Objective To construct an artificial neural network model to predict postoperative blurred vision in patients undergoing general anesthesia,aiming to provide a reference for clinical preventive measures.Methods A retrospective analysis was conducted on 997 patients who underwent general anesthesia at our hospital from March 2023 to July 2023.Basic information of the patients was collected,including age,gender,body mass index (BMI),American society of anesthesiologists (ASA) classification,history of hypertension,history of diabetes,duration of surgery,surgical position,operating room temperature,use of enflurane,use of atropine during surgery,whether blood pressure fluctuated severely,volume of fluid replacement during surgery,intraoperative blood loss,and whether pneumoperitoneum was used during surgery.These data were used to construct an artificial neural network prediction model,and the importance of each factor was assessed through the model.The specificity,sensitivity,and accuracy of the model were calculated,and the receiver operating characteristic (ROC) curve was plotted to calculate the area under the curve (AUC) value.Results An artificial neural network model was constructed with potential risk factors as the input layer and the occurrence of blurred vision after general anesthesia as the output layer,with a model structure of 27-9-2.The specificity,sensitivity,and accuracy of the model in the training set were 91.7％,65.7％,and 87.8％,respectively,while in the test set,they were 92.2％,74.4％,and 89.5％,respectively,with an AUC value of 0.889.The model predicted that the use of enflurane(0.193),surgical position(0.155),duration of surgery(0.120),and age(0.116)are important factors leading to blurred vision after general anesthesia(standardized importance>50％).Conclusion The artificial neural network model can be used effectively to predict the occurrence of blurred vision after general anesthesia,providing a scientific basis for clinical physicians to choose safe and reasonable preventive treatment plans.
万方数据
维普
