Application Of The Deep Learning In The Diagnosis Of Heart Disease

Deep learning is one of the important methods in the field of machine learning.Due to its powerful ability to build complex models,it has attracted widespread attention from the academic and industrial circles.Although deep learning models have huge advantages in terms of abstract features,the use of shallow neural networks as classifiers restricts the classification performance of deep learning.Ensemble learning is a general method to improve the controllability and generalization ability of models in the field of machine learning,but ensemble learning models have limited ability to express high-dimensional complex problems.Therefore,combining the advantages of deep learning and ensemble learning,a deep belief-boosting network ensemble classification model is designed and improved.Heart disease is one of the major diseases endangering human life and health in the world.In China,myocardial infarction alone accounts for about 600,000 people every year.The main cause of myocardial infarction and sudden death is abnormal heart activity,also known as arrhythmia.Among them,electrocardiogram is the cheapest non-invasive detection tool for diagnosing arrhythmia.However,in the current situation of massive ECG data and complicated causes of disease,the diagnosis of ECG by artificial vision can no longer meet people’s needs.In this case,automatic classification technology must be used for analysis.Therefore,this article deeply studies the relevant theories and technologies of ECG signal diagnosis,And realized an automatic arrhythmia detection method based on integrated classification model.The main research of the thesis includes the following two aspects:(1)Design and improve a deep belief-boosting network integration classification model.The ensemble classification model is composed of input layer,hidden layer and output layer.The input layer uses the ensemble deep belief network(Deep Belief Networks,DBN)as the feature extractor,while using feature selection to remove redundant features;the hidden layer is a deep promotion network based on the decision tree of the gradient promotion strategy,and the number of layers can be based on each layer The learning results are adaptive;the output layer adopts a relatively majority voting strategy to obtain the final classification results.Based on 5 public data sets,the model in this paper is compared with DBN,limit gradient boosting tree and DBN-XGBDT.The experimental results show that the classification accuracy of this model is improved by 13.8%,6.98% and 4.18% compared with the three models.Which can provide a new direction for the combination of ensemble learning and deep learning.(2)Automatic detection method of arrhythmia based on ensemble classification model.This article takes the detection and recognition of ECG signals as the key technology for diagnosing arrhythmia,and conducts research and analysis on the preprocessing,waveform extraction and classification of ECG signals.First,the ECG is preprocessed by ECG signal denoising,R peak location and ECG signal division.Secondly,the ensemble classification model is applied to arrhythmia classification,and specific experiments are designed for verification.The experimental results show that the automatic arrhythmia detection method based on the ensemble classification model not only has high accuracy,but also has strong stability.(3)Designed a prototype system for automatic arrhythmia analysis.In this paper,a prototype system for automatic arrhythmia analysis is established based on the trained arrhythmia automatic recognition model.It can automatically classify the uploaded ECG waveforms by uploading the user’s personal information and ECG data,and show the user a graphical arrhythmia analysis report in the form of a WeChat applet.