ECHOCARDIOGRAM ANALYSIS USING FUZZY-SETS AND RELATIONS

This paper describes part of a large project whose objective is to obtain completely automated analysis of a set of time-sequenced echocardiogram images, to yield such heart function parameters as chamber and stroke volumes, ejection fraction, and indices of regional myocardial function. The research reported here is concerned with identifying collections of local regions in different frames which have similar characteristics and hence comprise a global region, and those global regions which add together to form other known global regions. The local regions are described by a fuzzy vector whose components are fuzzy numbers. An equality-like relation is defined on the set of local regions and equivalence classes are used to define the global regions. A similar technique is used to decide which global regions add to larger global regions. The end result is used by a fuzzy expert system which classifies the global regions according to their anatomical significance or as artifacts. The results here reported suggest the utility of the fuzzy set techniques employed. The possible application of our methods to other pattern recognition problems is discussed.