Abstract
A brief summary of selected papers from an extensive literature is followed by an explanation of the underlying physiological processes that give rise to the electrocardigram. Data for use in an investigation of potential statistical techniques were obtained from simple recordings: Lead I from eight subjects under three conditions - sitting, standing and supine. All were recorded on analogue magnetic tape, with a bandwidth of DC-1kHz; sufficiently wide to permit a more detailed analysis if that should prove necessary - so far little use has been made of the bandwidth extremes. The data were digitised at 1kHz - 9 bits - and stored on disc. A base programme was written for the production of simple descriptive statistics for each subject/condition; it has as its basis a minor variation of the standard clinical definitions and it is with this analysis that other statistical techniques are compared. A second programme employed the parameters produced by the base programme to arrive at an analysis using Principal Components. The results demonstrated that discrimination between ecg's, both for subject and condition, was marginally improved in comparison with an Analysis of Variance on the base parameters. Next the raw data (125 samples/sec) were treated by principal component analysis. This provided a more satisfactory discrimination, with the result that 98% of the potential information in each ecg complex may be described by 6 values. The base programme was written in Fortran IV for a Harris 500 computer; it was later rewritten, with a much longer execution time, for a Cromemco microcomputer - Z80A and 64kbytes. All the techniques would benefit further development and suggestions are made for future work.