A real-time ST-segment monitoring algorithm for implantable devices.

Continuous ST-segment monitoring by implantable devices may lead to clarification of the substrate of arrhythmias, clarification of the origin of nonspecific chest pain, and titration or preventative application of established anti-ischemic therapies. Although ST-segment monitoring algorithms are available for surface electrocardiogram, the computational demand of algorithms for implantable devices must be minimized for considerations of device longevity. The new algorithm first locates a fiducial point (FPT) at the dominant peak of each QRS complex. The ST-segment deviation (measured at 2 rate-adaptive delays after FPT, eg, FPT + 96 ms and FPT + 152 ms at 60 BPM) with respect to the isoelectric level (measured at the minimum slope preceding the QRS) is then measured. The following features are also quantified by simple operations: R-R interval, R-wave slope, R-wave amplitude, ST-segment slope, and noise content during the isoelectric segment. Inconsistencies in these features relative to their adaptive normal ranges are used to reject noisy or ectopic beats and sudden morphology changes. Finally, the ST-segment deviation over time is filtered to reject rates of change that are not likely attributable to human ischemia. Performance of the algorithm was evaluated on the European Society of Cardiology ST-T Database, which contains 180 hours of ambulatory electrocardiogram with 250 expert-annotated ischemic episodes. The sensitivity was 79% [74% 84%] (mean [95% CI]) and positive predictivity was 81% [76% 86%]. This performance is statistically equivalent to that of published electrocardiogram algorithms that were validated on the same dataset. Estimates of computational burden suggest that the algorithm could process two channels of electrogram continuously for more than 5 years with current implanted device technology. In conclusion, we have developed an algorithm for ST-segment monitoring that can be implemented in current implantable devices with sensitivity and positive predictivity that are comparable with the state-of-the-art.