Risk Factors Of Postoperative Arrhythmia In Lung Cancer Patients Based On Machine Learning Method

Purpose:1.To study the status quo of postoperative arrhythmia in patients with lung cancer and explore its related risk factors;2.To explore the application value of machine learning algorithm model in predicting arrhythmia after lung cancer surgery,so as to provide effective means and tools for clinical risk assessment.Method:In this study,data related to 234 patients with non-small cell lung cancer(NSCLC)who met the inclusion and exclusion criteria in the thoracic surgery Department of the First Affiliated Hospital of Nanchang University from June 2022 to December 2022 were collected.SPSS 26.0 statistical software was used for univariate and multivariate analysis of each observation index collected.Univariate analysis,Wilcoxon rank sum test was used for measurement data,chi-square test was used for counting data.Multivariate analysis,binary Logistic regression algorithm was used to analyze the results,the independent risk factors of postoperative arrhythmia of lung cancer were output,and a simple machine learning Logistic regression risk classification prediction model was built.P<0.05 indicated that the difference was statistically significant.Based on python language of machine learning algorithm of decision tree(DT),random forests(RF)and support vector machine(SVM)classifying caused by cardiac arrhythmia after lung cancer,build three other risk classification prediction model.According to the model output prediction of risk classification,by calculating the Area under the ROC curve(Area under the curve,AUC)and accuracy(ACC),Precision(Precision),and the Recall rate(Recall)and f-score,to assess the predictive performance of the model.Results:1.Current situation analysisIn this study,through the analysis of the status quo of 234 patients with arrhythmia within 48-72 hours after lung cancer surgery,ECG report results showed that the number of patients with arrhythmia after lung cancer surgery was 44,accounting for 18.8%of the total number.Among them,the main types of postoperative arrhythmias include atrial fibrillation,atrial tachycardia,atrial premature beat,ventricular tachycardia,ventricular premature beat,bundle branch block,cardiac arrest and sinus arrhythmia.26 patients with atrial arrhythmia accounted for 11.1%of the total number,59.1%of the total number of arrhythmias,12 patients with atrial fibrillation accounted for 5.1%of the total number,27.3%of the total number of arrhythmias.2.Analysis of risk factorsUnivariate analysis showed that age≥60 years old,smoking,coronary heart disease,preoperative chemotherapy,preoperative electrocardiogram,intraoperative blood transfusion,postoperative electrolyte,surgical method,surgical scope,mediastinal lymph node dissection,lymph node metastasis,pathological type of squamous cell carcinoma,pathological stage,operation duration,BNP,and FEV1%,and there were statistically significant differences in 16 variables(P<0.05),is a risk factor associated with lung cancer after surgery.Multivariate Logistic regression analysis showed that age≥60 years old,coronary heart disease,preoperative electrocardiogram,intraoperative blood transfusion,operation duration,BNP,a total of 6 independent risk factors for postoperative arrhythmia of lung cancer(P<0.05),FEV1%was an independent protective factor for arrhythmia after lung cancer surgery.3.The accuracy(ACC)of Logistic regression model,decision tree model,random forest model and support vector machine model were 88%,81%,94%and 92%,respectively,and the Recall rates were 79%,0.76%,90%and 85%,respectively.The Precision was 83%,71%,86%,88%,and the area under receiver operating characteristic curve(ROC)was 0.90,0.83,0.95,and 0.94,respectively.The recognition accuracy of random forest and support vector machine is higher than that of Logistic regression and decision tree.The accuracy of random forest and support vector machine was higher than that of Logistic regression and decision tree in the prediction and classification model of arrhythmia after lung cancer surgery.The stochastic forest prediction model has the best performance and the decision tree has the worst performance.4.The high incidence of arrhythmia after lung cancer surgery is mainly related to age≥60 years old,coronary heart disease,preoperative electrocardiogram,intraoperative blood transfusion,operation duration,FEV1%,BNP and other factors.The random forest model is better than the support vector machine model,decision tree model and Logistic regression model in predicting the accuracy of arrhythmia after lung cancer surgery,and the Logistic regression model is better than other models in explaining the variables.Therefore,practical application,can cooperate with each other.Conclusion:1.Arrhythmia is one of the common complications after radical thoracic surgery for lung cancer.In this study,the incidence of postoperative arrhythmia in lung cancer patients was 18.8%,atrial arrhythmia was the most common,accounting for 11.1%of the total number,and the incidence of postoperative atrial fibrillation was 5.1%.2.Age≥60 years old,coronary heart disease,preoperative electrocardiogram,intraoperative blood transfusion,operation duration,BNP were independent risk factors for postoperative arrhythmia,and FEV1%was independent protective factor.3.Among the risk classification prediction models of machine learning algorithms used in this study,Logistic regression,decision tree,random forest and support vector machine can all accurately predict whether there is a complicated arrhythmia after lung cancer surgery.In terms of prediction accuracy,random forest model has the best performance,followed by support vector machine model,Logistic regression model and decision tree model respectively.4.Since the Logistic regression model can assign values to variables and output the predicted probabilities of each variable,it has the ability to explain risk variables,and can more intuitively describe the complex network risk mechanism between postoperative arrhythmia and risk factors in lung cancer patients.In clinic,Logistic regression model and random forest model can be combined to provide powerful and effective clinical evaluation tools and means for preoperative prediction and screening of lung cancer patients.Thus,this reduces the incidence of concurrent arrhythmia after lung cancer,reduces the rate of hospitalization of patients,reduces stroke and mortality,improves the long-term prognosis of patients,and has certain clinical significance.