Redundancy reduction for improved display and analysis of body surface potential maps. I. Spatial compression

Abstract

The Karhunen-Loeve technique of random process representation was investigated as a method of quantitatively characterizing body surface potential maps. One hundred ninety-two lead body surface potential maps from 124 normal subjects and 97 patients with independently documented heart disease were used in the study. Each map frame in QRS and ST-T of 34 maps in a test set was represented as a linear sum of orthonormal distributions derived from the covariance matrix estimated from all QRS frames in the 221 training maps. A 16:1 reduction in spatial data of the test set was achieved with rms errors of 45 and 21 μV in QRS and ST-T, respectively. Results suggest that 12 independent waveforms, derived from the 192 measured ECGs, may be used in place of those 192 ECGs. In addition to providing a convenient and familiar method of display for map data, the technique puts the data in an appropriate form for quantitative statistical analysis.