Research And Implementation Of The System Of Arrhythmia Intelligent Analysis

It is shown by many authoritative investigations that one third of the people die because of cardiovascular disease every year among the total deaths of the world people. As a matter of fact, there is a great significance for researching on arrhythmia intelligent classification algorithm and ECG monitoring device in order to save lives.The selected issue is from the coorperation project with Nuvoton Technology corp. called "single-lead ECG monitoring system" and the open project program of the national laboratory of pattern recognition called "Ambulatory ECG intelligent diagnosis method research and implementation". The main work includes arrhythmia feature extraction and classification algorithm study together with the arrhythmia intelligent diagnosis system hardware and software. The system consists of single-lead ECG monitoring and arrhythmia intelligent diagnosis software which integrates the entire algorithm into it.The paper has summarized and compared the domestic and foreign Ambulatory ECG (AECG) intelligent diagnosis algorithm. In order to remedy the shortage of the classic differential threshold, the author uses ambulatory threshold together with area information for R wave detection. The method effectively cuts down the false detection rate, and makes the detection rate above98%. Then according to the shortage of traditional feature extraction method, the author uses continuous wavelet transform algorithm to solve the problem which could easily describe P and QRS complex in shape and details. The simulating results show that the continuous wavelet transformation method not only apply less feature vectors to implement complete and fast classification process but also overcome the former method in accuracy. Finally, as the branch logic algorithm is out of style and the neural network algorithm is extremely depending on the training sample, the author uses principal and subordinate SVM to separate the heart beats into six categories which feature vector sampling scheme is perfect and the accuracy of classification for a single beat is the highest. The simulating result shows that the training dataset affect less in accrcy for SVM classifiers meanwhile the subordinate SVM could improve the classification result of the principal SVM, which ensures the classification accuracy step into over97%. In the process of single-lead ECG monitor designing, considering the elements of space, cost and power dissipation, the author selects Nuvoton ARM7TDMI NUC501as the CPU. It has the advantage of fast computing, low cost and low power dissipation. Compared with the other similar hardware systems, our hareware has the property of low cost, easy designing, extending and debugging. The arrhythmia intelligent diagnosis system had the function of intelligent analysis and human-computer operation interface. The functions include ECG signals denosing, intelligent classification, data analysis, statistic output. The software has also developed portals, which includes the function of data uploading, patients data management, ECG diagnosis report downloading in order to make the patients communicate with doctors from distance.After test performance and validation, the system have the features of fast computing, accurate and reliable.The whole system meets the needs of clinical demand, and acquires the approval by the first hospital of China Medical University cardiology expertise.