Artificial Intelligence In Predicting The Recurrence Of Atrial Fibrillation After Radiofrequency Catheter Ablation

Objective:Artificial intelligence was used to analyze the data of patients with atrial fibrillation(AF)who were admitted to our hospital and only underwent pulmonary vein isolation.And then logistic regression,random forest and convolution neural network will be established.This study compares the differences between different models and selects the best model for predicting the prognosis after radiofrequency catheter ablation(RFCA).Methods:This study included 364 patients with AF who only underwent pulmonary vein isolation in our hospital from June 1,2017 to December 1,2018.The follow-up is performed every 3 months after the RFCA,and the follow-up period is one year or until the patient has a positive result(such as AF,atrial flutter or atrial tachycardia).Logistic regression,random forest and convolution neural network are used to analyze the data.After the model is established,the performance of different models is evaluated by receiver operator characteristic curve(ROC curve)and area under curve(AUC).Results:In this study,364 patients with paroxysmal or persistent AF(duration <1 year)wereincluded.Pulmonary vein isolation was successfully completed in all patients.Among the 364 patients,57 were patients with persistent atrial fibrillation and 231 were male.The average age was 60.3 ± 9.87 years,and the average follow-up time was 331.7 ±170.4 days.A total of 84 patients had AF recurrence,and the AF recurrence rate was23.1%.This study used these 364 patients to establish a logistic regression model,a random forest model,and a convolutional neural network model.The AUC of these three different models in the test group are: the AUC of the Logistic regression model is 0.708(95CI%: 0.526-0.890),and the AUC of the random forest model is 0.624(95CI%: 0.408-0.841).The AUC of the convolutional neural network was 0.746(95%CI: 0.614-0.877).Delong test was used to compare the AUC of different models,and it was found that there was no difference between the random forest,the convolutional neural network and the logistic regression model.Conclusion:Convolutional neural network has a good performance in predicting the recurrence of AF after RFCA.There was no significant difference between these three models.However,convolutional neural networks has the advantage of automatically extracting image features for prediction,which saves a lot of human resources.With the development of artificial intelligence,convolutional neural network still has a great development space in the field of predicting the recurrence of AF after RFCA.