The Dialectical Model Of The Pain Level After Video-assisted Thoracoscopic Surgery Based On Naive Bayesian Classifier

Objective: Comprehensively considering the characteristics of the patients, the research aims to build the dialectical model of the pain level after video-assisted thoracoscopic surgery(VATS) under two different types of analgesia based on Naive Bayesian Classifier. The research will contribute to the decision of the anesthetist on the type of analgesia as the clinical guidance and reference.Methods: The research studied 141 adult patients with video-assisted thoracoscopic surgery(VATS), allocated randomly to receive either Epidural Analgesia(EA) or Intercostal Nerve Block(INB). In addition, the 1~2 root thoracic drainage tube was placed after the surgery. The median age of the patients was 67(range from 20 to 83) years old. Operation duration was 3.5±2 hours. The groups were also similar with respect to gender, body-mass index, performance atstus, and type of surgery. None of the patients had serious heart and lung diseases, liver and kidney dysfunction. The patients who abuse the alcohol, drugs and anesthesia drug were excluded. Preoperative examinations were completed and the indicators were normal. The patients were randomly separated into two groups with 71 cases in EA group and 70 cases in INB group. For both of the groups, same induction of anesthesia was opearated by using midazolam, etomidate, cisatracurium besylate and sufentanili and intraoperative anesthesia was maintained by the propofol, cisatracurium besylate and remifentanil using TCI. The BIS equipment was applied for monitoring depth of anesthesia. The characteristics of the patients were recorded as 10 variables including sex(male/female), age, weight, method of postoperativeanalgesia and so on. In order to eliminate mutual interference between the 10 variables and reduce the difficulties in analysis, the PCA(Principal Component Analysis) was adopted firstly to transform the variables into synthetic features with lower dimension. Based on these features and the corresponding VAS ratings at 6, 12 and 24 hours after surgery, the dialectical model of the pain level was built based on Naive Bayesian Classifier. The algorithm and the interface of the Naive Bayesian Classifier were programmed using matlab and C#.Results: Application of principal component analysis was used to process the original data to highlight the different surgery conditions. We got 5 principal components which are respectively related with Sex, BMI, BP, Age, and S-FEN. The corresponding postoperative analgesia for patients after operation mode and 6 hours after VAS scores was applied as input data of the Naive Bayesian classifierlearning process. To verify the error rate and reliability of the Naive Bayesian model, the application of Naive Bayesian model was used on 20 practical cases of patients with operation anddemonstration. Finally, results showed that 6 hours of VAS scores in postoperative patients are consistent with the model prediction results.Conclusion: The research was focused on the preliminary analysis on the prediction of the pain level after surgery using Naive Bayesian model. 5 principal components was extracted respectively: Sex feature, BMI feature, BP feature, Age feature, and S-FEN feature, and the mathematical model was built. Then, the dialectical model of the pain level after surgery based on Naive Bayesian classifier was built. Through verification, the model has low error rate and reasonable reliability. Therefore, it is believed that it can provide guidance and support to the decision-making process of doctors with a further research.