Welch Allyn Means ECG Physicians Manual for CP Series Electrocardiographs - User Manual - Manual (Page 4)

MEANS Physicians Manual
Welch Allyn
4
1 Introduction
Computers and humans interpret ECG signals in fundamentally different ways. The principal
difference is in the manner in which a computer “looks at” the signal. To be interpretable, a
continuous (analog) signal must be converted into numbers, i.e., digitized. The signals are
measured at short intervals, and the measured values (the samples) are stored as digital
numbers. On this set of numbers the analysis must take place. The sampling must be dense
enough to ensure sufficient fidelity in rendering the original analog signal. Current standards
for ECG recording recommend a sampling rate of 500 Hz or higher.
After collection of the data, the processing follows a number of successive stages:

Signal conditioning

Pattern recognition

Parameter extraction

Diagnostic classification
Each of these steps must be performed correctly to ensure a satisfactory final result. If, for
instance, the signals are not correctly cured of disturbances this may result in a faulty
waveform recognition. The diagnostic classification is then likely to come out wrong. The
successive steps will now be discussed more extensively.
1.1 Signal conditioning
The ECG signal can be disturbed in several ways:

Continuous noise of a single frequency, sometimes with higher harmonics, due to 50 or
60 Hz AC mains interference.

Drift: more or less gradual baseline shifts, e.g., caused by respiration.

Bursts of noise of mixed frequencies and various amplitudes due to electrical signals from
active muscles.

Sudden baseline jumps due to changes in electrode-skin impedance.

Spikes: isolated, large amplitude variations of short duration.

Amplitude saturation of the signal.
To correct these disturbances, several techniques have been used. Mains interference is
suppressed by an adaptive filter that estimates the coming noise estimates and subtracts the
estimates from the encountered signal. Baseline shift is corrected by simply connecting the
onsets of successive QRS complexes by straight lines and determining the signal amplitudes
with respect to these line segments. Beat selection and averaging (see below) help to reduce
disturbances of muscle noise. If a disturbance is detected that may affect the diagnostic
classification, the program issues a warning.