User Manual

ARRHYTHMIA ANALYSIS
56 PatientNet Operator’s Manual, v1.04, 10001001-00X, Draft
All information contained herein is subject to the rights and restrictions on the title page.
Step 4 - Template Creation
Learning the Reference Beat
The learn “mode” may be initiated by any one of many possible triggers. See “Learn and
Relearn Considerations” on page 70. The purpose of the learning period is to identify or learn
the patient’s dominant beat which is accomplished by identifying the dominant morphology or
the most frequently occurring morphology during the learn period. At the beginning of every
learn, all previously stored template information is cleared.
Up to six templates are formed during this phase. Each beat is compared to the existing tem-
plates and if the beat matches, it is averaged into the template. The learning method uses a
complex mathematical method to determine the patient’s dominant or “normal” beat. If all
beats are identical, learning would complete in 30 beats. However, if the patient has a wide
variety of ECG morphologies (rare) or the data is extremely noisy, learning may take as long
as 250 beats.
Remember, the learning phase is usually short but may be delayed due to ectopy or artifact
occurring during the learn phase. While the monitor is in the learning mode, arrhythmia alarms
and trend collection, except for heart rate information, are suspended. Two lethal alarms
(ASYSTOLE and V-FIB) and the High and Low Rate alarms will break through during a learn
phase. See “Learn and Relearn Considerations” on page 70.
Step 5 - Template Comparison
After the system has learned the patient’s “normal” or dominant morphology, and subsequent
beats are detected, the system classifies each beat as normal (N), aberrant normal (Q),
supraventricular ectopic (A), ventricular (V), or unrecognized (?).
Template Matching
A template stores information about the shape of a beat. Templates serve as a reference to
which all incoming beats will be compared. A simple cross correlation measure is used to
determine if a beat matches an existing template. If the beat is not a good match based on this
measure, additional measures are made and compared (see feature extraction below).
Fig. 17. Template Matching