ECG Signal Analysis - ECG Acquisition and Automated Remote Processing - page 17

individually promised 100% perfection level. In general, the steps for ECG signal
analysis can be represented by the block diagram shown in Fig. 2.2 .
The first step A is normally referred to as 'preprocessing of ECG.' In spite of
best practices and precaution during acquisition, ECG signal is corrupted by dif-
ferent types of noises, a brief description of which is provided in Sect. 3.3.1 .
Among all, power line interference (PLI) [ 2 ], baseline wander (BW), and muscle
tremor (EMG noise) are dominant in affecting the ECG wave shapes. A detailed
discussion on motion artifacts, power line interference, electrosurgical interfer-
ence, and their minimizing techniques are discussed in [ 3 ].
QRS being the sharpest and steep region in the ECG cycle is relatively easy to
detect, and hence, Step B is most important. Many cardiac signal analysis algo-
rithms start with this step in order to capture other wave peaks with respect to
already-detected R peaks. Once the fiducial points are detected, the wave signa-
tures can be computed by simple mathematical steps. Since the wave signatures of
a normal person also vary slightly among the beats in continuous recording, Step D
and E are used to identify the variance of clinical signatures among the different
beats by statistical calculation.

Fig. 2.2 Generalized steps
for ECG signal analysis

A

Filtering for noise removal (pre-processing)

Detection of cardiac intervals, i.e., beats through
detection of R-peaks

B

Detection of fiducial points (wave
peaks, onset and offset) of individual
beats

C

D

Estimation of ECG clinical
signatures from all beats

Computation of variance of ECG clinical signatures among
different beats through statistical measures

E

Recording and presentation to
the cardiologist for clinical
diagnosis

F