How to apply PCA (Principal Component Analysis) on ECG signals

Hello.

I have three ECG signals, called X1,X2,X3 for three different leads, and I want apply PCA (Principal Components Analysis) on all of them to find the component which has the least noise.

Could anybody help me?

Answer 1

[Image Analyst]

I assume you looked at the help and tried

coefficients = pca([X1, X2, X3]);

Why did that not work? What went wrong?

[djamaleddine djeldjli]

i tried it but it dose not work this is the size of the first signal hier i do not use time axis

>> whos X1

Name      Size              Bytes  Class     Attributes

X1      1x1477            11816  double

i tried it for one signal X1

load anes

coefficients = pca(X1);

Error in pca (line 23)

coefficients = pca(X1);

[Image Analyst]

But that's NOT what I said to do. How can you get PCA components when you have only one component as input? I said to use all 3 vectors:

coefficients = pca([X1, X2, X3]);

[djamaleddine djeldjli]

load('a1.mat')

load('a2.mat')

load('a3.mat')

coefficients = pca([vecta,vectz,vecte]);

all vectors are same size but i still have a problem

coefficients =

   Empty matrix: 4350-by-0

i tried coefficients = pca([vecta;vectz;vecte]);

 0.0260    0.0200
    0.0261    0.0087
    0.0262    0.0356
    0.0261    0.0324
    0.0262    0.0324
    0.0264    0.0549
    0.0264    0.0528
    0.0263    0.0459
    0.0262    0.0295
    0.0263    0.0189
    0.0263    0.0416

i think this is just two components and i need 3 output components because i have 3 input components,and how can i plot every component alone

thank you for everything

[Image Analyst]

You forgot to attach the .mat files so I can't do anything to help you other than to check the sizes of your vectors and make sure they're in columns and not empty.

I'm attaching an example of how to use PCA to determine principal components of the 3-D color gamut.

If Star (our resident physician) reads this he might tell you if this makes any sense. Even though you can do PCA on something you have to know how to interpret the components you get out. So whether a weighted sum of your different cardiac signal actually means anything meaningful is a good question. I think that, depending on which signals you combine, it may well be meaningless.