replacing values with nan does not work while other approaches work (nifti files)

I have three very similar pieces of code that operate on two matrices. Firt two versions of the code work, while nan function does not and I am wondering why

Code aims to use indexes of chosen values in one matrix to remove elements of another. Specifically, I am working on neurological images and I am trying use a binary mask to remove non-brain space from another image.

Here is the code

% load the .nii result file (79x95x68 matrix)
PB = niftiread('/MATLAB Drive/Phonology 005.nii');
PB = PB(:);

% load a binary brain mask (79x95x68 matrix, 0 = non-brain space)
mask = niftiread("mask_overlap70_thr50.hdr","mask_overlap70_thr50.img");
mask = mask(:);

%% remove non-brain space from PB
% this method works and removes number of vector elements in half
PB(mask == 0) = []; 

% this method to replace values also works 
PB(mask == 0) = 5; % I can then drop values of 5 from PB

% this method does not work
PB(mask == 0) = nan; 

I have double checked for nan values with isnan which returned a vector of 0s

So why does my nan function not work? I have found this approach from the forums

Thank you for your help

Answer 1

[Steven Lord]

My suspicion is that PB is an array of one of the eight integer types. One of the things that page states about the integer types is "If you convert a NaN value into an integer class, the result is a value of 0 in that integer class."

myPB = ones(4, 'int8')
myPB(5) = NaN

Since an int8 array can't store NaN, element 5 of myPB is set to 0 instead.