ECG Signal Analysis - ECG Acquisition and Automated Remote Processing - page 16

interval are considered as the most important in the ECG wave. The QT interval is
defined as the time from the start of the QRS complex (Q on ) to the end of the T
wave (i.e., T off ) and corresponds to the total duration of electrical activity (both
depolarization and repolarization) in the ventricles. Similarly, the PR interval is
defined as the time from the start of the P wave (i.e., P on ) to the start of the QRS
complex (i.e., Q on ) and corresponds to the time from the onset of atrial depolar-
ization to the onset of ventricular depolarization. QT interval can provide signif-
icant information for diseases related to ventricular activity. Similarly, PR interval
can provide information related to AV node conduction problems. Sometimes, the
wave shapes and their inclination- or curvature-related parameters also indicate
certain diseases, e.g., ST-segment elevation or depression may signify acute
myocardial infarction.
Detection of baseline is of paramount importance in case of ECG feature
extraction, since all the amplitude features (wave peak heights) are measured with
respect to baseline voltage.
The scope of computerized ECG analysis covers R-peak detection for rhythm
analysis and feature extraction for disease identification. Accurate detection of
ECG fiducial points, P, QRS, and T, and their respective onset and offset points for
computation of wave durations along with the heights of wave peaks are the prime
objectives of any ECG analysis software. For clinical diagnosis, the cardiologists
ask for at least 3-4 cardiac cycles of 12 lead ECG record for visual analysis. For
rhythm analysis, however, long-duration (sometimes 24-36 h) ECG records are
necessary. In principle, a cardiac cycle can be divided into a low-frequency (P and
T waves, and equipotential segments) and high-frequency (QRS complex) regions.
In principle, the detection of QRS complex is easier using statistical templates,
consisting of slope and amplitude measures. These methods are broadly classified
as event detection techniques [ 1 ]. In case of ECG signals, however, the presence of
different artifacts poses a challenge to the accurate estimation of the position of the
fiducial points. Moreover, the shape and magnitude of the ECG vary widely across
populations of different continents and determined by food habits, demographic
and hereditary factors. Hence, the different algorithms proposed over the years for
ECG signal analysis have achieved better accuracy over one another, but none has

Fig. 2.1 A typical ECG with
fiducial points

R

T

P

Q on

T on

Q

P off

P off

S off T off

S