Design And Implementation Of Multi-lead ECG Signal Discrimination Algorithm Based On Deep Learning

Cardiovascular disease(CVD)has always been the primary threat to human health.At present,doctors usually use electrocardiogram(ECG)for diagnosis.At the same time,judging whether the electrocardiogram is abnormal is a very professional work,so doctors with rich theoretical knowledge and rich clinical experience are needed,but the number of cardiovascular disease doctors is insufficient and the work intensity is large.Therefore,it is important to develop fast and accurate algorithms.Although many algorithms have been proposed to automatically discriminate ECGs,most of the default methods installed in ECG machines are not accurate enough,and some public high-precision algorithms are not suitable for use in hospitals because they rely on long-term ECG signals Data also has strict hardware requirements and high time costs.In addition,most existing methods ignore the specificity of the patient’s ECG,resulting in poor generalization ability.Aiming at these problems,in order to better discriminate data with different scales and shapes and improve the generalization ability of the algorithm,this paper proposes a multi-stream lead loop and convolutional neural network(MLRCNN),which can effectively use different Scale features.In a multi-stream network,there are multiple convolution kernels of different scales and sizes.Then we connect the features of each network together for final output.We conducted experiments using ECG signal data in the laboratory and some public real data sets,and the results showed that the method proposed in this paper has a significant improvement compared with the previous method.At the same time,this paper uses the model training methods of transfer learning and confidence ranking to improve the generalization performance of the model and make the model applicable to other ECG data.The experimental data we used in this article come from more than 20,000 ECG data collected in the laboratory.Among them,the accuracy of the best effect MLRCNN model reached 92.80%,and it also performed well in other experiments.Therefore,it can be considered that the MLRCNN model proposed in this paper is a successful improvement of existing methods,and has certain reference significance and application value.