Strucutural Comparison and Cluster Analysis of Time-Series Medical Data

In this paper we present a cluster analysis scheme for time series medical data. It allows us the structural comparison and hierarchical grouping of irregularly-sampled, irregular-length time series. The core technique is modified multiscale matching, which improves the segment parameter representation and dissimilarity measures in the multiscale structure matching so that the problem of shrinkage and mixture of multiple attributes in the dissimilarity can be solved. We examined the usefulness of the method on the platelet sequences in the chronic hepatitis dataset. The results demonstrated that the dissimilarity matrix produced by the proposed method, combined with conventional clustering techniques, lead to the successful clustering for both synthetic and real-world data