A clinical-radiomics nomogram based on multisequence MRI for predicting intraoperative vertebral artery injury in patients with primary cervical spine tumor: a diagnostic study

A clinical-radiomics nomogram based on multisequence MRI for predicting intraoperative vertebral artery injury in patients with primary cervical spine tumor: a diagnostic study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A clinical-radiomics nomogram based on multisequence MRI for predicting intraoperative vertebral artery injury in patients with primary cervical spine tumor: a diagnostic study Xiangzhi Ni, Jiayang Yan, Jiayi Zhang, Fukai Li, Guangwen Duan, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7218476/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 20 Jan, 2026 Read the published version in European Spine Journal → Version 1 posted 9 You are reading this latest preprint version Abstract Background Currently, reliable preoperative methods for predicting VA invasion are lacking. The authors develop a novel model based on MRI radiomic signatures combined with clinical and imaging features for predicting intraoperative vertebral artery injury in patients with primary cervical tumors. Methods Included in this retrospective study were 167 patients who received surgical resection for primary cervical tumors. They were randomly assigned to a training set (n = 116) and a test set (n = 51) set. Least absolute shrinkage and selection operator logistic regression was applied for feature selection and radiomic signature construction. A multilayer perceptron (MLP) model and 10 machine learning models were used to develop diverse prediction models. Independent risk factors of clinical variables were screened by Logistic regression, based on which a clinical model was constructed. A combined model was established by combining the radiomic signatures and clinical factors. The predictive performance of the combined model was evaluated in both training and test sets using Hosmer–Lemeshow test and decision curve analysis (DCA). Results According to the scoring system, the MLP model obtained the highest total score of 82, meaning that its prediction performance was the best of all evaluated models, so the MLP was selected to construct the radiomics model. The AUC of the combined model in the training and test cohorts was 0.952 and 0.932 respectively, and both were higher than that of the radiomics model (AUC 0.861 in training set, p = 0.005, AUC 0.773 in test set, p = 0.006) and the clinical model (AUC 0.777 in training set, p < 0.001, AUC 0.740 in test set, p = 0.002) alone. Conclusion The present study presents a nomogram that incorporates radiomic signatures and clinical features, which could be used to predict the risk of intraoperative VA injury in patients with primary cervical tumors. Primary cervical tumors MRI Radiomics Vertebral artery Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Cervical spine tumors, which are classified as primary or metastatic, often cause severe spinal cord compression and spinal instability, leading to incurable pain, loss of mobility, and neurological deficits [ 1 ]. Although metastatic cervical spine tumors are more common than primary cervical spine tumors, the former are often treated with palliative treatment due to limited survival time, while primary cervical spine tumors aim at cure and the treatment process is more complicated [ 2 ]. Chordomas, chondrosarcomas, giant cell tumors of bone and neurogenic tumors are the most common tumors in the cervical spine [ 3 , 4 ]. Due to the drug resistance of the tumor and the maximum tolerated dose of radiation to the spinal cord [ 5 ], surgery is often the primary treatment for primary tumors of the cervical spine [ 6 , 7 ]. A variety of studies have pointed out that En bloc resection of the vertebrae with adequate margins in an intact patient decreases the risk of local recurrence and tumor-related mortality [ 8 , 9 ]. Due to the complexity of cervical anatomy involving the vertebral artery (VA) and cervical nerve roots, it is challenging to achieve en bloc resection. The VA, arising as the first branches from the subclavian arteries, is partitioned into four distinct segments from V1 to V4 [ 10 ]. The bilateral VA enters the intracranial aspect, where they merge to form the basilar artery, which participates in the Willis circulation and supplies an important part of the posterior cerebral circulation [ 11 , 12 ]. Therefore, great attention should be paid to ligation of the unilateral VA. Knowing that sacrificing the bilateral VA may elicit a high risk of stroke or even death, surgery becomes a great challenge when the primary cervical tumor involves the VA [ 13 , 14 ]. Tumor invasion of the VA increases the probability of failure of VA dissection, which often leads to intraoperative vascular injury and tumor residue [ 15 , 16 ]. The dominant or bilateral VA injury needs to be remedied by vascular repair or vascular grafting. Therefore, preoperative prediction of VA injury is helpful to make individualized treatment plans and improve the safety of surgery. At present, the routine imaging preoperative examination for cervical spine tumors includes X-ray, CT and MRI etc. [ 17 – 19 ]. It is often difficult to predict the risk of VA injury based on visual features from preoperative examination. Traditional imaging features are usually visually assessed and qualitatively described by radiologists or nuclear medicine physicians, which creates the possibility of large intra-observer and inter-observer variability in these visual assessments [ 20 ]. In recent years, radiomics has developed rapidly and showed outstanding potential in many computer-aided diagnostic applications, also in predicting tumor invasion of blood vessels. Xu et al. successfully developed a new model for predicting microvascular invasion of hepatocellular carcinoma based on CT radiomics analysis, and the model showed good performance [ 21 ]. Yang et al. used PET-CT-based radiomics features to predict lymph vascular invasion in gastric cancer patients and provided effective predictors of patient survival outcomes [ 22 ]. Because of its high resolution, tissue contrast, and multiplanar capacity, MRI has innate advantages in the evaluation of cervical spine tumors [ 23 ]. Therefore, it is more meaningful to conduct radiomics research based on MRI for patients with primary cervical tumors. To the best of our knowledge from a review of the literature, there is no radiomics study on prediction of VA injury in surgery of cervical tumor. Therefore, the aim of this study is to develop a combined model based on MRI and clinical information to predict the probability of VA injury during surgery for the sake of increasing the safety of surgery in cervical tumors and providing a useful guide for selecting the strategy of surgical intervention. 2. Materials and Methods 2.1 Participants and image data sets This research project was approved by the Ethics Committee of our hospital, with the written informed consent waived due to the retrospective nature of the study. The work has been reported in line with the STARD (Standards for the Reporting of Diagnostic accuracy studies) criteria[ 24 ]. Patients who were diagnosed with cervical spine tumors and received gross total resection in our institution between January 2013 and June 2024 were reviewed retrospectively. The inclusion criteria were patients (1) who were diagnosed with primary cervical spine tumors; (2) who underwent preoperative MRI examination and the quality was qualified by the radiologist; (3) who achieved gross total resection; and (4) whose clinical data were complete. The exclusion criteria were patients (1) with metastatic tumors; (2) with low-quality MRI or incomplete clinical data; and (3) whose internal fixation in the cervical spine was interfered with. Finally, 579 patients were enrolled, and 167 patients were included in this study, including 56 patients with VA injury. They were randomly divided into a training set ( n = 116) and a test set ( n = 51) at a 7:3 ratio. The workflow for patient inclusion and exclusion is shown in Fig. 1 . All patients received MRI examinations using one of five MRI scanners, including 1.5T Siemens, 1.5T General Electric, 3.0T United Imaging uMR770,3.0T Philips and 3.0T GE premier. The imaging parameters from the five scanners are detailed in Table S1 . And all scans were performed consistently. Imaging sequences included in this study were T2WI and CET1 sequences. CET1 was performed after intravenous injection of the contrast medium (GD-DTPA Injection, Consun Pharma, Guangzhou, China) according to the recommended dose of 0.2 ml/kg. All sagittal, axial and coronal views were acquired and saved in DICOM format. In addition, the basic clinical information of the patients, including gender, age, pathology of the tumor, history of smoking, history of diabetes mellitus, history of alcohol consumption, body mass index (BMI), level of education and tumor invasion of VA segments, was collected through the electronic medical record system and PACS system. Two imaging experts with more than five years of imaging experience independently evaluated and measured the following information: 1) the angle of the tumor surrounding the VA measured in the cross-section of the enhanced image; 2) the length of the tumor invading the VA measured in the sagittal plane of the enhanced image; and 3) the degree of VA stenosis measured in the cross-section of enhanced images (the narrowest vessel circumference/the normal vessel circumference) [ 25 , 26 ]. 2.2 Treatment regimen The surgical indications are as follows: 1) malignant tumors were diagnosed by imaging or pathology, 2) patients with neurological symptoms, unbearable pain, and spinal instability. Gross total resection was used in all patients. All surgical procedures were performed by the same surgical team. For patients with large tumor invasion of the VA on MRI, preoperative balloon occlusion test was performed. Intraoperative tumor invasion of the VA was managed according to the following principles: 1) Attempts to isolate the VA were prioritized; 2) When VA injury occurred, priority is given to repair the VA; 3) when the VA could not be repaired and the contralateral VA had good circulation, ligation of the injured VA was considered; and 4) if the contralateral VA circulation was poor, vascular grafting was considered. The whole operation process comprised tumor excision, decompression of the spinal cord, reconstruction, and stabilization of the spine. 2.3 MRI segmentation and radiomic feature extraction The study flow diagram is shown in Fig. 2 . 3D image segmentation was manually performed by a senior resident in radiology with experience in musculoskeletal and oncologic imaging using the open-source software ITK-SNAP (version 4.2). The reader drew a region of interest (ROI) along tumor borders on each slice using an T2WI and CET1 sequence. Using the PyRadiomics platform, radiomic features of the volume of interest were extracted. The extracted features included first-order, gray level co-occurrence matrix, gray level run length matrix, gray level size zone matrix, gray level dependence matrix, and shape features. The Z score method was used to normalize the features and eliminate the difference in numerical scale. The optimal radiomics features were identified through a systematic three-step process. Initially, we performed a Student's t-test for features that met the criteria of normal distribution, while the Mann-Whitney U test was applied to those that did not follow this distribution. Radiomic features with a p-value of less than 0.05 were retained for further analysis. In the subsequent step, features exhibiting a Pearson correlation coefficient exceeding 0.9 were randomly removed to mitigate redundancy [ 27 , 28 ]. Lastly, the least absolute shrinkage and selection operator (LASSO) regression algorithm, along with the adjustment of penalty parameters, was employed to implement tenfold cross-validation. The radiomic model was developed by computing a radiomics score for each patient, derived from a linear combination of the selected features along with their corresponding weight coefficients. 2.4 Development, validation, and performance evaluation of the combined model Firstly, univariate and multivariate Logistic regression analyses of clinical characteristics were performed to define independent risk factors as meaningful risk factors, and a clinical model was constructed based on these factors. Second, patients in the training cohort were used to establish radiomics models, including logistic regression, NaiveBayes, Support Vector Machine, K-Nearest Neighbor, RandomForest, ExtraTrees, XGBoosting machine, Light Gradient Boosting Machine, GradientBoosting, AdaBoost, and Multilayer Perceptron (MLP). Patients in the test cohort were used to validate the model. Additionally, based on previous studies [ 29 , 30 ], this study used a scoring system to comprehensively evaluate the prediction performance of the models. Finally, a combined model was constructed based on the clinical model and the radiomics model, based on which a nomogram was generated to visualize the combined model, graphically evaluate variable importance, and calculate prediction accuracy. To validate the performance of the three sets of models, we used ROC curves and AUC to evaluate the performance of the prediction of the combined model, radiomics model, and clinical model. The calibration curves (Hosmer–Lemeshow test) were performed to evaluate the calibration of the nomogram. Decision curve analysis (DCA) was applied to evaluate the clinical practicability of the nomogram. 2.5 Statistical analysis Statistical analysis was performed by using the SPSS (SPSS 22.0; IBM Corp) and R Programming Language (version 3.0.1; http://www.Rproject.org ). Numeric data were presented as the mean ± standard deviation (SD). Differences in the distribution of categorical variables were compared using the chi-square test, and comparison of continuous variables was performed using Student’s t test and Mann-Whitney U test. Independent predictors were identified from the clinical variables by multivariate logistic regression. The R Programming Language was applied to conduct the LASSO regression analysis, ROC analysis, calibration curve analysis (with the Hosmer–Lemeshow test). And the “rmda” package was utilized for DCA. The AUC values between different models were compared by using the DeLong method. “ p ” value less than 0.05 was set as the statistically significant threshold. 3. Results 3.1 Clinical characteristics The baseline clinical characteristics of 167 included patients with primary cervical spine tumors are summarized in Table 1 . There was no significant difference in their distribution of the degree of VA stenosis, the length of VA invasion, injury of VA, age, gender, the segment of VA invasion, pathological diagnosis, and the angle of the tumor surrounding the VA between the overall, training and test cohorts. Univariate analysis showed that preoperative clinical information, including the degree of VA stenosis, the length of VA invasion, age, the segment of VA invasion, pathological diagnosis, was significantly associated with VA injury (Table 2 ). The multivariate analysis showed that the degree of VA stenosis and the length of VA invasion were independent risk factors for VA injury (Table 2 ). Table 1 The clinical characteristics of patients with vertebral artery injury and intact vertebral arteries in the training and test cohorts Features Overall (n = 167) Group p value Training cohort (n = 116) Test cohort (n = 51) Age (years) 0.891 < 47 92 (55.09%) 63 (54.31%) 29 (56.86%) ≥ 47 75 (44.91%) 53 (45.69%) 22 (43.14%) Sex 0.755 Male 119 (71.26%) 84 (72.41%) 35 (68.63%) Female 48 (28.74%) 32 (27.59%) 16 (31.37%) Smoking status 0.393 Nonsmoker 122 (73.05%) 87 (75.00%) 35 (68.63%) Current smoker 45 (26.95%) 29 (25.00%) 16 (31.37%) Diabetes 0.111 Yes 24 (14.37%) 20 (17.24%) 4 (7.84%) No 143 (85.63%) 96 (82.76%) 47 (92.16%) Drinking alcohol 0.699 Nondrinker 125 (74.85%) 88 (75.86%) 37 (72.55%) Current drinker 42 (25.15%) 28 (24.14%) 14 (27.45%) BMI 0.848 low weight 9 (5.39%) 7 (6.03%) 2 (3.92%) Normal weight 151 (90.42%) 104 (89.66%) 47 (92.16%) Overweight 7 (4.19%) 5 (4.31%) 2 (3.92%) Degree of education 0.480 High school and above 98 (58.68%) 66 (56.90%) 32 (62.75%) Below high school 69 (41.32%) 50 (43.10%) 19 (37.25%) VA injury 0.393 Yes 56 (33.53%) 36 (31.03%) 20 (39.22%) None 111 (66.47%) 80 (68.97%) 31 (60.78%) Pathological diagnosis 0.367 Malignant 89 (53.29%) 65 (56.03%) 24 (47.06%) Benign 78 (46.71%) 51 (43.97%) 27 (52.94%) Segment of VA 0.643 V1 + V2 55 (32.93%) 40 (34.48%) 15 (29.41%) V3 + V4 112 (67.07%) 76 (65.52%) 36 (70.59%) Angle (°) 0.765 < 180 66 (39.52%) 45 (38.79%) 21 (41.18%) ≤ 180 < 270 52 (31.14%) 35 (30.17%) 17 (33.33%) ≥ 270 49 (29.34%) 36 (31.03%) 13 (25.49%) Degree of VA stenosis 0.68 [0.50, 0.83] 0.67 [0.47, 0.82] 0.71 [0.55, 0.84] 0.340 Length of VA invasion (mm) 23.22 [16.83, 33.80] 24.35 [18.23, 35.14] 21.39 [15.81, 29.05] 0.080 Table 2 Univariate and multivariate logistic regression analysis were used to analyze the independent risk factors of intraoperative vertebral artery injury in patients with primary cervical tumor Variables Univariate analysis Multivariate analysis OR (95%CI) p value OR (95%CI) p value Age 0.395 (0.239–0.652) 0.002 1.047 (0.440–2.492) 0.930 Sex 0.600 (0.329–1.094) 0.162 Smoking status 0.803 (0.317–2.035) 0.643 Diabetes 0.943 (0.330–2.692) 0.912 Drinking alcohol 1.070 (0.429–2.668) 0.884 BMI 0.863 (0.254–2.930) 0.813 Degree of education 0.924 (0.418–2.042) 0.845 Pathological diagnosis 0.413 (0.264–0.647) 0.001 0.877 (0.377–2.040) 0.798 Segment of VA 0.429 (0.243–0.756) 0.014 0.733 (0.301–1.782) 0.565 Angle 1.011 (0.791–1.293) 0.940 Degree of VA stenosis 0.154 (0.087–0.272) < 0.001 0.003 (0.001–0.020) < 0.05 Length of VA invasion 0.987 (0.976–0.997) 0.031 1.095 (1.058–1.133) < 0.05 3.2 Feature selection and radiomics signature building A total of 2394 radiomic features were extracted. After reproducibility assessment and Pearson correlation analysis, 72 radiomic features were reserved, respectively. Finally, 13 radiomic features with non-zero coefficients were selected by LASSO regression (Fig. 3 ). By linearly combining those features after weighing by their corresponding coefficients, we constructed the radiomic signature. 3.3 Performance evaluation of the radiomics model Among the eleven models, the MLP model performed the best in terms of AUC (0.773, 95% CI: 0.643–0.902), and the XGBoosting Machine model only had the fourth AUC (0.760, 95% CI: 584–0.871), whereas the LightGBM model had a lower AUC of 0.672 (95% CI: 0.520–0.823) (Table 3 ). Using the scoring system, the MLP model achieved the highest total score of 82 (Fig. 4 ), indicating the best predictive performance. These results support MLP as the best model to evaluate whether VA injury occurred during surgery for cervical primary tumors. Therefore, we used the MLP model as the radiomics model in our study Table 3 Prediction performance of all models based on the test cohort’s validation Modle/Metrics Accuracy AUC(95% CI) Sensitivity Specificity PPV NPV Precision Recall F1 Total score LR 0.667 0.731 (0.589–0.872) 0.600 0.710 0.571 0.733 0.571 0.600 0.585 46 NaiveBayes 0.725 0.726 (0.583–0.869) 0.450 0.903 0.750 0.718 0.750 0.450 0.562 55 SVM 0.725 0.674 (0.508–0.840) 0.550 0.839 0.687 0.743 0.687 0.550 0.611 60 KNN 0.627 0.719 (0.582–0.856) 0.650 0.613 0.520 0.731 0.520 0.650 0.578 37 RandomForest 0.686 0.723 (0.571–0.874) 0.600 0.742 0.600 0.742 0.600 0.600 0.600 57 ExtraTrees 0.686 0.760 (0.627–0.894) 0.650 0.710 0.591 0.759 0.591 0.650 0.619 69 XGBoost 0.706 0.727 (0.584–0.871) 0.750 0.677 0.600 0.808 0.600 0.750 0.667 82 LightGBM 0.667 0.672 (0.520–0.823) 0.650 0.677 0.565 0.750 0.565 0.650 0.605 49 GradientBoosting 0.608 0.686 (0.536–0.837) 0.700 0.548 0.500 0.739 0.500 0.700 0.583 37 AdaBoost 0.588 0.678 (0.526–0.831) 0.750 0.484 0.484 0.750 0.484 0.750 0.588 42 MLP 0.706 0.773 (0.643–0.902) 0.650 0.742 0.619 0.767 0.619 0.650 0.634 82 To further verify the performance of the radiomics model, MLP was applied to the CET1 and T2WI sequences of MRI to establish the radiomics model. The results showed that the AUC of the radiomics model (0.773, 95% CI: 0.643–0.902) performed better than those of the CET1 model (0.673, 95% CI: 0.517–0.828) and T2WI model (0.773, 95% CI: 0.598–0.877) in the test cohort. 3.4 Construction and comparison of the clinical model, radiomic model, model, and combined model The clinical prediction model was constructed based on independent predictors of multivariate analysis, including length of the tumor invading the VA and degree of VA stenosis. A combined model was constructed by combining radiomics model and the independent risk factors of the above clinical features. The performance of the radiomic, clinical and combined models is shown in Fig. 5 and Table 4 . It was found that the performance of the combined model was better than that of the other two models in both the training set and test set. The AUC of the combined model (0.952, 95%CI, 0.919–0.986) was higher than that of the radiomic model (0.861, 95%CI 0.791–0.932, p = 0.005) and the clinical model (0.777, 95%CI 0.685–0.869, p < 0.001) in the training set. The AUC of the combined model (0.932, 95%CI, 0.868–0.996) was higher than that of the radiomic model (0.773, 95%CI 0.643–0.902, p = 0.006) and the clinical model (0.740, 95%CI 0.596–0.884, p = 0.002) in the test set. Table 4 Comparison of diagnostic performance of the radiomics model, clinical model, and combined model in the training and test cohorts Model AUC (95% CI) Accuracy Sensitivity Specificity PPV NPV Radiomics model Training cohort 0.861(0.791–0.932) 0.836 0.611 0.937 0.815 0.843 Test cohort 0.773(0.643–0.902) 0.706 0.650 0.742 0.619 0.767 Clinical model Training cohort 0.777(0.685–0.869) 0.716 0.778 0.687 0.528 0.873 Test cohort 0.740(0.596–0.884) 0.667 0.800 0.581 0.552 0.818 Combined model Training cohort 0.952(0.919–0.986) 0.853 0.889 0.837 0.711 0.944 Test cohort 0.932(0.868–0.996) 0.824 0.950 0.742 0.704 0.958 3.5 Development and performance of the nomogram For easier use by clinicians, we transformed the combined model into a nomogram (Fig. 6 ). Nomogram scores were given based on the weight of independent predictors, and the scale length of the nomogram variable was found to be positively correlated with its impact on the prediction of VA injury in patients with primary cervical spine tumors. According to the combined model, radiomics features had the greatest impact, followed by the length of the tumor invading the VA and degree of VA stenosis. The model calibration curve demonstrated good agreement between prediction and observation outcomes in the two cohorts (Fig. 6 ). The decision curves of the radiomics nomogram in the Training and Test sets are shown in Fig. 6 . According to DCA, when the probability threshold of a patient or physician's clinical decision was greater than 0.1, the combined model was more favorable than the other models in predicting the risk of intraoperative VA injury in the training and test set. An example of the nomogram in use is shown in Fig. 7 . We can read the top score scale upward from the predictors to determine the points associated with the length of the tumor invading the VA and degree of VA stenosis, and the Radscore. We can derive a total score based on the scores of each factor to determine the probability of VA injury during surgery for primary cervical tumors. 4. Discussion In this study, we developed a novel model for predicting interoperative VA injury for cervical primary tumors by analyzing the radiomics features of MRI and clinical features. To improve the accuracy of the model and facilitate the use by clinicians, a new combined model was constructed by combining radiomics features with clinical features. The results showed that the prediction performance of the multimodal model was significantly better than that of the radiomics model or the clinical feature model alone. Based on the above data, we finally constructed a nomogram to facilitate the use by clinicians. The result showed that the new combined model greatly reduced intraoperative uncertainty in patients with primary cervical tumors and VA invasion and provided a great deal of help for preoperative planning and perioperative nursing. The surgical treatment principles include gross total resection of the tumor and protection of the VA and important nerve roots [ 8 , 9 ]. However, the realization of this goal often faces formidable challenges. In addition to the anatomical complexity of the cervical spine, the larger reason is that surgical resection of cervical tumors can be a great challenge when the tumor surrounds the major blood vessels, especially the VA [ 31 – 33 ]. Studies have shown that the probability of VA injury during cervical spine surgery ranges from 0.03–1.2% [ 34 ], and this figure is increasing during cervical spine tumor surgery. Encapsulation of the tumor by major vessels often makes complete resection difficult and may cause serious complications such as fistulas, pseudoaneurysm, late-onset hemorrhage, thrombosis, embolism, cerebral ischemia, or even death [ 35 , 36 ]. However, incomplete resection of the primary tumor is often disastrous for the poor prognosis of patients due to tumor recurrence. When VA injury occurs, vascular repair, grafting or ligation is required. Therefore, it is very valuable to accurately predict the major risks that may occur during the operation. Our prediction model can predict in advance whether VA injury may occur or whether ligation is required during the operation of the primary cervical tumor, which could help surgeons to plan the operation strategy before operation and achieve total resection of the tumor under the premise of ensuring the safety of patients. Tumors are heterogeneous at gross and cellular levels as well as genetic and phenotypic levels, with spatial heterogeneity in cell density, angiogenesis, and necrosis [ 37 ]. This heterogeneity is often closely related to the treatment and prognosis of the patient. On the other hand, tumor heterogeneity also shows that different tumor tissues have different degrees of vascular invasion. It is usually difficult to assess tumor heterogeneity simply by using the conventional imaging tools [ 38 ](28898189). Compared with traditional imaging tools, radiomics can often extract the hidden information in the image [ 39 , 40 ]. Some radiomic metrics, particularly texture analysis metrics, have been reported to assess intratumoral heterogeneity in studies assessing the diagnosis, prognosis, and treatment response of cancer [ 41 , 42 ]. In our study, we constructed a radiomics signature consisting of four radiomics features, including first-order features, morphological features, texture features and wavelet features. None of these informative data can be obtained with the naked eye. Morphologic features describe the size, volume, and shape of the volume of interest, while first-order features mainly reflect the internal texture of the lesions. Textural features, including the gray level co-occurrence matrix and gray level dependence matrix, describe the spatial relationship between each pixel and its neighbors. Wavelet features mainly reflect the time frequency domain within the lesion [ 28 ]. Vascular imaging analysis has been extensively reported in the literature. A new model based on the study of texture and other features of CT images by Xu et al. also showed better performance than traditional features [ 21 ]. For new cerebral ischemic lesions after carotid artery stenting, Zhang et al. [ 43 ] constructed a new model based on radiomics features of high-resolution MRI of the vascular wall with better prediction performance than traditional features. The above studies showed that the prediction ability of models based on radiomics features could be significantly improved. In our study, we constructed a radiomics signature describing vascular and cervical spine tumors and combined it with clinical features to form a novel multimodal model. The result showed that the predictive performance of the new model was significantly better than that of the clinical model and the radiomics model. Many researchers have studied the relationship between tumors and surrounding large blood vessels, such as the relationship between pancreatic cancer and mesenteric arteries and veins. According to the American National Comprehensive Cancer Network (NCCN) consensus reporting guidelines, a tumor is resectable when the circumference of its contact with blood vessels is less than or equal to 180°, and unresectable when the circumference exceeds 180°. The degree of vascular stenosis in patients is often closely related to tumor compression [ 25 , 44 ]. To quantify the degree of vascular compression, Sugae et al. [ 26 ] developed the corresponding evaluation criteria, including the diameter of the stenosis/the diameter of the normal part of the vessel, and the length of vascular stenosis. It was found in our study that the degree of stenosis of the tumor-wrapped VA and the length of tumor invasion of the VA were independent risk factors for prognosis. This is generally consistent with literature reports. In general, the more tightly a tumor binds to the surrounding blood vessels, the deeper its invasion is, and the surrounding blood vessels are often narrowed by the tumor. The wrapping angle of the tumor and the VA were shown as significant factors in the univariate analysis of our study but shown no significance in multivariate analysis. This may be because the wrapping angle is not directly related to the degree of invasion, and on the other hand tumors with a large wrapping angle are often treated by combined anterior and posterior approaches. Our model may fill up a lacuna in the treatment of cervical primary tumors invading the VA and can be used to predict the probability of VA injury during surgery. Our model may fill up a lacuna in the treatment of cervical primary tumors invading the VA and can be used to predict the probability of VA injury during surgery [ 45 ]. Surgeons provide preoperative and postoperative safety safeguards for high-risk patients in advance. Before surgery, a VA balloon occlusion test can be performed to determine whether the contralateral VA is patent. Intraoperative surgical management of VA injury includes hemostatic tamponade, microvascular repair of the injured artery, vascular grafting and ligation of the VA. Therefore, acquisition of coordinative assistance from relevant departments in advance can not only achieve a rapid response to VA injury but also avoid the waste of medical resources. This study has some limitations. Firstly, selection bias could not be avoided because of the retrospective nature of the study. In addition, the sample size of patients with cervical primary tumors is not large enough. Finally, this is a single-center retrospective study without introducing external validation to ensure the unity of surgical methods and techniques. Multi-center and larger- sample studies are required to validate our findings and conclusions. 5. Conclusion We have developed a combined model to predict the presence or absence of VA injury during primary cervical tumor surgery by integrating radiomic-based features and clinical features (the length of the tumor invading the VA and degree of VA stenosis). This combined model may be able to provide some help in clinical decision making, and fill the gap in this field of research. Declarations Author Contribution Xiangzhi Ni, Jiayang Yan, and Jiayi Zhang contributed equally to this work. Fukai Li, Shuming Hou, Lingyun Shen, and Guangwen Duan conducted data acquisition, collection and interpretation. Fukai Li, Xiang Wang, and Jiayi Zhang made contribute in the data acquisition and annotation. Jiayang Yan made important contributions to the model designing and building. Jiayi Zhang and Hongbiao Sun made of image preprocessing and model interpretation. Xiangzhi Ni and Jiayang Yan made major contributions in manuscript drafting. Tielong Liu and Minglei Yang made contribution to the manuscript editing. Shiyuan Liu had final responsibility for the decision to submit for publication. All authors contributed to this study and article, and approved the submitted version. References Yang M, Zhong N, Dai Z, Ma X, Leng A, Zhou Y, Wang J, Jiao J, Xiao J (2024) Risks for prolonged mechanical ventilation and reintubation after cervical malignant tumor surgery: a nested case-control study. Eur Spine J 33:3069-3081. doi: 10.1007/s00586-024-08313-7 Lau D, Chan AK, Theologis AA, Chou D, Mummaneni PV, Burch S, Berven S, Deviren V, Ames C (2016) Costs and readmission rates for the resection of primary and metastatic spinal tumors: a comparative analysis of 181 patients. J Neurosurg Spine 25:366-378. doi: 10.3171/2016.2.SPINE15954 Kerr DL, Dial BL, Lazarides AL, Catanzano AA, Lane WO, Blazer DG, 3rd, Brigman BE, Mendoza-Lattes S, Eward WC, Erickson ME (2019) Epidemiologic and survival trends in adult primary bone tumors of the spine. Spine J 19:1941-1949. doi: 10.1016/j.spinee.2019.07.003 Chi JH, Bydon A, Hsieh P, Witham T, Wolinsky JP, Gokaslan ZL (2008) Epidemiology and demographics for primary vertebral tumors. Neurosurg Clin N Am 19:1-4. doi: 10.1016/j.nec.2007.10.005 Jiang P, Zhang X, Jiang W, Meng N, Aili A, Wang J (2020) Analysis of long-term outcome of image-guided volumetric modulated arc therapy (VMAT) for primary malignant tumor of the cervical spine. Cancer Biol Ther 21:623-628. doi: 10.1080/15384047.2020.1743149 Ferraresi V, Nuzzo C, Zoccali C, Marandino F, Vidiri A, Salducca N, Zeuli M, Giannarelli D, Cognetti F, Biagini R (2010) Chordoma: clinical characteristics, management and prognosis of a case series of 25 patients. BMC Cancer 10:22. doi: 10.1186/1471-2407-10-22 MacDonald IJ, Lin CY, Kuo SJ, Su CM, Tang CH (2019) An update on current and future treatment options for chondrosarcoma. Expert Rev Anticancer Ther 19:773-786. doi: 10.1080/14737140.2019.1659731 Mukherjee D, Chaichana KL, Parker SL, Gokaslan ZL, McGirt MJ (2013) Association of surgical resection and survival in patients with malignant primary osseous spinal neoplasms from the Surveillance, Epidemiology, and End Results (SEER) database. Eur Spine J 22:1375-1382. doi: 10.1007/s00586-012-2621-4 Ailon T, Torabi R, Fisher CG, Rhines LD, Clarke MJ, Bettegowda C, Boriani S, Yamada YJ, Kawahara N, Varga PP, Shin JH, Saghal A, Gokaslan ZL (2016) Management of Locally Recurrent Chordoma of the Mobile Spine and Sacrum: A Systematic Review. Spine (Phila Pa 1976) 41 Suppl 20:S193-S198. doi: 10.1097/BRS.0000000000001812 Lee MJ, Weaver DJ, El Dafrawy MH (2024) Extracranial Vertebral Artery Injuries. J Am Acad Orthop Surg 32:e899-e908. doi: 10.5435/JAAOS-D-24-00244 Eskander MS, Drew JM, Aubin ME, Marvin J, Franklin PD, Eck JC, Patel N, Boyle K, Connolly PJ (2010) Vertebral artery anatomy: a review of two hundred fifty magnetic resonance imaging scans. Spine (Phila Pa 1976) 35:2035-2040. doi: 10.1097/BRS.0b013e3181c9f3d4 Lebl DR, Bono CM, Velmahos G, Metkar U, Nguyen J, Harris MB (2013) Vertebral artery injury associated with blunt cervical spine trauma: a multivariate regression analysis. Spine (Phila Pa 1976) 38:1352-1361. doi: 10.1097/BRS.0b013e318294bacb Li D, Wu Z, Ren C, Hao SY, Wang L, Xiao XR, Tang J, Wang YG, Meng GL, Zhang LW, Zhang JT (2017) Foramen magnum meningiomas: surgical results and risks predicting poor outcomes based on a modified classification. J Neurosurg 126:661-676. doi: 10.3171/2016.2.JNS152873 Fatima N, Shin JH, Curry WT, Chang SD, Meola A (2021) Microsurgical resection of foramen magnum meningioma: multi-institutional retrospective case series and proposed surgical risk scoring system. J Neurooncol 153:331-342. doi: 10.1007/s11060-021-03773-z Westbroek EM, Pennington Z, Ehresman J, Ahmed AK, Gailloud P, Sciubba DM (2020) Vertebral Artery Sacrifice versus Skeletonization in the Setting of Cervical Spine Tumor Resection: Case Series. World Neurosurg 139:e601-e607. doi: 10.1016/j.wneu.2020.04.071 Peeters JB, Dessesard Olijnyk L, Janelle F, Shedid D, Bojanowski MW, Labidi M (2024) Surgical management of tumors of the cervical spine and craniovertebral junction involving the vertebral artery: A narrative review. Neurochirurgie 70:101550. doi: 10.1016/j.neuchi.2024.101550 Woertler K (2003) Benign bone tumors and tumor-like lesions: value of cross-sectional imaging. Eur Radiol 13:1820-1835. doi: 10.1007/s00330-003-1902-z Tang H, Ahlawat S, Fayad LM (2018) Multiparametric MR Imaging of Benign and Malignant Bone Lesions. Magn Reson Imaging Clin N Am 26:559-569. doi: 10.1016/j.mric.2018.06.010 Thakur NA, Daniels AH, Schiller J, Valdes MA, Czerwein JK, Schiller A, Esmende S, Terek RM (2012) Benign tumors of the spine. J Am Acad Orthop Surg 20:715-724. doi: 10.5435/JAAOS-20-11-715 Yip SS, Aerts HJ (2016) Applications and limitations of radiomics. Phys Med Biol 61:R150-166. doi: 10.1088/0031-9155/61/13/R150 Xu X, Zhang HL, Liu QP, Sun SW, Zhang J, Zhu FP, Yang G, Yan X, Zhang YD, Liu XS (2019) Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma. J Hepatol 70:1133-1144. doi: 10.1016/j.jhep.2019.02.023 Yang L, Chu W, Li M, Xu P, Wang M, Peng M, Wang K, Zhang L (2022) Radiomics in Gastric Cancer: First Clinical Investigation to Predict Lymph Vascular Invasion and Survival Outcome Using (18)F-FDG PET/CT Images. Front Oncol 12:836098. doi: 10.3389/fonc.2022.836098 Mansour TMM, El-Barody MM, Tammam H, Okasha A (2021) Role of diffusion-weighted MRI in differentiating between benign and malignant bone lesions: a prospective study. Clin Radiol 76:576-584. doi: 10.1016/j.crad.2021.02.030 Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, Lijmer JG, Moher D, Rennie D, de Vet HC, Kressel HY, Rifai N, Golub RM, Altman DG, Hooft L, Korevaar DA, Cohen JF, Group S (2015) STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 351:h5527. doi: 10.1136/bmj.h5527 Al-Hawary MM, Francis IR, Chari ST, Fishman EK, Hough DM, Lu DS, Macari M, Megibow AJ, Miller FH, Mortele KJ, Merchant NB, Minter RM, Tamm EP, Sahani DV, Simeone DM (2014) Pancreatic ductal adenocarcinoma radiology reporting template: consensus statement of the society of abdominal radiology and the american pancreatic association. Gastroenterology 146:291-304 e291. doi: 10.1053/j.gastro.2013.11.004 Zaky AM, Wolfgang CL, Weiss MJ, Javed AA, Fishman EK, Zaheer A (2017) Tumor-Vessel Relationships in Pancreatic Ductal Adenocarcinoma at Multidetector CT: Different Classification Systems and Their Influence on Treatment Planning. Radiographics 37:93-112. doi: 10.1148/rg.2017160054 Xiao ML, Wei Y, Zhang J, Jian JM, Song Y, Lin ZJ, Qian L, Zhang GF, Qiang JW (2022) MRI Texture Analysis for Preoperative Prediction of Lymph Node Metastasis in Patients with Nonsquamous Cell Cervical Carcinoma. Acad Radiol 29:1661-1671. doi: 10.1016/j.acra.2022.01.005 Zhou TH, Zhou XX, Ni J, Ma YQ, Xu FY, Fan B, Guan Y, Jiang XA, Lin XQ, Li J, Xia Y, Wang X, Wang Y, Huang WJ, Tu WT, Dong P, Li ZB, Liu SY, Fan L (2024) CT whole lung radiomic nomogram: a potential biomarker for lung function evaluation and identification of COPD. Mil Med Res 11:14. doi: 10.1186/s40779-024-00516-9 Cui Y, Shi X, Wang S, Qin Y, Wang B, Che X, Lei M (2022) Machine learning approaches for prediction of early death among lung cancer patients with bone metastases using routine clinical characteristics: An analysis of 19,887 patients. Front Public Health 10:1019168. doi: 10.3389/fpubh.2022.1019168 Shi X, Cui Y, Wang S, Pan Y, Wang B, Lei M (2024) Development and validation of a web-based artificial intelligence prediction model to assess massive intraoperative blood loss for metastatic spinal disease using machine learning techniques. Spine J 24:146-160. doi: 10.1016/j.spinee.2023.09.001 Origitano TC, al-Mefty O, Leonetti JP, DeMonte F, Reichman OH (1994) Vascular considerations and complications in cranial base surgery. Neurosurgery 35:351-362; discussion 362-353. doi: 10.1227/00006123-199409000-00001 Carpentier A, Blanquet A, George B (2001) Suboccipital and cervical chordomas: radical resection with vertebral artery control. Neurosurg Focus 10:E4. doi: 10.3171/foc.2001.10.3.5 Hoshino Y, Kurokawa T, Nakamura K, Seichi A, Mamada T, Saita K, Miyoshi K (1996) A report on the safety of unilateral vertebral artery ligation during cervical spine surgery. Spine (Phila Pa 1976) 21:1454-1457. doi: 10.1097/00007632-199606150-00011 Ball JR, Shelby T, Mertz K, Mills ES, Ton A, Alluri RK, Hah RJ (2024) The Incidence of Vertebral Artery Injury in Cervical Spine Surgery. World Neurosurg 181:e841-e847. doi: 10.1016/j.wneu.2023.10.144 Sharma A, Sharma A, Sharma R, Dewan Y (2017) Management of iatrogenic vertebral artery injury following cervical neurofibroma excision: a case report. Br J Neurosurg 31:478-480. doi: 10.1080/02688697.2016.1199780 Garcia Alzamora M, Rosahl SK, Lehmberg J, Klisch J (2005) Life-threatening bleeding from a vertebral artery pseudoaneurysm after anterior cervical spine approach: endovascular repair by a triple stent-in-stent method. Case report. Neuroradiology 47:282-286. doi: 10.1007/s00234-005-1343-2 Davnall F, Yip CS, Ljungqvist G, Selmi M, Ng F, Sanghera B, Ganeshan B, Miles KA, Cook GJ, Goh V (2012) Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice? Insights Imaging 3:573-589. doi: 10.1007/s13244-012-0196-6 Lubner MG, Smith AD, Sandrasegaran K, Sahani DV, Pickhardt PJ (2017) CT Texture Analysis: Definitions, Applications, Biologic Correlates, and Challenges. Radiographics 37:1483-1503. doi: 10.1148/rg.2017170056 Gillies RJ, Kinahan PE, Hricak H (2016) Radiomics: Images Are More than Pictures, They Are Data. Radiology 278:563-577. doi: 10.1148/radiol.2015151169 Varghese BA, Cen SY, Hwang DH, Duddalwar VA (2019) Texture Analysis of Imaging: What Radiologists Need to Know. AJR Am J Roentgenol 212:520-528. doi: 10.2214/AJR.18.20624 Li Z, Yu L, Wang X, Yu H, Gao Y, Ren Y, Wang G, Zhou X (2018) Diagnostic Performance of Mammographic Texture Analysis in the Differential Diagnosis of Benign and Malignant Breast Tumors. Clin Breast Cancer 18:e621-e627. doi: 10.1016/j.clbc.2017.11.004 Balagurunathan Y, Gu Y, Wang H, Kumar V, Grove O, Hawkins S, Kim J, Goldgof DB, Hall LO, Gatenby RA, Gillies RJ (2014) Reproducibility and Prognosis of Quantitative Features Extracted from CT Images. Transl Oncol 7:72-87. doi: 10.1593/tlo.13844 Zhang R, Zhang Q, Ji A, Lv P, Acosta-Cabronero J, Fu C, Ding J, Guo D, Teng Z, Lin J (2023) Prediction of new cerebral ischemic lesion after carotid artery stenting: a high-resolution vessel wall MRI-based radiomics analysis. Eur Radiol 33:4115-4126. doi: 10.1007/s00330-022-09302-4 Versteijne E, Suker M, Groothuis K, Akkermans-Vogelaar JM, Besselink MG, Bonsing BA, Buijsen J, Busch OR, Creemers GM, van Dam RM, Eskens F, Festen S, de Groot JWB, Groot Koerkamp B, de Hingh IH, Homs MYV, van Hooft JE, Kerver ED, Luelmo SAC, Neelis KJ, Nuyttens J, Paardekooper G, Patijn GA, van der Sangen MJC, de Vos-Geelen J, Wilmink JW, Zwinderman AH, Punt CJ, van Eijck CH, van Tienhoven G, Dutch Pancreatic Cancer G (2020) Preoperative Chemoradiotherapy Versus Immediate Surgery for Resectable and Borderline Resectable Pancreatic Cancer: Results of the Dutch Randomized Phase III PREOPANC Trial. J Clin Oncol 38:1763-1773. doi: 10.1200/JCO.19.02274 Peng CW, Chou BT, Bendo JA, Spivak JM (2009) Vertebral artery injury in cervical spine surgery: anatomical considerations, management, and preventive measures. Spine J 9:70-76. doi: 10.1016/j.spinee.2008.03.006 Additional Declarations No competing interests reported. Supplementary Files TableS1.docx Cite Share Download PDF Status: Published Journal Publication published 20 Jan, 2026 Read the published version in European Spine Journal → Version 1 posted Editorial decision: Revision requested 19 Nov, 2025 Reviews received at journal 18 Nov, 2025 Reviews received at journal 14 Nov, 2025 Reviewers agreed at journal 07 Nov, 2025 Reviewers agreed at journal 06 Nov, 2025 Reviewers invited by journal 07 Aug, 2025 Editor assigned by journal 01 Aug, 2025 Submission checks completed at journal 01 Aug, 2025 First submitted to journal 26 Jul, 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-7218476","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":497224190,"identity":"58832656-0b6b-43fb-9d68-2e599ec5d353","order_by":0,"name":"Xiangzhi Ni","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiangzhi","middleName":"","lastName":"Ni","suffix":""},{"id":497224191,"identity":"67faade7-0400-4069-a0e1-e054603eb655","order_by":1,"name":"Jiayang Yan","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jiayang","middleName":"","lastName":"Yan","suffix":""},{"id":497224192,"identity":"ee39f22e-2906-46dd-bf5f-31b7abfbd65e","order_by":2,"name":"Jiayi Zhang","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jiayi","middleName":"","lastName":"Zhang","suffix":""},{"id":497224193,"identity":"39a6c809-78b2-4579-8d58-03365cc25147","order_by":3,"name":"Fukai Li","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Fukai","middleName":"","lastName":"Li","suffix":""},{"id":497224194,"identity":"5c0b228e-a081-4945-bdca-68bca2e44c3d","order_by":4,"name":"Guangwen Duan","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Guangwen","middleName":"","lastName":"Duan","suffix":""},{"id":497224195,"identity":"4cf68b50-d42f-4abc-80d7-cb791663205b","order_by":5,"name":"Shuming Hou","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Shuming","middleName":"","lastName":"Hou","suffix":""},{"id":497224196,"identity":"786a5c76-42ca-4683-a3eb-8dcf3b55006a","order_by":6,"name":"Lingyun Shen","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lingyun","middleName":"","lastName":"Shen","suffix":""},{"id":497224197,"identity":"fcd0e2d3-c91a-40da-bdbd-737f217fba77","order_by":7,"name":"Hongbiao Sun","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hongbiao","middleName":"","lastName":"Sun","suffix":""},{"id":497224198,"identity":"dd14be1f-e8b6-48c1-ae5f-4772657ad9a7","order_by":8,"name":"Xiang Wang","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiang","middleName":"","lastName":"Wang","suffix":""},{"id":497224199,"identity":"02a92d03-5299-4e73-9571-ece2638cafbc","order_by":9,"name":"Minglei Yang","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Minglei","middleName":"","lastName":"Yang","suffix":""},{"id":497224200,"identity":"1cf59cd6-6b5e-42c9-9cb7-8dd12e6e8f65","order_by":10,"name":"Tielong Liu","email":"","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":false,"prefix":"","firstName":"Tielong","middleName":"","lastName":"Liu","suffix":""},{"id":497224201,"identity":"0f349693-fcb1-49f8-90ed-78c0f2cc9a31","order_by":11,"name":"Shiyuan Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA2ElEQVRIiWNgGAWjYHACgwMMBVDmBwMbOyK1GEBYjDMK0pKJ0sIA08LM8+EQYwMh9ebthzceALlnwwHmx59tDA4wM7AfProBnxaZM2kFB2cYpCVvOMBmYJxjcIePgSct7QY+LRIMOQaHeQwOJxsc4GFIzjF4xswgwWOGXwv/G4SWwxYGhxkbCGqRgNhiB9TC2MxAnJZnYL8kSB5gM2bsAXqKjaBf+JM3f/hQYWPPBwyxDz/+2Njxsx8+hlcLDCQuuP8AwmIjRjkI2Ms3EKt0FIyCUTAKRhwAAGO9SjdYo94QAAAAAElFTkSuQmCC","orcid":"","institution":"Changzheng Hospital of the Navy Medical University","correspondingAuthor":true,"prefix":"","firstName":"Shiyuan","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2025-07-26 04:53:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7218476/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7218476/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00586-025-09706-y","type":"published","date":"2026-01-20T15:57:22+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":88950791,"identity":"291b72a9-f605-42eb-b9c9-e7936af700bd","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2133009,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the patient selection procedure.\u003c/p\u003e","description":"","filename":"Fig1.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/9d5dab59df102205f3c7ee86.jpg"},{"id":88950786,"identity":"a39c4e4a-b56a-4499-8d07-3a648533abe4","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":5462857,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow of radiomics nomogram for intraoperative vertebral artery injury risk in patients with primary cervical spine tumors\u003c/p\u003e","description":"","filename":"Fig2.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/439122e6888ced551762393b.jpg"},{"id":88950788,"identity":"3e53aa67-eeaf-4197-9f35-ffb3a3a44c97","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":309527,"visible":true,"origin":"","legend":"\u003cp\u003eLASSO regression model-based radiomic feature screening. a. The LASSO coefficient profiles of the 299 radiomics features. b. Tuning parameter (λ) selection for the LASSO model according to minimal criteria using tenfold cross-validation. c. Thirteen features with nonzero coefficients are displayed, along with the contribution of features to our signature.\u003c/p\u003e","description":"","filename":"Fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/30234481b20420ec2a005273.jpg"},{"id":88950783,"identity":"eb953168-8c5f-4c7d-9fcb-8dce28b1cacf","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2519279,"visible":true,"origin":"","legend":"\u003cp\u003eHeatmap of the scoring system for comprehensively evaluating the prediction performance of all models. The scoring system incorporated 10 evaluation metrics, and each metric was rated on a scale of 1 to 11, where higher scores denoted superior predictive performance.\u003c/p\u003e","description":"","filename":"Fig4.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/71cd65faa754065b70d4f6ea.jpg"},{"id":88950805,"identity":"a01f15a8-4300-4fec-b220-e83a296c51fe","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1266719,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of the radiomic model, clinical model, and combined model in predicting vertebral artery injury in primary cervical tumor surgery. a. Training cohort, b. Test cohort.\u003c/p\u003e","description":"","filename":"Fig5.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/aad5cd87298d94e233254bdf.jpg"},{"id":88950796,"identity":"3556d907-c3bf-4dd9-97c5-66d00fb46b84","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":3561395,"visible":true,"origin":"","legend":"\u003cp\u003eDevelopment and performance of combined model nomogram. a. combined model nomogram developed to predict vertebral artery injury in primary cervical tumor surgery. b. Calibration curve between the predicted and actual incidences of vertebral artery injury in primary cervical tumor surgery. c. Decision curve analysis compares three models in predicting the risk of vertebral artery injury in primary cervical tumor surgery.\u003c/p\u003e","description":"","filename":"Fig6.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/cc15d451d687320963b8e97e.jpg"},{"id":88950790,"identity":"71b0de65-6d4f-457b-9e06-a0af14662ff8","added_by":"auto","created_at":"2025-08-13 05:50:06","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":3318827,"visible":true,"origin":"","legend":"\u003cp\u003eThe dynamic nomogram was applied to one example. 23-year-old male patient diagnosed with osteoblastoma. a.The patient's preoperative MRI images.b. The left vertebral artery was ligated during the operation. c.The dynamic nomogram was utilized to estimate the risk of intraoperative vertebral artery injury. The prediction was based on the following parameters: Stenosis = 0.4, Length = 29.04, and Radscore = 0.58. The calculated total score was approximately 160 points, corresponding to a predicted probability of 99% for vertebral artery injury during the procedure.\u003c/p\u003e","description":"","filename":"Fig7.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/c78f76ecb1ce80722e9f2f23.jpg"},{"id":101151816,"identity":"4c7966df-856e-4f8d-8bb7-29399812e76f","added_by":"auto","created_at":"2026-01-26 16:06:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":19680422,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/4b233b91-1d3a-4d1f-8d86-1f007bcf9863.pdf"},{"id":88952025,"identity":"67d5534f-e785-41df-ae00-cb0d15e537f0","added_by":"auto","created_at":"2025-08-13 05:58:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":14816,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7218476/v1/b1f76bdd854b7d9c8934cd62.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A clinical-radiomics nomogram based on multisequence MRI for predicting intraoperative vertebral artery injury in patients with primary cervical spine tumor: a diagnostic study","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eCervical spine tumors, which are classified as primary or metastatic, often cause severe spinal cord compression and spinal instability, leading to incurable pain, loss of mobility, and neurological deficits [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Although metastatic cervical spine tumors are more common than primary cervical spine tumors, the former are often treated with palliative treatment due to limited survival time, while primary cervical spine tumors aim at cure and the treatment process is more complicated [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Chordomas, chondrosarcomas, giant cell tumors of bone and neurogenic tumors are the most common tumors in the cervical spine [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Due to the drug resistance of the tumor and the maximum tolerated dose of radiation to the spinal cord [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], surgery is often the primary treatment for primary tumors of the cervical spine [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eA variety of studies have pointed out that En bloc resection of the vertebrae with adequate margins in an intact patient decreases the risk of local recurrence and tumor-related mortality [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Due to the complexity of cervical anatomy involving the vertebral artery (VA) and cervical nerve roots, it is challenging to achieve en bloc resection. The VA, arising as the first branches from the subclavian arteries, is partitioned into four distinct segments from V1 to V4 [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. The bilateral VA enters the intracranial aspect, where they merge to form the basilar artery, which participates in the Willis circulation and supplies an important part of the posterior cerebral circulation [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Therefore, great attention should be paid to ligation of the unilateral VA. Knowing that sacrificing the bilateral VA may elicit a high risk of stroke or even death, surgery becomes a great challenge when the primary cervical tumor involves the VA [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Tumor invasion of the VA increases the probability of failure of VA dissection, which often leads to intraoperative vascular injury and tumor residue [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. The dominant or bilateral VA injury needs to be remedied by vascular repair or vascular grafting. Therefore, preoperative prediction of VA injury is helpful to make individualized treatment plans and improve the safety of surgery.\u003c/p\u003e\u003cp\u003eAt present, the routine imaging preoperative examination for cervical spine tumors includes X-ray, CT and MRI etc. [\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. It is often difficult to predict the risk of VA injury based on visual features from preoperative examination. Traditional imaging features are usually visually assessed and qualitatively described by radiologists or nuclear medicine physicians, which creates the possibility of large intra-observer and inter-observer variability in these visual assessments [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. In recent years, radiomics has developed rapidly and showed outstanding potential in many computer-aided diagnostic applications, also in predicting tumor invasion of blood vessels. Xu et al. successfully developed a new model for predicting microvascular invasion of hepatocellular carcinoma based on CT radiomics analysis, and the model showed good performance [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Yang et al. used PET-CT-based radiomics features to predict lymph vascular invasion in gastric cancer patients and provided effective predictors of patient survival outcomes [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Because of its high resolution, tissue contrast, and multiplanar capacity, MRI has innate advantages in the evaluation of cervical spine tumors [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Therefore, it is more meaningful to conduct radiomics research based on MRI for patients with primary cervical tumors.\u003c/p\u003e\u003cp\u003eTo the best of our knowledge from a review of the literature, there is no radiomics study on prediction of VA injury in surgery of cervical tumor. Therefore, the aim of this study is to develop a combined model based on MRI and clinical information to predict the probability of VA injury during surgery for the sake of increasing the safety of surgery in cervical tumors and providing a useful guide for selecting the strategy of surgical intervention.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Participants and image data sets\u003c/h2\u003e\u003cp\u003e This research project was approved by the Ethics Committee of our hospital, with the written informed consent waived due to the retrospective nature of the study. The work has been reported in line with the STARD (Standards for the Reporting of Diagnostic accuracy studies) criteria[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Patients who were diagnosed with cervical spine tumors and received gross total resection in our institution between January 2013 and June 2024 were reviewed retrospectively. The inclusion criteria were patients (1) who were diagnosed with primary cervical spine tumors; (2) who underwent preoperative MRI examination and the quality was qualified by the radiologist; (3) who achieved gross total resection; and (4) whose clinical data were complete. The exclusion criteria were patients (1) with metastatic tumors; (2) with low-quality MRI or incomplete clinical data; and (3) whose internal fixation in the cervical spine was interfered with. Finally, 579 patients were enrolled, and 167 patients were included in this study, including 56 patients with VA injury. They were randomly divided into a training set (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;116) and a test set (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;51) at a 7:3 ratio. The workflow for patient inclusion and exclusion is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAll patients received MRI examinations using one of five MRI scanners, including 1.5T Siemens, 1.5T General Electric, 3.0T United Imaging uMR770,3.0T Philips and 3.0T GE premier. The imaging parameters from the five scanners are detailed in \u003cb\u003eTable S1\u003c/b\u003e. And all scans were performed consistently. Imaging sequences included in this study were T2WI and CET1 sequences. CET1 was performed after intravenous injection of the contrast medium (GD-DTPA Injection, Consun Pharma, Guangzhou, China) according to the recommended dose of 0.2 ml/kg. All sagittal, axial and coronal views were acquired and saved in DICOM format.\u003c/p\u003e\u003cp\u003eIn addition, the basic clinical information of the patients, including gender, age, pathology of the tumor, history of smoking, history of diabetes mellitus, history of alcohol consumption, body mass index (BMI), level of education and tumor invasion of VA segments, was collected through the electronic medical record system and PACS system. Two imaging experts with more than five years of imaging experience independently evaluated and measured the following information: 1) the angle of the tumor surrounding the VA measured in the cross-section of the enhanced image; 2) the length of the tumor invading the VA measured in the sagittal plane of the enhanced image; and 3) the degree of VA stenosis measured in the cross-section of enhanced images (the narrowest vessel circumference/the normal vessel circumference) [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Treatment regimen\u003c/h2\u003e\u003cp\u003eThe surgical indications are as follows: 1) malignant tumors were diagnosed by imaging or pathology, 2) patients with neurological symptoms, unbearable pain, and spinal instability. Gross total resection was used in all patients. All surgical procedures were performed by the same surgical team. For patients with large tumor invasion of the VA on MRI, preoperative balloon occlusion test was performed. Intraoperative tumor invasion of the VA was managed according to the following principles: 1) Attempts to isolate the VA were prioritized; 2) When VA injury occurred, priority is given to repair the VA; 3) when the VA could not be repaired and the contralateral VA had good circulation, ligation of the injured VA was considered; and 4) if the contralateral VA circulation was poor, vascular grafting was considered. The whole operation process comprised tumor excision, decompression of the spinal cord, reconstruction, and stabilization of the spine.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 MRI segmentation and radiomic feature extraction\u003c/h2\u003e\u003cp\u003eThe study flow diagram is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. 3D image segmentation was manually performed by a senior resident in radiology with experience in musculoskeletal and oncologic imaging using the open-source software ITK-SNAP (version 4.2). The reader drew a region of interest (ROI) along tumor borders on each slice using an T2WI and CET1 sequence.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eUsing the PyRadiomics platform, radiomic features of the volume of interest were extracted. The extracted features included first-order, gray level co-occurrence matrix, gray level run length matrix, gray level size zone matrix, gray level dependence matrix, and shape features. The Z score method was used to normalize the features and eliminate the difference in numerical scale.\u003c/p\u003e\u003cp\u003eThe optimal radiomics features were identified through a systematic three-step process. Initially, we performed a Student's t-test for features that met the criteria of normal distribution, while the Mann-Whitney U test was applied to those that did not follow this distribution. Radiomic features with a p-value of less than 0.05 were retained for further analysis. In the subsequent step, features exhibiting a Pearson correlation coefficient exceeding 0.9 were randomly removed to mitigate redundancy [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Lastly, the least absolute shrinkage and selection operator (LASSO) regression algorithm, along with the adjustment of penalty parameters, was employed to implement tenfold cross-validation. The radiomic model was developed by computing a radiomics score for each patient, derived from a linear combination of the selected features along with their corresponding weight coefficients.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Development, validation, and performance evaluation of the combined model\u003c/h2\u003e\u003cp\u003eFirstly, univariate and multivariate Logistic regression analyses of clinical characteristics were performed to define independent risk factors as meaningful risk factors, and a clinical model was constructed based on these factors. Second, patients in the training cohort were used to establish radiomics models, including logistic regression, NaiveBayes, Support Vector Machine, K-Nearest Neighbor, RandomForest, ExtraTrees, XGBoosting machine, Light Gradient Boosting Machine, GradientBoosting, AdaBoost, and Multilayer Perceptron (MLP). Patients in the test cohort were used to validate the model. Additionally, based on previous studies [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e], this study used a scoring system to comprehensively evaluate the prediction performance of the models. Finally, a combined model was constructed based on the clinical model and the radiomics model, based on which a nomogram was generated to visualize the combined model, graphically evaluate variable importance, and calculate prediction accuracy. To validate the performance of the three sets of models, we used ROC curves and AUC to evaluate the performance of the prediction of the combined model, radiomics model, and clinical model. The calibration curves (Hosmer\u0026ndash;Lemeshow test) were performed to evaluate the calibration of the nomogram. Decision curve analysis (DCA) was applied to evaluate the clinical practicability of the nomogram.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Statistical analysis\u003c/h2\u003e\u003cp\u003eStatistical analysis was performed by using the SPSS (SPSS 22.0; IBM Corp) and R Programming Language (version 3.0.1; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.Rproject.org\u003c/span\u003e\u003cspan address=\"http://www.Rproject.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Numeric data were presented as the mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD). Differences in the distribution of categorical variables were compared using the chi-square test, and comparison of continuous variables was performed using Student\u0026rsquo;s t test and Mann-Whitney U test. Independent predictors were identified from the clinical variables by multivariate logistic regression. The R Programming Language was applied to conduct the LASSO regression analysis, ROC analysis, calibration curve analysis (with the Hosmer\u0026ndash;Lemeshow test). And the \u0026ldquo;rmda\u0026rdquo; package was utilized for DCA. The AUC values between different models were compared by using the DeLong method. \u0026ldquo;\u003cem\u003ep\u003c/em\u003e\u0026rdquo; value less than 0.05 was set as the statistically significant threshold.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Clinical characteristics\u003c/h2\u003e\u003cp\u003eThe baseline clinical characteristics of 167 included patients with primary cervical spine tumors are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. There was no significant difference in their distribution of the degree of VA stenosis, the length of VA invasion, injury of VA, age, gender, the segment of VA invasion, pathological diagnosis, and the angle of the tumor surrounding the VA between the overall, training and test cohorts. Univariate analysis showed that preoperative clinical information, including the degree of VA stenosis, the length of VA invasion, age, the segment of VA invasion, pathological diagnosis, was significantly associated with VA injury (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The multivariate analysis showed that the degree of VA stenosis and the length of VA invasion were independent risk factors for VA injury (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe clinical characteristics of patients with vertebral artery injury and intact vertebral arteries in the training and test cohorts\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"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\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eFeatures\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eOverall (n\u0026thinsp;=\u0026thinsp;167)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e\u003cp\u003eGroup\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ep value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTraining cohort (n\u0026thinsp;=\u0026thinsp;116)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTest cohort\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;51)\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)\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.891\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e92 (55.09%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e63 (54.31%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e29 (56.86%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026ge;\u0026thinsp;47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e75 (44.91%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e53 (45.69%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e22 (43.14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.755\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e119 (71.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e84 (72.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e35 (68.63%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e48 (28.74%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e32 (27.59%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e16 (31.37%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmoking status\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.393\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNonsmoker\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e122 (73.05%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e87 (75.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e35 (68.63%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCurrent smoker\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e45 (26.95%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e29 (25.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e16 (31.37%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.111\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\u003e24 (14.37%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e20 (17.24%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4 (7.84%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e143 (85.63%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e96 (82.76%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e47 (92.16%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrinking alcohol\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.699\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNondrinker\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e125 (74.85%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e88 (75.86%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e37 (72.55%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCurrent drinker\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e42 (25.15%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e28 (24.14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e14 (27.45%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.848\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003elow weight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e9 (5.39%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7 (6.03%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2 (3.92%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNormal weight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e151 (90.42%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e104 (89.66%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e47 (92.16%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOverweight\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e7 (4.19%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5 (4.31%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2 (3.92%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDegree of education\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.480\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHigh school and above\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e98 (58.68%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e66 (56.90%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e32 (62.75%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBelow high school\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e69 (41.32%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e50 (43.10%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e19 (37.25%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVA injury\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.393\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\u003e56 (33.53%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e36 (31.03%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e20 (39.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e111 (66.47%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e80 (68.97%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e31 (60.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePathological diagnosis\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.367\u003c/p\u003e\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\u003e89 (53.29%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e65 (56.03%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e24 (47.06%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBenign\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e78 (46.71%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e51 (43.97%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e27 (52.94%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSegment of VA\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.643\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eV1\u0026thinsp;+\u0026thinsp;V2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e55 (32.93%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40 (34.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e15 (29.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eV3\u0026thinsp;+\u0026thinsp;V4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e112 (67.07%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e76 (65.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e36 (70.59%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAngle (\u0026deg;)\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=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.765\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;180\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e66 (39.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e45 (38.79%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e21 (41.18%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026le;\u0026thinsp;180\u0026thinsp;\u0026lt;\u0026thinsp;270\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e52 (31.14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e35 (30.17%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e17 (33.33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026ge;\u0026thinsp;270\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e49 (29.34%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e36 (31.03%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e13 (25.49%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDegree of VA stenosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.68 [0.50, 0.83]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.67 [0.47, 0.82]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.71 [0.55, 0.84]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.340\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLength of VA invasion (mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e23.22 [16.83, 33.80]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e24.35 [18.23, 35.14]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e21.39 [15.81, 29.05]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.080\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\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\u003eUnivariate and multivariate logistic regression analysis were used to analyze the independent risk factors of intraoperative vertebral artery injury in patients with primary cervical tumor\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"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\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eVariables\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003eUnivariate analysis\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003eMultivariate analysis\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOR (95%CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ep value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eOR (95%CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003ep 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\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.395 (0.239\u0026ndash;0.652)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.047 (0.440\u0026ndash;2.492)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.930\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.600 (0.329\u0026ndash;1.094)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.162\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmoking status\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.803 (0.317\u0026ndash;2.035)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.643\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.943 (0.330\u0026ndash;2.692)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.912\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrinking alcohol\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.070 (0.429\u0026ndash;2.668)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.884\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.863 (0.254\u0026ndash;2.930)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.813\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDegree of education\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.924 (0.418\u0026ndash;2.042)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.845\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePathological diagnosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.413 (0.264\u0026ndash;0.647)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.877 (0.377\u0026ndash;2.040)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.798\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSegment of VA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.429 (0.243\u0026ndash;0.756)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.014\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.733 (0.301\u0026ndash;1.782)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.565\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAngle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.011 (0.791\u0026ndash;1.293)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.940\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\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDegree of VA stenosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.154 (0.087\u0026ndash;0.272)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.003 (0.001\u0026ndash;0.020)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.05\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLength of VA invasion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.987 (0.976\u0026ndash;0.997)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.031\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.095 (1.058\u0026ndash;1.133)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.05\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Feature selection and radiomics signature building\u003c/h2\u003e\u003cp\u003eA total of 2394 radiomic features were extracted. After reproducibility assessment and Pearson correlation analysis, 72 radiomic features were reserved, respectively. Finally, 13 radiomic features with non-zero coefficients were selected by LASSO regression (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). By linearly combining those features after weighing by their corresponding coefficients, we constructed the radiomic signature.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Performance evaluation of the radiomics model\u003c/h2\u003e\u003cp\u003eAmong the eleven models, the MLP model performed the best in terms of AUC (0.773, 95% CI: 0.643\u0026ndash;0.902), and the XGBoosting Machine model only had the fourth AUC (0.760, 95% CI: 584\u0026ndash;0.871), whereas the LightGBM model had a lower AUC of 0.672 (95% CI: 0.520\u0026ndash;0.823) (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Using the scoring system, the MLP model achieved the highest total score of 82 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), indicating the best predictive performance. These results support MLP as the best model to evaluate whether VA injury occurred during surgery for cervical primary tumors. Therefore, we used the MLP model as the radiomics model in our study\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\u003ePrediction performance of all models based on the test cohort\u0026rsquo;s validation\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"11\"\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\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModle/Metrics\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC(95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003eF1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c11\"\u003e\u003cp\u003eTotal score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.731 (0.589\u0026ndash;0.872)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.710\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.571\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.733\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.571\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.585\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e46\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNaiveBayes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.725\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.726 (0.583\u0026ndash;0.869)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.450\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.903\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.718\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.450\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.562\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e55\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.725\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.674 (0.508\u0026ndash;0.840)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.839\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.687\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.743\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.687\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.611\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e60\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.627\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.719 (0.582\u0026ndash;0.856)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.613\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.520\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.731\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.520\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.578\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e37\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandomForest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.686\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.723 (0.571\u0026ndash;0.874)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eExtraTrees\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.686\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.760 (0.627\u0026ndash;0.894)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.710\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.591\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.759\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.591\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.619\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e69\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.706\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.727 (0.584\u0026ndash;0.871)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.677\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.808\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.600\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e82\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.672 (0.520\u0026ndash;0.823)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.677\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.565\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.565\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.605\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e49\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGradientBoosting\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.608\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.686 (0.536\u0026ndash;0.837)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.700\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.548\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.739\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.700\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.583\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e37\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAdaBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.588\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.678 (0.526\u0026ndash;0.831)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.484\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.484\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.484\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.588\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e42\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMLP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.706\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.773 (0.643\u0026ndash;0.902)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.619\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.767\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.619\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.634\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e\u003cp\u003e82\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\u003cp\u003eTo further verify the performance of the radiomics model, MLP was applied to the CET1 and T2WI sequences of MRI to establish the radiomics model. The results showed that the AUC of the radiomics model (0.773, 95% CI: 0.643\u0026ndash;0.902) performed better than those of the CET1 model (0.673, 95% CI: 0.517\u0026ndash;0.828) and T2WI model (0.773, 95% CI: 0.598\u0026ndash;0.877) in the test cohort.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Construction and comparison of the clinical model, radiomic model,\u003c/h2\u003e\u003cp\u003emodel, and combined model\u003c/p\u003e\u003cp\u003eThe clinical prediction model was constructed based on independent predictors of multivariate analysis, including length of the tumor invading the VA and degree of VA stenosis. A combined model was constructed by combining radiomics model and the independent risk factors of the above clinical features.\u003c/p\u003e\u003cp\u003eThe performance of the radiomic, clinical and combined models is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. It was found that the performance of the combined model was better than that of the other two models in both the training set and test set. The AUC of the combined model (0.952, 95%CI, 0.919\u0026ndash;0.986) was higher than that of the radiomic model (0.861, 95%CI 0.791\u0026ndash;0.932, p\u0026thinsp;=\u0026thinsp;0.005) and the clinical model (0.777, 95%CI 0.685\u0026ndash;0.869, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) in the training set. The AUC of the combined model (0.932, 95%CI, 0.868\u0026ndash;0.996) was higher than that of the radiomic model (0.773, 95%CI 0.643\u0026ndash;0.902, p\u0026thinsp;=\u0026thinsp;0.006) and the clinical model (0.740, 95%CI 0.596\u0026ndash;0.884, p\u0026thinsp;=\u0026thinsp;0.002) in the test set.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparison of diagnostic performance of the radiomics model, clinical model, and combined model in the training and test cohorts\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"7\"\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\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRadiomics model\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\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.861(0.791\u0026ndash;0.932)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.836\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.611\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.937\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.815\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.843\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTest cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.773(0.643\u0026ndash;0.902)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.706\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.619\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.767\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClinical model\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\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.777(0.685\u0026ndash;0.869)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.716\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.778\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.687\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.528\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.873\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTest cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.740(0.596\u0026ndash;0.884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.581\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.552\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCombined model\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\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTraining cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.952(0.919\u0026ndash;0.986)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.853\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.889\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.837\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.711\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.944\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTest cohort\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.932(0.868\u0026ndash;0.996)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.824\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.950\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.704\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.958\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=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.5 Development and performance of the nomogram\u003c/h2\u003e\u003cp\u003eFor easier use by clinicians, we transformed the combined model into a nomogram (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Nomogram scores were given based on the weight of independent predictors, and the scale length of the nomogram variable was found to be positively correlated with its impact on the prediction of VA injury in patients with primary cervical spine tumors. According to the combined model, radiomics features had the greatest impact, followed by the length of the tumor invading the VA and degree of VA stenosis. The model calibration curve demonstrated good agreement between prediction and observation outcomes in the two cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). The decision curves of the radiomics nomogram in the Training and Test sets are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. According to DCA, when the probability threshold of a patient or physician's clinical decision was greater than 0.1, the combined model was more favorable than the other models in predicting the risk of intraoperative VA injury in the training and test set.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAn example of the nomogram in use is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e. We can read the top score scale upward from the predictors to determine the points associated with the length of the tumor invading the VA and degree of VA stenosis, and the Radscore. We can derive a total score based on the scores of each factor to determine the probability of VA injury during surgery for primary cervical tumors.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eIn this study, we developed a novel model for predicting interoperative VA injury for cervical primary tumors by analyzing the radiomics features of MRI and clinical features. To improve the accuracy of the model and facilitate the use by clinicians, a new combined model was constructed by combining radiomics features with clinical features. The results showed that the prediction performance of the multimodal model was significantly better than that of the radiomics model or the clinical feature model alone. Based on the above data, we finally constructed a nomogram to facilitate the use by clinicians. The result showed that the new combined model greatly reduced intraoperative uncertainty in patients with primary cervical tumors and VA invasion and provided a great deal of help for preoperative planning and perioperative nursing.\u003c/p\u003e\u003cp\u003eThe surgical treatment principles include gross total resection of the tumor and protection of the VA and important nerve roots [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. However, the realization of this goal often faces formidable challenges. In addition to the anatomical complexity of the cervical spine, the larger reason is that surgical resection of cervical tumors can be a great challenge when the tumor surrounds the major blood vessels, especially the VA [\u003cspan additionalcitationids=\"CR32\" citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Studies have shown that the probability of VA injury during cervical spine surgery ranges from 0.03\u0026ndash;1.2% [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], and this figure is increasing during cervical spine tumor surgery. Encapsulation of the tumor by major vessels often makes complete resection difficult and may cause serious complications such as fistulas, pseudoaneurysm, late-onset hemorrhage, thrombosis, embolism, cerebral ischemia, or even death [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. However, incomplete resection of the primary tumor is often disastrous for the poor prognosis of patients due to tumor recurrence. When VA injury occurs, vascular repair, grafting or ligation is required. Therefore, it is very valuable to accurately predict the major risks that may occur during the operation. Our prediction model can predict in advance whether VA injury may occur or whether ligation is required during the operation of the primary cervical tumor, which could help surgeons to plan the operation strategy before operation and achieve total resection of the tumor under the premise of ensuring the safety of patients.\u003c/p\u003e\u003cp\u003eTumors are heterogeneous at gross and cellular levels as well as genetic and phenotypic levels, with spatial heterogeneity in cell density, angiogenesis, and necrosis [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. This heterogeneity is often closely related to the treatment and prognosis of the patient. On the other hand, tumor heterogeneity also shows that different tumor tissues have different degrees of vascular invasion. It is usually difficult to assess tumor heterogeneity simply by using the conventional imaging tools [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e](28898189). Compared with traditional imaging tools, radiomics can often extract the hidden information in the image [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. Some radiomic metrics, particularly texture analysis metrics, have been reported to assess intratumoral heterogeneity in studies assessing the diagnosis, prognosis, and treatment response of cancer [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. In our study, we constructed a radiomics signature consisting of four radiomics features, including first-order features, morphological features, texture features and wavelet features. None of these informative data can be obtained with the naked eye. Morphologic features describe the size, volume, and shape of the volume of interest, while first-order features mainly reflect the internal texture of the lesions. Textural features, including the gray level co-occurrence matrix and gray level dependence matrix, describe the spatial relationship between each pixel and its neighbors. Wavelet features mainly reflect the time frequency domain within the lesion [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Vascular imaging analysis has been extensively reported in the literature. A new model based on the study of texture and other features of CT images by Xu et al. also showed better performance than traditional features [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. For new cerebral ischemic lesions after carotid artery stenting, Zhang et al. [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e] constructed a new model based on radiomics features of high-resolution MRI of the vascular wall with better prediction performance than traditional features. The above studies showed that the prediction ability of models based on radiomics features could be significantly improved. In our study, we constructed a radiomics signature describing vascular and cervical spine tumors and combined it with clinical features to form a novel multimodal model. The result showed that the predictive performance of the new model was significantly better than that of the clinical model and the radiomics model.\u003c/p\u003e\u003cp\u003eMany researchers have studied the relationship between tumors and surrounding large blood vessels, such as the relationship between pancreatic cancer and mesenteric arteries and veins. According to the American National Comprehensive Cancer Network (NCCN) consensus reporting guidelines, a tumor is resectable when the circumference of its contact with blood vessels is less than or equal to 180\u0026deg;, and unresectable when the circumference exceeds 180\u0026deg;. The degree of vascular stenosis in patients is often closely related to tumor compression [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. To quantify the degree of vascular compression, Sugae et al. [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] developed the corresponding evaluation criteria, including the diameter of the stenosis/the diameter of the normal part of the vessel, and the length of vascular stenosis. It was found in our study that the degree of stenosis of the tumor-wrapped VA and the length of tumor invasion of the VA were independent risk factors for prognosis. This is generally consistent with literature reports. In general, the more tightly a tumor binds to the surrounding blood vessels, the deeper its invasion is, and the surrounding blood vessels are often narrowed by the tumor. The wrapping angle of the tumor and the VA were shown as significant factors in the univariate analysis of our study but shown no significance in multivariate analysis. This may be because the wrapping angle is not directly related to the degree of invasion, and on the other hand tumors with a large wrapping angle are often treated by combined anterior and posterior approaches.\u003c/p\u003e\u003cp\u003eOur model may fill up a lacuna in the treatment of cervical primary tumors invading the VA and can be used to predict the probability of VA injury during surgery. Our model may fill up a lacuna in the treatment of cervical primary tumors invading the VA and can be used to predict the probability of VA injury during surgery [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. Surgeons provide preoperative and postoperative safety safeguards for high-risk patients in advance. Before surgery, a VA balloon occlusion test can be performed to determine whether the contralateral VA is patent. Intraoperative surgical management of VA injury includes hemostatic tamponade, microvascular repair of the injured artery, vascular grafting and ligation of the VA. Therefore, acquisition of coordinative assistance from relevant departments in advance can not only achieve a rapid response to VA injury but also avoid the waste of medical resources.\u003c/p\u003e\u003cp\u003eThis study has some limitations. Firstly, selection bias could not be avoided because of the retrospective nature of the study. In addition, the sample size of patients with cervical primary tumors is not large enough. Finally, this is a single-center retrospective study without introducing external validation to ensure the unity of surgical methods and techniques. Multi-center and larger- sample studies are required to validate our findings and conclusions.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eWe have developed a combined model to predict the presence or absence of VA injury during primary cervical tumor surgery by integrating radiomic-based features and clinical features (the length of the tumor invading the VA and degree of VA stenosis). This combined model may be able to provide some help in clinical decision making, and fill the gap in this field of research.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eXiangzhi Ni, Jiayang Yan, and Jiayi Zhang contributed equally to this work. Fukai Li, Shuming Hou, Lingyun Shen, and Guangwen Duan conducted data acquisition, collection and interpretation. Fukai Li, Xiang Wang, and Jiayi Zhang made contribute in the data acquisition and annotation. Jiayang Yan made important contributions to the model designing and building. Jiayi Zhang and Hongbiao Sun made of image preprocessing and model interpretation. Xiangzhi Ni and Jiayang Yan made major contributions in manuscript drafting. Tielong Liu and Minglei Yang made contribution to the manuscript editing. Shiyuan Liu had final responsibility for the decision to submit for publication. All authors contributed to this study and article, and approved the submitted version.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eYang M, Zhong N, Dai Z, Ma X, Leng A, Zhou Y, Wang J, Jiao J, Xiao J (2024) Risks for prolonged mechanical ventilation and reintubation after cervical malignant tumor surgery: a nested case-control study. Eur Spine J 33:3069-3081. doi: 10.1007/s00586-024-08313-7\u003c/li\u003e\n\u003cli\u003eLau D, Chan AK, Theologis AA, Chou D, Mummaneni PV, Burch S, Berven S, Deviren V, Ames C (2016) Costs and readmission rates for the resection of primary and metastatic spinal tumors: a comparative analysis of 181 patients. J Neurosurg Spine 25:366-378. doi: 10.3171/2016.2.SPINE15954\u003c/li\u003e\n\u003cli\u003eKerr DL, Dial BL, Lazarides AL, Catanzano AA, Lane WO, Blazer DG, 3rd, Brigman BE, Mendoza-Lattes S, Eward WC, Erickson ME (2019) Epidemiologic and survival trends in adult primary bone tumors of the spine. Spine J 19:1941-1949. doi: 10.1016/j.spinee.2019.07.003\u003c/li\u003e\n\u003cli\u003eChi JH, Bydon A, Hsieh P, Witham T, Wolinsky JP, Gokaslan ZL (2008) Epidemiology and demographics for primary vertebral tumors. Neurosurg Clin N Am 19:1-4. doi: 10.1016/j.nec.2007.10.005\u003c/li\u003e\n\u003cli\u003eJiang P, Zhang X, Jiang W, Meng N, Aili A, Wang J (2020) Analysis of long-term outcome of image-guided volumetric modulated arc therapy (VMAT) for primary malignant tumor of the cervical spine. Cancer Biol Ther 21:623-628. doi: 10.1080/15384047.2020.1743149\u003c/li\u003e\n\u003cli\u003eFerraresi V, Nuzzo C, Zoccali C, Marandino F, Vidiri A, Salducca N, Zeuli M, Giannarelli D, Cognetti F, Biagini R (2010) Chordoma: clinical characteristics, management and prognosis of a case series of 25 patients. BMC Cancer 10:22. doi: 10.1186/1471-2407-10-22\u003c/li\u003e\n\u003cli\u003eMacDonald IJ, Lin CY, Kuo SJ, Su CM, Tang CH (2019) An update on current and future treatment options for chondrosarcoma. Expert Rev Anticancer Ther 19:773-786. doi: 10.1080/14737140.2019.1659731\u003c/li\u003e\n\u003cli\u003eMukherjee D, Chaichana KL, Parker SL, Gokaslan ZL, McGirt MJ (2013) Association of surgical resection and survival in patients with malignant primary osseous spinal neoplasms from the Surveillance, Epidemiology, and End Results (SEER) database. Eur Spine J 22:1375-1382. doi: 10.1007/s00586-012-2621-4\u003c/li\u003e\n\u003cli\u003eAilon T, Torabi R, Fisher CG, Rhines LD, Clarke MJ, Bettegowda C, Boriani S, Yamada YJ, Kawahara N, Varga PP, Shin JH, Saghal A, Gokaslan ZL (2016) Management of Locally Recurrent Chordoma of the Mobile Spine and Sacrum: A Systematic Review. Spine (Phila Pa 1976) 41 Suppl 20:S193-S198. doi: 10.1097/BRS.0000000000001812\u003c/li\u003e\n\u003cli\u003eLee MJ, Weaver DJ, El Dafrawy MH (2024) Extracranial Vertebral Artery Injuries. J Am Acad Orthop Surg 32:e899-e908. doi: 10.5435/JAAOS-D-24-00244\u003c/li\u003e\n\u003cli\u003eEskander MS, Drew JM, Aubin ME, Marvin J, Franklin PD, Eck JC, Patel N, Boyle K, Connolly PJ (2010) Vertebral artery anatomy: a review of two hundred fifty magnetic resonance imaging scans. Spine (Phila Pa 1976) 35:2035-2040. doi: 10.1097/BRS.0b013e3181c9f3d4\u003c/li\u003e\n\u003cli\u003eLebl DR, Bono CM, Velmahos G, Metkar U, Nguyen J, Harris MB (2013) Vertebral artery injury associated with blunt cervical spine trauma: a multivariate regression analysis. Spine (Phila Pa 1976) 38:1352-1361. doi: 10.1097/BRS.0b013e318294bacb\u003c/li\u003e\n\u003cli\u003eLi D, Wu Z, Ren C, Hao SY, Wang L, Xiao XR, Tang J, Wang YG, Meng GL, Zhang LW, Zhang JT (2017) Foramen magnum meningiomas: surgical results and risks predicting poor outcomes based on a modified classification. J Neurosurg 126:661-676. doi: 10.3171/2016.2.JNS152873\u003c/li\u003e\n\u003cli\u003eFatima N, Shin JH, Curry WT, Chang SD, Meola A (2021) Microsurgical resection of foramen magnum meningioma: multi-institutional retrospective case series and proposed surgical risk scoring system. J Neurooncol 153:331-342. doi: 10.1007/s11060-021-03773-z\u003c/li\u003e\n\u003cli\u003eWestbroek EM, Pennington Z, Ehresman J, Ahmed AK, Gailloud P, Sciubba DM (2020) Vertebral Artery Sacrifice versus Skeletonization in the Setting of Cervical Spine Tumor Resection: Case Series. World Neurosurg 139:e601-e607. doi: 10.1016/j.wneu.2020.04.071\u003c/li\u003e\n\u003cli\u003ePeeters JB, Dessesard Olijnyk L, Janelle F, Shedid D, Bojanowski MW, Labidi M (2024) Surgical management of tumors of the cervical spine and craniovertebral junction involving the vertebral artery: A narrative review. Neurochirurgie 70:101550. doi: 10.1016/j.neuchi.2024.101550\u003c/li\u003e\n\u003cli\u003eWoertler K (2003) Benign bone tumors and tumor-like lesions: value of cross-sectional imaging. Eur Radiol 13:1820-1835. doi: 10.1007/s00330-003-1902-z\u003c/li\u003e\n\u003cli\u003eTang H, Ahlawat S, Fayad LM (2018) Multiparametric MR Imaging of Benign and Malignant Bone Lesions. Magn Reson Imaging Clin N Am 26:559-569. doi: 10.1016/j.mric.2018.06.010\u003c/li\u003e\n\u003cli\u003eThakur NA, Daniels AH, Schiller J, Valdes MA, Czerwein JK, Schiller A, Esmende S, Terek RM (2012) Benign tumors of the spine. J Am Acad Orthop Surg 20:715-724. doi: 10.5435/JAAOS-20-11-715\u003c/li\u003e\n\u003cli\u003eYip SS, Aerts HJ (2016) Applications and limitations of radiomics. Phys Med Biol 61:R150-166. doi: 10.1088/0031-9155/61/13/R150\u003c/li\u003e\n\u003cli\u003eXu X, Zhang HL, Liu QP, Sun SW, Zhang J, Zhu FP, Yang G, Yan X, Zhang YD, Liu XS (2019) Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma. J Hepatol 70:1133-1144. doi: 10.1016/j.jhep.2019.02.023\u003c/li\u003e\n\u003cli\u003eYang L, Chu W, Li M, Xu P, Wang M, Peng M, Wang K, Zhang L (2022) Radiomics in Gastric Cancer: First Clinical Investigation to Predict Lymph Vascular Invasion and Survival Outcome Using (18)F-FDG PET/CT Images. Front Oncol 12:836098. doi: 10.3389/fonc.2022.836098\u003c/li\u003e\n\u003cli\u003eMansour TMM, El-Barody MM, Tammam H, Okasha A (2021) Role of diffusion-weighted MRI in differentiating between benign and malignant bone lesions: a prospective study. Clin Radiol 76:576-584. doi: 10.1016/j.crad.2021.02.030\u003c/li\u003e\n\u003cli\u003eBossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig L, Lijmer JG, Moher D, Rennie D, de Vet HC, Kressel HY, Rifai N, Golub RM, Altman DG, Hooft L, Korevaar DA, Cohen JF, Group S (2015) STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 351:h5527. doi: 10.1136/bmj.h5527\u003c/li\u003e\n\u003cli\u003eAl-Hawary MM, Francis IR, Chari ST, Fishman EK, Hough DM, Lu DS, Macari M, Megibow AJ, Miller FH, Mortele KJ, Merchant NB, Minter RM, Tamm EP, Sahani DV, Simeone DM (2014) Pancreatic ductal adenocarcinoma radiology reporting template: consensus statement of the society of abdominal radiology and the american pancreatic association. Gastroenterology 146:291-304 e291. doi: 10.1053/j.gastro.2013.11.004\u003c/li\u003e\n\u003cli\u003eZaky AM, Wolfgang CL, Weiss MJ, Javed AA, Fishman EK, Zaheer A (2017) Tumor-Vessel Relationships in Pancreatic Ductal Adenocarcinoma at Multidetector CT: Different Classification Systems and Their Influence on Treatment Planning. Radiographics 37:93-112. doi: 10.1148/rg.2017160054\u003c/li\u003e\n\u003cli\u003eXiao ML, Wei Y, Zhang J, Jian JM, Song Y, Lin ZJ, Qian L, Zhang GF, Qiang JW (2022) MRI Texture Analysis for Preoperative Prediction of Lymph Node Metastasis in Patients with Nonsquamous Cell Cervical Carcinoma. Acad Radiol 29:1661-1671. doi: 10.1016/j.acra.2022.01.005\u003c/li\u003e\n\u003cli\u003eZhou TH, Zhou XX, Ni J, Ma YQ, Xu FY, Fan B, Guan Y, Jiang XA, Lin XQ, Li J, Xia Y, Wang X, Wang Y, Huang WJ, Tu WT, Dong P, Li ZB, Liu SY, Fan L (2024) CT whole lung radiomic nomogram: a potential biomarker for lung function evaluation and identification of COPD. Mil Med Res 11:14. doi: 10.1186/s40779-024-00516-9\u003c/li\u003e\n\u003cli\u003eCui Y, Shi X, Wang S, Qin Y, Wang B, Che X, Lei M (2022) Machine learning approaches for prediction of early death among lung cancer patients with bone metastases using routine clinical characteristics: An analysis of 19,887 patients. Front Public Health 10:1019168. doi: 10.3389/fpubh.2022.1019168\u003c/li\u003e\n\u003cli\u003eShi X, Cui Y, Wang S, Pan Y, Wang B, Lei M (2024) Development and validation of a web-based artificial intelligence prediction model to assess massive intraoperative blood loss for metastatic spinal disease using machine learning techniques. Spine J 24:146-160. doi: 10.1016/j.spinee.2023.09.001\u003c/li\u003e\n\u003cli\u003eOrigitano TC, al-Mefty O, Leonetti JP, DeMonte F, Reichman OH (1994) Vascular considerations and complications in cranial base surgery. Neurosurgery 35:351-362; discussion 362-353. doi: 10.1227/00006123-199409000-00001\u003c/li\u003e\n\u003cli\u003eCarpentier A, Blanquet A, George B (2001) Suboccipital and cervical chordomas: radical resection with vertebral artery control. Neurosurg Focus 10:E4. doi: 10.3171/foc.2001.10.3.5\u003c/li\u003e\n\u003cli\u003eHoshino Y, Kurokawa T, Nakamura K, Seichi A, Mamada T, Saita K, Miyoshi K (1996) A report on the safety of unilateral vertebral artery ligation during cervical spine surgery. Spine (Phila Pa 1976) 21:1454-1457. doi: 10.1097/00007632-199606150-00011\u003c/li\u003e\n\u003cli\u003eBall JR, Shelby T, Mertz K, Mills ES, Ton A, Alluri RK, Hah RJ (2024) The Incidence of Vertebral Artery Injury in Cervical Spine Surgery. World Neurosurg 181:e841-e847. doi: 10.1016/j.wneu.2023.10.144\u003c/li\u003e\n\u003cli\u003eSharma A, Sharma A, Sharma R, Dewan Y (2017) Management of iatrogenic vertebral artery injury following cervical neurofibroma excision: a case report. Br J Neurosurg 31:478-480. doi: 10.1080/02688697.2016.1199780\u003c/li\u003e\n\u003cli\u003eGarcia Alzamora M, Rosahl SK, Lehmberg J, Klisch J (2005) Life-threatening bleeding from a vertebral artery pseudoaneurysm after anterior cervical spine approach: endovascular repair by a triple stent-in-stent method. Case report. Neuroradiology 47:282-286. doi: 10.1007/s00234-005-1343-2\u003c/li\u003e\n\u003cli\u003eDavnall F, Yip CS, Ljungqvist G, Selmi M, Ng F, Sanghera B, Ganeshan B, Miles KA, Cook GJ, Goh V (2012) Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice? Insights Imaging 3:573-589. doi: 10.1007/s13244-012-0196-6\u003c/li\u003e\n\u003cli\u003eLubner MG, Smith AD, Sandrasegaran K, Sahani DV, Pickhardt PJ (2017) CT Texture Analysis: Definitions, Applications, Biologic Correlates, and Challenges. Radiographics 37:1483-1503. doi: 10.1148/rg.2017170056\u003c/li\u003e\n\u003cli\u003eGillies RJ, Kinahan PE, Hricak H (2016) Radiomics: Images Are More than Pictures, They Are Data. Radiology 278:563-577. doi: 10.1148/radiol.2015151169\u003c/li\u003e\n\u003cli\u003eVarghese BA, Cen SY, Hwang DH, Duddalwar VA (2019) Texture Analysis of Imaging: What Radiologists Need to Know. AJR Am J Roentgenol 212:520-528. doi: 10.2214/AJR.18.20624\u003c/li\u003e\n\u003cli\u003eLi Z, Yu L, Wang X, Yu H, Gao Y, Ren Y, Wang G, Zhou X (2018) Diagnostic Performance of Mammographic Texture Analysis in the Differential Diagnosis of Benign and Malignant Breast Tumors. Clin Breast Cancer 18:e621-e627. doi: 10.1016/j.clbc.2017.11.004\u003c/li\u003e\n\u003cli\u003eBalagurunathan Y, Gu Y, Wang H, Kumar V, Grove O, Hawkins S, Kim J, Goldgof DB, Hall LO, Gatenby RA, Gillies RJ (2014) Reproducibility and Prognosis of Quantitative Features Extracted from CT Images. Transl Oncol 7:72-87. doi: 10.1593/tlo.13844\u003c/li\u003e\n\u003cli\u003eZhang R, Zhang Q, Ji A, Lv P, Acosta-Cabronero J, Fu C, Ding J, Guo D, Teng Z, Lin J (2023) Prediction of new cerebral ischemic lesion after carotid artery stenting: a high-resolution vessel wall MRI-based radiomics analysis. Eur Radiol 33:4115-4126. doi: 10.1007/s00330-022-09302-4\u003c/li\u003e\n\u003cli\u003eVersteijne E, Suker M, Groothuis K, Akkermans-Vogelaar JM, Besselink MG, Bonsing BA, Buijsen J, Busch OR, Creemers GM, van Dam RM, Eskens F, Festen S, de Groot JWB, Groot Koerkamp B, de Hingh IH, Homs MYV, van Hooft JE, Kerver ED, Luelmo SAC, Neelis KJ, Nuyttens J, Paardekooper G, Patijn GA, van der Sangen MJC, de Vos-Geelen J, Wilmink JW, Zwinderman AH, Punt CJ, van Eijck CH, van Tienhoven G, Dutch Pancreatic Cancer G (2020) Preoperative Chemoradiotherapy Versus Immediate Surgery for Resectable and Borderline Resectable Pancreatic Cancer: Results of the Dutch Randomized Phase III PREOPANC Trial. J Clin Oncol 38:1763-1773. doi: 10.1200/JCO.19.02274\u003c/li\u003e\n\u003cli\u003ePeng CW, Chou BT, Bendo JA, Spivak JM (2009) Vertebral artery injury in cervical spine surgery: anatomical considerations, management, and preventive measures. Spine J 9:70-76. doi: 10.1016/j.spinee.2008.03.006\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":"european-spine-journal","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"esjo","sideBox":"Learn more about [European Spine Journal](http://link.springer.com/journal/586)","snPcode":"586","submissionUrl":"https://submission.springernature.com/new-submission/586/3","title":"European Spine Journal","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Primary cervical tumors, MRI, Radiomics, Vertebral artery","lastPublishedDoi":"10.21203/rs.3.rs-7218476/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7218476/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eCurrently, reliable preoperative methods for predicting VA invasion are lacking. The authors develop a novel model based on MRI radiomic signatures combined with clinical and imaging features for predicting intraoperative vertebral artery injury in patients with primary cervical tumors.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eIncluded in this retrospective study were 167 patients who received surgical resection for primary cervical tumors. They were randomly assigned to a training set (n\u0026thinsp;=\u0026thinsp;116) and a test set (n\u0026thinsp;=\u0026thinsp;51) set. Least absolute shrinkage and selection operator logistic regression was applied for feature selection and radiomic signature construction. A multilayer perceptron (MLP) model and 10 machine learning models were used to develop diverse prediction models. Independent risk factors of clinical variables were screened by Logistic regression, based on which a clinical model was constructed. A combined model was established by combining the radiomic signatures and clinical factors. The predictive performance of the combined model was evaluated in both training and test sets using Hosmer\u0026ndash;Lemeshow test and decision curve analysis (DCA).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eAccording to the scoring system, the MLP model obtained the highest total score of 82, meaning that its prediction performance was the best of all evaluated models, so the MLP was selected to construct the radiomics model. The AUC of the combined model in the training and test cohorts was 0.952 and 0.932 respectively, and both were higher than that of the radiomics model (AUC 0.861 in training set, p\u0026thinsp;=\u0026thinsp;0.005, AUC 0.773 in test set, p\u0026thinsp;=\u0026thinsp;0.006) and the clinical model (AUC 0.777 in training set, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, AUC 0.740 in test set, p\u0026thinsp;=\u0026thinsp;0.002) alone.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe present study presents a nomogram that incorporates radiomic signatures and clinical features, which could be used to predict the risk of intraoperative VA injury in patients with primary cervical tumors.\u003c/p\u003e","manuscriptTitle":"A clinical-radiomics nomogram based on multisequence MRI for predicting intraoperative vertebral artery injury in patients with primary cervical spine tumor: a diagnostic study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-13 05:50:01","doi":"10.21203/rs.3.rs-7218476/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-19T10:13:49+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-19T03:29:13+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-15T01:24:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"7347955676356537482945710980655187944","date":"2025-11-07T06:11:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"130671268606224020307649489833963842639","date":"2025-11-07T02:23:08+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-07T12:12:55+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-01T06:05:01+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-01T06:03:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"European Spine Journal","date":"2025-07-26T04:49:41+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"european-spine-journal","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"esjo","sideBox":"Learn more about [European Spine Journal](http://link.springer.com/journal/586)","snPcode":"586","submissionUrl":"https://submission.springernature.com/new-submission/586/3","title":"European Spine Journal","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"560782f9-24ff-4549-b3f1-8ff2e0b73d90","owner":[],"postedDate":"August 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-01-26T16:02:14+00:00","versionOfRecord":{"articleIdentity":"rs-7218476","link":"https://doi.org/10.1007/s00586-025-09706-y","journal":{"identity":"european-spine-journal","isVorOnly":false,"title":"European Spine Journal"},"publishedOn":"2026-01-20 15:57:22","publishedOnDateReadable":"January 20th, 2026"},"versionCreatedAt":"2025-08-13 05:50:01","video":"","vorDoi":"10.1007/s00586-025-09706-y","vorDoiUrl":"https://doi.org/10.1007/s00586-025-09706-y","workflowStages":[]},"version":"v1","identity":"rs-7218476","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7218476","identity":"rs-7218476","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}