Biomedical Engineering Reference
In-Depth Information
(or p min ), it is concluded that the amplitude threshold-based search might have
captured a T wave. Hence, the next search is initiated from 200 samples downside
along ECG array. The algorithm steps are given in the end of the chapter Appendix 3.
Tables 2.5 and 2.6 represent some of the test results with ptb-db data and mit-db data,
respectively. Figure 2.12 a and b represent R-peak detection, baseline modulation
correction using ptb-db and mit-db data, respectively.
2.4.2 Baseline Detection
The baseline point in each R-R interval is determined in the TP segment, which is the
equipotential region between the T-offset point of current beat and P-onset point of
following beat. For baseline point detection, a typical RR interval can be assumed to
be consisting of two halves, with the left half containing T wave of the current beat
and right half containing P wave of the following beat. The TP segment is nominally
considered to be spanned between ±200 samples on either side of hypothetical mid-
RR line. Since the algorithm is tested on noisy samples from Physionet, the average
slope approach is used for minimizing abrupt fluctuations in the data. In the selected
TP segment, baseline point is taken as the index with minimum absolute average
slope over ±15 points. The baseline modulation is then applied to all samples
between two successive baseline points by shifting them vertically up or down by an
empirically determined formula, given in Eq. 2.16 . The formula is based on the fact
that a sample, positioned between two baseline points, which are vertically offset, is
to be shifted in proportion to a fixed reference. To apply the correction throughout the
dataset, this reference point is taken as first baseline point for the ECG dataset.
o k ¼½ x ð k Þ blp ð i Þ D
L
ð 2 : 16 Þ
where o k is the vertical correction to be applied to a sample, x ð k Þ; blp ð i Þ is the
reference point, 'D' is the vertical offset, and 'L' is the index difference, respec-
tively between two successive baseline points. The baseline correction is illus-
trated with the help of Fig. 2.13 .
2.4.3 Determination of Fiducial Points
The detection of fiducial points from the ECG data is performed by a window search
method, starting form the R peak of the respective cycle. For data analysis, we
considered the following window sizes for different wave durations and intervals.
P wave: 100 ms
QRS: 120 ms
T wave: 160 ms
ST segment: 120 ms.
Search WWH ::




Custom Search