Continue to Site

# [Moved]: ECG features Extraction

Status
Not open for further replies.

#### Humaira

##### Newbie level 4
Hey everyone!
I am working on sleep apnea detection using ECG. the issue is now that I have done RR peak detection RR interval calculation but please can anyone help me, in how to extract RR features? my working environment is MATLAB.

Re: ECG features Extraction

An ECG circuit measures the heart rate and its waveform, not the breathing rate.
The sleep apnea condition is when a sleeping person stops breathing for a moment that has not much to do with detecting the heart function, except the heart stops beating when the patient with severe apnea dies.

Re: ECG features Extraction

But my approved thesis is apnea detection using classifiers and for that I need RR features

What is a "classifier"?

Did you read the definition of sleep apnea?

Where I have used the word of ECG cricuit?
I am working in MATLAB R is the peak in ECG signal I think u are totally unfamiliar with this and I know very well about apnea I know its pause in breathing
I have chosen thesis after complete literature

Humaira,

Apnea monitoring and ECG monitoring use different circuits and emploies different sciences on their analysis. the ONLY relation between both is that once one definitelly stops, the other too.

I don't know how to explain my problem

ECG is one of many different methods used to predict Obstructive Sleep Apnea (OSA)
Feature extraction can be done in many ways such as DFT as well done in this **broken link removed**.

It is basically is the process of normalization, differential computations over various time frames, histograms of various bin sizes, adaptive and fixed, to transform the SNR of irrelevant data into a SNR of useful data , whose features that are unique to this problem.

The concept requires measuring and statistical analysis of Microsoft standard datasets to determine the False Positive and False Negative error rates for predicting OSA before it occurs. ( missing a true OSA signature or detecting a false OSA signature)

There may be other telltale features in the QRS wavelet. There a great many thesis on this topic.

I would approach it from a method of conversion to phonems such as the method used for voice recognition.

Good luck

It looks like this thesis has already been done. Are you allowed to copy it or repeat it and call it "your" thesis??

I have to modified the algorithm by using different classifiers and features and then comparison

Re: [Moved]: ECG features Extraction

Humaira
I think I know little bit of what is being discussed.
You have ECG data (time series).
You isolated the waves containing QRS complex.
and measure the inter beat ((RR) interval from R peak temporal separation.

The waves would look something like this.
if the baseline wandering is removed and temporally aligned with R peaks they would almost superimpose near R peak and some what deviate else where.
What is the level of research M.Phil or Ph.D, based on that some classification approach could be planned to fit the requirement and the time frame.
Simplest way is to use your RR interval as time series and find out mean, mode, median, Standard Deviation, these are the features you are looking for, to begin with.

Last edited:
Sally_A and Humaira

Points: 2

### Sally_A

Points: 2
Re: [Moved]: ECG features Extraction

Thanks a lot my thesis level is of MS Engineering thesis

Regarding features what you were looking for is clear and your purpose is served or what?
Other features from frequency domain etc are bit difficult to handle.
SD1 and SD2 from poincare also perform well in classification.

#### Attachments

• Poincare 1475-925X-8-17.pdf
464.3 KB · Views: 82
Last edited:

Other features from frequency domain etc are bit difficult to handle.
SD1 and SD2 from poincare also perform well in classification.

Although this technique brings promising results on detection of the cardiac arrhythmias concerning to the RR intervals, it is in essence a post analysis, which requires a big amount of samples enough to infer a reliable result. It is also particularly suitable to either detect if the patient is experiencing physiological or psychological stress.

The apnea disease on the other hand, requires an immediate attention, therefore the application of this algorithm should be considered not to provide life support, but as an additional feature.

I am working on sleep apnea detection using ECG. the issue is now that I have done RR peak detection

Perfect

The only useful contribution made by me here is an attempt to provide a direction for features.
Other possibilities will have to be examined, as to whether this works, if so to what extent!

andre_luis

### andre_luis

Points: 2
Ready data from standard sources available on net is used or you generate your data, if so, what ECG device and what format?
A Raoof Khan

Hey everyone!
I am working on sleep apnea detection using ECG. the issue is now that I have done RR peak detection RR interval calculation but please can anyone help me, in how to extract RR features? my working environment is MATLAB.

First take the diff of data
this will give you R peaks.
if you can do this then RR interval will be easy to find
Good Luck

Status
Not open for further replies.