How to calculate the dice similarity coefficient

Dear All,

Anyone know how to calculate the dice similarity my image data set as attached.

Answer 1

[Rik]

Let's first see what you have in those mat files:

s1=load('prediction.mat')

s1 = struct with fields:
    allBW: [130×130×90 logical]

s2=load('groundtruth.mat')

s2 = struct with fields:
    allBW: [130×130×90 logical]

So you have two equal-sized binary arrays.

Ten seconds of Googling ('wiki dice coefficient') can give you this formula:

Now you only have to implement this in Matlab.

X = s1.allBW;
Y = s2.allBW;
X = X(:);Y = Y(:); % linearize to make notation clearer
DSC = (2*sum(X&Y))/(sum(X)+sum(Y))

DSC = 0.8806

We can check this implementation by using the other notation:

% (for the calculation X and Y can either be the ground truth, the DSC will
% have the same value)
TP = sum( X& Y);
FP = sum( X&~Y);
FN = sum(~X& Y);
DSC = 2*TP/(2*TP+FP+FN)

DSC = 0.8806