Research On Classifying ST Segment In ECG Using Modified Fuzzy ARTMAP

Parameters and shapes of ST segment in Ambulatory Electrocardiograph (AECG) are main reference indexes in diagnosing myocardial ischemia. However, in early times, parameter labels and shape analysis of ST segment are manually implemented by doctors, in addition, AECG data are large numbers. These limitations debase diagnosing reliability. Since neural networks are parallel distributive processors based on human thinking, it is available to achieve this function using neural networks.In this paper, based on the research of Fuzzy ARTMAP, a new classifier used to assort the shapes of ST segment is proposed. The new classifier, called Modified Simplified Fuzzy ARTMAP (MSFAM), is a modified and simplified version of the well-known Fuzzy ARTMAP. The main contribution and innovation of this dissertation are showed as follows:1. Profound dynamical characteristic analysis and mechanism analysis of fuzzy ARTMAP classifier. The analysis includes the choice of some parameters, such as choice parameter and vigilance parameter. And geometric interpretation of fuzzy ART classifier is also presented.2. Simplify and modify the fuzzy ARTMAP classifier. The structure of fuzzy ARTMAP classifier is too complex if it is just used to classify, so a new classifier, MSFAM classifier is proposed on the base of fuzzy ARTMAP. The modified SFAM algorithm only uses one membership function to calculate the similarity between the input vector and the weight of winner neuron. And every neuron in the MSFAM system has its own vigilance parameter.3 On the base of ST segment being extracted from ECG using wavelet transform, these three classifiers, MSFAM, SFAM and Fuzzy ARTMAP, are used to recognize and categorize the shape of ST segment in ECG, respectively. The result shows that the recognition rate of each classifier is high.