Title :
Efficient hemodynamic event detection utilizing relational databases and wavelet analysis
Author :
Saeed, M. ; Mark, RG
Author_Institution :
Harvard-MIT, Cambridge, MA, USA
Abstract :
Development of a temporal query framework for time-oriented medical databases has hitherto been a challenging problem. We describe a novel method for the detection of hemodynamic events in multiparameter trends utilizing wavelet coefficients in a MySQL relational database. Storage of the wavelet coefficients allowed for a compact representation of the trends, and provided robust descriptors for the dynamics of the parameter time series. A data model was developed to allow for simplified queries along several dimensions and time scales. Of particular importance, the data model and wavelet framework allowed for queries to be processed with minimal table-join operations. A web-based search engine was developed to allow for user-defined queries. Typical queries required between 0.01 and 0.02 seconds, with at least two orders of magnitude improvement in speed over conventional queries. This powerful and innovative structure will facilitate research on largescale time-oriented medical databases
Keywords :
SQL; haemodynamics; medical diagnostic computing; query processing; relational databases; temporal databases; time series; wavelet transforms; MySQL relational database; data model; hemodynamic events; large-scale time-oriented medical databases; minimal table-join operations; multiparameter trends; parameter time series; robust descriptors; temporal query framework; time-oriented medical databases; wavelet coefficients; wavelet framework; web-based search engine; Data analysis; Data models; Data structures; Event detection; Heart rate; Hemodynamics; Relational databases; Search engines; Time series analysis; Wavelet analysis;
Conference_Titel :
Computers in Cardiology 2001
Conference_Location :
Rotterdam
Print_ISBN :
0-7803-7266-2
DOI :
10.1109/CIC.2001.977614
