Construction Of Prediction Model For Rupture Risk Of Anterior Communicating Aneurysm Based On Nomogram And Machine Learning

Objective:By rupture were retrospectively analyzed before the traffic before aneurysm patients with unruptured aneurysms in patients with general clinical data and the morphological features of aneurysms using Logistic regression method and machine learning methods,obtain independent risk factors for the development of traffic before aneurysm rupture,and establish the former traffic aneurysm rupture risk prediction model,the decision to provide help and reference for clinical treatment.Methods:The information of patients with anterior communicating aneurysm who visited the Affiliated Hospital of Qingdao University from December 2012 to January 2021 was retrieved by using the big data platform of clinical scientific research.Then,the general clinical characteristics of the patients were queried and measured using the electronic medical record system,and then the CTA imaging morphological characteristics related to anterior communicating aneurysm were measured and collected on the hospital imaging system.R software(version 4.0.4)was used to process and analyze the collected data of clinical and morphological characteristics of patients.The patients’ characteristic data were compared and analyzed in general data,and clinical characteristics and morphological characteristics were included as predictive factors.The traditional logistic regression,classification tree in machine learning,random forest,artificial neural network and integrated model were used respectively to establish a prediction model for anterior communicating aneurysm rupture.The accuracy,sensitivity,specificity and other indexes of the prediction models were calculated.Receiver operating characteristic(ROC)curves were drawn respectively,and the area under the curve(AUC)of several prediction models was compared.Results: A total of 436 patients with anterior communicating aneurysm that met the criteria were included in this study.The patients included in the study were divided into the ruptured group and the unruptured group according to whether the aneurysm ruptured or not.A total of 263 patients with ruptured anterior communicating aneurysm causing subarachnoid hemorrhage and 173 patients with unruptured anterior communicating aneurysm were included in the ruptured group.There was no statistical significance in the following characteristics between the two groups: sex ratio,smoking history,history of hypertension,aneurysm orientation,and history of subarachnoid hemorrhage.There were statistically significant differences in age,aneurysm height,aneurysm neck width,irregular shape and A1 dominant sign between the two groups,which were also independent risk factors for anterior communicating aneurysm rupture after multivariate Logistic regression analysis.The age of patients in the ruptured group was 57.30±10.18 years,and that in the non-ruptured group was 60.53±9.80 years.Patients in the ruptured group were younger than those in the non-ruptured group,P=0.001.The height of aneurysm in the ruptured group was 5.74±2.77 mm,while that in the non-ruptured group was4.68±2.59 mm.The height of aneurysm in the ruptured group was significantly higher than that in the non-ruptured group,P<.0.001.The neck width of aneurysm aneurysm aneurysm was smaller in ruptured group than in non-ruptured group(2.71±1.06 mm in ruptured group;2.97±1.48 mm in the unruptured group),P=0.033.Among the ruptured group,there were 150 patients with irregular aneurysm shape,accounting for 57.0%,and 52 patients with irregular aneurysm shape in the unruptured group,accounting for30.1%.The proportion of patients with irregular aneurysm shape in the ruptured group was significantly higher than that in the unruptured group,P<0.001.There were176 patients with A1 dominant sign in the rupture group,accounting for 66.9%,and81 patients with A1 dominant sign in the non-rupture group,accounting for 46.8%.A higher proportion of patients with A1 dominant sign in the rupture group,P<0.001,the difference was statistically significant.The enrolled patients were divided into training set and test set in a 7:3 ratio according to the random sampling method.The prediction model was established in the training set data and verified in the test set.After logistic regression established the prediction model and applied it to the test set,the area under the ROC curve(AUC)was 0.781(95%CI: 0.690-0.871),and the model had the best performance at the cut-off value of 0.436.Under this cut-off value,the accuracy of prediction in the test set was 0.782(95%CI: 0.699-0.851),the sensitivity was 0.833,the specificity was0.711,and the Kappa value of consistency test was 0.549.According to the relevant research,the classification tree,random forest,XGBoost and artificial neural network methods in machine learning were selected to build the prediction model,and applied to the test set.The areas under the ROC curve(AUC)were 0.716,0.733,0.774 and 0.733,respectively.By comparison,in the test set,the prediction model established by the above machine learning method and the prediction model established by the logistic regression method have similar performance.Then several underlying models were included to build an integrated model,which was also verified in the test set.The area under the ROC curve(AUC)of the prediction model established by the integrated model method was 0.810(95%CI:0.725-0.894),and the optimal truncation value was 0.331.Under this cutoff value,the accuracy,sensitivity,specificity and consistency test Kappa value in the test set were0.774(95%CI: 0.690,0.844),0.722,0.846,and 0.550.The area under the ROC curve(AUC)was compared using the De Long test.Compared with the traditional logistic regression prediction model,the integrated model established by machine learning method had better performance,and the difference was statistically significant(P<0.05).Conclusion: Age,height of aneurysm,neck width,A1 dominant sign and irregular shape of aneurysm are independent risk factors for rupture of anterior communicating aneurysm.The establishment of a prediction model by multivariate Logistic regression and the drawing of a column chart can better predict the rupture risk of anterior communicating aneurysm.Using the classification tree and artificial neural network method in machine learning,the model performance is similar to that of Logistic regression.The integrated model method in machine learning can be used to construct the prediction model with better efficiency than the traditional logistic regression method.In the face of unruptured anterior communicating aneurysm,this study can provide some reference and help for treatment decision.