ECG Compression - ECG Acquisition and Automated Remote Processing - page 99

An important criterion for ECG compression is the clinical validity of the
reconstructed signal for diagnosis. In [ 33 ], some distortion measures based for
QRS complex, P and T waves are proposed. These include duration and shape of
QRS, P and T waves, their durations, ST segment elevation, etc. In [ 34 ], the
authors proposed a total of 18 distortion measures in decompressed ECG signal,
which included magnitude, slope, and shape of all constituent wave and segments
of ECG. A critical discussion on traditional distortion measures (like PRD) has
been carried out in [ 35 ] which advocates for wavelet-based diagnostic distortion
measure to obtain superior results. The same authors proposed a multiscale
entropy-based wavelet distortion measures in [ 36 ].

5.3 Proposed ECG Compression Scheme

In the following sections, an ECG encoding technique based on direct data
compression (DDC) is described. The scheme is developed for off-line compres-
sion of single-channel ECG using 8-bit resolution and an adaptation of delta
modulation. The technique is computationally simple and found equally suitable
for tele-ECG applications as well as archiving of ECG data.
A typical ECG cycles can be divided into two zones, viz. QRS, where fluctu-
ation is high, and non-QRS, where fluctuation is low. Now, if the first difference
(also called successive sample difference, or SSD) of the ECG sample is computed
and compared with the original ECG plot (Fig. 5.1 ), it is seen that the fluctuations
in the SSD array is confined to almost 15% of the total array, in QRS regions, as
indicated by red color as shown in Fig. 5.1 .
The fluctuations in the non-QRS zones are either minimum (P and T waves) or
almost nil (isoelectric segments), of the order of 10 -2

in millivolt scale. In the

Fig. 5.1 ECG with first difference (SSD) plot