Qrs detector (a brief study)

Centro Politécnico Superior Universidad de Zaragoza

QRS detector
Author: Alejandro Alcaine Otin Subject: Biological Signal Processing Telecomunications Engineering University of Zaragoza Course 2009/10

Centro Politécnico Superior Universidad de Zaragoza

Index
1- Introduction .................................................................................. Pag. 2 2- SingleLead QRS Detector ............................................................Pag. 2 3- Multi Lead QRS Detector ..............................................................Pag. 4 4- Single Lead Matched QRS Detector ............................................. ag. 5 P 5- Single Lead Matched Squared QRS Detector ............................... ag. 7 P 6- Final Conclusions......................................................................... Pag. 9

QRS Detector

1

Centro Politécnico Superior Universidad de Zaragoza

1- Introduction:
The QRS detection may be the main significant block into a Electrocardiogram processing device and applications. In this final subject work we developed a QRS detector based in the tree typical stages of a QRS detector, linear filtering, nonlinear transformation and thresholding decision rule. In order to evaluate our algorithm we have three leads (X, Y, Z) recorded with a 250Hz sampling rate during 8.8 minutes with contains 537 QRS complexes.

2- Single Lead QRS Detector:
The first aproximation of this QRS detector consist of developing a QRS detector by a linear filtering based in a bandpass filter centered in the main frequency of theQRS shape of the ECG signal, squaring like a nonlinear transformation, filtering by a matching filter to the typical shape of this nonlinear transformation and a final decision rule based in a fixed threshold.


The bandpass filtering may be implemented in several ways, we decide to implement a bandpass filter based in two steps, first we designed a Hamming window based Highpass filter with a cuttofffrequency of 0.5Hz in order to eliminate the DC component and baseline wander, then we designed a Hamming window based Lowpass filter with a cuttoff frequency of 50Hz in order to eliminate high frequency that not corresponds with the signal we want.

QRS Detector

2

Centro Politécnico Superior Universidad de Zaragoza

Figure 2.1 - Amplitude response of the implemented Highpass andLowpass filters

The idea of a matchig filter to the square transformation is augmenting the amplitude of the events that can be a QRS wave from the rest. This is our basic block diagram:

Figure 2.2 - Basic block diagram of a Single Lead QRS Detector

So with this procedure we need a double thresholding, the first, to identificate the locations of the first nonlinear transformation peaks, in order toimplement a matched filter to this shape, and the second is used to determinate te possible locations of a QRS wave in the ECG signal. Both of this thresholds are tunned by visual inspection of the minimum peaks of the squared signal and its matched filtered version, the results are the following:

QRS Detector

3

Centro Politécnico Superior Universidad de Zaragoza

X lead True PositivesFalse Positives False Negatives Sensibility Positive predictive value Used Threshodls 535 2 42 0,9962 0,9272 2.1 E4 1 E9

Y lead 535 2 62 0,9962 0,8961 8 E3 1,92 E8

Z lead 537 0 6 1 0,9889 5.33 E4 5.54 E9

Table 2.1 - Results of the Single Lead QRS Detector algorithm

3 - Multi Lead QRS Detector:
Now, we generalize the previous idea to the available three leads, so, the output of thenonlinear transformation is the sumation of the squaring individual signal, now in this case we using only a threshold in the final decision because the first threshold see in the previous point is now fixed. The block diagram of this implementation is:

Figure 3.1 - Block diagram of a Multi Lead QRS Detector

QRS Detector

4

Centro Politécnico Superior Universidad de Zaragoza

In this...