Establishment And Evaluation Of Computer Convolutional Neural Network Models For Premature Ventricular Contractions In Children

Background and Objectives With the development of computer chip capability and the coming of big data Era,Artificial intelligence models applying computer deep learning have been able to simulate the neural network structure of the human brain.The models can process the input data,compute the data with all the layers,and finally output the prediction based on the understanding of input data.In recent years,the deep neural network models have been gradually applied in the medical field and the application of deep learning in ECG diagnosis has just begun.In this research we preliminary establish three convolutional neural network models for automatic diagnosis of premature ventricular contractions in children applying the method of computer deep learning,and evaluate their diagnostic value.Methods ECGs of 1200 children with premature ventricular contractionscollected in Shanghai Children’s Hospital were used as PVC group and ECGs of 1200 normal children in the same age and sex were taken as normal control group.Eliminate few ECGs that are not suitable for model training then randomly extract 800 samples in two groups,applying the method of computer deep learning,train and establish three convolutional neural network models for automatic diagnosis of premature ventricular contractions in children.200 other samples were extracted from each group to test and verify the performance of each model.The diagnosis of electrocardiogram expert group of Department of Cardiology in Shanghai Children’s Hospital is regarded as "gold standard".Using medical statistics,calculate the index such as sensitivity,specificity,missed diagnosis rate,misdiagnosis rate,positive predictive value,negative predictive value,accuracy rate,Kappa value and evaluate the reliability and validity of the models.Results Applying the method of computer deep learning,2D CNN model and fine-tuned Inception-V3 model are trained with ECG waveform image,1D CNN model are trained with ECG time-series data.The sensitivity of the 2D CNN model is 65%,the specificity 71.5%,the missed diagnosis rate 35%,the misdiagnosis rate 28.5%,the positive predictive value 69.5%,the negative predictive value 67.1%,the accuracy rate 68.2%,and the Kappa value 0.365.The sensitivity of the V3 model is 82%,specificity 85%,missed diagnosisrate 18%,misdiagnosis rate 15%,positive predictive value 84.5%,negative predictive value 82.5%,accuracy 83.5%,Kappa 0.670;The sensitivity of 1D CNN model is 87.5%,specificity 89.5%,missed diagnosis rate 12.5%,misdiagnosis rate 10.5%,positive predictive value 89.3%,negative predictive value 87.7%,accuracy rate 88.5%,Kappa 0.770.V3 model and 1D CNN model perform well,and the reliability and validity of 1D CNN model are especially good,Kappa value 0.77 indicates highly consistent with the diagnosis of electrocardiogram expert group of Department of Cardiology in Shanghai Children’s Hospital.Conclusions Three convolutional neural network models for automatic diagnosis of premature ventricular contractions in children have been established using computer deep learning.Among them,the 1D CNN model is expected to be applied in clinical practice in the future to improve the clinical diagnostic efficiency and diagnostic accuracy of premature ventricular contractions in children and also provides a basis for establishing an automatic ECG model for 24 hours electrocardiogram and treadmill exercise test.