Hilbert transform processing of electrocardiograms

Abstract

The use of the Discrete Hilbert Transform (DHT) in the area of Electrocardiogram (EGG) analysis is investigated. It is shown that by using DHT processing, time alignment of individual heart-beats is not necessary, thus allowing the use of a pattern recognition method previously untried in EGG processing. It is possible to perform morphological recognition by mapping the Hilbert Transformed EGG into a two dimensional recognition matrix and achieve abnormality discrimination by applying a generalised form of the method of modified potential functions in association with a Bhattacharyya distance measure. Other topics examined include the mathematics and implementation of a Hilbert Transformer, display of ECGs, especially P- and T-wave enhancements, QRS detection and waveform delineation, and abnormality recognition (both morphology end rhythms). A suggested clinical system that consists of a 16-bit microprocessor host system capable of basic EGG processing, data storage, and report generation is outlined. The EGG processing subsystem is comprised of a hardware module using bit-slice components to accomplish the DHT and auxiliary filtering, and a custom VLSI I.C. to perform the morphological recognition. Experimental results are included in addition to a brief history of EGG computer systems, techniques , and traditional pattern recognition approaches.