8
$\begingroup$

I have a project about extracting features from a homemade ECG. I need the PQRST points from the signal of the heart beat. The ECG I'm using has 3 electrodes that attach to the chest.

The problem is I don't know where I should define the PQRST points because the result does not look much like an ECG. I want to try to create a program later that can find those points.

Here are the raw ECG traces data taken from 3 subjects that were alcohol-intoxicated (0.2 to 0.3 promille):

enter image description here enter image description here

I'm guessing the red and orange points here. And I'm pretty sure the green points are accurate located:

enter image description here

Green points is easy to locate but the orange is a little harder and the red ones are really hard to pin point out. Those points can be used to define the length of different complexes, intervals and segments of PQRST.

Which points is safe to extract? Where exactly are those points located in the graph?

Is the data too inaccurate for finding abnormalities?

Is there any other useful feature, beside heart rate, I can extract from my graph?

$\endgroup$
5
$\begingroup$

First bump is called P point, middle is QRS wave (as in Q-down, R-up, S-down), and last bump is T. Every feature is representation of electrical activity in certain region of heart. E.g. P bump is contraction of atrium. See more here. Bumps are P/QRS/T.

The reason why you see so many different recordings is that it was taken with many-lead ECG. Heart has very distinct shape, its atrium is on top of ventricle, it is a bit tilted to the right, etc. Your recording is a projection of electrical activity vector onto vector that connects two leads used for given recording. Pick two other electrodes and you will have different picture just because axis of projection changed.

heart ECG vectors

via NIH

It seems to me that recording you show are from different pairs of electrodes, but it shouldn't matter. As you can see, different pairs of electrodes will "pick up" some processes better that the other.

Now, because ECG is electrical activity recording, heart rate is least useful information to extract from it. Measuring pulse will be less of a hassle and as accurate. But what you can do is estimate biophysicals properties of given heart. For example, distance between P and QRS (PR interval) complexes is time that it takes for heart to contract ventricle after atrial contraction. Any abnormalities will cause lengthening of PR or other change. Lengthening of PR interval often indicates 1st degree AV block (AV node is not conducting as usual).

There is bunch of information out there on ECG and abnormalities, you can find useful this page from U Utah with examples of many severe conduction abnormalities.

$\endgroup$
  • $\begingroup$ Can you please point out any algorithm/pseudocode which can find P / QRS / T in channels, please. I think there exist many of them but I would like to understand some basics how to do it. $\endgroup$ – Léo Léopold Hertz 준영 Oct 26 '16 at 16:18
  • $\begingroup$ @Masi what do you mean $\endgroup$ – aaaaaa Oct 28 '16 at 5:26
  • $\begingroup$ Your answer is too generic. I would like to get some examples how you really analyse ECG with basic examples of false negatives etc QRS complexes with suspected PVCs in the immediate surrounding. $\endgroup$ – Léo Léopold Hertz 준영 Oct 28 '16 at 9:17
  • $\begingroup$ @masi well, then I would refer you to medical literature. Or some online knowledge bases, like ecglibrary.com/ecghome.php $\endgroup$ – aaaaaa Oct 28 '16 at 20:54
2
$\begingroup$

The only abnormality which can be seen on the ECG curve is a prolonged PR Interval but only when acute alcohol poisoning has been done.

In the referenced paper the patient had consumed 3.7 g/l (3.7 per mille) which resulted in a PR prolongation of 200-300 ms. The normal PR interval is different for everybody but it normally is 0.12 to 0.20 seconds. A long PR interval will result in a first degree heart block. In your case (0.2 per mille) I doubt that you are going to see any abnormalities. But if the ECG sensor has a good sampling rate then maybe there is a chance.

The PR interval reference 1 is defined as shown here: enter image description here

PR interval reference 2:
enter image description here

Further reading:

  1. Normal ECG
  2. Third-degree atrioventricular block in an adolescent following acute alcohol intoxication. van Cleef AN1, Schuurman MJ, Busari JO.
$\endgroup$
  • $\begingroup$ I think you should avoid the word only here. I would really like to see any pseudocode/software how you can detect those properies and waves in each basic ECG signals. $\endgroup$ – Léo Léopold Hertz 준영 Oct 26 '16 at 16:20
1
$\begingroup$

Beautiful work!

Whether the PR is prolonged depends on timescale om x axis (which i cannot find...).

I didn't really get the scope of this task. If you are attempting to create a computer automated interpretation of the crucial ECG intervals, waveforms and points, you are in for a very tough task (if you want it to be clinically safe). This is the subject of intense research with thousands of patents applying to it. There are also open-source software (i think MIT and Matlab has some resources on it.

Some good resources:

http://circep.ahajournals.org/content/6/1/2.full

http://circ.ahajournals.org/content/115/10/1306.abstract?ijkey=f624706cb1aa52e719908ba6fb39726f4434bcd9&keytype2=tf_ipsecsha

Www.lifeinthefastlane.com

www.ecgwaves.com (my site)

Www.bem.fi (good om electricity aspects).

$\endgroup$
  • 1
    $\begingroup$ This does not answer the question - answers are expected to stand alone and have links and references for further reading. Additionally one could see the promotion of your own website as critical. $\endgroup$ – Chris Oct 17 '16 at 12:15
  • $\begingroup$ you're right. i've modified the answer and included a reference to Kligfield et al which is a good starting point. $\endgroup$ – Adam Robinsson Oct 17 '16 at 12:34
  • $\begingroup$ Please, no link-only answers without explaining specifically what you include. Your answer is pretty hard to read. $\endgroup$ – Léo Léopold Hertz 준영 Oct 26 '16 at 16:21
0
$\begingroup$

Additional resources for EKG tracings with analysis:

Practical Clinical Skills
EKG Academy (my site)

Bill

$\endgroup$
  • $\begingroup$ This is not commercial as far as I see. $\endgroup$ – Chris May 24 '17 at 14:23
  • 1
    $\begingroup$ @Chris Isn't it considered spam. $\endgroup$ – user237650 May 24 '17 at 15:34
  • 2
    $\begingroup$ @Mesentery I would simply avoid that - the cases where people post own websites are often spam... $\endgroup$ – Chris May 24 '17 at 16:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.