change resolution of DICOM

i am trying to reduce the DICOM image resolution. i though of using imresize, but imresize does not determine a mean, it samples the nearest grid point. which is not the best way for DICOM. is there another way to do it ???

the code below i used in the first time and trying to find better way

clear
image_list=dir('MicroFinal*.dcm');
[slice_no,y]=size(image_list);

for i=1:slice_no
rph_ct=dicomread(image_list(i).name);
image_stuck(:,:,i)=dicomread(image_list(i).name); % this line to get the 3d matrix 
imgHdr = dicominfo(image_list(i).name);
rph_ctd_hu = imresize(rph_ct,[41,41]); 
rph_ctd_huu = int16(rph_ctd_hu);
dicomwrite(rph_ctd_huu, ['F' num2str(i) '.dcm'], imgHdr,'CreateMode','Copy')


end

Answer 1

[Walter Roberson]

imresize() does not normally sample the nearest grid point (but it can.) The default is bicubic interpolation. You can change the interpolation method; https://www.mathworks.com/help/images/ref/imresize.html#buxswkh-1-method

Answer 2

[Rik]

It looks like you want to interpolate a 3D image. As most image processing methods assume a 2D image, you might be better off looking at tools like interpn. See the documentation for the options. The default is an n-linear iterpolation (so trilinear for 3D).

[Walter Roberson]

I think you misread the code, Rik:

rph_ct=dicomread(image_list(i).name);
image_stuck(:,:,i)=dicomread(image_list(i).name); % this line to get the 3d matrix 

Same content is read into rph_ct and image_stuck(:,:,i) . We do not know the array dimensions for the first one, but the second one fails unless the data being read in is 2D (not 3D), so we can deduce that rph_ct is 2D since it is the same content.

imgHdr = dicominfo(image_list(i).name);
rph_ctd_hu = imresize(rph_ct,[41,41]); 

The data that we have proves that it must be a 2D array being resized, not the image stack being resized.

[Ahmad Abbas]

the data are 2D and the dimensions are 481x481

the number of slices are 391

[Walter Roberson]

Your existing code looks okay, but you might want to experiment with changing interpolation methods for imresize(). Though as I indicated earlier, the default is not to take nearest grid element.

[Rik]

For 2D images imresize should be what you need, so you should have a look at the answer by Walter.

If you want to take into account the fact that you have a 3D image stack, you can use my suggestion. You don't need to change the number of slices, but this will allow the interpollation method to consider values from neighboring slices.