Applying a mask to 3D Data

Hi everyone, I'm trying to generate a histogram of some data and I want to apply a 3D mask to the data. I have written as below, where fw is the data set that I'm trying to mask:

nbins = 20;

mask = find(mask);
selected_voxels = fw(mask); % ?
counts = histcounts(selected_voxels, nbins);

range = linspace(0, 1, nbins);
bar(range, counts)

I've chosen to use bar instead of MATLAB's histogram because I want to be able to place multiple bars beside one another and I've not found a way to do that using the basic histogram function. I used the find function since the values in my "mask" are not necessarily binary. However, a large number of voxels are being put in the highest bin.

The correct histogram should look as below, which I've plotted with Excel:

Does anyone have any suggestions on how to fix my histogram? Notably, the highest bin should not have a peak and the number of voxels included in the histogram should be significantly lower. I have included both fw and the mask in .mat format.

Thank you,

Warren

Answer 1

[Anton Kogios]

nbins = 20;
mask = mask==1;
selected_voxels = fw(mask); % ?
counts = histcounts(selected_voxels, nbins);
range = linspace(0, 1, nbins);
bar(range, counts)

[Warren Boschen]

The distribution looks better, but this only includes values of 1 in the mask, whereas I want to use any value that is non-zero. How would I change what you have to do that?

Answer 2

[Walter Roberson]

Your fw is 128 by 128 by 72 which is 1179648 elements.

Your mask is 186456 x 1. That is less than 1/6th of the size of fw.

Are you sure the mask is the right size?? MATLAB will effectively pad out the mask with false to be the same size as the array, but it seems very odd that you would pull out that particular part of the original data.

[Warren Boschen]

Sorry about that! The imported mask should also be 128x128x72. I included the correct mask as a part of this comment. The mask is not the right size, which is why I asked this question in the first place. Since find(mask) produces a much smaller set than the original size, I can't reshape it to the correct 128x128x72. Is there a better way to get about this?

[Walter Roberson]

A lot of your selected voxels are exactly 1. As long as your mask is correct, there should be a large number in the last bin in the plot.

nbins = 20;

load fw
load mask_correct
selected_voxels = fw(logical(mask));

nnz(selected_voxels == 1)

ans = 12626

counts = histcounts(selected_voxels, nbins);
range = linspace(0, 1, nbins);
bar(range, counts)