An Automatic ECG Identification Method For Myocardial Infarction Based On Fusion Features

Myocardial Infarction(MI)is a serious cardiovascular condition characterized by ischemia and anoxia in the coronary artery supply zone,which can occur suddenly or over time.If the patient is not diagnosed and treated in time,a large area of myocardial cells will die,resulting in irreversible damage or death in severe cases.Therefore,a timely and accurate diagnosis of MI is crucial.Therefore,timely and accurate diagnosis of MI is crucial.Electrocardiogram(ECG)is one of the commonly used diagnostic tools for myocardial infarction.The main abnormal waveform of MI in ECG are ST segment abnormality,T wave abnormality and pathological Q wave.Whether patients with myocardial infarction can receive timely and accurate treatment is crucial to the prognosis of MI patients.Therefore,a growing number of scholars focus on the automatic interpretation of MI-ECG with the help of machine learning to improve the speed of MI diagnosis.In this context,this paper proposes a new automatic interpretation method of MI-ECG based on fusion features.Firstly,targeting at the T-wave specific manifestation in MI-ECG,a new automatic location algorithm T peak and starting and ending points is designed.Secondly,on the basis of feature wave localization,different features are extracted for ST segment abnormalities and specific manifestation of MI-ECG.Finally,the effective features are extracted according to the typical performance of ST segment anomalies,and then combining with Softmax regression to achieve abnormal ST segment identification.Finally,combined with the abnormal performance of other waves on the MI-ECG,and fuse them together,then the automatic interpretation of MI-ECG is realized by Light GBM algorithm.This paper uses Chinese physiological signals challenge 2018(CPSC2018)database and the PTB diagnostic database to verify the proposed ST segment anomaly identification method and MI-ECG automatic interpretation method is effective.The numerical experiment results show that the average accuracy,specificity and sensitivity of the proposed automatic abnormal ST segment identification method in this paper are 92.47%,96.05% and 90.94%,respectively.The average accuracy,specificity and sensitivity of MI-ECG automatic interpretation method in this paper are are 97.82%,99.03% and 96.51%,respectively.