Data structure-guided development of electrocardiographic signal characterization and classification

摘要

ObjectiveThe study introduces and elaborates on a certain perspective of biomedical data analysis where data structure is revealed through fuzzy clustering. The key objective of the study is to develop a characterization of the content of the clusters by offering a number of their descriptors established on the basis of membership grades of patterns included there, as well as on the basis of their class membership. Next, a design of a cluster-based classifier is presented in which the structure of the classifier is based on a collection of clusters. The structure also exploits the descriptors of the clusters as well as aggregates their characteristics with the activation levels of the associated clusters formed in the feature space in which QRS complexes are represented.