K means clustering in DICOM files

Hi guys. I attempt to do k mean clustering to detect some small areas of very bright or very dark spots inside a DICOM file. The problem is that I can not find the right format for my DICOM images to perform the k-mean function in Matlab. My images are let's say 256*256 and when I do perform k mean clustering I get back a 256*1 matrix which I am not sure how to interpret. Any ideas would be really much appreciated! Best wishes, George

Answer 1

[Hamoon]

It'll be better idea if you do feature extraction on your image before using kmeans, but If you simply want to perform kmeans on pixel intensities of your gray level image, you can try this:

 im=imread('cameraman.tif');  % an exemplar image (256*256)
 imData = reshape(im,[],1); % convert it to an array (65536*1)
 imData = double(imData);      %  convert from uint8 to double
 IDX = kmeans(imData,2);      % perform kmeans with two clusters
 imIDX = reshape(IDX,size(im));  % a (256*256) matrix of indexes