An intelligent electrocardiogram classification method

Abstract

The invention discloses an intelligent electrocardiogram classification method. The invention performs 3 Hz and 4 Hz sampling on the raw data of I lead ECG waveform sampling points with a sampling rate of 500 Hz in 30 seconds, and performs 3 Hz sampling on the raw data of the I lead ECG waveform sampling points. 45 Hz band-pass filtering, R-peak recognition algorithm based on wavelet transform isused to extract R-peak position, extracting a heart beat template of 195 sampling points based on the R peak position, and then extracting the maximum amplitude of the ECG waveform, Minimum, Mean, Logarithmic entropy, amplitude and position of PQRS complex and other traditional ECG signal characteristics, using the original ECG waveform, corresponding classification labeling, gradient descent training deep residual network superposition long-short time memory network to extract the deep-level features of ECG waveform, and finally input all the extracted features into the stochastic forest model for classification and diagnosis. The invention can largely remove noise interference, reduce the dependence on artificial feature identification, and greatly improve the speed and accuracy of electrocardiogram classification.

Description

technical field [0001] The invention belongs to the technical field of intelligent electrocardiogram classification, and in particular relates to an intelligent electrocardiogram classification method based on machine learning and signal processing. Background technique [0002] With the continuous improvement of living standards, people pay more and more attention to the health of the heart, and the electrocardiogram has become an important reference for knowing the state of the heart because it contains a lot of information, is non-invasive, and is cheap. needs are becoming more and more urgent. The traditional electrocardiogram classification is analyzed by the naked eye, which relies heavily on the existing practical electrocardiogram classification experience. In the early days, it also tried to identify the characteristic waveforms by using traditional signal processing methods, and then performed auxiliary analysis through statistical analysis. However, some important...

AI Technical Summary

Problems solved by technology

However, the shallowness of features, lack of real ECG data, noise interference, and the diversity of abnormal ECG have resulted in no significant breakthroughs in ECG intelligent classification methods. How to use machine learning to extract deep features in ECG has become a major difficulty.

Images