Methods for discriminating pre-ectopic sinus beats

One hour, high resolution 12-lead ECG recordings were digitally obtained from 10 patients with frequent isolated premature ventricular contraction (PVCs >100/hr). Recordings were processed and single beats were automatically annotated. For each lead one record per beat was generated with 102 quantitative attributes of ECG waveform For each recording, two sets of beats were randomly sampled: 200 isolated regular beats (S1) and 200 isolated premature ventricular beats (S2). In isolated premature ventricular beats the 3 preceding and 3 following beats were normal sinus beats. For each of the 200 ectopic beats, a new record having as attributes the maximum, minimum and average value of every original attribute of the 3 preceding sinus beats. Likewise, we randomly sampled 200 sinus beats such that the 3 preceding and following beats were regular as well, and, for each of the 200, created a record in the same fashion. Every time a beat was sampled, the 3 beats preceding it and the 3 beats following it were deleted from the data set, as well as the sampled beat. Thus, every record of the first set represents a pre-ectopic beat, whereas every record of the second set a non pre-ectopic beat. A two-tailed unpaired Studentpsilas T-test and other experiments not only showed that there is a statistically significant difference between the two sets, but also that, for every patient, one attribute of the new data set allows to discriminate between pre-ectopic beats and non pre-ectopic beats in 80% of the cases. These results indicate that detectable signs of incoming PVC are present in pre-ectopic sinus beats.