Application Of Nonlinear Dynamical Techniques To The Ventricular Fibrillation Identification And Its Implementation

Ventricular fibrillation (VF) is a malignant arrhythmia, and is one of theimportant factors for cardiac arrest and sudden death induced. If the patient doesnot have to be provided effective treatment such as defibrillation within4-6minutes, it will lead to the patient is to go to their own lives within a few minutes.However, if normal sinus rhythm (NSR) is incorrectly detected as shockable, andimplementation of unnecessary shocks to the patient, it will cause fatal damageto the human heart function. Thus, the timely and accurate detection of the theshockable signal (including the VF, heart rate is greater than160bpmVT, VFL) isvery important.According to the needs of Research ventricular fibrillation algorithm. First,introduce requirements for ventricular fibrillation detection algorithm of AED inthe standard ANSI/AAMI DF80:2003. A description of the ECG signal isdivided into shockable signal in AHA database, VF database as well as the CUdatabase.Second, according to the needs of enterprise engineering applications,nonlinear ventricular fibrillation detection algorithm including EMD algorithm,delay algorithms, SVM algorithm as well as the Hurst exponent algorithm, thetime-domain ventricular fibrillation detection algorithm including TCSCalgorithm and frequency domain ventricular algorithm including DWT algorithmis researched, and is implemented by Matlab with AHADB, VFDB, CUDB. Theresult of experiment and the time complexity of the algorithm is researched andanalyzed. According to the experimental results, it is found that for nonlinearventricular fibrillation detection algorithm, the accuracy of the EMD algorithmand the Hurst index algorithm detecting the shockable signal and thenonshockable signal is not high, and their sensitivity and specificity are moredifference with the requirements for ventricular fibrillation detection algorithmin standard of AED. SVM algorithm has high specificity, but its sensitivity isrelatively low. In other words, it will some shockable signal is detected as anon-shockable, and time complexity of SVM algorithm is the highest of all thealgorithms. The PSR algorithm has high sensitivity but its specificity is not high,that is, a part of a non-shockable signal is erroneously detected as a shock signal.For the time-domain ventricular fibrillation detection algorithm, TCSCalgorithms to detect heart rate small ventricular fibrillation will be incorrectly,but the accuracy of the detection ventricular flutter and heart rate greaterventricular tachycardia will be better. And its time complexity is smaller than other experiment algorithms. For the frequency domain ventricular algorithm, theaccuracy of DWT algorithm detecting shockable signal is lower. And for all thealgorithms, the detection of ventricular asystole is relatively poor.Finally, taking into account the ECG signal can be classified as non-linear.This paper propose a new algorithm, which is based on nonlinear re constructionalgorithm and combined with the improved the TCSC algorithm. This algorithmincludes the detection of ventricular asystole in advance, in order to improve theaccuracy of the ventricular fibrillation detection, and reduces the length of thetime window of the detection to six seconds, thereby reducing the time ofdetecting VF in AED. And the sensitivity and specificity of this algorithm ishigher.Based on the study and implementation of existing ventricular fibrillationdetection algorithm, this paper proposed a new algorithm which major based onnonlinear algorithm and combined with the time-domain algorithm ventricularfibrillation detection algorithm. The sensitivity and specificity of the proposedalgorithm are all higher than requirements for ventricular fibrillation detectionalgorithm in AED Standard (ANSI/AAMI DF80:2003).