A Study On Life-threatening Arrhythmias Detection By Pulse-to-pulse Intervals Analysis

Life-threatening arrhythmias(LTA)are the main symptoms before the onset of acute cardiovascular diseases such as myocardial infarction and cerebral infarction,and timely and accurate recognition of LTA is crucial.Pulse signals contain rich physiological and pathological information about the cardiovascular system,and the pulse-to-pulse intervals(PPIs)extracted from pulse signals can reflect changes in the rhythm of the human cardiovascular system,which can be utlized for LTA recognition.Therefore,this study proposed the recognition method and detection system of LTA based on the PPIs.By extracting significant change features in PPIs signals and using machine learning methods,LTA can be identified.The main research results are as follows:1.A PPIs extraction technique is proposed that combines pulse signal filtering,peak extraction and motion artifacts discrimination to accurately extract PPIs under high noise and interference background.Firstly,noise and interference are wiped out from the pulse signal by an integer coefficient notch filter.Then,an improved automatic multiscale-based peak detection method is proposed to segment the pulse signal period.Finally,a motion artifacts discrimination method based on the decision tree(DT)is proposed to evaluate the quality of the segmented pulse wave and obtain accurate PPIs.Using this method to extract PPIs from 109 sets of pulse signals in the "MIMIC/Fantasia" and "the 2015 Physiology Net/Cin C Challenge" databases,the results show that the proposed method has an accuracy of over 99% for PPIs extraction.2.A fast feature extraction method based on the data updating characteristics of microprocessors is proposed to quickly extract significant change features in PPIs.The calculation processes of eight significant change features,including mean,standard deviation,root mean square difference,normalized root mean square difference,the percentage of time intervals greater than 40 ms between adjacent sampling points of PPIs signal(PNN40),approximate entropy,sample entropy,and permutation entropy,are analyzed.The calculation processes of these features are improved according to the data sliding window update process of the microprocessor.The results show that compared with the unimproved method,the improved method can achieve online extraction of features with a single sampling point step length while ensuring accuracy.3.An intelligent method based on the analysis of non-emergency segment PPIs signals of LTA is proposed to address the intelligent recognition problem of LTA.Firstly,significant changes in the PPIs are extracted as features.Then,BP neural network,extreme learning machine,and DT are exploited for LTA identification.The results show that the DT classifier has the best performance in identifying LTA,with an average accuracy rate of over 97%.4.A prototype sample for detecting LTA was developed and the proposed method was validated online.The sample mainly consists of a lower computer and an upper computer.The lower computer is responsible for pulse signal acquisition and online filtering,while the upper computer is responsible for PPIs and significant change features extraction,and detects the health status of the human body based on the trained DT model.The results show that the proposed method can be engaged for online identification of LTA.