Artificial neural networks predict the incidence of deep venous thrombosis in hospitalized patient

Artificial neural networks predict the incidence of deep venous thrombosis in hospitalized patient | 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 Artificial neural networks predict the incidence of deep venous thrombosis in hospitalized patient Zhongbin Zhou, Yuan Yao, Hanyu Zhou, Yuanyuan Qiao, Zhihan Gao, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4564132/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective: To construct and validate artificial neural networks (ANNs) for predicting the occurrence of deep venous thrombosis(DVT) and compare the predictive performance of the ANNs model with that of logistic regression(LR)model,linear discriminant analysis(LDA) model,and simple artificial neural network (SANNs) model. Methods: 1295 cases were selected, including 729 patients with DVT and 566 patients without. 75% of the cases (993 cases) are randomly selected as the training set for model construction, and the remaining 25% of the cases (302 cases) are used as the testing set to verify the prediction performance. After deep learning of the training data, the ANNs model with different numbers of hidden_nodes and epochs was constructed. The prediction efficiency of the ANNs model was tested by comparing the results of LR,LDA,and SANNs model as the benchmark afterwards. Results: When the number of hidden_nodes was 8 and the number of epochs was 800 in ANNS model, the Acc reached the highest,which the Acc, Youden index was 81.84%, 0.6450 respectively.The prediction performance of this model was higher than that of LR,LDA ,and SANNs. Conclusions: This study provided good evidence for the application of ANNs to predict DVT in a large number of data. However, more research will be needed to confirm its application in the prediction of DVT. Deep venous thrombosis artificial neural network logistic regression discriminant analysis forecasting Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction DVT is a venous reflux disorder caused by abnormal coagulation of blood in the deep veins, and it mostly occurs in lower limb thrombosis [ 1 ] . Venous wall injury, slow blood flow, and hypercoagulability of blood might easily lead to DVT. The majority of patients with DVT have an insidious thrombosis, without obvious symptoms, while a few patients could have symptomatic deep vein thrombosis, mainly manifested by sudden swelling, stiffness, pain, tactile abnormalities, increased soft tissue tension, aggravated after activity and alleviated after the elevation of the affected limb. Pulmonary embolism (PE) and DVT are called venous thromboembolism (VTE), which have the same susceptibility factors and are two clinical manifestations of VTE at different sites and at different stages. Researches showed that 40%~60% of surgical patients and non-surgical patients were at risk of VTE, while the prevention proportion of high-risk groups was very low, even lower in Asian countries [ 2 – 5 ] . Using various methods to early determine whether patients are at high risk of VTE and take timely preventive measures can reduce the incidence of VTE in patients [ 6 ] . The prevention protocols of VTE in hospitals suggests the use of Caprini score scale or Padua score scale for VTE risk assessment for inpatients, especially inpatients in high-risk VTE departments. The two methods are simple and convenient to operate and play a great role in VTE prevention but slightly rough when compared with the computer-aided big data calculation. A recent study had shown that area under curve(AUC) (95%CI), sensitivity(Sen), specificity(Spe), accuracy(Acc), Youden index of Caprini Risk Assessment Scale were 0.596 (0.552, 0.638), 26.1%,96.5%,0.596, 0.23 respectively, but that of the VTE risk warning model were 0.960 (0.940, 0.976), 92.6%,91.8%,0.960,0.84 respectively, with statistically significant differences (Z = 14.521, P < 0.0001) [ 7 ] . The risk warning model outperformed the Caprini Risk Assessment Scale in predicting the occurrence of VTE in hospitalized patients. Traditional risk warning models includes logistic regression, linear regression, and discriminant analysis, while ANNs represents the latest alternative methods to traditional modeling techniques [ 8 – 10 ] . In recent years ANN has been widely applied in medical areas [ 11 – 18 ] . The ANNs systems has been extensively used to address complex real world problems, particularly when dealing with non-linear models or when the definition of underlying mechanisms is incomplete. Only a few studies applied the artificial neural network algorithm model for DVT prediction [ 15 , 19 ] . Among them, the number of hidden_nodes and epochs in the model had fixed values, and how the changes of the number these parameters affected the prediction effect was not discussed. The purpose of this study was to develop the ANN-based predictive model to predict high-risk patients with DVT. Furthermore, we compared the prediction performance of ANNs with logistic regression, discriminant analysis, and SANNs 2. Materials and methods 2.1. Materials Hospital cases in our hospital from 1 January 2011 to 31 December 2020 were reviewed. Cases diagnosed with DVT during this hospitalization were considered as positive cases, and cases not diagnosed with DVT as negative cases. The diagnosis of DVT was mainly based on Doppler ultrasonography and venography. Referring to the previous research [ 20 – 21 ] and considering whether it could be directly extracted from the hospital information system, this study took 11 variables of patient sex, age, disease diagnosis, test, and surgical information as the variables of this study. The specific variables and data treatment were shown in Table 1 . There were 729 positive cases, all of which were used for model building and performance evaluation. There were 233,620 negative cases, from which 1,300 were randomly selected. Records with missing test data due to lack of lab tests during this hospitalization were excluded, and 566 cases were left for model building and performance evaluation. In total, 1295 data were included in this study, and the distribution of each variable among DVT and non-DVT patients is shown in Table 2 . The hospital Ethics Committee had approved the protocol with the ethics approval number of HZKY-PJ-2022-21. And the study entailed no intervention in patient care, informed consent was not required. The study was carried out according to the principles of The Declaration of Helsinki. Table 1 Coding of variables. Variable Variable description Coding Gender Gender 0.01:female;1:male Age(years) age 0.01:=60 DD(ng/ml) D-Dimer Antigen 0.01:<200;0.5:=500 FIB(g/L) Fibrinogen 0.01:<2;0.5:=4 HBP Hypertension 0.01:no;1:yes CHD Coronary heart disease 0.01:no;1:yes MT Malignant tumor 0.01:no;1:yes TRO Trauma or operation 0.01:no;1:yes DM Diabetes Mellitus 0.01:no;1:yes CHEMO Chemotherapy 0.01:no;1:yes COPD Chronic obstructive pulmonary disease 0.01:no;1:yes Table 2 Distribution of influencing factors variables without DVT(n = 566) with DVT(n = 729) coding = 0.01 coding = 0.5 coding = 1 coding = 0.01 coding = 0.5 coding = 1 Gender 246 320 223 506 Age 293 273 331 398 DD 310 116 140 91 107 531 FIB 12 383 171 29 391 309 HBP 393 173 412 317 CHD 447 119 546 183 MT 431 135 528 201 TRO 365 201 243 486 DM 484 82 581 148 CHEMO 565 1 715 14 COPD 497 69 478 251 2.2. Setting training data and testing data 75% of the cases (993 cases) were randomly selected as the training set for model construction, and the remaining 25% of the cases (302 cases) were used as the testing set to verify the prediction performance. The categorical percentage of each variable was calculated separately in the training set and testing set. Whether there were significant differences between the training set and testing set were assessed for a significant association by either Chi-square statistics or Fisher’s exact test. 2.3. Logistic regression(LR) The LR equations were constructed by the stepwise process. The first was to introduce the variables of the training set into the equation one by one, requiring its Chi-square value to be significant at the SLENTRY(Specifies the significance level threshold for the variables introduced in the forward selection method) level. After introducing a variable, all variables already included in the equation were tested, removing variables that were not significant at the SLSTAY(Specifies the significance level threshold for retained variables in the backward removal method) level. All nonsignificant variables were removed before reintroducing new variables. When all variables outside the equation were not significant at the SLENTRY level and all variables within the equation with Wald Chi-Square were significant at the SLSTAY level, the stepwise process were stopped. In addition, variables that were just removed were again introduced into the equation, and the stepwise process was also stopped. Both the SLENTRY and SLSTAY values in this stepwise process were set to 0.15. After the regression equations were constructed, the equation prediction performance was evaluated by testing set data using MedCalc. 2.4. Discriminant analysis The entry model variables were first determined by stepwise discriminant analysis(SDA), and then linear discriminant analysis(LDA) was used to construct the prediction equation. SDA introduced variables into the model one by one, requiring the variable WILKS'S LAMBDA(The ratio of the within-group sum of squares to the total sum of squares) value to be significant at the SLENTRY level. As the variable gradually increased in the model, the discriminative power of the variables introduced earlier could also change, removing the variable if the variable in the model was not significant at the SLSTAY level. This process was repeated until all variables of the model were significant at the SLSTAY level and the others did not meet the criteria for entering the model. Both the SLENTRY and SLSTAY values in this process were set to 0.15. After the regression equations were constructed, the equation prediction performance was evaluated by testing set data using MedCalc . 2.5. Artificial neural networks(ANNs) The ANNs is divided into three layers, one is the input layer, one is hidden layer, and the last one is the final output layer. Each layer has one or more nodes, as shown in Fig. 1 . The possible influencing factors of DVT are input in the input layer. Each node of the input layer is connected to each node of the hidden layer, and the connection weights between nodes are different. The input layer inputs information to the hidden layer through the connection between nodes. The information inputs to the hidden layer is processed by the sigmoid activation function before the information is output. The sigmoid function is to suppress the output signal if the combined signal input from the input layer to the hidden layer is not strong enough. But if it is strong enough,the effect of sigmoid function is to stimulate the output signal. Each node of the hidden layer is also connected to each node of the output layer. The information of the hidden layer is input to the output layer through connections with different weights. The information of the output layer is also processed by the sigmoid activation function to get the risk of having a DVT. After building the model architectures, the next step is to determine the connection weights between the input layer and the hidden layer, the hidden layer and the output layer through the training process. The training process is divided in two steps. In the first step, an initial random weight is created (sampling weight using a normal probability distribution, where the mean is 0 and the standard variance is the square root of the number of nodes) and the model calculates the output for a given training sample. In the second step, the difference between the model output and the original data of the training sample guides the connection weights to update by the gradient descent method. The following is the weight update matrix algorithm. The training set (including 11 variables) was input into the input layer to construct the ANNs model. Testing set was used to verify the prediction accuracy of the constructed neural network model and to select the optimal neural network model. 2.6. Simple artificial neural networks(SANNs) To study the prediction performance of the ANNs model with a small number of input nodes, the six variables selected by the LR model were used as input variables to develop the “simplified” ANNs model. The data set and training process were the same as the ANNs model with all eleven input variables. 2.7. Statistical analysis In this research ,data analysis and statistics were performed with the commercial software SAS (SAS for Windows, Version 9.3, SAS Inc.). Significant differences between different groups were evaluated by chi-square analysis and unpaired Student t-test and ROC analysis were performed by using the medcalc software(Version 19.0.4). P values < 0.05 were considered statistically significant. The ANNs models were performed with Python 3.6. Prediction indicators included Sen, Spe, Acc, Youden index, AUC. Acc = (Sen + Spe) / (Sen + Spe + Fpr (false positive rate) + Fnr(false negative rate)) Youden index=(Sen + Spe) -1. 3. Result 3.1. Distribution of variables in the training set and testing set The incidence of DVT in the training and testing sets were 57.3% and 53.0% respectively, with no significant(p = 0.1850) differences between the two sets. 11 variables also showed no significant difference in the training and testing sets. It was shown in Table 3 . It meant that the two sets were well balanced in the distribution of clinical characteristics. Table 3 Characteristics of patients. Training set (n = 993) Testing set (n = 302) p-Value DVT(%) 57.3 53.0 0.1850 Gender: male(%) 54.2 59.6 0.0968 Age: >=60(%) 60.2 59.9 0.9287 DD: >500(%),4(%),<2(%) 38.0,3.0 34.1,3.6 0.4450 HBP(%) 38.0 37.4 0.8633 CHD(%) 23.4 23.2 0.9470 MT(%) 26.2 25.2 0.7239 TRO(%) 53.9 50.3 0.2796 DM(%) 17.2 20.2 0.2055 CHEMO(%) 1.21 0.99 0.2396 COPD(%) 25.8 21.2 0.1055 3.2. Results of LR The results of LR analysis were summarized in Table 4 . There were six variables included in the final LR model, the probability of VTE could be calculated by the following logistic equations: = 2.2656–0.4409×(HBP) + 0.3428×(MT)-1.2073×(TRO)-1.3056×(COPD)-2.0844×(DD)-0.4817×(AGE)。 Table 4 The results of LR Parameter DF Estimate Standard Error Wald Chi-Square Pr> ChiSq Intercept 1 2.2656 0.1973 131.9156 < .0001 HBP 1 -0.4409 0.1722 6.5542 0.0105 MT 1 0.3428 0.1826 3.5236 0.0605 TRO 1 -1.2073 0.1630 54.8816 < .0001 COPD 1 -1.3056 0.2012 40.1013 < .0001 DD 1 -2.0844 0.1866 124.8458 < .0001 AGE 1 -0.4817 0.1705 7.9779 0.0047 Analyzing the ROC curve of the LR model, the Sen index was 91.87%, the Spe index was 69.72%, and the AUC index was 0.870. It was shown in Fig. 2 . 3.3. Results of LDA The results of LDA were summarized in Table 5 . There were six variables included in the final SDA model. The probability of VTE could be calculated by the following LDA equations: Probability=-4.1665 + 1.4459×(HBP) + 0.3877×(MT) + 2.8786×(TRO) + 1.6916×(COPD) + 4.4186×(DD) + 2.2095×(AGE) Analyzing the ROC curve of the LDA model, the Sen was 88.12%, the Spe was 73.94%, and AUC was 0.872. It was shown in Fig. 3 . Table 5 The results of discriminant analysis Parameter Partial R-Square F value Pr > F Wilks’ lambda Pr ASCC DD 0.2231 284.54 < 0.0001 0.7769 < 0.0001 0.2231 < 0.0001 COPD 0.0447 46.33 < 0.0001 0.7422 < 0.0001 0.2578 < 0.0001 TRO 0.0508 52.959 < 0.0001 0.7045 < 0.0001 0.2955 < 0.0001 HBP 0.0142 14.27 0.0002 0.6944 < 0.0001 0.3055 < 0.0001 AGE 0.0069 6.85 0.0090 0.6897 < 0.0001 0.3103 < 0.0001 MT 0.0035 3.47 0.0627 0.6827 < 0.0001 0.3128 < 0.0001 3.4. Results of SANNs The prediction accuracy of the SANNs model at different hidden_nodes and epochs was shown in Fig. 4 . When the number of hidden_nodes was 7 and the number of epochs was 800, the accuracy reached the highest,with an average of 80.11% . 3.5. Results of ANNs The prediction accuracy of the ANNs model at different hidden_nodes and epochs was shown in Fig. 5 . When the number of hidden_nodes was 8 and the number of epochs was 800, the accuracy reached the highest, with an average of 81.84%. 3.6. Comparison of the various prediction results There were high Sen but low Spe in LR and LDA model,the Sen of SANNs model was lower than that of LDA and LR model,but the Spe was higher than both of them.The Sen and Spe of ANNs model were higher than that of SANNs. The Youden index and Acc of ANNs model was higher than all other prediction models. The prediction performance of each model was shown in Fig. 6 and Table 6 . Table 6 The results of ROC curve analysis Sen (%) Spe (%) Youden index Acc (%) AUC (mean ± S.D.) LR 91.87 69.72 0.6159 79.47 0.869 ± 0.021 LDA 88.12 73.94 0.6207 80.18 0.872 ± 0.020 SANNs 84.37 76.76 0.6114 80.11 0.869 ± 0.020 ANNs 85.62 78.87 0.6450 81.84 0.870 ± 0.021 4. Discussion There were many factors affecting the occurrence of DVT, including age,malignant tumor, previous medical history, family medical history, BMI, trauma, surgical operation, abnormal pulmonary function, bedridden, severe pulmonary disease, uncontrolled hypertension etc. Hopefully, appropriate methods can be applied to construct the DVT prediction models to identify high-risk patients at the early stage and take timely preventive measures for them to reduce the time to cure and reduce patient mortality. In recent years, some scholars had studied the occurrence of DVT. Qiang Li et al. found old age, longer operation time, and arthroplasty were independent risk factors, physical labor and postoperative exercises were protective factors for DVT in patients with lower extremity fractures [ 20 ] .Wang Fei Li et al. found age, smoking history, operation time and body mass index are independent risk factors for venous thromboembolism [ 21 ] . This study found age, d-dimer, Trauma or operation, Chronic obstructive pulmonary disease, Hypertension, and Malignant tumor were risk factors for DVT, which is basically consistent with the findings of other scholars. For the high-risk patients of DVT predicted by the model,doctors can take basic prevention (such as early going out of bed, avoiding dehydration, etc), drug prevention, mechanical prevention, placement of recyclable vena cava filter to prevent their occurrence.They can also develop more detailed DVT prevention methods based on the ANNS model. Few reports have focused on predicting the occurrence of VTE. Two recent studies have used logistic regression to predict whether VTE occurs.Juhua Li et al. built a simple and efficient diagnostic factor for DVT in patients after Neurosurgery, which was a reliable predictive index for DVT in patients [ 22 ] .Chen Shen et al. developed VTE risk warning model,which had high accuracy in predicting VTE occurrence in hospitalized patients [ 7 ] . Laleh Agharezaei et al. built ANNs to help experts diagnose and predict the risk level of pulmonary embolism in patients [ 19 ] .In this study, we compared the abilities of LR model, LDA model and ANNs model to predict the occurrence of VTE. We evaluated the performance of these models by calculating Youden index and Acc. LR is a widely used algorithm, efficient and fast, unnecessary to scale the input features.But its disadvantage is that it can not be used to solve nonlinear problems.Press and Willson concluded by comparing LR and LDA that when these variables were not multivariate normal distribution in the class,LR was preferable to LDA. However ,if the independent variable belonged to multivariate normal distribution, LR was more inefficient. Efron gave two normal populations with common covariance, the efficiency of the LR was only between 1/2 or 2/3 of the efficiency of LDA [ 23 ] with about the same misjudgment rate. Compared with the two methods.ANNs have no requirements for independent variable and good ability to identify complex nonlinear relationships among variables. It is particularly suitable for processing information that should take many factors and conditions into considerations at the same time [ 24 ] . The area under the ROC curve (AUC) is a global measure to discriminate whether a specific condition is present or not present.In general, if its AUC value is greater than 0.85,indicates that the model has a better predictive efficacy.In this study, the AUC values of the LR ,LDA, SANNs, and ANNs models were greater than 0.85, indicating that the predictive efficacy of all the models was good.When selecting an optimal threshold (or cut-off point) of the ROC curve, we need to consider the aims of the diagnostic test, considering the significance and costs of a false-positive or false-negative interpretation. A commonly used approach when selecting a cut-off point is to give equal weight to the importance of sensitivity and specificity by choosing the point nearest to the top-left most corner of the ROC curve. This point is also known as the Youden Index [ 25 ] . This research showed that the Youden index and Acc of the ANNs model were 0.6450 and 81.84%, which were both higher than the other models. ANNS model is better predicted than the other models. ANNs model construction requires a process of setting hidden_node and epochs of the neural network, and the number of hidden_nodes and epochs has a great influence on the prediction efficiency of the model. Very few reports had focused on it. We explored the number setting of hidden_nodes and epochs of the neural network in this paper.The number of hidden_nodes of the neural network is related to the number of input_nodes and output_nodes, the complexity of the problem, and the distribution of the data.The basic methods to determine the number of hidden_nodes is to take the most compact structure as possible while satisfying the accuracy requirements, that is, to take the number of hidden_nodes as few as possible. Following this methods, the number of hidden_nodes was 8 in this study. Increasing the number of epochs of the neural network can improve the accuracy, but after reaching a limited value, if it continues to increase, there will be overfitting, and the accuracy may also decrease. When the number of epoch in this study reached around 800, the accuracy is not improved. In this study, the prediction accuracy of SANNs model and Youden index were both lower than ANNs. It indicated that the variables screened by the LR model did not contribute significantly(p > 0.15) to the DVT, but some of the information could be identified by the ANNs model, which improved the accuracy of their prediction. ANNs is an analysis method with high prediction efficient with good variable identification ability. It is able to process multifactorial, imprecise, and ambiguous information simultaneously. Due to the limitations of this research, part possible influencing factors were not included in the ANNs model for DVT prediction. However, in practice, the influencing factors could be added or constantly adjusted, and the prediction accuracy should become higher. This research provided good evidence for applying ANNs to predict DVT. ANNs will be promising in DVT prediction applications. Declarations Ethics approval and consent to participate The Medical Ethics Committee of the 6th Medical Center of PLA General Hospital approved this study with a specific ID HZKY-PJ-2022-21. The Medical Ethics Committee of the 6th Medical Center of PLA General Hospital waived informed consent given the lack of intervention and the anonymity of the data. Consent for publication Not applicable Availability of data and materials The data used to support the findings of this study are restricted by The 6th Medical Center of PLA General Hospital in order to protect patient privacy. Data are available from The 6th Medical Center of PLA General Hospital for researchers who meet the criteria for access to confidential data. Competing Interests The authors declare that they have no competing interests. Funding No funding. Authors’contributions Zhongbin Zhou wrote the main manuscript text,Hanyu Zhou processed and analyzed data，All authors reviewed the manuscript. Acknowledgements This wok was not supported by outside funds. References Abdel-Razeq, H, Qari, M, Kristensen, J, et al. Guidelines for diagnosis and treatment of deep venous thrombosis and pulmonary embolism. Methods Mol Med. 2004; 93 267-92. doi: 10.1385/1-59259-658-4:267 Zhang, Y, Yang, Y, Chen, W, et al. Prevalence and associations of VTE in patients with newly diagnosed lung cancer. CHEST. 2014; 146 CHEST. doi: 10.1378/chest.13-2379 Hill, J, Treasure, T. Reducing the risk of venous thromboembolism in patients admitted to hospital: summary of NICE guidance. BMJ. 2010; 340 c95. doi: 10.1136/bmj.c95 Qu, H, Li, Z, Zhai, Z, et al. Predicting of Venous Thromboembolism for Patients Undergoing Gynecological Surgery. MEDICINE. 2015; 94 (39): e1653. doi: 10.1097/MD.0000000000001653 Cohen, AT, Tapson, VF, Bergmann, JF, et al. Venous thromboembolism risk and prophylaxis in the acute hospital care setting (ENDORSE study): a multinational cross-sectional study. LANCET. 2008; 371 (9610): 387-94. doi: 10.1016/S0140-6736(08)60202-0 Zhai, Z, Wang, C. [Establishing and improving venous thromboembolism prevention and management system in hospital]. Zhonghua Yi Xue Za Zhi. 2015; 95 (30): 2417-8. PMID: 26711198 Shen, C, Ge, B, Liu, X, et al. Predicting the occurrence of venous thromboembolism: construction and verification of risk warning model. BMC Cardiovasc Disord. 2020; 20 (1): 249. doi: 10.1186/s12872-020-01519-9 Xiaoyu, D, Kai, C, Zhihui, H, et al. Predictive value of preoperative erythrocyte electrophoresis exponent for acute deep vein thrombosis after total knee arthroplasty in patients with knee osteoarthritis. J Orthop Surg Res. 2020; 15 (1): 496. doi: 10.1186/s13018-020-02020-x Zhang, Y, Cao, M, Ren, J. NLR value and IL-18 level and their clinical significance in patients with deep vein thrombosis after receiving the surgery for spinal degeneration. Am J Transl Res. 2021; 13 (6): 7156-7163. PMID: 34306476 Maniruzzaman, M, Kumar, N, Menhazul Abedin, M, et al. Comparative approaches for classification of diabetes mellitus data: Machine learning paradigm. COMPUT METH PROG BIO. 2017; 152 23-34. doi: 10.1016/j.cmpb.2017.09.004 Mohammadfam, I, Soltanzadeh, A, Moghimbeigi, A, et al. Use of Artificial Neural Networks (ANNs) for the Analysis and Modeling of Factors That Affect Occupational Injuries in Large Construction Industries. Electron Physician. 2015; 7 (7): 1515-22. doi: 10.19082/1515 Fei, Y, Hu, J, Li, WQ, et al. Artificial neural networks predict the incidence of portosplenomesenteric venous thrombosis in patients with acute pancreatitis. J THROMB HAEMOST. 2017; 15 (3): 439-445. doi: 10.1111/jth.13588 Fei, Y, Gao, K, Li, WQ. Artificial neural network algorithm model as powerful tool to predict acute lung injury following to severe acute pancreatitis. PANCREATOLOGY. 2018; 18 (8): 892-899. doi: 10.1016/j.pan.2018.09.007 Fei, Y, Gao, K, Li, WQ. Prediction and evaluation of the severity of acute respiratory distress syndrome following severe acute pancreatitis using an artificial neural network algorithm model. HPB. 2018; 21 (7): 891-897. doi: 10.1016/j.hpb.2018.11.009 Willan, J, Katz, H, Keeling, D. The use of artificial neural network analysis can improve the risk-stratification of patients presenting with suspected deep vein thrombosis. BRIT J HAEMATOL. 2019; 185 (2): 289-296. doi: 10.1111/bjh.15780 Damiani, G, Conic, RRZ, Pigatto, PDM, et al. Predicting Secukinumab Fast-Responder Profile in Psoriatic Patients: Advanced Application of Artificial-Neural-Networks (ANNs). J DRUGS DERMATOL. 2020; 19 (12): 1241-1246. doi: 10.36849/JDD.2020.5006 Qiu, Y, Wang, T, Yang, X, et al. Development and Validation of Artificial Neural Networks for Survival Prediction Model for Patients with Spontaneous Hepatocellular Carcinoma Rupture After Transcatheter Arterial Embolization. Cancer Manag Res. 2021; 13 7463-7477. doi: 10.2147/CMAR.S328307 Umar, M, Sabir, Z, Raja, MAZ, et al. Numerical Investigations through ANNs for Solving COVID-19 Model. Int J Environ Res Public Health. 2021; 18 (22): doi: 10.3390/ijerph182212192 Agharezaei, L, Agharezaei, Z, Nemati, A, et al. The Prediction of the Risk Level of Pulmonary Embolism and Deep Vein Thrombosis through Artificial Neural Network Acta Inform Med. 2016; 24 (5): 274. doi: 10.5455/aim.2016.24.274-279 Li, Q, Chen, X, Wang, Y, et al. Analysis of the occurrence of deep venous thrombosis in lower extremity fractures: A clinical study. PAK J MED SCI. 2018; 34 (4): 828-832. doi: 10.12669/pjms.344.14752 Fei, W, Jian, Z, Zhi, G, et al. Analysis of risk factors for venous thromboembolism in patients after thoracic surgery: A clinical study of 167 cases. TURK GOGUS KALP DAMA. 2018; 26 (1): 93-98. doi: 10.5606/tgkdc.dergisi.2018.14980 Li, J, Ren, X, Zhu, X, et al. Clinical Predictive Factors of Lower Extremity Deep Vein Thrombosis in Relative High-Risk Patients after Neurosurgery: A Retrospective Study. DIS MARKERS. 2020; 2020 5820749. doi: 10.1155/2020/5820749 h.Gao, User manual for SAS/STAT software of SAS system ,29. Gismondi, RC, Almeida, RM, Infantosi, AF. Artificial neural networks for infant mortality modelling. COMPUT METH PROG BIO. 2002; 69 (3): 237-47. doi: 10.1016/s0169-2607(02)00006-8. Hoo, ZH, Candlish, J, Teare, D. What is an ROC curve? EMERG MED J. 2017; 34 (6): 357-359. doi: 10.1136/emermed-2017-206735 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4564132","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":323317996,"identity":"a965784a-0878-40ba-9bc3-172547624a6b","order_by":0,"name":"Zhongbin Zhou","email":"","orcid":"","institution":"The 6 th Medical Center of PLA General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zhongbin","middleName":"","lastName":"Zhou","suffix":""},{"id":323317997,"identity":"133e6a06-a204-4506-bd8e-4bb93a41f755","order_by":1,"name":"Yuan Yao","email":"","orcid":"","institution":"Hospital Management Institute of General Hospital, PLA","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Yao","suffix":""},{"id":323317998,"identity":"475a7397-ccd8-47d9-9df1-78b1c98d5088","order_by":2,"name":"Hanyu Zhou","email":"","orcid":"","institution":"University of Melbourne","correspondingAuthor":false,"prefix":"","firstName":"Hanyu","middleName":"","lastName":"Zhou","suffix":""},{"id":323317999,"identity":"e96cb490-09a3-4e94-bf40-24f9fbb3143d","order_by":3,"name":"Yuanyuan Qiao","email":"","orcid":"","institution":"The 6 th Medical Center of PLA General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yuanyuan","middleName":"","lastName":"Qiao","suffix":""},{"id":323318000,"identity":"6d983832-74e7-43d4-9237-5842f21b4a2d","order_by":4,"name":"Zhihan Gao","email":"","orcid":"","institution":"The 6 th Medical Center of PLA General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zhihan","middleName":"","lastName":"Gao","suffix":""},{"id":323318001,"identity":"db6d34dd-fea6-4777-9cc2-e304df056f7a","order_by":5,"name":"Yutao Guo","email":"","orcid":"","institution":"The 6 th Medical Center of PLA General Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yutao","middleName":"","lastName":"Guo","suffix":""},{"id":323318002,"identity":"0d7a0a4a-e57f-46b7-9a4b-bf8d3676d3a0","order_by":6,"name":"Ying Yang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIie3PsYrCQBCA4Qkra7MkbWLjPcJAIBwY4qsoga0WySNsZeUDXHEPkUfY3HBeo9imsBNsTBE7CwsVznaTUnD/cplvhwFwuV42BODjgzEtpll/EoCcV1+FzPsvirSKSbQ/nu78/29LJ1Hss9JskFI0DIb0W1rJZiEnAo95WS0LUrj3QUhZW4lRSSyQciRW3smRQSgSO9k1/2QNSJ9Inu4ktYoPd5JFqwES9CFR3STeN9IsCPmsWqHMedct/k7FbXOlKQ8ZtZdrmgVDWlvJhwE+EgBz/XzhtvFHYw3sfAGYdg26XC7XG3cDFWtT399iM/gAAAAASUVORK5CYII=","orcid":"","institution":"The 6 th Medical Center of PLA General Hospital","correspondingAuthor":true,"prefix":"","firstName":"Ying","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2024-06-11 12:36:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4564132/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4564132/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":60618961,"identity":"b6627563-0037-477b-8289-0c0e6aaa4286","added_by":"auto","created_at":"2024-07-18 20:40:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":72407,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe architecture of Artificial neural network.The 11 variable data were input into the input layer, and after the data processing, the information was input into the hidden layer, and then the hidden layer data was processed to output layer.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/5a2b916cdef4997c81ae772c.png"},{"id":60618958,"identity":"6db3325c-786c-4efb-a774-0d7add2fae7d","added_by":"auto","created_at":"2024-07-18 20:40:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":36321,"visible":true,"origin":"","legend":"\u003cp\u003eROC curve of LR model\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/08f861ee3e648197661cd96e.png"},{"id":60618957,"identity":"da54a4f8-a2c4-4297-b4f0-ffdbfcf687b9","added_by":"auto","created_at":"2024-07-18 20:40:51","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":36801,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eROC curve of LDR model\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/70a9d1a2283fd7ae1cbefc90.png"},{"id":60620964,"identity":"c471fa45-f299-4889-8236-7ef735ebebf1","added_by":"auto","created_at":"2024-07-18 20:56:51","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":45666,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe prediction accuracy of the SANNs model at different hidden_nodes and epochs\u003c/em\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/60c21ac4dbcd39de1c72190f.png"},{"id":60620025,"identity":"2ca988e2-17ef-432e-a4bd-4e31921591e9","added_by":"auto","created_at":"2024-07-18 20:48:51","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":42771,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe prediction accuracy of the ANNs model at different hidden_nodes and epochs\u003c/em\u003e\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/248e65df122cf159ed271bd1.png"},{"id":60618963,"identity":"29de5283-653b-488b-9429-76d6579fbd45","added_by":"auto","created_at":"2024-07-18 20:40:51","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":40833,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe Youden index of optimal threshold is higher than the other models.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/f1b335ab4852eb38bf8031a1.png"},{"id":65617783,"identity":"2f6f9075-de9f-40aa-8114-eac7ffeeb999","added_by":"auto","created_at":"2024-09-30 14:33:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":965684,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4564132/v1/efc2995e-4555-4044-be9d-7247b1bd1e73.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Artificial neural networks predict the incidence of deep venous thrombosis in hospitalized patient","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eDVT is a venous reflux disorder caused by abnormal coagulation of blood in the deep veins, and it mostly occurs in lower limb thrombosis\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Venous wall injury, slow blood flow, and hypercoagulability of blood might easily lead to DVT. The majority of patients with DVT have an insidious thrombosis, without obvious symptoms, while a few patients could have symptomatic deep vein thrombosis, mainly manifested by sudden swelling, stiffness, pain, tactile abnormalities, increased soft tissue tension, aggravated after activity and alleviated after the elevation of the affected limb.\u003c/p\u003e \u003cp\u003ePulmonary embolism (PE) and DVT are called venous thromboembolism (VTE), which have the same susceptibility factors and are two clinical manifestations of VTE at different sites and at different stages. Researches showed that 40%~60% of surgical patients and non-surgical patients were at risk of VTE, while the prevention proportion of high-risk groups was very low, even lower in Asian countries\u003csup\u003e[\u003cspan additionalcitationids=\"CR3 CR4\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Using various methods to early determine whether patients are at high risk of VTE and take timely preventive measures can reduce the incidence of VTE in patients\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. The prevention protocols of VTE in hospitals suggests the use of Caprini score scale or Padua score scale for VTE risk assessment for inpatients, especially inpatients in high-risk VTE departments. The two methods are simple and convenient to operate and play a great role in VTE prevention but slightly rough when compared with the computer-aided big data calculation. A recent study had shown that area under curve(AUC) (95%CI), sensitivity(Sen), specificity(Spe), accuracy(Acc), Youden index of Caprini Risk Assessment Scale were 0.596 (0.552, 0.638), 26.1%,96.5%,0.596, 0.23 respectively, but that of the VTE risk warning model were 0.960 (0.940, 0.976), 92.6%,91.8%,0.960,0.84 respectively, with statistically significant differences (Z\u0026thinsp;=\u0026thinsp;14.521, P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001)\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. The risk warning model outperformed the Caprini Risk Assessment Scale in predicting the occurrence of VTE in hospitalized patients.\u003c/p\u003e \u003cp\u003eTraditional risk warning models includes logistic regression, linear regression, and discriminant analysis, while ANNs represents the latest alternative methods to traditional modeling techniques\u003csup\u003e[\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. In recent years ANN has been widely applied in medical areas\u003csup\u003e[\u003cspan additionalcitationids=\"CR12 CR13 CR14 CR15 CR16 CR17\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. The ANNs systems has been extensively used to address complex real world problems, particularly when dealing with non-linear models or when the definition of underlying mechanisms is incomplete.\u003c/p\u003e \u003cp\u003eOnly a few studies applied the artificial neural network algorithm model for DVT prediction\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. Among them, the number of hidden_nodes and epochs in the model had fixed values, and how the changes of the number these parameters affected the prediction effect was not discussed. The purpose of this study was to develop the ANN-based predictive model to predict high-risk patients with DVT. Furthermore, we compared the prediction performance of ANNs with logistic regression, discriminant analysis, and SANNs\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e2.1. Materials\u003c/h2\u003e\n \u003cp\u003eHospital cases in our hospital from 1 January 2011 to 31 December 2020 were reviewed. Cases diagnosed with DVT during this hospitalization were considered as positive cases, and cases not diagnosed with DVT as negative cases. The diagnosis of DVT was mainly based on Doppler ultrasonography and venography. Referring to the previous research \u003csup\u003e[\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e and considering whether it could be directly extracted from the hospital information system, this study took 11 variables of patient sex, age, disease diagnosis, test, and surgical information as the variables of this study. The specific variables and data treatment were shown in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. There were 729 positive cases, all of which were used for model building and performance evaluation. There were 233,620 negative cases, from which 1,300 were randomly selected. Records with missing test data due to lack of lab tests during this hospitalization were excluded, and 566 cases were left for model building and performance evaluation. In total, 1295 data were included in this study, and the distribution of each variable among DVT and non-DVT patients is shown in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. The hospital Ethics Committee had approved the protocol with the ethics approval number of HZKY-PJ-2022-21. And the study entailed no intervention in patient care, informed consent was not required. The study was carried out according to the principles of The Declaration of Helsinki.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003e\u003cem\u003eCoding of variables.\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"3\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003cp\u003edescription\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCoding\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:female;1:male\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge(years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:\u0026lt;60;1:\u0026gt;=60\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDD(ng/ml)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eD-Dimer Antigen\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:\u0026lt;200;0.5:\u0026lt;500;1:\u0026gt;=500\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFIB(g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFibrinogen\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:\u0026lt;2;0.5:\u0026lt;4;1:\u0026gt;=4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHBP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypertension\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCHD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCoronary heart disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMalignant tumor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTRO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTrauma or operation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDiabetes Mellitus\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCHEMO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChemotherapy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCOPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eChronic obstructive pulmonary disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.01:no;1:yes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"char\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003e\u003cem\u003eDistribution of influencing factors\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"8\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003evariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003ewithout DVT(n\u0026thinsp;=\u0026thinsp;566)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003ewith DVT(n\u0026thinsp;=\u0026thinsp;729)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ecoding\u0026thinsp;=\u0026thinsp;0.01\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ecoding\u0026thinsp;=\u0026thinsp;0.5\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ecoding\u0026thinsp;=\u0026thinsp;1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ecoding\u0026thinsp;=\u0026thinsp;0.01\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ecoding\u0026thinsp;=\u0026thinsp;0.5\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ecoding\u0026thinsp;=\u0026thinsp;1\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e246\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e320\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e223\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e506\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e293\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e273\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e331\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e398\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e310\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e116\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e140\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e107\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e531\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFIB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e383\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e171\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e391\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e309\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHBP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e393\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e173\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e412\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e317\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCHD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e447\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e119\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e546\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e183\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e431\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e135\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e528\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e201\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTRO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e365\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e201\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e243\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e486\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e484\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e581\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e148\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCHEMO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e565\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e715\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCOPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e497\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e478\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e251\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e2.2. Setting training data and testing data\u003c/h2\u003e\n \u003cp\u003e75% of the cases (993 cases) were randomly selected as the training set for model construction, and the remaining 25% of the cases (302 cases) were used as the testing set to verify the prediction performance. The categorical percentage of each variable was calculated separately in the training set and testing set. Whether there were significant differences between the training set and testing set were assessed for a significant association by either Chi-square statistics or Fisher\u0026rsquo;s exact test.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e2.3. Logistic regression(LR)\u003c/h2\u003e\n \u003cp\u003eThe LR equations were constructed by the stepwise process. The first was to introduce the variables of the training set into the equation one by one, requiring its Chi-square value to be significant at the SLENTRY(Specifies the significance level threshold for the variables introduced in the forward selection method) level. After introducing a variable, all variables already included in the equation were tested, removing variables that were not significant at the SLSTAY(Specifies the significance level threshold for retained variables in the backward removal method) level. All nonsignificant variables were removed before reintroducing new variables. When all variables outside the equation were not significant at the SLENTRY level and all variables within the equation with Wald Chi-Square were significant at the SLSTAY level, the stepwise process were stopped. In addition, variables that were just removed were again introduced into the equation, and the stepwise process was also stopped. Both the SLENTRY and SLSTAY values in this stepwise process were set to 0.15. After the regression equations were constructed, the equation prediction performance was evaluated by testing set data using MedCalc.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e2.4. Discriminant analysis\u003c/h2\u003e\n \u003cp\u003eThe entry model variables were first determined by stepwise discriminant analysis(SDA), and then linear discriminant analysis(LDA) was used to construct the prediction equation. SDA introduced variables into the model one by one, requiring the variable WILKS\u0026apos;S LAMBDA(The ratio of the within-group sum of squares to the total sum of squares) value to be significant at the SLENTRY level. As the variable gradually increased in the model, the discriminative power of the variables introduced earlier could also change, removing the variable if the variable in the model was not significant at the SLSTAY level. This process was repeated until all variables of the model were significant at the SLSTAY level and the others did not meet the criteria for entering the model. Both the SLENTRY and SLSTAY values in this process were set to 0.15. After the regression equations were constructed, the equation prediction performance was evaluated by testing set data using MedCalc .\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003e2.5. Artificial neural networks(ANNs)\u003c/h2\u003e\n \u003cp\u003eThe ANNs is divided into three layers, one is the input layer, one is hidden layer, and the last one is the final output layer. Each layer has one or more nodes, as shown in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. The possible influencing factors of DVT are input in the input layer. Each node of the input layer is connected to each node of the hidden layer, and the connection weights between nodes are different. The input layer inputs information to the hidden layer through the connection between nodes. The information inputs to the hidden layer is processed by the sigmoid activation function \u003cimg src=\"data:image/png;base64,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\"\u003e\u0026nbsp;before the information is output. The sigmoid function is to suppress the output signal if the combined signal input from the input layer to the hidden layer is not strong enough. But if it is strong enough,the effect of sigmoid function is to stimulate the output signal. Each node of the hidden layer is also connected to each node of the output layer. The information of the hidden layer is input to the output layer through connections with different weights. The information of the output layer is also processed by the sigmoid activation function to get the risk of having a DVT.\u003c/p\u003e\n \u003cp\u003eAfter building the model architectures, the next step is to determine the connection weights between the input layer and the hidden layer, the hidden layer and the output layer through the training process. The training process is divided in two steps. In the first step, an initial random weight is created (sampling weight using a normal probability distribution, where the mean is 0 and the standard variance is the square root of the number of nodes) and the model calculates the output for a given training sample. In the second step, the difference between the model output and the original data of the training sample guides the connection weights to update by the gradient descent method. The following is the weight update matrix algorithm.\u003c/p\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/span\u003e\u003c/p\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003c/span\u003eThe training set (including 11 variables) was input into the input layer to construct the ANNs model. Testing set was used to verify the prediction accuracy of the constructed neural network model and to select the optimal neural network model.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e2.6. Simple artificial neural networks(SANNs)\u003c/h2\u003e\n \u003cp\u003eTo study the prediction performance of the ANNs model with a small number of input nodes, the six variables selected by the LR model were used as input variables to develop the \u0026ldquo;simplified\u0026rdquo; ANNs model. The data set and training process were the same as the ANNs model with all eleven input variables.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e2.7. Statistical analysis\u003c/h2\u003e\n \u003cp\u003eIn this research ,data analysis and statistics were performed with the commercial software SAS (SAS for Windows, Version 9.3, SAS Inc.). Significant differences between different groups were evaluated by chi-square analysis and unpaired Student t-test and ROC analysis were performed by using the medcalc software(Version 19.0.4). P values\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered statistically significant. The ANNs models were performed with Python 3.6. Prediction indicators included Sen, Spe, Acc, Youden index, AUC.\u003c/p\u003e\n \u003cp\u003eAcc = (Sen\u0026thinsp;+\u0026thinsp;Spe) / (Sen\u0026thinsp;+\u0026thinsp;Spe\u0026thinsp;+\u0026thinsp;Fpr (false positive rate)\u0026thinsp;+\u0026thinsp;Fnr(false negative rate))\u003c/p\u003e\n \u003cp\u003eYouden index=(Sen\u0026thinsp;+\u0026thinsp;Spe) -1.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Result","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1. Distribution of variables in the training set and testing set\u003c/h2\u003e\n \u003cp\u003eThe incidence of DVT in the training and testing sets were 57.3% and 53.0% respectively, with no significant(p\u0026thinsp;=\u0026thinsp;0.1850) differences between the two sets. 11 variables also showed no significant difference in the training and testing sets. It was shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e. It meant that the two sets were well balanced in the distribution of clinical characteristics.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003e\u003cem\u003eCharacteristics of patients.\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTraining set\u003c/p\u003e\n \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;993)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTesting set\u003c/p\u003e\n \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;302)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-Value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDVT(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e57.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e53.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1850\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender: male(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e59.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0968\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge: \u0026gt;=60(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e59.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.9287\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDD: \u0026gt;500(%),\u0026lt;200(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e52.4,29.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50.0,34.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2238\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFIB: \u0026gt;4(%),\u0026lt;2(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e38.0,3.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e34.1,3.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.4450\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHBP(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e38.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.8633\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCHD(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.9470\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMT(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.7239\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTRO(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e53.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2796\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDM(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e17.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2055\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCHEMO(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2396\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCOPD(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e21.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1055\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2. Results of LR\u003c/h2\u003e\n \u003cp\u003eThe results of LR analysis were summarized in Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. There were six variables included in the final LR model, the probability of VTE could be calculated by the following logistic equations:\u003c/p\u003e\n \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n \u003cp\u003e=\u0026thinsp;2.2656\u0026ndash;0.4409\u0026times;(HBP)\u0026thinsp;+\u0026thinsp;0.3428\u0026times;(MT)-1.2073\u0026times;(TRO)-1.3056\u0026times;(COPD)-2.0844\u0026times;(DD)-0.4817\u0026times;(AGE)。\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003e\u003cem\u003eThe results of LR\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"6\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eParameter\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDF\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eEstimate\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eStandard\u003c/p\u003e\n \u003cp\u003eError\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eWald\u003c/p\u003e\n \u003cp\u003eChi-Square\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePr\u0026gt;\u003c/p\u003e\n \u003cp\u003eChiSq\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntercept\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.2656\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1973\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e131.9156\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHBP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.4409\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.5542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0105\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.3428\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1826\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.5236\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0605\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTRO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-1.2073\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1630\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e54.8816\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCOPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-1.3056\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2012\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40.1013\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-2.0844\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1866\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e124.8458\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAGE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.4817\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1705\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7.9779\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0047\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eAnalyzing the ROC curve of the LR model, the Sen index was 91.87%, the Spe index was 69.72%, and the AUC index was 0.870. It was shown in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3. Results of LDA\u003c/h2\u003e\n \u003cp\u003eThe results of LDA were summarized in Table \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e. There were six variables included in the final SDA model. The probability of VTE could be calculated by the following LDA equations:\u003c/p\u003e\n \u003cp\u003eProbability=-4.1665\u0026thinsp;+\u0026thinsp;1.4459\u0026times;(HBP)\u0026thinsp;+\u0026thinsp;0.3877\u0026times;(MT)\u0026thinsp;+\u0026thinsp;2.8786\u0026times;(TRO)\u0026thinsp;+\u0026thinsp;1.6916\u0026times;(COPD)\u0026thinsp;+\u0026thinsp;4.4186\u0026times;(DD)\u0026thinsp;+\u0026thinsp;2.2095\u0026times;(AGE)\u003c/p\u003e\n \u003cp\u003eAnalyzing the ROC curve of the LDA model, the Sen was 88.12%, the Spe was 73.94%, and AUC was 0.872. It was shown in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab5\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003e\u003cem\u003eThe results of discriminant analysis\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"8\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eParameter\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePartial R-Square\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eF value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePr\u0026thinsp;\u0026gt;\u0026thinsp;F\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eWilks\u0026rsquo;\u003c/p\u003e\n \u003cp\u003elambda\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePr\u0026lt;\u003c/p\u003e\n \u003cp\u003eLambda\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eASCC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePr\u0026gt;\u003c/p\u003e\n \u003cp\u003eASCC\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2231\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e284.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.7769\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2231\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCOPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0447\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e46.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.7422\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2578\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTRO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0508\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e52.959\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.7045\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2955\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHBP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0142\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6944\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.3055\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAGE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0069\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0090\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6897\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.3103\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0035\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0627\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6827\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.3128\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e3.4. Results of SANNs\u003c/h2\u003e\n \u003cp\u003eThe prediction accuracy of the SANNs model at different hidden_nodes and epochs was shown in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. When the number of hidden_nodes was 7 and the number of epochs was 800, the accuracy reached the highest,with an average of 80.11% .\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e3.5. Results of ANNs\u003c/h2\u003e\n \u003cp\u003eThe prediction accuracy of the ANNs model at different hidden_nodes and epochs was shown in Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e. When the number of hidden_nodes was 8 and the number of epochs was 800, the accuracy reached the highest, with an average of 81.84%.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e3.6. Comparison of the various prediction results\u003c/h2\u003e\n \u003cp\u003eThere were high Sen but low Spe in LR and LDA model,the Sen of SANNs model was lower than that of LDA and LR model,but the Spe was higher than both of them.The Sen and Spe of ANNs model were higher than that of SANNs. The Youden index and Acc of ANNs model was higher than all other prediction models. The prediction performance of each model was shown in Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e and Table \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab6\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003e\u003cem\u003eThe results of ROC curve analysis\u003c/em\u003e\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"6\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSen\u003c/p\u003e\n \u003cp\u003e(%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpe\u003c/p\u003e\n \u003cp\u003e(%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eYouden index\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAcc\u003c/p\u003e\n \u003cp\u003e(%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003cp\u003e(mean\u0026thinsp;\u0026plusmn;\u0026thinsp;S.D.)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e91.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e69.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6159\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e79.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.869\u0026thinsp;\u0026plusmn;\u0026thinsp;0.021\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLDA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e88.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e73.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6207\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e80.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.872\u0026thinsp;\u0026plusmn;\u0026thinsp;0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSANNs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e84.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e76.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6114\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e80.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.869\u0026thinsp;\u0026plusmn;\u0026thinsp;0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANNs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e85.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e78.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.6450\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e81.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.870\u0026thinsp;\u0026plusmn;\u0026thinsp;0.021\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThere were many factors affecting the occurrence of DVT, including age,malignant tumor, previous medical history, family medical history, BMI, trauma, surgical operation, abnormal pulmonary function, bedridden, severe pulmonary disease, uncontrolled hypertension etc. Hopefully, appropriate methods can be applied to construct the DVT prediction models to identify high-risk patients at the early stage and take timely preventive measures for them to reduce the time to cure and reduce patient mortality.\u003c/p\u003e \u003cp\u003eIn recent years, some scholars had studied the occurrence of DVT. Qiang Li et al. found old age, longer operation time, and arthroplasty were independent risk factors, physical labor and postoperative exercises were protective factors for DVT in patients with lower extremity fractures\u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e.Wang Fei Li et al. found age, smoking history, operation time and body mass index are independent risk factors for venous thromboembolism\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. This study found age, d-dimer, Trauma or operation, Chronic obstructive pulmonary disease, Hypertension, and Malignant tumor were risk factors for DVT, which is basically consistent with the findings of other scholars. For the high-risk patients of DVT predicted by the model,doctors can take basic prevention (such as early going out of bed, avoiding dehydration, etc), drug prevention, mechanical prevention, placement of recyclable vena cava filter to prevent their occurrence.They can also develop more detailed DVT prevention methods based on the ANNS model.\u003c/p\u003e \u003cp\u003eFew reports have focused on predicting the occurrence of VTE. Two recent studies have used logistic regression to predict whether VTE occurs.Juhua Li et al. built a simple and efficient diagnostic factor for DVT in patients after Neurosurgery, which was a reliable predictive index for DVT in patients \u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e.Chen Shen et al. developed VTE risk warning model,which had high accuracy in predicting VTE occurrence in hospitalized patients\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. Laleh Agharezaei et al. built ANNs to help experts diagnose and predict the risk level of pulmonary embolism in patients \u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e.In this study, we compared the abilities of LR model, LDA model and ANNs model to predict the occurrence of VTE. We evaluated the performance of these models by calculating Youden index and Acc.\u003c/p\u003e \u003cp\u003eLR is a widely used algorithm, efficient and fast, unnecessary to scale the input features.But its disadvantage is that it can not be used to solve nonlinear problems.Press and Willson concluded by comparing LR and LDA that when these variables were not multivariate normal distribution in the class,LR was preferable to LDA. However ,if the independent variable belonged to multivariate normal distribution, LR was more inefficient. Efron gave two normal populations with common covariance, the efficiency of the LR was only between 1/2 or 2/3 of the efficiency of LDA\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e with about the same misjudgment rate. Compared with the two methods.ANNs have no requirements for independent variable and good ability to identify complex nonlinear relationships among variables. It is particularly suitable for processing information that should take many factors and conditions into considerations at the same time\u003csup\u003e[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe area under the ROC curve (AUC) is a global measure to discriminate whether a specific condition is present or not present.In general, if its AUC value is greater than 0.85,indicates that the model has a better predictive efficacy.In this study, the AUC values of the LR ,LDA, SANNs, and ANNs models were greater than 0.85, indicating that the predictive efficacy of all the models was good.When selecting an optimal threshold (or cut-off point) of the ROC curve, we need to consider the aims of the diagnostic test, considering the significance and costs of a false-positive or false-negative interpretation. A commonly used approach when selecting a cut-off point is to give equal weight to the importance of sensitivity and specificity by choosing the point nearest to the top-left most corner of the ROC curve. This point is also known as the Youden Index\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e. This research showed that the Youden index and Acc of the ANNs model were 0.6450 and 81.84%, which were both higher than the other models. ANNS model is better predicted than the other models.\u003c/p\u003e \u003cp\u003eANNs model construction requires a process of setting hidden_node and epochs of the neural network, and the number of hidden_nodes and epochs has a great influence on the prediction efficiency of the model. Very few reports had focused on it. We explored the number setting of hidden_nodes and epochs of the neural network in this paper.The number of hidden_nodes of the neural network is related to the number of input_nodes and output_nodes, the complexity of the problem, and the distribution of the data.The basic methods to determine the number of hidden_nodes is to take the most compact structure as possible while satisfying the accuracy requirements, that is, to take the number of hidden_nodes as few as possible. Following this methods, the number of hidden_nodes was 8 in this study. Increasing the number of epochs of the neural network can improve the accuracy, but after reaching a limited value, if it continues to increase, there will be overfitting, and the accuracy may also decrease. When the number of epoch in this study reached around 800, the accuracy is not improved.\u003c/p\u003e \u003cp\u003eIn this study, the prediction accuracy of SANNs model and Youden index were both lower than ANNs. It indicated that the variables screened by the LR model did not contribute significantly(p\u0026thinsp;\u0026gt;\u0026thinsp;0.15) to the DVT, but some of the information could be identified by the ANNs model, which improved the accuracy of their prediction.\u003c/p\u003e \u003cp\u003eANNs is an analysis method with high prediction efficient with good variable identification ability. It is able to process multifactorial, imprecise, and ambiguous information simultaneously. Due to the limitations of this research, part possible influencing factors were not included in the ANNs model for DVT prediction. However, in practice, the influencing factors could be added or constantly adjusted, and the prediction accuracy should become higher. This research provided good evidence for applying ANNs to predict DVT. ANNs will be promising in DVT prediction applications.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Medical Ethics Committee of the 6th Medical Center of PLA General Hospital approved this study with a specific ID HZKY-PJ-2022-21. The Medical Ethics Committee of the 6th Medical Center of PLA General Hospital waived informed consent given the lack of intervention and the anonymity of the data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data used to support the findings of this study are restricted by The 6th Medical Center of PLA General Hospital in order to protect patient privacy. Data are available from The 6th Medical Center of PLA General Hospital for researchers who meet the criteria for access to confidential data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNo funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eZhongbin Zhou \u0026nbsp; wrote the main manuscript text,Hanyu Zhou \u0026nbsp;processed and analyzed data，All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis wok was not supported by outside funds.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAbdel-Razeq, H, Qari, M, Kristensen, J, et al. Guidelines for diagnosis and treatment of deep venous thrombosis and pulmonary embolism. Methods Mol Med. 2004; 93 267-92. doi: 10.1385/1-59259-658-4:267\u003c/li\u003e\n \u003cli\u003eZhang, Y, Yang, Y, Chen, W, et al. Prevalence and associations of VTE in patients with newly diagnosed lung cancer. CHEST. 2014; 146 CHEST. doi: 10.1378/chest.13-2379\u003c/li\u003e\n \u003cli\u003eHill, J, Treasure, T. Reducing the risk of venous thromboembolism in patients admitted to hospital: summary of NICE guidance. BMJ. 2010; 340 c95. doi: 10.1136/bmj.c95\u003c/li\u003e\n \u003cli\u003eQu, H, Li, Z, Zhai, Z, et al. Predicting of Venous Thromboembolism for Patients Undergoing Gynecological Surgery. MEDICINE. 2015; 94 (39): e1653. doi: 10.1097/MD.0000000000001653\u003c/li\u003e\n \u003cli\u003eCohen, AT, Tapson, VF, Bergmann, JF, et al. Venous thromboembolism risk and prophylaxis in the acute hospital care setting (ENDORSE study): a multinational cross-sectional study. LANCET. 2008; 371 (9610): 387-94. doi: 10.1016/S0140-6736(08)60202-0\u003c/li\u003e\n \u003cli\u003eZhai, Z, Wang, C. [Establishing and improving venous thromboembolism prevention and management system in hospital]. Zhonghua Yi Xue Za Zhi. 2015; 95 (30): 2417-8. PMID: 26711198\u003c/li\u003e\n \u003cli\u003eShen, C, Ge, B, Liu, X, et al. Predicting the occurrence of venous thromboembolism: construction and verification of risk warning model. BMC Cardiovasc Disord. 2020; 20 (1): 249. doi: 10.1186/s12872-020-01519-9\u003c/li\u003e\n \u003cli\u003eXiaoyu, D, Kai, C, Zhihui, H, et al. Predictive value of preoperative erythrocyte electrophoresis exponent for acute deep vein thrombosis after total knee arthroplasty in patients with knee osteoarthritis. J Orthop Surg Res. 2020; 15 (1): 496. doi: 10.1186/s13018-020-02020-x\u003c/li\u003e\n \u003cli\u003eZhang, Y, Cao, M, Ren, J. NLR value and IL-18 level and their clinical significance in patients with deep vein thrombosis after receiving the surgery for spinal degeneration. Am J Transl Res. 2021; 13 (6): 7156-7163. PMID: 34306476\u003c/li\u003e\n \u003cli\u003eManiruzzaman, M, Kumar, N, Menhazul Abedin, M, et al. Comparative approaches for classification of diabetes mellitus data: Machine learning paradigm. COMPUT METH PROG BIO. 2017; 152 23-34. doi: 10.1016/j.cmpb.2017.09.004\u003c/li\u003e\n \u003cli\u003eMohammadfam, I, Soltanzadeh, A, Moghimbeigi, A, et al. Use of Artificial Neural Networks (ANNs) for the Analysis and Modeling of Factors That Affect Occupational Injuries in Large Construction Industries. Electron Physician. 2015; 7 (7): 1515-22. doi: 10.19082/1515\u003c/li\u003e\n \u003cli\u003eFei, Y, Hu, J, Li, WQ, et al. Artificial neural networks predict the incidence of portosplenomesenteric venous thrombosis in patients with acute pancreatitis. J THROMB HAEMOST. 2017; 15 (3): 439-445. doi: 10.1111/jth.13588\u003c/li\u003e\n \u003cli\u003eFei, Y, Gao, K, Li, WQ. Artificial neural network algorithm model as powerful tool to predict acute lung injury following to severe acute pancreatitis. PANCREATOLOGY. 2018; 18 (8): 892-899. doi: 10.1016/j.pan.2018.09.007\u003c/li\u003e\n \u003cli\u003eFei, Y, Gao, K, Li, WQ. Prediction and evaluation of the severity of acute respiratory distress syndrome following severe acute pancreatitis using an artificial neural network algorithm model. HPB. 2018; 21 (7): 891-897. doi: 10.1016/j.hpb.2018.11.009\u003c/li\u003e\n \u003cli\u003eWillan, J, Katz, H, Keeling, D. The use of artificial neural network analysis can improve the risk-stratification of patients presenting with suspected deep vein thrombosis. BRIT J HAEMATOL. 2019; 185 (2): 289-296. doi: 10.1111/bjh.15780\u003c/li\u003e\n \u003cli\u003eDamiani, G, Conic, RRZ, Pigatto, PDM, et al. Predicting Secukinumab Fast-Responder Profile in Psoriatic Patients: Advanced Application of Artificial-Neural-Networks (ANNs). J DRUGS DERMATOL. 2020; 19 (12): 1241-1246. doi: 10.36849/JDD.2020.5006\u003c/li\u003e\n \u003cli\u003eQiu, Y, Wang, T, Yang, X, et al. Development and Validation of Artificial Neural Networks for Survival Prediction Model for Patients with Spontaneous Hepatocellular Carcinoma Rupture After Transcatheter Arterial Embolization. Cancer Manag Res. 2021; 13 7463-7477. doi: 10.2147/CMAR.S328307\u003c/li\u003e\n \u003cli\u003eUmar, M, Sabir, Z, Raja, MAZ, et al. Numerical Investigations through ANNs for Solving COVID-19 Model. Int J Environ Res Public Health. 2021; 18 (22): doi: 10.3390/ijerph182212192\u003c/li\u003e\n \u003cli\u003eAgharezaei, L, Agharezaei, Z, Nemati, A, et al. The Prediction of the Risk Level of Pulmonary Embolism and Deep Vein Thrombosis through Artificial Neural Network Acta Inform Med. 2016; 24 (5): 274. doi: 10.5455/aim.2016.24.274-279\u003c/li\u003e\n \u003cli\u003eLi, Q, Chen, X, Wang, Y, et al. Analysis of the occurrence of deep venous thrombosis in lower extremity fractures: A clinical study. PAK J MED SCI. 2018; 34 (4): 828-832. doi: 10.12669/pjms.344.14752\u003c/li\u003e\n \u003cli\u003eFei, W, Jian, Z, Zhi, G, et al. Analysis of risk factors for venous thromboembolism in patients after thoracic surgery: A clinical study of 167 cases. TURK GOGUS KALP DAMA. 2018; 26 (1): 93-98. doi: 10.5606/tgkdc.dergisi.2018.14980\u003c/li\u003e\n \u003cli\u003eLi, J, Ren, X, Zhu, X, et al. Clinical Predictive Factors of Lower Extremity Deep Vein Thrombosis in Relative High-Risk Patients after Neurosurgery: A Retrospective Study. DIS MARKERS. 2020; 2020 5820749. doi: 10.1155/2020/5820749\u003c/li\u003e\n \u003cli\u003eh.Gao, User manual for SAS/STAT software of SAS system ,29.\u003c/li\u003e\n \u003cli\u003eGismondi, RC, Almeida, RM, Infantosi, AF. Artificial neural networks for infant mortality modelling. COMPUT METH PROG BIO. 2002; 69 (3): 237-47. doi: 10.1016/s0169-2607(02)00006-8.\u003c/li\u003e\n \u003cli\u003eHoo, ZH, Candlish, J, Teare, D. What is an ROC curve? EMERG MED J. 2017; 34 (6): 357-359. doi: 10.1136/emermed-2017-206735\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Deep venous thrombosis, artificial neural network, logistic regression, discriminant analysis, forecasting","lastPublishedDoi":"10.21203/rs.3.rs-4564132/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4564132/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective: \u003c/strong\u003eTo construct and validate artificial neural networks (ANNs) for predicting the occurrence of deep venous thrombosis(DVT) and compare the predictive performance of the ANNs model with that of logistic regression(LR)model,linear discriminant analysis(LDA) model,and simple artificial neural network (SANNs) model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003e1295 cases were selected, including 729 patients with DVT and 566 patients without. 75% of the cases (993 cases) are randomly selected as the training set for model construction, and the remaining 25% of the cases (302 cases) are used as the testing set to verify the prediction performance. After deep learning of the training data, the ANNs model with different numbers of hidden_nodes and epochs was constructed. The prediction efficiency of the ANNs model was tested by comparing the results of LR,LDA,and SANNs model as the benchmark afterwards.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eWhen the number of hidden_nodes was 8 and the number of epochs was 800 in ANNS model, the Acc reached the highest,which the Acc, Youden index was 81.84%, 0.6450 respectively.The prediction performance of this model was higher than that of LR,LDA ,and SANNs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions: \u003c/strong\u003eThis study provided good evidence for the application of ANNs to predict DVT in a large number of data. However, more research will be needed to confirm its application in the prediction of DVT.\u003c/p\u003e","manuscriptTitle":"Artificial neural networks predict the incidence of deep venous thrombosis in hospitalized patient","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-18 20:40:46","doi":"10.21203/rs.3.rs-4564132/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"45f62f02-a4df-4280-b952-d8387247ed19","owner":[],"postedDate":"July 18th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-01-15T09:38:23+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-18 20:40:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4564132","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4564132","identity":"rs-4564132","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}