Find clusters with specific value in 3D matrix

Hello,

I have an image of the size (91*109*91). The image contains voxels with values from 0 to 1. Now I would like to discard all voxels instead of those meeting the following conditions:

1. They have a value smaller than .05
2. they are part of cluster (connected voxels) of minimum 30 voxels with value < .05

At the end, I would like to have a 91*109*91 matrix/image containing only clusters of voxels with a value < 0.05.

I can figure out the first condition, what’s tricky for me is the second condition.

[Matt J]

At the end, I would like to have a 91*109*91 matrix/image containing only clusters of voxels with a value < 0.05.

And all discarded voxel values set to what? Zero? NaN?

Answer 1

[Matt J]

S=regionprops(yourImage<0.05,'Area','PixelIdxList');
S=S([S.Area]>=30); %edited

result=zeros(size(yourImage));

for i=1:numel(S)
    idx=S(i).PixelIdxList;
    result(idx)=yourImage(idx); 
    
end

[nivons]

regionprops3 probably does the trick, however the indexing to the table doesn't seem to work in your script.

After S=S(S.Volume>=30) I get the following error:

You cannot subscript a table using linear indexing (one subscript) or multidimensional indexing (three or more subscripts). Use a row subscript and a variable subscript.

Do you have any idea how to solve this?

[Matt J]

Never mind regionprops3.Here's a version with regionprops

S=regionprops(yourImage<0.05,'Area','PixelIdxList');
S=S([S.Volume]>=30);

result=zeros(size(yourImage));
for i=1:numel(S)
    idx=S(i).PixelIdxList;
    result(idx)=yourImage(idx); 
    
end

[nivons]

Great! That works, thanks!

small error though: it has to be S=S([S.Area]>=30);