A trend detecting algorithm for ST-segment and rate abnormalities

A method is presented for processing beat-by-beat measurements of ST-segment deviation and rate to detect abnormal events of sufficient duration to trigger an alarm on a portable ECG monitor. The method removes the problem of rate dependency from the event-detection process. The algorithm maintains a sorted list of the 64 samples most recent of an 8-beat average of the measurement, taken once per second. A value from a particular list element is compared to a preprogrammed threshold to determine the presence of an alarm condition. The position of the list element used is determined by a programmed alarm window time and the prior presence of the alarm condition. This technique overcomes several difficulties: (1) cardiac-rate dependency is reduced by sampling an average value at a fixed clock rate; (2) event chopping due to fluctuations of a measurement above and below a threshold is eliminated by changing the sample point once an event is in progress; (3) sensitivity to outliers is reduced both by using 8-beat averages and by requiring multiple entries in the sorted list to cross the threshold before triggering or ending an event; (4) a programmable time window of acquisition for ST events is supported by making the sample slot offset variable