Ventricular Fibrillation Detection Algorithms Based On Time-frequency Domain

In recent years, the incidence of cardiovascular disease increases over the years,seriously threatening the lives of people. The sudden cardiac death is the most serioussymptom of arrhythmia, if defibrillation is not timely, survival probability of the patientwill reduce with the time. We know that the sudden cardiac death results fromventricular fibrillation (VF) or sustained ventricular tachycardia (VT) in mostconditions. VF is the most deadly, and VT will quickly turn into VF, which leads tosudden death. Therefore, it is particularly important to detect ventricular fibrillationrapidly and accurately.The VF detection algorithms are analyzed in time and frequency domain in thispaper. We find that the TCSC algorithm does not consider the shape of the ECG signal,for that reason it fails to separate VT from VF successfully. Meanwhile, the standardexponential algorithm is simple in time domain, which searches the absolute maximumvalue of the investigated sequence of the signal. It therefore forms a graduallydecreasing exponential curve, and we then count the number of crossing points of theECG signal with the exponential curve to classify different types of ECG signals. But itsresult is not satisfactory. Then we analyze the Lempel-Ziv complexity algorithm, whichtransforms the ECG signal into a binary sequence and searches for repeated patterns todetect VT and VF. When the heart changes fast and then exhibits arrhythmia, thecorresponding complexity become higher, so it is difficult to detect VT and VF signals.Last but not the least we study empirical mode decomposition which originates from thefrequency domain. Our realization of such algorithm separates the VF from the normalsinus rhythm (NSR) which shows almost100%accuracy rate. However, when othertypes of pathology besides the NSR, such as the left bundle branch block beat, rightbundle branch block beat, supraventricular premature beat, premature ventricularcontraction beat are presented, the performance of this technique becomes poor.Above algorithms are analyzed alone in either time domain or frequency domain,and their results are not satisfactory and cannot reach the practical requirement. In orderto improve the accuracy of the algorithm, combination of several techniques from bothtime domain and frequency domain is prefers. In this paper we utlize the discretewavelet transform algorithm, whose inputs come from the statistical feature extracted bythe previous methods. We merely utilize three parameters of wavelet decomposition, D4,D6and D7to classify VT, supraventricular tachycardia (SVT), VF and ventricularflutter signals (VFL). In comparison with the conventional wavelet algorithm, ouralgorithm greatly reduces the computational cost. Regarding to the identificationperformance, the discrete wavelet transform algorithm can effectively detect VF.In order to validate the reliability of the defibrillation algorithms, finally we complete the statistic analysis of the medical instrument clinical trials. Medical statisticsis important to the medical instrument clinical trials Here we apply matching design ofmathematical statistics to the trials. We will compare ECG-1260which made in Biocareelectronic company with MAC5500of General Electric Company. The measurementdata is analyzed by means of the homologous matching t test and count data is analyzedby the pairing four tableχ2test. Our results show that statistical results of ECG-1260are consistent with those of MAC5500. Based on this statistical analysis, we can extendthe matching design to clinical trials of the defibrillator.