Syntactic recognition of common cardiac arrhythmias

Outlines a syntactic approach to the recognition of common cardiac arrhythmias within a single ambulatory ECG trace. This methodology essentially involves the annotation of an electrocardiogram trace in terms of the syntax primitives and the subsequent parsing of these annotations into various syntactic forms that describe their appropriate arrhythmia. The syntax primitives, which the authors collectively term arrlets, are a set of curves, which are modelled as a series expansion of orthonormal hermite basis functions. By using as features the parameters of this model, a probabilistic neural network is then employed to detect the occurrences of arrlets within an ECG trace. The approach was evaluated using data from the MIT-BIH arrhythmia database