Real-time serial analysis of infarctional changes in the vectorcardiogram

Abstract

Infarct size evolution can be readily assessed from spatial QRS and ST segment changes. A computer system for real-time serial analysis of QRS and ST segment changes in the vectorcardiogram has been developed. QRS detection is done by multichannel template matching, the template being the running average complex. Detection of QRS onset and end is done by a two-dimensional pattern recognition technique on the spatial velocity function using fixed templates. Tested on 216 recordings from 11 patients undergoing myocardial infarction 92.9% of 6050 beats were correctly detected with 0 false detections. The differences between program and referee estimates of QRS onset and of QRS end were computed in 1015 average complexes from the same group of patients. The median (0.05 and 0.95 fractiles) differences were 0.6 msec (−1.9 and 3.9 msec) for QRS onset and −2.0 msec (−16.7 and 3.7 msec) for QRS end. A similar test was made on the annotated CSE database. The results were −2.2 msec (−6.9 and 10.6 msec) for QRS onset and −1.6 msec (−14.6 and 5.2 msec) for QRS end. Systematic errors in QRS onset or QRS end of ±8 msec will produce 3–4% error of the average infarct size estimated by QRS vector changes. In conclusion, the present system can provide instantaneous information about the development of myocardial infarction which may be of importance in the management of infarct-limiting treatment.