Tachyarrhythmia therapy utilizing electrical stimulation: a review

Arrhythmia recognition strategies and hardware decisions for the implantable cardioverter-defibrillator--a review

The avoidance of inappropriate shocks from the implantable cardioverter-defibrillator (ICD), together with its need to apply antitachycardia pacing to either atria or ventricles, demands considerable sophistication in the design of algorithms to interpret electrical or other cardiac signals in real-time. Methods based on rate and using single short-gap bipolar leads lack discrimination. Right ventricular electrogram morphology algorithms offer improvement but no universal algorithm exists; however, for any given patient an optimum algorithm of this type might be found. One improvement would be to provide atrial information in addition, by employing more than one electrode or a long-gap single bipolar lead. Alternatively, transducer signals could be included, once their efficacy and reliability have been improved. A different approach would be to use the much more sophisticated algorithms at present being tried with surface electrocardiograms. Integrated Circuit technology is reaching the point where this could be done but the requirement for exceptionally high reliability means that special system structures, such as a Memory Intensive Computer Architecture, may be required. When decisions on these approaches are to be made, it must also be remembered that ICDs will soon be implanted and programmed as a routine rather than a highly specialized procedure.

Statistical discriminant analysis of arrhythmias using intracardial electrograms

The problem of classifying ventricular arrhythmias from intracardial electrograms is considered. Standard statistical discrimination procedures are applied using a simple parametric model for the shape of the pulse near its peak. This approach makes simultaneous use of the model parameters, has well known statistical properties, and involves computations that can be carried out efficiently. Preliminary analyses of real data sets, using both linear and quadratic discrimination functions, yield promising results.