Create a 3D cell with vector elements of size 1x5 from five 3D matrices

Hello!

I have a structure of five 3D images (double,size 181x213x41). I would like to create a cell from this data, of same size as the 3D images (181x213x41) where each element will be a vector of size 1x5. I.e. the first vector element comes from 3D image 1, the second element from 3D image 2 and so on.

Example:

cellImage{i,j,k} = [1 2 4 1 0];

where

structure(1).image(i,j,k) = 1

structure(2).image(i,j,k) = 2

structure(3).image(i,j,k) = 4

structure(4).image(i,j,k) = 1

structure(5).image(i,j,k) = 0

I can do this by looping:

for i = 1:5
     for j = 1:181
         for k = 1:213
             for l = 1:41
                 cellImage{j,k,l}(i) = structure(i).image(j,k,l);
             end
         end
     end
end

This takes a very long time however. I want something similar to below (but that doesn't work).

for i = 1:5
    cellImage{:,:,:}(i) = structure(i).image(:,:,:);
end

Does anyone know of a better way of doing this?

Thank you!

Answer 1

[Stephen23]

This can be done without any loops which should be a lot faster. The basic concept is to join all of your data together into one numeric array, and then split it again to get the final vectors.

% Fake data: 1x2 struct, 2x3 numeric arrays, all unique values:
A(1).data = [9,8,7;6,5,NaN];
A(2).data = [0,1,2;3,4,Inf];
% Convert to a simple numeric array:
B = cat(4,A.data); % join along the fourth dimension :)
% Create dimension "indices":
C = arrayfun(@(n)ones(1,n),size(B),'UniformOutput',false);
C{4} = size(B,4);
% Use the "indices" to split the numeric array into a cell:
D = mat2cell(B,C{:});
% Optional: tidy-up by removing superfluous dimensions:
D = cellfun(@squeeze, D, 'UniformOutput',false);

Save this in a script and run it. Then in the command window you can check that it has grouped the data together corresponding to the same location in each of the original separate numeric arrays:

>> D{1,1}
ans =
     9
     0

>> D{1,3}
ans =
     7
     2

>> D{2,3}
ans =
     NaN
     Inf

Because I concatenated along the fourth dimension, it should also work for your 3D arrays.

Important: Do not use i or j as variable names, as these are both names for the inbuilt imaginary unit .

[Ida]

Hi Stephen and thank you for your response!

Your suggestion worked perfectly! =) Although the step

D = mat2cell(B,C{:});

still takes a bit of time, but is ok. Thank you again!