How to find Q and S point in QRS complex of ECG signal?

How to find Q and R points of ECG signal shown below?

Code:

 load 16265m.mat;
 plot(val(1,:))

[Explorer]

This is exactly what I want to do. How do I find Q and S points for my signal.

The image is taken from MATLAB Documentation

[Explorer]

My attempt to code:

D=load('16265m.mat');
t=1:1280;
ECG_data=D.val(1,:);
% plot(t,ECG_data)
% grid on,

[~,locs_Rwave] = findpeaks(ECG_data,'MinPeakHeight',0.5,...
                                    'MinPeakDistance',200);

   ECG_inverted = -ECG_data;
  [~,locs_Swave] = findpeaks(ECG_inverted,'MinPeakHeight',0.5,...
                                          'MinPeakDistance',200);

figure
hold on
plot(t,ECG_data);
plot(locs_Rwave,ECG_data(locs_Rwave),'rv','MarkerFaceColor','r');
plot(locs_Swave,ECG_data(locs_Swave),'rs','MarkerFaceColor','b');
%axis([0 1850 -1.1 1.1]); 
grid on;
legend('ECG Signal','R-waves','S-waves');
xlabel('Samples'); ylabel('Voltage(mV)')
title('R-wave and S-wave in Noisy ECG Signal')

smoothECG = sgolayfilt(ECG_data,7,21);

figure
plot(t,ECG_data,'b',t,smoothECG,'r'); grid on
axis tight;
xlabel('Samples'); ylabel('Voltage(mV)');
legend('Noisy ECG Signal','Filtered Signal')
title('Filtering Noisy ECG Signal')

[~,min_locs] = findpeaks(-smoothECG,'MinPeakDistance',40);

% Peaks between -0.2mV and -0.5mV
locs_Qwave = min_locs(smoothECG(min_locs)>-0.5 & smoothECG(min_locs)<-0.2);

figure
hold on
plot(t,smoothECG);
plot(locs_Qwave,smoothECG(locs_Qwave),'rs','MarkerFaceColor','g');
plot(locs_Rwave,smoothECG(locs_Rwave),'rv','MarkerFaceColor','r');
plot(locs_Swave,smoothECG(locs_Swave),'rs','MarkerFaceColor','b');
grid on
title('Thresholding Peaks in Signal')
xlabel('Samples'); ylabel('Voltage(mV)')
ax = axis; 
%axis([0 1850 -1.1 1.1])
legend('Smooth ECG signal','Q-wave','R-wave','S-wave');

Still points are not detected correctly. Please help me in improving this code.

[anas khan]

D.val error show why sir plz guid me

Answer 1

[Star Strider]

The first trace doesn’t have any S-waves, so don’t look for them. The EKG morphology is dependent on the lead placement and heart health. The second one is a very sick heart, with a left bundle-branch block pattern. It also has an RSR' pattern, so you have to be careful to not detect the first R-deflection.

My code:

q = load('Explorer 16265m');
EKG1 = q.val(1,:);
EKG2 = q.val(2,:);
Fs   = 128;
t    = [0:size(EKG1,2)-1]/Fs;

[R1,TR1]  = findpeaks( EKG1, t, 'MinPeakHeight',200);
[Q1,TQ1]  = findpeaks(-EKG1, t, 'MinPeakHeight',100);              % NOTE: No ‘S’ Waves In EKG1

[R2,TR2]   = findpeaks( EKG2, t, 'MinPeakHeight', 50);
[QS2,TQS2] = findpeaks(-EKG2, t, 'MinPeakHeight', 75);

figure(1)
subplot(2,1,1)
plot(t, EKG1)
hold on
plot(TR1, R1, '^r')
plot(TQ1, -Q1, 'vg')
hold off
grid
axis([0  2    ylim])
legend('EKG', 'R', 'Q')
subplot(2,1,2)
plot(t, EKG2)
hold on
plot(TR2, R2, '^r')
plot(TQS2(1:2:end), -QS2(1:2:end), 'vg')
plot(TQS2(2:2:end), -QS2(2:2:end), 'vb')
hold off
grid
axis([0  2    ylim])
legend('EKG', 'R', 'Q', 'S')

My plot:

[Explorer]

@ Star Strider

I am getting error after line # 7 of your code i.e.

 [R1,TR1]  = findpeaks( EKG1, t, 'MinPeakHeight',200);

Error:

 Error using uddpvparse (line 122)
 Invalid Parameter/Value pairs.

 Error in findpeaks>parse_inputs (line 84)
 hopts = uddpvparse('dspopts.findpeaks',varargin{:});

 Error in findpeaks (line 59)
 [X,Ph,Pd,Th,Np,Str,infIdx] = parse_inputs(X,varargin{:});

I faced this error earlier too and asked about this error here but did not get answer.

[Star Strider]

It ran without error for me (in R2015b).

What version of MATLAB are you using? I am aware that there are version differences in findpeaks. Also, MATLAB is case-sensitive, and in some earlier versions, upper-case arguments are required, for example:

[R1,TR1]  = findpeaks( EKG1, t, 'MINPEAKHEIGHT',200);

[Explorer]

MATLAB Version: 8.1.0.604 (R2013a)

[Star Strider]

The R2013a documentation for findpeaks tells me that your findpeaks call should be:

[R1,TR1_Idx]  = findpeaks( EKG1, 'MINPEAKHEIGHT',200);
TR1 = t(TR1_Idx);

and so for the others. My code should then work with R2013a. I don’t see any other potential incompatibilities.

The current online documentation is for R2015b, so it is not guaranteed to be applicable to or work with prior versions.

[Explorer]

Thanks! Yes, line # 7 is running fine. I tried to do same for line # 8 but again facing same error.

 [Q1,TQ1_Idx]  = findpeaks(-EKG1, t, 'MINPEAKHEIGHT',100);              % NOTE: No ‘S’ Waves In EKG1
 TQ1 = t(TQ1_Idx);

Am I missing something?

[Star Strider]

Yes. You have to take out the ‘t’ argument in every one of your findpeaks calls for it to work in R2013a.

[Explorer]

Thank you very much!

[Star Strider]

My pleasure!

[Explorer]

@ Star Strider

Don't you think that first trace must also have S wave as this trace is of Normal Sinus Rhythm? And all recordings of NSR must have all wave. Is there somethings wrong with PhysioNet Dataset of NSR?

[Star Strider]

No, I don’t.

The cardiac electrophysiologic cycle traces out three loops in 3D space and time corresponding to the P-wave, QRS complex, and T-wave. A reconstruction of this loop from the information in the scalar EKG is known as the vector EKG, or vectorcardiogram. The scalar EKG (the 12 traces of amplitude as a funciton of time in a normal full EKG record) are the projections of that loop on lines connecting the various leads, and on the augmented (derived) ‘A’ and precordial ‘V’ leads. So depending on the lead, various parts of those loops will appear more or less prominently as projected on it. Of the QRS complex, the R-wave always appears in every lead, but the entire QRS complex (as distinct individual components) does not have to appear in any given lead. (So the S-wave does not always have to distinctly appear in a given lead.) The morphology of the QRS complex depends on heart health, anatomic position of the heart (characteristic of individual thoracic anatomy), and the particular lead of the scalar EKG you are examining. Lead placement is fairly standard, so that is not usually an issue.

Nothing is wrong with any of the 6-lead PhysioNet Datasets that I’ve looked at, although some 12-lead records have confusing V-lead recordings with no obvious R-wave progression, a distinct abnormality that would never be seen in a correctly-recorded EKG. You’re looking at one of the standard limb leads (I believe Lead II), and all of them I’ve seen are of high quality.

That you are not seeing the S-wave in one of your recordings is simply a normal variant. I remember that particular tracing, and was not concerned about it.

[Explorer]

Okay, thank you for detailed answer.

I was finding Q R and S points to find slope of QR wave and RS wave.

In trace 1, there are Q and R points. So slope of QR can be found using slope formula i.e. (y2-y1)/(x2-x1)

But the problem here I am encountering is number of Q points and R points are not equal. What should I do now?

Probably this is because false Q points are detected especially after 300 (x-axis).

 axis([300  330    ylim])

[Star Strider]

My pleasure. Electrocardiography and the principles behind it are not trivial to learn.

I didn’t look at that signal in detail, but the number of Q-waves and R-waves should not differ by more than 1, if the EKG trace was cut off in the middle of a QRS complex. Otherwise, experiment with different values of 'MINPEAKDISTANCE' in your findpeaks call to be sure the peaks it returns aren’t too close to each other. Do that for both the Q-wave and R-wave findpeak calls. You can use the same value for 'MINPEAKDISTANCE' for both once you detetermine the best value for it.

Answer 2

[Shatha Jazar]

please can you predict the R wave in my signal ?

Answer 3

[tursun wali]

Dear @Star Strider, ----------------------------------- Answer by Star Strider on 9 Feb 2016 Accepted answer ----------------------------------- Your result is impressive and what I want. I want to find height of Q, R and S peaks ; so far find R and S not Q peak I have tested you code (I've changed one line:q = load('16265m.mat'); ) It shows this error message: ----------------------------------- Error using uddpvparse (line 122) Invalid Parameter/Value pairs.

Error in findpeaks>parse_inputs (line 84) hopts = uddpvparse('dspopts.findpeaks',varargin{:});

Error in findpeaks (line 59) [X,Ph,Pd,Th,Np,Str,infIdx] = parse_inputs(X,varargin{:});

Error in QRS_detect_tursun6 (line 6) [R1,TR1] = findpeaks( EKG1, t, 'MinPeakHeight',200); -----------------------------------

I would be very happy if you tell me how to fix it? Well actually My signal contains only one EKG (attachment: person6.mat): I want to find height of Q, R and S peaks ; so far find R and S not Q peak

[Star Strider]

I am not certain that you can ‘fix’ it. There may be normal Q-waves, but it is difficult to tell.

Your EKG is very noisy, so you would have to filter it (with a bandpass filter to eliminate the high-frequency noise as well as the varying baseline) using a filter designed with a sampling frequency derived from a time vector you did not provide.

It is a distinctly abnormal EKG with each ‘normal’ QRS complex followed by a wide complex of some sort.

The ‘complexes’ could be peaked T-waves of the sort seen in hyperkalemia (or the effect of an unknown drug), but in the absence of an accurate time base (in seconds) and amplitude scale (in volts), I cannot determine the heart rate or the amplitudes and intervals necessary determine what the components of your EKG are.

Designing the filter is not difficult with the designfilt function. It would be best to begin by examining the spectrum of your signal by taking the fft (link) of it. That would give you the best information as to what are your signal frequencies and what are your noise frequencies. Use a passband of 1 Hz for the low passband, and at least 50 Hz to at most 100 Hz for the high passband, depending on what gives the best results. Use the filtfilt function to do the actual filtering.

[sania urooj]

@Star Strider I have used your above code for detecting QRS peaks. It's working but my task is to detect peaks of multiple ECG signal that are already stored in one mat file. How can I do it as each signal may have different peaks ? Please help me.

[Christopher McCausland]

@sania urooj I would suggest creating a new question. However to run any code on diffrent arrays a simple 'for' loop of the lenght/width (depending on the orentation of the ecg leads) of the data should surfice. Remeber to include an array within the loop to store loop data in and a second array outside the loop to write the loop data too once the loop completes.

Hope this helps!

For 1:width(ecgchanneldata)
    % do this code

    end