Setting Background to Zero Messes Up Image Segmentation

Hi Guys,

I am using some code written by someone else to segment medical image.

The code can be found here and the initial results when run on one of the medical images is attached.

The segmentation works well initially, however I have written an extra section into the code that removes the background of the image:

im = double(im2gray(imread('LN034_T95.png')));
im = imresize(im,[512,512]);
figure(2)
imshow(im,[])
windowWidth = 15; %Sets blur level
kernel = ones(windowWidth) / windowWidth ^ 2;
imblur = imfilter(im,kernel); %blurs image
imshow(imblur,[])
level = 30;
BW = imbinarize(imblur,level);
BWFill = imfill(BW, 'holes'); %fills 'holes' inside tissue
imshowpair(im,BWFill,'montage')
for y = 1:length(im)
    for x = 1:length(im)
        if BWFill(x,y) == 0
            im(x,y) = BW(x,y);
        end
    end
end
imshow(im,[])

The problem is, that when the background is removed from the image, it seems to mess up the segmentation. See the attached images from after the background is removed (set to 0).

I have got absolutely no idea why this is happening - surely the background should all become one segment?

Thanks in advance.