Mean of many changing "for loop" 3D matrices

Hi,

I have 200 3D matrices with same dimensions. I'd like to create one 3d matrix that would contain mean of my 200 matrices by elements → (1,1,1) mean of 200 positions in my 3D matrices and so on ...

Here's my code:

for tt = 1:200;
      M =      ...         (tt)                                            loads the 3D data
  end
  mean()

Can you help me please?

Thanks.

[dpb]

mean of what terms in the 3D arrays? It doesn't take a 3D array to hold the overall mean, so what do you actually want.

Whichever it is, simply use

mn(whichever_dimensions_needed)=mean(M,dimensionOverWhichToAverage);

[Patricia Klobusiakova]

Each number in one 3D matrix represents one voxel. I'd like to take these 200 matrices and calculate the mean of each voxel in order to create one 3D matrix that could represent these 200. I was thinking about using concatenate function to create 4th dimension→create one huge 4D matrix from these 200 3D (although I cannot imagine such a matrix) and then calculate mean along 4th dimension, but I don't know how to do that when I have a for loop there and matrix M changes with every iteration.

[Patricia Klobusiakova]

for ii = 1:nsub

    for tt = 1:200;
        M = spm_read_vols(spm_vol(files(tt).name(ii,:)));
    end
    mean(M,4)

Answer 1

[dpb]

Well, you could do that (store in a 4D array, that is) but that requires memory for all 200 at once and there's no need when computing a mean to have all the data at once...just accumulate the totals and then divide after the loop is finished...

    M = spm_read_vols(spm_vol(files(1).name));  % read the first into accumulator array
    for tt = 2:200
       M=M+spm_read_vols(spm_vol(files(tt).name));  % accumulate the total
    end
    M=M/200;    % average

[Patricia Klobusiakova]

That's cool. I didn't think of that. Thank you very much!

Answer 2

[Kwin]

If you have your matrices all stored in one (four dimensional) matrix, for example M(:,:,:,k) would return the kth 3D matrix, but another order of the index would be fine as well, such as M(k,:,:,:), then, as dpb already mentioned, you can get the mean by simply using:

mean(M,location_of_k)

so for example for M(:,:,:,k), location_of_k=4 and for M(k,:,:,:), location_of_k=1.

[Patricia Klobusiakova]

Thanks for your answer! Could you take a look on the comment I wrote under my question? The problem is I am not able to create a 4D matrix because of a for loop I have in the code.