How to calculate Dice Similarity for 3D images

Hi all,

i have 3D image gray scale. The code is below.

[spect map]=dicomread('I-131sphere10nisbah1');
info = dicominfo('I-131sphere10nisbah1');
%gp=info.SliceThickness;
spect=(squeeze(spect));%smooth3
aa=size(spect);aa=aa(3);

The I binarize it using this code.

seedR = 58; seedC = 76; seedP = 45;
W = graydiffweight(spect, seedC, seedR, seedP , 'GrayDifferenceCutoff', 2134);
thresh = 0.00459;
[BW, D] = imsegfmm(W, seedC, seedR, seedP, thresh);
figure, imshow3D(BW)
T = regionprops('table', BW,'Area','Centroid')

Then I used this function to calculate the dice similarity,

similarity = dice(BW, spect);

But the error like below

Error using dice (line 117)
Expected input number 2, B, to be one of these types:

logical, double, categorical

Instead its type was uint16.

Answer 1

[Image Analyst]

If it's for 2-D images then you can do your own. It's a pretty simple algorithm, like the overlap (intersection) divided by the union or something similar.

[mohd akmal masud]

yah that is for 2 D images. But if I adopt into 3D images, I have to do one by one. any ways to calculate the whole 3D images?

[Image Analyst]

Not sure what you mean. Can't you just do something like

unionImage = image1 | image2;
overlapImage = image1 & image2;
diceCoefficient = sum(overlapImage(:)) / sum(unionImage(:));