinguishing benign from malignant ovarian lesions. Study have demonstrated that radiomics based on transvaginal ultrasonography and serum cancer antigen 125 (CA125) can predict the risk of malignancy of ovarian lesions with an accuracy of 88%[ 15 ]. Here, we aimed to develop and validate a model as an non-invasive tool for a personalized diagnostic of OC patients. The present study developed a radiomics score by extracting and selecting features from the ovarian lesion US images. Least absolute shrinkage and selection operator (LASSO) regression, a statistical technique used to study the effects of variables on outcome predictions, was used to select radiomics features. Eventually, a nomogram comprising clinical information, the radiomics score, and the O-RADS US category was developed to predict the malignancy of ovarian lesions. The diagnostic performance of this model was evaluated, and its clinical utility was demonstrated. Materials and methods Ethics The retrospective study was approved by the institutional ethical committee board of the Fujian Cancer Hospital and Nanping First Hospital Affiliated to Fujian Medical University (K2021-128-01 and NPSY202201015). Due to the retrospective nature of the study, all patients waived the informed consent. The research methods were conducted in accordance with the relevant guidelines and regulations. Study population We retrospectively studied 981 consecutive patients (with 981 ovarian masses) who were pathologically confirmed with ovarian tumor between January 2017 and December 2021 from two tertiary hospitals: Fujian Cancer Hospital (FJZL, n = 749) and Nanping First Hospital Affiliated to Fujian Medical University (NPFH, n = 232). All patients were included in this study based on surgery pathological results and received an ultrasound with sufficient image quality before surgery. The clinical characteristics comprising age, maximum tumor diameter, location, ascites, and CA125 were recorded in this study. Maximum tumor diameter and ascites were evaluated via the US and then recorded. Maximum tumor diameter was obtained by measuring the maximum diameter of the mass in any plane. Ascites are fluid extending above the uterine fundus beyond the pouch of Douglas or fluid anterior/superior to the uterus [ 16 ]. The study flow diagram of the population is shown in Fig. 1 . US Image Acquisition Preoperative US examinations were performed by experienced radiologists and acquired standardized ultrasound images. The ultrasonic equipment included LOGIQ E11 (GE Healthcare, United States; convex array probes measuring 1–5 MHz; transvaginal probes measuring 2–9 MHz) and Philips IU 22 (Philips Bothell, United States; convex array probes with 2–6 MHz; transvaginal probes measuring 4–8 MHz). In cases of multiple or bilateral lesions, the lesion with the highest O-RADS US score was selected for further analysis. For each ovarian lesion, slices were acquired, and the larger mass or mass with the worst morphology was selected. Representative ultrasound image for each patient was selected by two radiologists with more than 6 years of ultrasound operating experience who were blinded for the outcome, and all recorded images were classified using O-RADS US. Any dispute was resolved with consensus. The classification criteria of O-RADS US were based on the ACR Ovarian-Adnexal Reporting and Data System Committee[ 16 ]. Radiomics analysis Radiomics workflow comprised tumor image segmentation and feature extraction; feature selection and radscore establishment; and model construction and evaluation. Tumor image segmentation All enrolled patients were randomly divided at a proportion of 7:3 between a training cohort (n = 686) and a validation cohort (n = 295). For each lesion, a region of interest (ROI) of the tumor boundaries was manually delineated using Labelme software ( https://github.com/wkentaro/labelme ) by ultrasound radiologists (WTX, with 7 years of experience). Subsequently, 794 radiomics features of ovarian tumor ROIs were extracted using PyRadiomic (version 2.2.0, https://pyradiomics.readthedocs.io/en/latest )[ 17 ]. The radiomics features are subdivided into eight classes, and the detailed description is described in Supplementary Table 1. Two independent radiologists (twice by WTX with an interval of 4 weeks and once by YQW with 15 years of experience) delineated ROIs on 295 randomly chosen images to evaluate inter-observer and intra-observer reproducibility. The inter-observer and intra-observer reproducibility were assessed using the intraclass and interclass correlation coefficients (ICCs). Features with an ICC of >0.90 were used in subsequent analyses. Feature extraction and selection After eliminating the constant term, 635 features from each patient were used for further selection. Normality analysis was performed for each radiomics feature and then matched in pairs for correlation analysis. Pearson correlation analysis was implemented to evaluate the pair-wise feature satisfied normality; otherwise, spearman correlation analysis was performed to calculate features that did not satisfy the normal analysis. Finally, LASSO regression were used to select and reduce the number of the extracted radiomics features. Development, validation, and clinical application of the nomogram We performed univariate and multivariate logistic regression to analyze the significant risk factors for OC. p < 0.05 was considered significant in both univariate and multivariate analysis. The multivariable logistic regression method established predictive nomogram models based on Radscore, clinically significant risk factors, and O-RADS US. Receiver operating curves (ROCs) were plotted to evaluate the differentiation efficiency of nomograms in the training and validation cohorts. Calibration curves were used to explore the predictive accuracy of the models. Decision curve analysis (DCA) was performed to determine the clinical usefulness of the nomograms in the validation cohort. Statistical analysis IBM SPSS Statistics (version 25.0; SPSS Inc., Chicago, IL, USA) and R (version 4.0.3) software were used for statistical analyses. All clinical independent variables were generated using univariable and multivariable regression analyses. LASSO regression, ROCs, and nomogram model were built using the “glmnet”, “pROC”, and rms R packages, respectively. The area under the curves (AUCs), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to measure the model’s performance. The DeLong’s test was used to compare the difference between the AUCs. Results Clinical characteristics A total of 981 patients with ovarian masses were retrospectively enrolled in this study. Supplementary Table 2 summarizes the characteristics of recruited patients with two centers. Final pathological outcome for all women who underwent surgery is shown in Table 1 . Among these patients, 499 (50.9%) were benign and 482 (49.1%) were malignant, respectively. The demographic characteristics of total recruited patients are reported in Table 2 . There were no significant differences in age, tumor size, CA125, location, ascites, O-RADS US, and Radscore between the training and validation datasets. Table 1 Final pathological diagnosis of 981 adnexal masses from the two centers Pathologic diagnosis No.(%) Benign adnexal masses 499(50.86) Mature teratoma 129(13.15) Endometrioid cyst 116(11.83) Mucinous cystadenoma 85(8.66) Serous cystadenoma 76(7.76) Thecoma-fibroma 41(4.18) Follicular cyst 17(1.73) Struma ovarii 10(1.02) Hemorrhagic cyst 7(0.71) Serous adenofibroma 6(0.61) Corpus luteum cyst 4(0.41) Brenner tumor 2(0.20) Stromal tumor 2(0.20) Mucinous adenofibroma 2(0.20) Inflammation 2(0.20) Malignant adnexal masses 482(49.13) Serous cystadenocarcinoma 271(27.62) Borderline mucinous cystadenoma 45(4.59) Borderline serous cystadenoma 38(3.87) Clear cell carcinoma 25(2.55) Metastatic carcinoma 39(3.98) Endometrioid carcinoma 17(1.73) Malignant Muellerian tube mixed tumor 8(0.82) Granulosa cell tumour 8(0.82) Immature teratoma 7(0.71) Mucinous cystadenocarcinoma 7(0.71) Borderline endometrioid tumor 6(0.61) Yolksactumor 3(0.32) Ovarian dysgerminoma 2(0.20) Sertoli-Leydig cell tumor 2(0.20) Borderline Brenner Tumor 2(0.20) Small cell carcinoma of the ovary 2(0.20) Data are given as n (%) Table 2 Clinicopathological characteristics in the training and the validation datasets. Characteristic Training dataset (n = 686) Validation dataset (n = 295 ) p value Age,years (Q.25-Q.75) 48.0 (37.0–57.0) 47.0 (36.0–57.0) 0.420 Tumor size, mm (Q.25-Q.75) 85.0 (58.2–124) 89.0 (58.0-130) 0.286 CA125,U/ml (Q.25-Q.75) 48.5 (16.2–362) 41.2 (15.0-248) 0.334 Location 0.490 Left 288 (42.0) 136 (46.1) Right 265 (38.6) 106 (35.9) Bilateral 133 (19.4) 53 (18.0) Ascites 0.640 Yes 167 (24.3) 67 (22.7) No 519 (75.7) 228 (77.3) O-RADS 0.377 2 150 (21.9) 52 (17.6) 3 140 (20.4) 60 (20.3) 4 155 (22.6) 78 (26.4) 5 241 (35.1) 105 (35.6) Pathological 0.999 Benign 349 (50.9) 150 (50.8) Malignant 337 (49.1) 145 (49.2) Radscore -0.15(-2.13- 2.95) 0.05 (-2.14- 3.99) 0.417 median (interquartile range)Q.25 25% quantile; Q.75 75% quantile Inter-observer and intra-observer reproducibility of feature extraction A total of 295 patients were randomly selected to evaluate the inter-observer and intra-observer reproducibility of feature extraction. The median intra-observer ICCs of the first reader’s two extractions was 0.998. The median inter-observer ICCs of radiomics features extraction between the two radiologists was 0.959. Subsequent calculations were based on the radiomics features of the first reader. Radscore establishment After analysis, 794 extracted features were reduced to seven potential radiomics features based on the training set (Fig. 2 ). Table 3 summarizes the explanation and relevance of these features for the diagnosis of ovarian cancer. Finally, seven radiomics features were incorporated into the radiomics score formula. Radiomics score (Radscore) = − 0.0355681 - (0.2541203 × original_firstorder_Maximum) + (0.1468945 × original_glszm_ZoneEntropy) + (0.0499583 × original_gldm_DependenceVariance) + (0.0034785 × CoLIAGe2D_WindowSize9_Entropy_firstorder_Variance) – (0.0077041 × wavelet.LHL_lbp.3D.m2_firstorder_InterquartileRange) – (0.0183370 × wavelet.HLH_lbp.3D.k_firstorder_Minimum) – (0.0002613 × wavelet.LLL_lbp.3D.m2_firstorder_Range). Table 3 A summary of the explanation and relevance to ovarian cancer diagnosis of the features extracted using PyRadiomics in the radiomics score formula. Feature name Description Relevance to ovarian cancer diagnosis original_firstorder_Maximum The maximum gray level intensity within the ROI Associated with certain high-density areas within the tumor which is helpful to identify benign and malignant tumors original_glszm_ZoneEntropy Measures the uncertainty/randomness in the distribution of zone sizes and gray levels A higher value indicates more heterogeneous in the texture patterns original_gldm_DependenceVariance Measures the variance in dependence size in the image Higher dependence variance may indicate heterogeneity of tumor tissue CoLIAGe2D_WindowSize9_Entropy_firstorder_Variance The variance of the entropy of co-occurrence local intensity angle gradient entropy within a window size of 9 Higher variance may indicate significant heterogeneity within the tumor region wavelet.LHL_lbp.3D.m2_firstorder_InterquartileRange The range from the 25 th and 75 th percentile of the image array A higher interquartile range may indicate a larger range of gray values within the tumor region wavelet.HLH_lbp.3D.k_firstorder_Minimum The minimum gray level intensity in the wavelet-transformed image The lowest gray value in different scales and directions, which may be related to the darkest region in the tumor area wavelet.LLL_lbp.3D.m2_firstorder_Range The range of gray values in the ROI A large range may indicate a wide range of density changes within the tumor Univariate and multivariate analyses of clinical information, O-RADS, and Radscore Table 4 displays the results of the univariate and multivariate logistic regression analyses for discriminating benign from malignant ovarian lesions in the training dataset. The odd ratios of age, CA125, O-RADS US, and radscore were found to be significant ( p <0.05). The O-RADS US + Radscore + Clinical nomogram was developed by incorporating the aforementioned independent risk factors (Fig. 3 e). The Radscore + Clinical model was developed with age, CA125, location, ascites, and Radscore (Fig. 3 c, Supplementary Table 3). The O-RADS US + Clinical model was developed with age, CA125, and O-RADS US (Figur. 3a, Supplementary Table 4). Table 4 Results of the univariate and multivariate analyses of O-RADS + Radscore + Clinical model based on the training set. Characteristic Univariate analysis Multivariate analysis OR (95% CI) p value OR (95% CI) p value Age, years 1.057(1.043,1.07) <0.001 1.024(1.002,1.047) 0.033 Tumor size,mm 1.008(1.005,1.011) <0.001 CA125(U/ml) 1.005(1.004,1.007) <0.001 1.001(1.000,1.002) 0.039 Location Left 0.689(0.508,0.934) 0.0168 Right 0.535(0.391,0.73) <0.001 Bilateral 5.473(3.520,8.774) <0.001 Ascites 1.44(3.85,14.163) <0.001 O-RADS 2.642(5.124,16.556) <0.001 11.15(7.030,18.92) <0.001 Radscore 1.77(1.619,1.955) <0.001 14.58(4.10,54.88) <0.001 Modeling and Evaluation of the Nomograms The radiomics features of each patient were extracted from US images and selected using LASSO. A predictive model—based on radiomics scores, clinical characteristics, and O-RADS US—was finally generated with the multivariable logistic regression method. Table 5 shows the diagnosis performance of the O-RADS US + Clinical, Radscore + Clinical, and O-RADS US + Radscore + Clinical models in differentiating benign from malignant ovarian neoplasm. Table 5 Performance of the nomograms for differentiating benign from malignant tumors. Variable O-RADS US ＋ Clinical model Radscore ＋ Clinical model O-RADS US+Radscore+Clinical ＋ Clinical model Training dataset Validation dataset Training dataset Validation dataset Training dataset Validation dataset AUC 0.830* (0.801, 0.861) 0.815** (0.767, 0.864) 0.876 ＆ (0.850, 0.904) 0.867 ＆＆ (0.826, 0.909) 0.967 (0.954, 0.980) 0.951 (0.929, 0.974) Sensitivity, % 0.756(0.707,0.806) 0.744(0.666,0.824) 0.792(0.742,0.831) 0.724(0.686,0.821) 0.928(0.896,0.953) 0.800(0.730,0.858) Specificity, % 0.773(0.725,0.822) 0.693(0.607, 0.780) 0.839(0.800,0.879) 0.860(0.810,0.920) 0.865(0.827,0.898) 0.926(0.873,0.961) PPV, % 0.763(0.714,0.813) 0.701(0.615,0.788) 0.826(0.783,0.867) 0.833(0.783,0.905) 0.869(0.831,0.901) 0.913()0.853,0.953 NPV, % 0.767(0.718,0.816) 0.737(0.659,0.816) 0.807(0.763,0.843) 0.763(0.724,0.844) 0.926(0.893,0.951) 0.827(0.763,0.879) PLR 3.342(2.647,4.222) 2.428(1.812,3.256) 4.937(3.844,6.566) 5.172(3.812,9.285) 6.896(5.28,8.99) 10.909(6.11,19.48) NLR 0.314(0.258,0.383) 0.368(0.275,0.492) 0.247(0.198,0.316) 0.320(0.203,0.376) 0.082(0.054,0.125) 0.215(0.153,0.304) Accuracy, % 0.765(0.732,0.798) 0.718(0.661,0.776) 0.816(0.788,0.843) 0.793(0.773,0.855) 0.896(0.872,0.917) 0.864(0.823,0.899) PPV, Positive predictive value; NPV, Negative predictive value; PLR, Positive likelihood ratio; NLR, Negative likelihood ratio; AUC, area under the receiver operating characteristic curve. *indicates a significant difference compared with that of O-RADS ＋ Radiomics ＋ Clinical model in the training cohort, p = 7.925e-31. ** indicates a significant difference compared with that of O-RADS ＋ Radiomics ＋ Clinical model in the validation cohort, p =1.07e-15. ＆ indicates a significant difference compared with that of O-RADS ＋ Radiomics ＋ Clinical model in the training cohort, p =7.76e-16. ＆＆ indicates a significant difference compared with that of O-RADS ＋ Radiomics ＋ Clinical model in the training cohort, p =1.886e-06. In the training dataset, the O-RADS US + Radscore + Clinical model achieved a sensitivity of 92.8%, a specificity of 86.5%, and an AUC value of 0.967, significantly higher than that of Radscore + Clinical (AUC = 0.876) and O-RADS US + Clinical (AUC = 0.830). In the validation cohort, the O-RADS US + Radscore + Clinical model also obtained an excellent result, with a significantly higher AUC value (AUC = 0.951) compared with that of the Radscore + Clinical (AUC = 0.867) and O-RADS US + Clinical (AUC = 0.815) models (both p < 0.05). The ROCs of all nomograms in the training and validation cohorts are plotted in Fig. 4 . The O-RADS US + Radscore + Clinical model exhibited outstanding discrimination performance. Calibration and clinical usefulness All three nomograms showed good agreement in predicting ovarian neoplasm; the calibration plots are shown in Figs. 3 b, 3 d, and 3 f. DCA was used to evaluate the clinical usefulness of the three nomograms (Fig. 5 ). The O-RADS US + Radscore + Clinical nomogram obtained the best clinical benefit for predicting OC, followed by the Radscore + Clinical and O-RADS US + Clinical nomograms in the DCA curves. Discussion This study focused on investigating the use of radiomics combining O-RADS US and clinical information for classification of ovarian masses. Therefore, we investigated whether radiomics could be used to identify patients with OC from ovarian masses and whether combining O-RADS US with radiomics could improve the diagnostic performance of OC. In the present study, we applied radiomics features to ultrasonographic images, and preoperative nomogram models were developed, integrating O-RADS US, radiomics, and clinical information to predict the malignant risk of ovarian lesions. Age and CA125 were identified as independent factors through multivariate analysis. Our results revealed that O-RADS US and radiomics could predict ovarian malignancy lesions with high accuracy. Combined with the aforementioned clinical information, the diagnostic AUCs of the Radscore + Clinical model significantly exceeded the O-RADS US + Clinical model in both the training (0.876 vs. 0.830) and validation cohorts (0.867 vs. 0.815). The AUCs of the O-RADS US + Radscore + Clinical model in the training and validation group were 0.967 and 0.951, respectively. The O-RADS US + Radscore + Clinical model offered significantly higher AUCs than the O-RADS US + Clinical and Radscore + Clinical model in both the training ( p < 0.001) and validation cohorts ( p < 0.001), demonstrating that radiomics plus O-RADS US imaging model in preoperative exhibited a high potential to discern OC from ovarian tumors patients. Ultrasound is a non-invasive tool in the diagnosis and treatment of ovarian tumors in clinical practice[ 18 ]. O-RADS US is an effective method to distinguish benign ovarian masses from malignant ones. Cao et al. reported that the diagnostic performance of O-RADS US was good, with an AUC was 0.960[ 19 ]. Similarly, Chen et al. compared O-RADS and deep learning to predict ovarian malignancy and revealed that the deep learning method and O-RADS US exhibited comparable diagnostic performance for classifying malignant from benign ovarian tumors with AUCs of 0.93 and 0.92, respectively[ 20 ]. A recent meta-analysis indicated that the pooled estimated sensitivity and specificity of the O-RADS US system for diagnostic adnexal masses were 97% (95% confidence interval (CI) = 94–98%) and 77% (95% CI = 68–84%), respectively[ 21 ]. However, Yuan et al. reported the diagnostic performance of O-RADS US with an AUC of 0.71 in the validation group[ 22 ]. Zhou et al. found that the AUC of O-RADS US was 0.86 in predicting malignant adnexal masses when comparing the diagnostic efficiency between subjective assessment and the O-RADS US[ 23 ]. The discrepant in diagnostic efficiency of O-RADS US could be attributed to the level of the sonographer and the proportion of malignant and benign ovarian tumor cases between studies. Radiomics is a computer-aided technology for diagnosing and predicting OC[ 24 ]. Previous studies have demonstrated that radiomics based on ultrasound images can differentiate between benign and malignant ovarian tumors[ 25 , 26 ]. Besides, Tang et al. found that ultrasound-based radiomics exhibited a good differential diagnosis of type I and type II epithelial OC before surgery[ 27 ]. YAO et al. reported that first-order statistics, GLSZM, GLRLM, GLCM, and NGTDM were significantly correlated with ovarian lesions, and their US radiomics model successfully differentiated type I and type II EOC [ 28 ]. Our study selected seven optimal radiomics of the 794 features to establish the radiomics score, including five features of first-order, one of GLSZM, and one of GLDM. These features reflected heterogeneity, non-uniformity, and variability, and their potential association with the characteristics of ovarian lesions. The radiomics score was an independent predictor for ovarian malignancy, and the Radscore + Clinical model exhibited good diagnostic performance for identifying benign and malignant ovarian tumors (with AUCs of 0.876 and 0.867 in the training and validation groups, respectively), suggesting that radiomics provided additional value for individualized malignant prediction. Consistent with our results, Qi et al. reported that combining clinical index and radiomics signatures performed the best AUC for differentiating benign from malignant ovarian serous tumors[ 29 ]. However, Radiomics methods share an essential limitation, where sonographers rely on handcrafted ROI[ 30 ]. Both imaging features and O-RADS US alone are often insufficient to determine the malignant risk of ovarian neoplasms. Hence, clinicians also consider all factors, including clinical information and sonographers’ subjective assessments make the diagnosis[ 31 ]. By contrast, models integrating O-RADS US, radiomics features, and clinical features could provide added diagnostic value, allowing more comprehensive evaluations to distinguish OC from benign ovarian lesions. As expected, our present study found that the combination nomogram model achieved satisfactory diagnostic performance, indicating the effectiveness of this combined model in preoperatively predicting the malignant risk of ovarian masses and in assisting individualized management in patients with adnexal lesions. Moreover, our study indicated that US-based radiomics can be used as an important supplement to O-RADS US in the classification of ovarian masses. As far as we know, only one study focused on the use of radiomics for discriminating between benign and malignant ovarian tumors according to O-RADS US, and achieved good diagnostic performance with an AUC of 0.93[ 32 ]. In this study, the clinical decision curve demonstrates that all models are valuable in predicting OC among various ovarian tumors, and the O-RADS US + Radscore + Clinical model had the highest net return. However, our work has some limitations. First, the retrospective study design may have case selection and verification bias. The O-RADS US category of reviews can also be considered a limitation. Considering these limitations, large multi-center prospective research will be required for further study. Second, only ultrasound images with larger lesions or lesions with the worst morphology were used to extract features. Although these images can represent the characteristics of the ovarian lesions, they did not completely represent the entire tumor. Therefore, future research requires ultrasound video images that can reflect the entire lesion. Third, the reliability and reproducibility of the ROIs drawn by radiologists did not completely avoid the tumor’s surrounding tissue. As reported by A. Das et al , deep learning can be employed to improve the identification of ovarian cancer[ 33 ]. Thus, deep learning combined with O-RADS US to differentiate between malignant and benign ovarian tumors will be required for further study. Fourth, images obtained by different equipment can be seen as a limitation of this study. The same ultrasound imaging parameters, such as frequency, dynamic range, and gain settings should be carried out in future work to reduce the impact on the extraction of radiomic features. Five, the distribution of benign and malignant cases might affect the accuracy of the model. Non-operated cases with presumed benign diagnosis followed up by ultrasound can be enrolled in further study. Additionally, a limitation of radiomics studies is that they require special software and statistical skills, which limits their application in clinical practice. This model could be turned into an online tool or software within the ultrasound device to help junior radiologists in diagnosing the malignancy of ovarian masses, which should be taken into consideration in future work. Conclusion In conclusion, we constructed a combination nomogram model that combines O-RADS US, Radscore, and clinical information, and we then validated the effective diagnostic performance of the model for the diagnosis of ovarian masses. The combination model could be a reliable tool to help radiologists differentiate ovarian tumors preoperative and thus may help in clinical decision-making for precision personality treatment. Abbreviations OC Ovarian cancer O-RADS Ovarian-Adnexal Reporting and Data System Ultrasound CA125 Carbohydrate antigen 125 LASSO Least Absolute Shrinkage and Selection Operator Rad-Score Radiomics score ROI Region of interest ROC Receiver operating characteristic curves AUC Area under the curve DCA Decision curve analysis Declarations Consent for publication Not applicable Competing interests The authors declare that they have no competing interests. Funding This study has been sponsored by Fujian Provincial Health Technology Project(Grant No.2022QNA044), the Startup Fund for Scientific Research, Fujian Medical University (Grant No.2019QH1194), Fujian Provincial Natural Science Foundation of China (Grant No.2023J011240), and Sciences Foundation of Fujian Cancer Hospital (Grant No.2023YN13). Author Contribution LNT and ZLW designed the study; YQW, YJC, ZSD and XHK performed the ultrasound examination; TFW and WTX analyzed the data; WTX wrote the manuscript; All authors reviewed the manuscript. Acknowledgements Not applicable Data Availability Data availability statement Data is provided within the manuscript or supplementary information files. References Zheng, L. et al. Incidence and mortality of ovarian cancer at the global, regional, and national levels, 1990–2017, Gynecologic oncology 159, 239-47, doi:10.1016/j.ygyno.2020.07.008(2020). Siegel, R. L. et al. Cancer statistics, 2022. CA: A Cancer Journal for Clinicians 72, 7-33,doi:10.3322/caac.21708(2022). Xiao, Y. et al. Multi-omics approaches for biomarker discovery in early ovarian cancer diagnosis. eBioMedicine 79, 104001, doi:10.1016/j.ebiom.2022.104001(2022). Sisodia, R.C. et al. Lesions of the Ovary and Fallopian Tube. New England Journal of Medicine 387, 727-736, doi:10.1056/NEJMra2108956(2022). Dang, Thi. Minh. N. et al. IOTA simple rules: An efficient tool for evaluation of ovarian tumors by non-experienced but trained examiners - A prospective study. Heliyon, 10, e24262, doi:10.1016/j.heliyon.2024.e24262(2024). Basha, M.A.A. et al. Comparison of O-RADS, GI-RADS, and IOTA simple rules regarding malignancy rate, validity, and reliability for diagnosis of adnexal masses. European radiology 31, 674-684, doi:10.1007/s00330-020-07143-7(2021). Andreotti, R.F. et al. Ovarian-Adnexal Reporting Lexicon for Ultrasound: A White Paper of the ACR Ovarian-Adnexal Reporting and Data System Committee. Journal of the American College of Radiology:JACR 15, 1415-1429, doi:10.1016/j.jacr.2018.07.004(2018). Strachowski, L.M. et al. O-RADS US v2022: An Update from the American College of Radiology's Ovarian-Adnexal Reporting and Data System US Committee. Radiology 308, e230685, doi:10.1148/radiol.230685(2023). Xie, W. T. et al. Efficacy of IOTA simple rules, O-RADS, and CA125 to distinguish benign and malignant adnexal masses. Journal of ovarian research 15, 15, doi:10.1186/s13048-022-00947-9(2022). Hack, K. et al. External Validation of O-RADS US Risk Stratification and Management System. Radiology 304, 114-120, doi:10.1148/radiol.211868(2022). Ponsiglione, A et al. Ovarian imaging radiomics quality score assessment: an EuSoMII radiomics auditing group initiative.European radiology 33, 2239-2247, doi:10.1007/s00330-022-09180-w(2022). Shrestha, P. et al. A systematic review on the use of artificial intelligence in gynecologic imaging–Background, state of the art, and future directions. Gynecologic oncology 166, 596-605, doi:10.1016/j.ygyno.2022.07.024(2022). Crispin-Ortuzar, M. et al. Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer. Nat Commun 14, 6756, doi:10.1038/s41467-023-41820-7(2023). Zuo, R. et al. Prediction of ovarian cancer prognosis using statistical radiomic features of ultrasound images. Phys Med Bio 69, doi:10.1088/1361-6560/ad4a02(2024). Chiappa, V. et al. A decision support system based on radiomics and machine learning to predict the risk of malignancy of ovarian masses from transvaginal ultrasonography and serum CA-125. Eur Radiol Exp 5, 28, doi:10.1186/s41747-021-00226-0(2021). Andreotti, R.F. et al. O-RADS US Risk Stratification and Management System: A Consensus Guideline from the ACR Ovarian-Adnexal Reporting and Data System Committee. Radiology 294, 168-185, doi:10.1148/radiol.2019191150(2020). Van, Griethuysen. J.J.M. et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Research 77, e104-e7, doi:10.1158/0008-5472.CAN-17-0339(2017). Wheeler, V. et al. Adnexal Masses: Diagnosis and Management. American family physician 108,580-587, (2023). Cao, L. et al. Validation of American College of Radiology Ovarian-Adnexal Reporting and Data System Ultrasound (O-RADS US): Analysis on 1054 adnexal masses. Gynecologic oncology 162, 107-112, doi:10.1016/j.ygyno.2021.04.031(2021). Chen, H. et al. Deep Learning Prediction of Ovarian Malignancy at US Compared with O-RADS and Expert Assessment. Radiology 304,106-113, doi:10.1148/radiol.211367(2022). Vara, J. et al. Ovarian Adnexal Reporting Data System (O-RADS) for Classifying Adnexal Masses: A Systematic Review and Meta-Analysis.Cancers 14, 3151, doi:10.3390/cancers14133151(2022). Yuan, K. et al. Contrast-enhanced US to Improve Diagnostic Performance of O-RADS US Risk Stratification System for Malignancy. Radiology 308, doi:10.1148/radiol.223003 (2023). Zhou, S. et al. Comparison of the diagnostic efficiency between the O-RADS US risk stratification system and doctors’ subjective judgment. BMC Medical Imaging 23, doi:10.1186/s12880-023-01153-9 (2023). Adusumilli, P. et al. Radiomics in the evaluation of ovarian masses—a systematic review. Insights into Imaging 14, 165, doi:10.1186/s13244-023-01500-y(2023). Mitchell, S et al. Artificial Intelligence in Ultrasound Diagnoses of Ovarian Cancer: A Systematic Review and Meta-Analysis. Cancers, 16, 422, doi:10.3390/cancers16020422(2024). Wang, Y. et al. Advances in artificial intelligence for the diagnosis and treatment of ovarian cancer (Review). Oncology reports 51, 46, doi:10.3892/or.2024.8705(2024). Tang, Z. et al. Ultrasound-based radiomics for predicting different pathological subtypes of epithelial ovarian cancer before surgery. BMC Medical Imaging 22, 147, doi:10.1186/s12880-022-00879-2(2022). Yao, F. et al. Nomogram based on ultrasound radiomics score and clinical variables for predicting histologic subtypes of epithelial ovarian cancer. The British Journal of Radiology 95, doi:10.1259/bjr.20211332 (2022). Qi, L. et al. Diagnosis of Ovarian Neoplasms Using Nomogram in Combination With Ultrasound Image-Based Radiomics Signature and Clinical Factors. Frontiers in genetics 12,753948,doi:10.3389/fgene.2021.753948(2021). Chen, J. et al. Diagnostic value of a CT-based radiomics nomogram for discrimination of benign and early stage malignant ovarian tumors. European Journal of Medical Research 28, 609, doi:10.1186/s40001-023-01561-1(2023). Hatamikia, S. et al. Ovarian cancer beyond imaging: integration of AI and multiomics biomarkers. European Radiology Experimental 7, 50, doi:10.1186/s41747-023-00364-7(2023). Liu, L. et al. Ultrasound image-based nomogram combining clinical, radiomics, and deep transfer learning features for automatic classification of ovarian masses according to O-RADS. Frontiers in oncology 14, doi:10.3389/fonc.2024.1377489 (2024). Das, A. et al. DeepOvaNet: A Comprehensive Deep Learning Framework for Predicting and Diagnosing Ovarian Cancer in Women Across Menopausal Transitions. 2024 Fourth International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies (ICAECT), 1-7, doi:10.1109/ICAECT60202.2024.10469613(2024). Additional Declarations No competing interests reported. Supplementary Files SupplementaryTable1.docx SupplementaryTable2.docx SupplementaryTable3.docx SupplementaryTable4.docx Cite Share Download PDF Status: Published Journal Publication published 02 Jun, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Accepted 15 May, 2025 Reviews received at journal 05 May, 2025 Reviewers agreed at journal 13 Apr, 2025 Reviewers invited by journal 27 Mar, 2025 Submission checks completed at journal 27 Mar, 2025 First submitted to journal 26 Mar, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5468347","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":434856675,"identity":"b64e2c97-0bf7-4fd3-adc4-e75bf184005c","order_by":0,"name":"Wenting Xie","email":"","orcid":"","institution":"Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wenting","middleName":"","lastName":"Xie","suffix":""},{"id":434856677,"identity":"a356620e-0c4d-47f4-aa08-f5aeb095ddde","order_by":1,"name":"Yaoqin Wang","email":"","orcid":"","institution":"Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yaoqin","middleName":"","lastName":"Wang","suffix":""},{"id":434856678,"identity":"0e199fc5-2503-48da-8afd-1eb84b8f17e8","order_by":2,"name":"Zhongshi Du","email":"","orcid":"","institution":"Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zhongshi","middleName":"","lastName":"Du","suffix":""},{"id":434856679,"identity":"2dcfb3e1-e8ac-4ebe-90b7-ab8294754e25","order_by":3,"name":"Yijie Chen","email":"","orcid":"","institution":"Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yijie","middleName":"","lastName":"Chen","suffix":""},{"id":434856680,"identity":"27828092-5d2a-4504-a0c2-06ded7da138b","order_by":4,"name":"Xiaohui Ke","email":"","orcid":"","institution":"Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaohui","middleName":"","lastName":"Ke","suffix":""},{"id":434856681,"identity":"e518b041-9eed-4ae9-912f-e51cf1ffa516","order_by":5,"name":"Tingfan Wu","email":"","orcid":"","institution":"United Imaging Healthcare Group Co., Ltd","correspondingAuthor":false,"prefix":"","firstName":"Tingfan","middleName":"","lastName":"Wu","suffix":""},{"id":434856682,"identity":"30293ac8-66b6-4326-9e47-8848fd4d49cc","order_by":6,"name":"Zhilan Wang","email":"","orcid":"","institution":"Department of Ultrasound, Nanping First Hospital Affiliated to Fujian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhilan","middleName":"","lastName":"Wang","suffix":""},{"id":434856685,"identity":"efd44396-1112-4baf-93a7-0d1f589525ec","order_by":7,"name":"Lina Tang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIie2Qv2sCMRTHnwTuloNbz0X/AiHl4NrBev/KCwdOR+noeFDI6to/Qygc7RYJ3NTuDkItgjg4xK2Dg8/g4JKzo9B8IC8/yIeXbwA8nltlxQFiphB+Ktqx00nL7ehUkJSuRARhFXLwqkKDN1StAleUPPzSW3xe9tMG1ka8D/P7MJ4bhFFvUDm6RE/jB+Sbu7qh94nPsfh4YSxBKNJMuR5WZhy57tTfFaKQmtYMSFGidinxzir5mwRUpOSksN9WJSnTFSliFlB4UjozzYL2LotdRiF08dpAAULaLBmlc2cJp2VqzEE/TmVUmL2kH4vn64WZjHouhQiSc0O0E7+oLpg5N1R/uOzxeDz/kSNu9WBOrgqGXAAAAABJRU5ErkJggg==","orcid":"","institution":"Clinical Oncology School of Fujian Medical University, Fujian Cancer Hospital","correspondingAuthor":true,"prefix":"","firstName":"Lina","middleName":"","lastName":"Tang","suffix":""}],"badges":[],"createdAt":"2024-11-17 05:38:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5468347/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5468347/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-02776-4","type":"published","date":"2025-06-02T15:57:40+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":79550796,"identity":"f1ad6239-24e8-4f68-9f3f-8acbfd482e76","added_by":"auto","created_at":"2025-03-31 06:35:54","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":626770,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the patient selection process.\u003c/p\u003e","description":"","filename":"Figure1.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/7444931b04c34b390e48e154.jpg"},{"id":79549776,"identity":"77107bf8-b23e-46be-8c07-2084f37e41a4","added_by":"auto","created_at":"2025-03-31 06:27:54","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":193819,"visible":true,"origin":"","legend":"\u003cp\u003eRadiomics feature selection using the least absolute shrinkage and selection operator regression (LASSO). a, selection of optimal penalization coefficient; the optimal penalization coefficient lambda (λ) was determine through ten-fold cross-validation in the LASSO model. b, distribution of coefficients in Lasso regression. Bottom X-axes represent the value of λ in the Lasso regression model.\u003c/p\u003e","description":"","filename":"Figure2.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/003c8bae2b666e92c5a9d07b.jpg"},{"id":79549781,"identity":"a4894eb3-61f2-44ba-9baa-fcb261779629","added_by":"auto","created_at":"2025-03-31 06:27:54","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":505246,"visible":true,"origin":"","legend":"\u003cp\u003eNomograms and calibration curves for predicting the malignant risk in patients with ovarian masses. a, the O-RADS US+clinical nomogram. b, calibration plots for the O-RADS US+clinical nomogram; c, the Radscore+clinical nomogram; d, calibration plots for the Radscore+clinical nomogram; e, the O-RADS US+Radscore+Clinical nomogram; f, calibration curves for the O-RADS US+Radscore+Clinical nomogram. The calibration curves depict the calibration of the models in terms of the agreement between the predicted probabilities of malignant and observed outcomes.\u003c/p\u003e","description":"","filename":"Figure3.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/be7a5d41921705a1371749c3.jpg"},{"id":79549783,"identity":"acec7667-8ece-4f13-83dd-4f41a4ec3f64","added_by":"auto","created_at":"2025-03-31 06:27:54","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":263830,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of the classification results for nomogram models. a, ROC for the training cohort; b, ROC for the validation cohort.\u003c/p\u003e","description":"","filename":"Figure4.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/75b84c5ad4c1b014673007b6.jpg"},{"id":79550804,"identity":"5e33b925-1c8e-41cc-a06b-766e0590fa11","added_by":"auto","created_at":"2025-03-31 06:35:55","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":93445,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curves analysis was used to assess the clinical significance of the model. The red curve represents the O-RADS US+Radscore+Clinical model. The blue curve represents the Radscore+Clinical model. The green curve represents the O-RADS US+Clinical model. The gray line represents the assumption that all masses were malignant (the treat-all scheme). The black line represents the assumption that all masses were benign (the treat-none scheme). All models exhibited higher clinical benefit values.\u003c/p\u003e","description":"","filename":"Figure5.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/ed119949e6a9ec601b7c04f9.jpg"},{"id":84242584,"identity":"7979eb91-6312-4e57-aa11-fdf77dee9849","added_by":"auto","created_at":"2025-06-09 16:09:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2674706,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/d91442a5-e83a-4d12-9229-2f3311750c78.pdf"},{"id":79549785,"identity":"feadd56f-468f-409a-b76b-b5309d4c2b5e","added_by":"auto","created_at":"2025-03-31 06:27:54","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":12100,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable1.docx","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/a87537059b56f9bd19ae0c15.docx"},{"id":79549779,"identity":"7421cd31-5592-4b14-abd7-42b55a713457","added_by":"auto","created_at":"2025-03-31 06:27:54","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":12956,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable2.docx","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/d622e9ad414be783b8248ea2.docx"},{"id":79549784,"identity":"53ca02f8-4a22-4695-b65e-7563d3fc7e2f","added_by":"auto","created_at":"2025-03-31 06:27:54","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":14468,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable3.docx","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/1bdf44589fdfca8c2e37db5f.docx"},{"id":79552052,"identity":"c8fd547d-5b93-455b-b09a-e00da1b452af","added_by":"auto","created_at":"2025-03-31 06:43:54","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":14100,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable4.docx","url":"https://assets-eu.researchsquare.com/files/rs-5468347/v1/696fa9bc5310973b194a11e5.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A nomogram combining clinical features, O-RADS US, and radiomics based on ultrasound imaging for diagnosing ovarian cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eOvarian cancer (OC) is a major malignant tumor worldwide and the fifth leading cause of death among women in the United States, with over 28,6100 incident cases and 17,6000 deaths in 2017[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Given the lack of early accurate screening tools, rarity of symptoms in early stages, and the rapid dissemination of the disease, the 5-year survival rate of patients with ovarian cancer ranges from 30\u0026ndash;50%[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. As the treatment planning for benign and malignant ovarian masses is quite different, preoperative and accurate discrimination of ovarian lesions is crucial for tailoring therapeutic procedures and preserving patients\u0026rsquo; fertility[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eUltrasound (US) is the primary first-line imaging modality for detecting and diagnosing ovarian masses. It has been widely used to differentiate benign from malignant ovarian tumors. As ovarian tumors are heterogeneous, different OC pathologies can exhibit similar US characteristics, and similar pathological types can exhibit different US features. Many models, including International Ovarian Tumour Analysis Simple Rules (IOTA) and the Gynecologic Imaging Report and data system (GI-RADS), have been proposed to improve the discrimination of adnexal masses[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Despite the high diagnostic accuracy of these models[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], their application in clinical practice has been limited internationally because of their rejection in certain nations. In 2018, the American College of Radiology (ACR) released Ovarian-Adnexal Reporting and Data System Ultrasound (O-RADS US) to improve the standardization of describing adnexal masses ultrasound features and reports[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. The system included six risk classification categories (O-RADS 0\u0026ndash;5). O-RADS category 0 is considered an incomplete evaluation; O-RADS category 1 is the normal ovary with 0% risk of malignancy; O-RADS category 2 is almost certainly benign lesions (\u0026lt;1% chance of malignancy); O-RADS category 3 referring to low risk (1% to\u0026lt;10% chance of malignancy); O-RADS category 4 is intermediate risk (10% to\u0026lt;50% chance of malignancy) and O-RADS category 5 is high risk (\u0026ge;\u0026thinsp;50% chance of malignancy)[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Our previous study evaluated the diagnostic performance of O-RADS US to differentiate benign from malignant tumors[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Although O-RADS US showed a higher ability to differentiate benign from malignant ovarian masses, the diversity and complexity of ovarian tumors still pose a formidable challenge for radiologists[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. An efficient and reproducible method for accurate diagnosis of ovarian masses is urgently needed to assist inexperienced radiologists or medical resource-lacking regions in diagnosing the malignancy of ovarian masses.\u003c/p\u003e \u003cp\u003eRecently, radiomics has garnered attention for its ability to accurately diagnose OC through computer algorithms to extract thousands of quantitative features from radiology images[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Studies have shown that radiomics nomogram analysis can help diagnose and predict the outcome of OC[\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Radiomics can be avoided inconsistent inter-observer and intra-observer circumstances for improving diagnostic accuracy in distinguishing benign from malignant ovarian lesions. Study have demonstrated that radiomics based on transvaginal ultrasonography and serum cancer antigen 125 (CA125) can predict the risk of malignancy of ovarian lesions with an accuracy of 88%[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHere, we aimed to develop and validate a model as an non-invasive tool for a personalized diagnostic of OC patients. The present study developed a radiomics score by extracting and selecting features from the ovarian lesion US images. Least absolute shrinkage and selection operator (LASSO) regression, a statistical technique used to study the effects of variables on outcome predictions, was used to select radiomics features. Eventually, a nomogram comprising clinical information, the radiomics score, and the O-RADS US category was developed to predict the malignancy of ovarian lesions. The diagnostic performance of this model was evaluated, and its clinical utility was demonstrated.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eEthics\u003c/h2\u003e \u003cp\u003e The retrospective study was approved by the institutional ethical committee board of the Fujian Cancer Hospital and Nanping First Hospital Affiliated to Fujian Medical University (K2021-128-01 and NPSY202201015). Due to the retrospective nature of the study, all patients waived the informed consent. The research methods were conducted in accordance with the relevant guidelines and regulations.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStudy population\u003c/h3\u003e\n\u003cp\u003eWe retrospectively studied 981 consecutive patients (with 981 ovarian masses) who were pathologically confirmed with ovarian tumor between January 2017 and December 2021 from two tertiary hospitals: Fujian Cancer Hospital (FJZL, n\u0026thinsp;=\u0026thinsp;749) and Nanping First Hospital Affiliated to Fujian Medical University (NPFH, n\u0026thinsp;=\u0026thinsp;232). All patients were included in this study based on surgery pathological results and received an ultrasound with sufficient image quality before surgery. The clinical characteristics comprising age, maximum tumor diameter, location, ascites, and CA125 were recorded in this study. Maximum tumor diameter and ascites were evaluated via the US and then recorded. Maximum tumor diameter was obtained by measuring the maximum diameter of the mass in any plane. Ascites are fluid extending above the uterine fundus beyond the pouch of Douglas or fluid anterior/superior to the uterus [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. The study flow diagram of the population is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eUS Image Acquisition\u003c/h3\u003e\n\u003cp\u003ePreoperative US examinations were performed by experienced radiologists and acquired standardized ultrasound images. The ultrasonic equipment included LOGIQ E11 (GE Healthcare, United States; convex array probes measuring 1\u0026ndash;5 MHz; transvaginal probes measuring 2\u0026ndash;9 MHz) and Philips IU 22 (Philips Bothell, United States; convex array probes with 2\u0026ndash;6 MHz; transvaginal probes measuring 4\u0026ndash;8 MHz). In cases of multiple or bilateral lesions, the lesion with the highest O-RADS US score was selected for further analysis. For each ovarian lesion, slices were acquired, and the larger mass or mass with the worst morphology was selected. Representative ultrasound image for each patient was selected by two radiologists with more than 6 years of ultrasound operating experience who were blinded for the outcome, and all recorded images were classified using O-RADS US. Any dispute was resolved with consensus. The classification criteria of O-RADS US were based on the ACR Ovarian-Adnexal Reporting and Data System Committee[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eRadiomics analysis\u003c/h3\u003e\n\u003cp\u003eRadiomics workflow comprised tumor image segmentation and feature extraction; feature selection and radscore establishment; and model construction and evaluation.\u003c/p\u003e\n\u003ch3\u003eTumor image segmentation\u003c/h3\u003e\n\u003cp\u003eAll enrolled patients were randomly divided at a proportion of 7:3 between a training cohort (n\u0026thinsp;=\u0026thinsp;686) and a validation cohort (n\u0026thinsp;=\u0026thinsp;295). For each lesion, a region of interest (ROI) of the tumor boundaries was manually delineated using Labelme software (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/wkentaro/labelme\u003c/span\u003e\u003cspan address=\"https://github.com/wkentaro/labelme\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) by ultrasound radiologists (WTX, with 7 years of experience). Subsequently, 794 radiomics features of ovarian tumor ROIs were extracted using PyRadiomic (version 2.2.0, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pyradiomics.readthedocs.io/en/latest\u003c/span\u003e\u003cspan address=\"https://pyradiomics.readthedocs.io/en/latest\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The radiomics features are subdivided into eight classes, and the detailed description is described in Supplementary Table\u0026nbsp;1.\u003c/p\u003e \u003cp\u003eTwo independent radiologists (twice by WTX with an interval of 4 weeks and once by YQW with 15 years of experience) delineated ROIs on 295 randomly chosen images to evaluate inter-observer and intra-observer reproducibility. The inter-observer and intra-observer reproducibility were assessed using the intraclass and interclass correlation coefficients (ICCs). Features with an ICC of \u0026gt;0.90 were used in subsequent analyses.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eFeature extraction and selection\u003c/h2\u003e \u003cp\u003eAfter eliminating the constant term, 635 features from each patient were used for further selection. Normality analysis was performed for each radiomics feature and then matched in pairs for correlation analysis. Pearson correlation analysis was implemented to evaluate the pair-wise feature satisfied normality; otherwise, spearman correlation analysis was performed to calculate features that did not satisfy the normal analysis. Finally, LASSO regression were used to select and reduce the number of the extracted radiomics features.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDevelopment, validation, and clinical application of the nomogram\u003c/h3\u003e\n\u003cp\u003eWe performed univariate and multivariate logistic regression to analyze the significant risk factors for OC. \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05 was considered significant in both univariate and multivariate analysis. The multivariable logistic regression method established predictive nomogram models based on Radscore, clinically significant risk factors, and O-RADS US. Receiver operating curves (ROCs) were plotted to evaluate the differentiation efficiency of nomograms in the training and validation cohorts. Calibration curves were used to explore the predictive accuracy of the models. Decision curve analysis (DCA) was performed to determine the clinical usefulness of the nomograms in the validation cohort.\u003c/p\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eIBM SPSS Statistics (version 25.0; SPSS Inc., Chicago, IL, USA) and R (version 4.0.3) software were used for statistical analyses. All clinical independent variables were generated using univariable and multivariable regression analyses. LASSO regression, ROCs, and nomogram model were built using the \u0026ldquo;glmnet\u0026rdquo;, \u0026ldquo;pROC\u0026rdquo;, and rms R packages, respectively. The area under the curves (AUCs), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to measure the model\u0026rsquo;s performance. The DeLong\u0026rsquo;s test was used to compare the difference between the AUCs.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eClinical characteristics\u003c/h2\u003e \u003cp\u003eA total of 981 patients with ovarian masses were retrospectively enrolled in this study. Supplementary Table\u0026nbsp;2 summarizes the characteristics of recruited patients with two centers. Final pathological outcome for all women who underwent surgery is shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Among these patients, 499 (50.9%) were benign and 482 (49.1%) were malignant, respectively. The demographic characteristics of total recruited patients are reported in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. There were no significant differences in age, tumor size, CA125, location, ascites, O-RADS US, and Radscore between the training and validation datasets.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eFinal pathological diagnosis of 981 adnexal masses from the two centers\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePathologic diagnosis\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo.(%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBenign adnexal masses\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e499(50.86)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMature teratoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e129(13.15)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEndometrioid cyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e116(11.83)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMucinous cystadenoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e85(8.66)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSerous cystadenoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e76(7.76)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThecoma-fibroma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e41(4.18)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFollicular cyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e17(1.73)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStruma ovarii\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e10(1.02)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHemorrhagic cyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7(0.71)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSerous adenofibroma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e6(0.61)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorpus luteum cyst\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4(0.41)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBrenner tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStromal tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMucinous adenofibroma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInflammation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMalignant adnexal masses\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e482(49.13)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSerous cystadenocarcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e271(27.62)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBorderline mucinous cystadenoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e45(4.59)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBorderline serous cystadenoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e38(3.87)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClear cell carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e25(2.55)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetastatic carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e39(3.98)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEndometrioid carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e17(1.73)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMalignant Muellerian tube mixed tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8(0.82)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGranulosa cell tumour\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8(0.82)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImmature teratoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7(0.71)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMucinous cystadenocarcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7(0.71)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBorderline endometrioid tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e6(0.61)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYolksactumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3(0.32)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOvarian dysgerminoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSertoli-Leydig cell tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBorderline Brenner Tumor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmall cell carcinoma of the ovary\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2(0.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"2\"\u003eData are given as n (%)\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClinicopathological characteristics in the training and the validation datasets.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraining dataset\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;686)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eValidation dataset\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;295 )\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge,years (Q.25-Q.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e48.0 (37.0\u0026ndash;57.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e47.0 (36.0\u0026ndash;57.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.420\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor size, mm (Q.25-Q.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e85.0 (58.2\u0026ndash;124)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e89.0 (58.0-130)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.286\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCA125,U/ml (Q.25-Q.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e48.5 (16.2\u0026ndash;362)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e41.2 (15.0-248)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.334\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLocation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.490\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLeft\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e288 (42.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e136 (46.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e265 (38.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e106 (35.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBilateral\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e133 (19.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e53 (18.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAscites\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.640\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e167 (24.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e67 (22.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e519 (75.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e228 (77.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eO-RADS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.377\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e150 (21.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e52 (17.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e140 (20.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e60 (20.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e155 (22.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e78 (26.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e241 (35.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e105 (35.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePathological\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.999\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBenign\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e349 (50.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e150 (50.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMalignant\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e337 (49.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e145 (49.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadscore\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.15(-2.13- 2.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.05 (-2.14- 3.99)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.417\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003emedian (interquartile range)Q.25 25% quantile; Q.75 75% quantile\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eInter-observer and intra-observer reproducibility of feature extraction\u003c/h2\u003e \u003cp\u003eA total of 295 patients were randomly selected to evaluate the inter-observer and intra-observer reproducibility of feature extraction. The median intra-observer ICCs of the first reader\u0026rsquo;s two extractions was 0.998. The median inter-observer ICCs of radiomics features extraction between the two radiologists was 0.959. Subsequent calculations were based on the radiomics features of the first reader.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eRadscore establishment\u003c/h2\u003e \u003cp\u003eAfter analysis, 794 extracted features were reduced to seven potential radiomics features based on the training set (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e summarizes the explanation and relevance of these features for the diagnosis of ovarian cancer. Finally, seven radiomics features were incorporated into the radiomics score formula. Radiomics score (Radscore)\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.0355681 - (0.2541203 \u0026times; original_firstorder_Maximum) + (0.1468945 \u0026times; original_glszm_ZoneEntropy) + (0.0499583 \u0026times; original_gldm_DependenceVariance) + (0.0034785 \u0026times; CoLIAGe2D_WindowSize9_Entropy_firstorder_Variance) \u0026ndash; (0.0077041 \u0026times; wavelet.LHL_lbp.3D.m2_firstorder_InterquartileRange) \u0026ndash; (0.0183370 \u0026times; wavelet.HLH_lbp.3D.k_firstorder_Minimum) \u0026ndash; (0.0002613 \u0026times; wavelet.LLL_lbp.3D.m2_firstorder_Range).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eA summary of the explanation and relevance to ovarian cancer diagnosis of the features extracted using PyRadiomics in the radiomics score formula.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFeature name\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRelevance to ovarian cancer diagnosis\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eoriginal_firstorder_Maximum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe maximum gray level intensity within the ROI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAssociated with certain high-density areas within the tumor which is helpful to identify benign and malignant tumors\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eoriginal_glszm_ZoneEntropy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMeasures the uncertainty/randomness in the distribution of zone sizes and gray levels\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eA higher value indicates more heterogeneous in the texture patterns\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eoriginal_gldm_DependenceVariance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMeasures the variance in dependence size in the image\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHigher dependence variance may indicate heterogeneity of tumor tissue\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoLIAGe2D_WindowSize9_Entropy_firstorder_Variance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe variance of the entropy of co-occurrence local intensity angle gradient entropy within a window size of 9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHigher variance may indicate significant heterogeneity within the tumor region\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ewavelet.LHL_lbp.3D.m2_firstorder_InterquartileRange\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe range from the 25\u003csup\u003eth\u003c/sup\u003e and 75\u003csup\u003eth\u003c/sup\u003e percentile of the image array\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eA higher interquartile range may indicate a larger range of gray values within the tumor region\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ewavelet.HLH_lbp.3D.k_firstorder_Minimum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe minimum gray level intensity in the wavelet-transformed image\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eThe lowest gray value in different scales and directions, which may be related to the darkest region in the tumor area\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ewavelet.LLL_lbp.3D.m2_firstorder_Range\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe range of gray values in the ROI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eA large range may indicate a wide range of density changes within the tumor\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eUnivariate and multivariate analyses of clinical information, O-RADS, and Radscore\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e displays the results of the univariate and multivariate logistic regression analyses for discriminating benign from malignant ovarian lesions in the training dataset. The odd ratios of age, CA125, O-RADS US, and radscore were found to be significant (\u003cem\u003ep\u003c/em\u003e\u0026lt;0.05). The O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical nomogram was developed by incorporating the aforementioned independent risk factors (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee). The Radscore\u0026thinsp;+\u0026thinsp;Clinical model was developed with age, CA125, location, ascites, and Radscore (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec, Supplementary Table\u0026nbsp;3). The O-RADS US\u0026thinsp;+\u0026thinsp;Clinical model was developed with age, CA125, and O-RADS US (Figur. 3a, Supplementary Table\u0026nbsp;4).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eResults of the univariate and multivariate analyses of O-RADS\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model based on the training set.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eUnivariate analysis\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eMultivariate analysis\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOR (95% CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ep value\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR (95% CI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003ep value\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge, years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.057(1.043,1.07)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.024(1.002,1.047)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.033\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor size,mm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.008(1.005,1.011)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCA125(U/ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.005(1.004,1.007)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.001(1.000,1.002)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLocation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLeft\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.689(0.508,0.934)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0168\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.535(0.391,0.73)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBilateral\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.473(3.520,8.774)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAscites\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.44(3.85,14.163)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eO-RADS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.642(5.124,16.556)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11.15(7.030,18.92)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadscore\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.77(1.619,1.955)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e14.58(4.10,54.88)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eModeling and Evaluation of the Nomograms\u003c/h2\u003e \u003cp\u003eThe radiomics features of each patient were extracted from US images and selected using LASSO. A predictive model\u0026mdash;based on radiomics scores, clinical characteristics, and O-RADS US\u0026mdash;was finally generated with the multivariable logistic regression method. Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows the diagnosis performance of the O-RADS US\u0026thinsp;+\u0026thinsp;Clinical, Radscore\u0026thinsp;+\u0026thinsp;Clinical, and O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical models in differentiating benign from malignant ovarian neoplasm.\u003c/p\u003e\n\u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eTable 5\u0026nbsp;\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003ePerformance of the nomograms for differentiating benign from malignant tumors.\u003c/span\u003e\u003c/p\u003e\n\u003ctable style=\"border-collapse: collapse;border: none;width: 568px;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border-right: none;border-bottom: none;border-left: none;border-image: initial;border-top: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eVariable\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 101.65pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eO-RADS US\u003c/span\u003e\u003cspan style=\"font-family:SimSun;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eClinical model\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border-right: none;border-bottom: none;border-left: none;border-image: initial;border-top: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 106.6pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eRadscore\u003c/span\u003e\u003cspan style=\"font-family:SimSun;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family: \"Times New Roman\",serif;'\u003eClinical model\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border-right: none;border-bottom: none;border-left: none;border-image: initial;border-top: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 122.6pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eO-RADS US+Radscore+Clinical\u0026nbsp;\u003c/span\u003e\u003cspan style=\"font-family:SimSun;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family: \"Times New Roman\",serif;'\u003eClinical model\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eTraining dataset\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eValidation dataset\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eTraining\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003edataset\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eValidation dataset\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eTraining\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003edataset\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eValidation dataset\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eAUC\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.830*\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e(0.801,\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.861)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.815**\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e(0.767,\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.864)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.876\u003c/span\u003e\u003csup\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＆\u003c/span\u003e\u003c/sup\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e(0.850,\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.904)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.867\u003c/span\u003e\u003csup\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＆＆\u003c/span\u003e\u003c/sup\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e(0.826,\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.909)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.967\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e(0.954,\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.980)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.951\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e(0.929,\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e0.974)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eSensitivity, %\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.756(0.707,0.806)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.744(0.666,0.824)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.792(0.742,0.831)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.724(0.686,0.821)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.928(0.896,0.953)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.800(0.730,0.858)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eSpecificity, %\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.773(0.725,0.822)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.693(0.607, 0.780)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.839(0.800,0.879)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.860(0.810,0.920)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.865(0.827,0.898)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.926(0.873,0.961)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003ePPV, %\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.763(0.714,0.813)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.701(0.615,0.788)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.826(0.783,0.867)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.833(0.783,0.905)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.869(0.831,0.901)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.913()0.853,0.953\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eNPV, %\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.767(0.718,0.816)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.737(0.659,0.816)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.807(0.763,0.843)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.763(0.724,0.844)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.926(0.893,0.951)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.827(0.763,0.879)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003ePLR\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e3.342(2.647,4.222)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e2.428(1.812,3.256)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e4.937(3.844,6.566)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e5.172(3.812,9.285)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e6.896(5.28,8.99)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e10.909(6.11,19.48)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eNLR\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.314(0.258,0.383)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.368(0.275,0.492)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.247(0.198,0.316)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.320(0.203,0.376)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.082(0.054,0.125)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border: none;padding: 0in 5.4pt;height: 17.05pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.215(0.153,0.304)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 71.25pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003eAccuracy, %\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 47.55pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.765(0.732,0.798)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.1pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.718(0.661,0.776)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 52.25pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.816(0.788,0.843)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54.35pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.793(0.773,0.855)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 12pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63.85pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.896(0.872,0.917)\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58.75pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;text-align:left;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:red;'\u003e0.864(0.823,0.899)\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;'\u003ePPV, Positive predictive value; NPV, Negative predictive value; PLR, Positive likelihood ratio; \u003cspan style=\"color:black;\"\u003eNLR, Negative likelihood ratio; AUC, area under the receiver operating characteristic curve.\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e*indicates a significant difference compared with that of O-RADS\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eRadiomics\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eClinical model in the training cohort, \u003cem\u003ep\u003c/em\u003e = 7.925e-31.\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003e** indicates a significant difference compared with that of O-RADS\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eRadiomics\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eClinical model \u0026nbsp; in the validation cohort, \u003cem\u003ep\u003c/em\u003e =1.07e-15.\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003csup\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＆\u003c/span\u003e\u003c/sup\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eindicates a significant difference compared with that of O-RADS\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eRadiomics\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eClinical model in the training cohort, \u003cem\u003ep\u0026nbsp;\u003c/em\u003e=7.76e-16.\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;text-align:justify;font-size:14px;font-family:\"Calibri\",sans-serif;'\u003e\u003csup\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＆＆\u003c/span\u003e\u003c/sup\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eindicates a significant difference compared with that of O-RADS\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eRadiomics\u003c/span\u003e\u003cspan style=\"font-family:SimSun;color:black;\"\u003e＋\u003c/span\u003e\u003cspan style='font-family:\"Times New Roman\",serif;color:black;'\u003eClinical model in the training cohort, \u003cem\u003ep\u003c/em\u003e =1.886e-06.\u003c/span\u003e\u003c/p\u003e \u003cp\u003eIn the training dataset, the O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model achieved a sensitivity of 92.8%, a specificity of 86.5%, and an AUC value of 0.967, significantly higher than that of Radscore\u0026thinsp;+\u0026thinsp;Clinical (AUC\u0026thinsp;=\u0026thinsp;0.876) and O-RADS US\u0026thinsp;+\u0026thinsp;Clinical (AUC\u0026thinsp;=\u0026thinsp;0.830). In the validation cohort, the O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model also obtained an excellent result, with a significantly higher AUC value (AUC\u0026thinsp;=\u0026thinsp;0.951) compared with that of the Radscore\u0026thinsp;+\u0026thinsp;Clinical (AUC\u0026thinsp;=\u0026thinsp;0.867) and O-RADS US\u0026thinsp;+\u0026thinsp;Clinical (AUC\u0026thinsp;=\u0026thinsp;0.815) models (both \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The ROCs of all nomograms in the training and validation cohorts are plotted in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model exhibited outstanding discrimination performance.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eCalibration and clinical usefulness\u003c/h2\u003e \u003cp\u003eAll three nomograms showed good agreement in predicting ovarian neoplasm; the calibration plots are shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb, \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed, and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef. DCA was used to evaluate the clinical usefulness of the three nomograms (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical nomogram obtained the best clinical benefit for predicting OC, followed by the Radscore\u0026thinsp;+\u0026thinsp;Clinical and O-RADS US\u0026thinsp;+\u0026thinsp;Clinical nomograms in the DCA curves.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study focused on investigating the use of radiomics combining O-RADS US and clinical information for classification of ovarian masses. Therefore, we investigated whether radiomics could be used to identify patients with OC from ovarian masses and whether combining O-RADS US with radiomics could improve the diagnostic performance of OC.\u003c/p\u003e \u003cp\u003eIn the present study, we applied radiomics features to ultrasonographic images, and preoperative nomogram models were developed, integrating O-RADS US, radiomics, and clinical information to predict the malignant risk of ovarian lesions. Age and CA125 were identified as independent factors through multivariate analysis. Our results revealed that O-RADS US and radiomics could predict ovarian malignancy lesions with high accuracy. Combined with the aforementioned clinical information, the diagnostic AUCs of the Radscore\u0026thinsp;+\u0026thinsp;Clinical model significantly exceeded the O-RADS US\u0026thinsp;+\u0026thinsp;Clinical model in both the training (0.876 vs. 0.830) and validation cohorts (0.867 vs. 0.815). The AUCs of the O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model in the training and validation group were 0.967 and 0.951, respectively. The O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model offered significantly higher AUCs than the O-RADS US\u0026thinsp;+\u0026thinsp;Clinical and Radscore\u0026thinsp;+\u0026thinsp;Clinical model in both the training (\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001) and validation cohorts (\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001), demonstrating that radiomics plus O-RADS US imaging model in preoperative exhibited a high potential to discern OC from ovarian tumors patients.\u003c/p\u003e \u003cp\u003eUltrasound is a non-invasive tool in the diagnosis and treatment of ovarian tumors in clinical practice[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. O-RADS US is an effective method to distinguish benign ovarian masses from malignant ones. Cao \u003cem\u003eet al.\u003c/em\u003e reported that the diagnostic performance of O-RADS US was good, with an AUC was 0.960[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Similarly, Chen \u003cem\u003eet al.\u003c/em\u003e compared O-RADS and deep learning to predict ovarian malignancy and revealed that the deep learning method and O-RADS US exhibited comparable diagnostic performance for classifying malignant from benign ovarian tumors with AUCs of 0.93 and 0.92, respectively[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. A recent meta-analysis indicated that the pooled estimated sensitivity and specificity of the O-RADS US system for diagnostic adnexal masses were 97% (95% confidence interval (CI)\u0026thinsp;=\u0026thinsp;94\u0026ndash;98%) and 77% (95% CI\u0026thinsp;=\u0026thinsp;68\u0026ndash;84%), respectively[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. However, Yuan \u003cem\u003eet al.\u003c/em\u003e reported the diagnostic performance of O-RADS US with an AUC of 0.71 in the validation group[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Zhou \u003cem\u003eet al.\u003c/em\u003e found that the AUC of O-RADS US was 0.86 in predicting malignant adnexal masses when comparing the diagnostic efficiency between subjective assessment and the O-RADS US[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. The discrepant in diagnostic efficiency of O-RADS US could be attributed to the level of the sonographer and the proportion of malignant and benign ovarian tumor cases between studies.\u003c/p\u003e \u003cp\u003eRadiomics is a computer-aided technology for diagnosing and predicting OC[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Previous studies have demonstrated that radiomics based on ultrasound images can differentiate between benign and malignant ovarian tumors[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Besides, Tang \u003cem\u003eet al.\u003c/em\u003e found that ultrasound-based radiomics exhibited a good differential diagnosis of type I and type II epithelial OC before surgery[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. YAO \u003cem\u003eet al.\u003c/em\u003e reported that first-order statistics, GLSZM, GLRLM, GLCM, and NGTDM were significantly correlated with ovarian lesions, and their US radiomics model successfully differentiated type I and type II EOC [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Our study selected seven optimal radiomics of the 794 features to establish the radiomics score, including five features of first-order, one of GLSZM, and one of GLDM. These features reflected heterogeneity, non-uniformity, and variability, and their potential association with the characteristics of ovarian lesions. The radiomics score was an independent predictor for ovarian malignancy, and the Radscore\u0026thinsp;+\u0026thinsp;Clinical model exhibited good diagnostic performance for identifying benign and malignant ovarian tumors (with AUCs of 0.876 and 0.867 in the training and validation groups, respectively), suggesting that radiomics provided additional value for individualized malignant prediction. Consistent with our results, Qi \u003cem\u003eet al.\u003c/em\u003e reported that combining clinical index and radiomics signatures performed the best AUC for differentiating benign from malignant ovarian serous tumors[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. However, Radiomics methods share an essential limitation, where sonographers rely on handcrafted ROI[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Both imaging features and O-RADS US alone are often insufficient to determine the malignant risk of ovarian neoplasms. Hence, clinicians also consider all factors, including clinical information and sonographers\u0026rsquo; subjective assessments make the diagnosis[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. By contrast, models integrating O-RADS US, radiomics features, and clinical features could provide added diagnostic value, allowing more comprehensive evaluations to distinguish OC from benign ovarian lesions. As expected, our present study found that the combination nomogram model achieved satisfactory diagnostic performance, indicating the effectiveness of this combined model in preoperatively predicting the malignant risk of ovarian masses and in assisting individualized management in patients with adnexal lesions. Moreover, our study indicated that US-based radiomics can be used as an important supplement to O-RADS US in the classification of ovarian masses. As far as we know, only one study focused on the use of radiomics for discriminating between benign and malignant ovarian tumors according to O-RADS US, and achieved good diagnostic performance with an AUC of 0.93[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn this study, the clinical decision curve demonstrates that all models are valuable in predicting OC among various ovarian tumors, and the O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model had the highest net return.\u003c/p\u003e \u003cp\u003eHowever, our work has some limitations. First, the retrospective study design may have case selection and verification bias. The O-RADS US category of reviews can also be considered a limitation. Considering these limitations, large multi-center prospective research will be required for further study. Second, only ultrasound images with larger lesions or lesions with the worst morphology were used to extract features. Although these images can represent the characteristics of the ovarian lesions, they did not completely represent the entire tumor. Therefore, future research requires ultrasound video images that can reflect the entire lesion. Third, the reliability and reproducibility of the ROIs drawn by radiologists did not completely avoid the tumor\u0026rsquo;s surrounding tissue. As reported by A. Das \u003cem\u003eet al\u003c/em\u003e, deep learning can be employed to improve the identification of ovarian cancer[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Thus, deep learning combined with O-RADS US to differentiate between malignant and benign ovarian tumors will be required for further study. Fourth, images obtained by different equipment can be seen as a limitation of this study. The same ultrasound imaging parameters, such as frequency, dynamic range, and gain settings should be carried out in future work to reduce the impact on the extraction of radiomic features. Five, the distribution of benign and malignant cases might affect the accuracy of the model. Non-operated cases with presumed benign diagnosis followed up by ultrasound can be enrolled in further study. Additionally, a limitation of radiomics studies is that they require special software and statistical skills, which limits their application in clinical practice. This model could be turned into an online tool or software within the ultrasound device to help junior radiologists in diagnosing the malignancy of ovarian masses, which should be taken into consideration in future work.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, we constructed a combination nomogram model that combines O-RADS US, Radscore, and clinical information, and we then validated the effective diagnostic performance of the model for the diagnosis of ovarian masses. The combination model could be a reliable tool to help radiologists differentiate ovarian tumors preoperative and thus may help in clinical decision-making for precision personality treatment.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eOC Ovarian cancer\u003c/p\u003e\n\u003cp\u003eO-RADS Ovarian-Adnexal Reporting and Data System Ultrasound\u003c/p\u003e\n\u003cp\u003eCA125 Carbohydrate antigen 125\u003c/p\u003e\n\u003cp\u003eLASSO Least Absolute Shrinkage and Selection Operator\u003c/p\u003e\n\u003cp\u003eRad-Score Radiomics score\u003c/p\u003e\n\u003cp\u003eROI Region of interest\u003c/p\u003e\n\u003cp\u003eROC Receiver operating characteristic curves\u003c/p\u003e\n\u003cp\u003eAUC Area under the curve\u003c/p\u003e\n\u003cp\u003eDCA Decision curve analysis\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConsent for publication\u003c/h2\u003e \u003cp\u003eNot applicable\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis study has been sponsored by Fujian Provincial Health Technology Project(Grant No.2022QNA044), the Startup Fund for Scientific Research, Fujian Medical University (Grant No.2019QH1194), Fujian Provincial Natural Science Foundation of China (Grant No.2023J011240), and Sciences Foundation of Fujian Cancer Hospital (Grant No.2023YN13).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eLNT and ZLW designed the study; YQW, YJC, ZSD and XHK performed the ultrasound examination; TFW and WTX analyzed the data; WTX wrote the manuscript; All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eNot applicable\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData availability statement Data is provided within the manuscript or supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eZheng, L. et al. Incidence and mortality of ovarian cancer at the global, regional, and national levels, 1990\u0026ndash;2017, Gynecologic oncology 159, 239-47, doi:10.1016/j.ygyno.2020.07.008(2020).\u003c/li\u003e\n\u003cli\u003eSiegel, R. L. et al. Cancer statistics, 2022. CA: A Cancer Journal for Clinicians 72, 7-33,doi:10.3322/caac.21708(2022).\u003c/li\u003e\n\u003cli\u003eXiao, Y. et al. Multi-omics approaches for biomarker discovery in early ovarian cancer diagnosis. eBioMedicine 79, 104001, doi:10.1016/j.ebiom.2022.104001(2022).\u003c/li\u003e\n\u003cli\u003eSisodia, R.C. et al. Lesions of the Ovary and Fallopian Tube. New England Journal of Medicine 387, 727-736, doi:10.1056/NEJMra2108956(2022).\u003c/li\u003e\n\u003cli\u003eDang, Thi. Minh. N. et al. IOTA simple rules: An efficient tool for evaluation of ovarian tumors by non-experienced but trained examiners - A prospective study. Heliyon, 10, e24262, doi:10.1016/j.heliyon.2024.e24262(2024).\u003c/li\u003e\n\u003cli\u003eBasha, M.A.A. et al. Comparison of O-RADS, GI-RADS, and IOTA simple rules regarding malignancy rate, validity, and reliability for diagnosis of adnexal masses. European radiology 31, 674-684, doi:10.1007/s00330-020-07143-7(2021).\u003c/li\u003e\n\u003cli\u003eAndreotti, R.F. et al. Ovarian-Adnexal Reporting Lexicon for Ultrasound: A White Paper of the ACR Ovarian-Adnexal Reporting and Data System Committee. Journal of the American College of Radiology:JACR 15, 1415-1429, doi:10.1016/j.jacr.2018.07.004(2018).\u003c/li\u003e\n\u003cli\u003eStrachowski, L.M. et al. O-RADS US v2022: An Update from the American College of Radiology\u0026apos;s Ovarian-Adnexal Reporting and Data System US Committee. Radiology 308, e230685, doi:10.1148/radiol.230685(2023).\u003c/li\u003e\n\u003cli\u003eXie, W. T. et al. Efficacy of IOTA simple rules, O-RADS, and CA125 to distinguish benign and malignant adnexal masses. Journal of ovarian research 15, 15, doi:10.1186/s13048-022-00947-9(2022).\u003c/li\u003e\n\u003cli\u003eHack, K. et al. External Validation of O-RADS US Risk Stratification and Management System. Radiology 304, 114-120, doi:10.1148/radiol.211868(2022).\u003c/li\u003e\n\u003cli\u003ePonsiglione, A et al. Ovarian imaging radiomics quality score assessment: an EuSoMII radiomics auditing group initiative.European radiology 33, 2239-2247, doi:10.1007/s00330-022-09180-w(2022).\u003c/li\u003e\n\u003cli\u003eShrestha, P. et al. A systematic review on the use of artificial intelligence in gynecologic imaging\u0026ndash;Background, state of the art, and future directions. Gynecologic oncology 166, 596-605, doi:10.1016/j.ygyno.2022.07.024(2022).\u003c/li\u003e\n\u003cli\u003eCrispin-Ortuzar, M. et al. Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer. Nat Commun 14, 6756, doi:10.1038/s41467-023-41820-7(2023).\u003c/li\u003e\n\u003cli\u003eZuo, R. et al. Prediction of ovarian cancer prognosis using statistical radiomic features of ultrasound images. Phys Med Bio 69, doi:10.1088/1361-6560/ad4a02(2024). \u003c/li\u003e\n\u003cli\u003eChiappa, V. et al. A decision support system based on radiomics and machine learning to predict the risk of malignancy of ovarian masses from transvaginal ultrasonography and serum CA-125. Eur Radiol Exp 5, 28, doi:10.1186/s41747-021-00226-0(2021).\u003c/li\u003e\n\u003cli\u003eAndreotti, R.F. et al. O-RADS US Risk Stratification and Management System: A Consensus Guideline from the ACR Ovarian-Adnexal Reporting and Data System Committee. Radiology 294, 168-185, doi:10.1148/radiol.2019191150(2020).\u003c/li\u003e\n\u003cli\u003eVan, Griethuysen. J.J.M. et al. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Research 77, e104-e7, doi:10.1158/0008-5472.CAN-17-0339(2017).\u003c/li\u003e\n\u003cli\u003eWheeler, V. et al. Adnexal Masses: Diagnosis and Management. American family physician 108,580-587, (2023).\u003c/li\u003e\n\u003cli\u003eCao, L. et al. Validation of American College of Radiology Ovarian-Adnexal Reporting and Data System Ultrasound (O-RADS US): Analysis on 1054 adnexal masses. Gynecologic oncology 162, 107-112, doi:10.1016/j.ygyno.2021.04.031(2021).\u003c/li\u003e\n\u003cli\u003eChen, H. et al. Deep Learning Prediction of Ovarian Malignancy at US Compared with O-RADS and Expert Assessment. Radiology 304,106-113, doi:10.1148/radiol.211367(2022). \u003c/li\u003e\n\u003cli\u003eVara, J. et al. Ovarian Adnexal Reporting Data System (O-RADS) for Classifying Adnexal Masses: A Systematic Review and Meta-Analysis.Cancers 14, 3151, doi:10.3390/cancers14133151(2022). \u003c/li\u003e\n\u003cli\u003eYuan, K. et al. Contrast-enhanced US to Improve Diagnostic Performance of O-RADS US Risk Stratification System for Malignancy. Radiology 308, doi:10.1148/radiol.223003 (2023).\u003c/li\u003e\n\u003cli\u003eZhou, S. et al. Comparison of the diagnostic efficiency between the O-RADS US risk stratification system and doctors\u0026rsquo; subjective judgment. BMC Medical Imaging 23, doi:10.1186/s12880-023-01153-9 (2023).\u003c/li\u003e\n\u003cli\u003eAdusumilli, P. et al. Radiomics in the evaluation of ovarian masses\u0026mdash;a systematic review. Insights into Imaging 14, 165, doi:10.1186/s13244-023-01500-y(2023).\u003c/li\u003e\n\u003cli\u003eMitchell, S et al. Artificial Intelligence in Ultrasound Diagnoses of Ovarian Cancer: A Systematic Review and Meta-Analysis. Cancers, 16, 422, doi:10.3390/cancers16020422(2024).\u003c/li\u003e\n\u003cli\u003eWang, Y. et al. Advances in artificial intelligence for the diagnosis and treatment of ovarian cancer (Review). Oncology reports 51, 46, doi:10.3892/or.2024.8705(2024).\u003c/li\u003e\n\u003cli\u003eTang, Z. et al. Ultrasound-based radiomics for predicting different pathological subtypes of epithelial ovarian cancer before surgery. BMC Medical Imaging 22, 147, doi:10.1186/s12880-022-00879-2(2022).\u003c/li\u003e\n\u003cli\u003eYao, F. et al. Nomogram based on ultrasound radiomics score and clinical variables for predicting histologic subtypes of epithelial ovarian cancer. The British Journal of Radiology 95, doi:10.1259/bjr.20211332 (2022).\u003c/li\u003e\n\u003cli\u003eQi, L. et al. Diagnosis of Ovarian Neoplasms Using Nomogram in Combination With Ultrasound Image-Based Radiomics Signature and Clinical Factors. Frontiers in genetics 12,753948,doi:10.3389/fgene.2021.753948(2021).\u003c/li\u003e\n\u003cli\u003eChen, J. et al. Diagnostic value of a CT-based radiomics nomogram for discrimination of benign and early stage malignant ovarian tumors. European Journal of Medical Research 28, 609, doi:10.1186/s40001-023-01561-1(2023).\u003c/li\u003e\n\u003cli\u003eHatamikia, S. et al. Ovarian cancer beyond imaging: integration of AI and multiomics biomarkers. European Radiology Experimental 7, 50, doi:10.1186/s41747-023-00364-7(2023).\u003c/li\u003e\n\u003cli\u003eLiu, L. et al. Ultrasound image-based nomogram combining clinical, radiomics, and deep transfer learning features for automatic classification of ovarian masses according to O-RADS. Frontiers in oncology 14, doi:10.3389/fonc.2024.1377489 (2024).\u003c/li\u003e\n\u003cli\u003eDas, A. et al. DeepOvaNet: A Comprehensive Deep Learning Framework for Predicting and Diagnosing Ovarian Cancer in Women Across Menopausal Transitions. 2024 Fourth International Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies (ICAECT), 1-7, doi:10.1109/ICAECT60202.2024.10469613(2024).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"ovarian cancer, Ultrasound, Ovarian-Adnexal Reporting and Data System Ultrasound, radiomics, Nomogram","lastPublishedDoi":"10.21203/rs.3.rs-5468347/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5468347/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eWe aimed to develop and validate a nomogram for diagnosing ovarian cancer from ovarian masses based on clinical information, O-RADS US, and radiomics.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA total of 981 patients with ovarian masses from two centers were randomly divided into the training cohort (n\u0026thinsp;=\u0026thinsp;686) and the validation cohort (n\u0026thinsp;=\u0026thinsp;295). We defined the region of interest (ROI) of the tumor by manually drawing the tumor contour on the ultrasound image of the lesion. The radiomics features were extracted from ultrasound images, and the radiomics score was then calculated. O-RADS US characteristics, radiomics score, and clinical features selected using the LASSO algorithm were used to develop O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical, Radscore\u0026thinsp;+\u0026thinsp;Clinical, and O-RADS US\u0026thinsp;+\u0026thinsp;Clinical models, respectively. Receiver operating characteristic (ROC), decision curve analysis, and calibration curve were used to evaluate the performance of the nomogram models.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAge, CA125, O-RADS US, and radiomics score were related to ovarian malignancy through univariate and multivariate logistic regression analyses. In the training and validation datasets, the areas under the ROC curve (AUC) of O-RADS US\u0026thinsp;+\u0026thinsp;Clinical model were 0.830 and 0.815, respectively, and those for the Radscore\u0026thinsp;+\u0026thinsp;Clinical model were 0.876 and 0.867, respectively. The O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical nomogram model presented improved AUC values of 0.967 in the training group and 0.951 in the validation group, significantly higher than that of Radscore\u0026thinsp;+\u0026thinsp;Clinical and O-RADS US\u0026thinsp;+\u0026thinsp;Clinical models. The calibration curve and the clinical decision curve analysis demonstrated that the nomogram models had high clinical benefits. The O-RADS US\u0026thinsp;+\u0026thinsp;Radscore\u0026thinsp;+\u0026thinsp;Clinical model had the highest net return.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eCombination nomogram model that integrates clinical features, O-RADS US, and radiomics based on ultrasound image analysis could predict ovarian malignancy with high diagnostic accuracy, indicating that this model might have a role in preoperative diagnosis for differentiating benign and malignant ovarian tumors.\u003c/p\u003e","manuscriptTitle":"A nomogram combining clinical features, O-RADS US, and radiomics based on ultrasound imaging for diagnosing ovarian cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-31 06:27:49","doi":"10.21203/rs.3.rs-5468347/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Accepted","date":"2025-05-15T11:37:33+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-05T15:58:50+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"12661119753973763751564981369691061345","date":"2025-04-13T13:07:26+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-03-27T12:39:44+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-03-27T06:57:57+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-03-26T21:44:27+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"bfe7a738-a220-46ee-810f-7148e474b5eb","owner":[],"postedDate":"March 31st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":46303964,"name":"Biological sciences/Cancer"},{"id":46303966,"name":"Biological sciences/Cancer/Cancer imaging"},{"id":46303968,"name":"Biological sciences/Cancer/Gynaecological cancer"}],"tags":[],"updatedAt":"2025-06-09T16:03:12+00:00","versionOfRecord":{"articleIdentity":"rs-5468347","link":"https://doi.org/10.1038/s41598-025-02776-4","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-06-02 15:57:40","publishedOnDateReadable":"June 2nd, 2025"},"versionCreatedAt":"2025-03-31 06:27:49","video":"","vorDoi":"10.1038/s41598-025-02776-4","vorDoiUrl":"https://doi.org/10.1038/s41598-025-02776-4","workflowStages":[]},"version":"v1","identity":"rs-5468347","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5468347","identity":"rs-5468347","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}