How do i calculate Tp, Fp, Dice coefficient for a histopathological image of breast cancer. My code gives accurate results for ultrasound images, but it gives wrong results for tp, fp, and dice? please help me in this.

i am attaching the evaluation code and image with its ground truth,

function [Jaccard,Dice,rfp,rfn]=sevaluate(m,o)
% gets label matrix for one tissue in segmented and ground truth 
% and returns the similarity indices
% m is a tissue in gold truth
% o is the same tissue in segmented image
% rfp false pasitive ratio
% rfn false negative ratio
m=m(:);
o=o(:);
common=sum(m & o); 
union=sum(m | o); 
cm=sum(m); % the number of voxels in m
co=sum(o); % the number of voxels in o
Jaccard=common/union;
Dice=(2*common)/(cm+co);
rfp=(co-common)/cm;
rfn=(cm-common)/cm;

[KALYAN ACHARJYA]

Can you share the ultrasound image, where you have tested the code? The result may vary, the same algo doesn't give the accurate results for all types of the input image, therefore the selection of to ve develop code should be based on the type of input images.

[chetna kaushal]

Respected Sir Kalyan, Thankyou so much for your reply. I got the issue resolved. It was problem with logical and uint conversion of images

[KALYAN ACHARJYA]

Welcome ...OK

[KALYAN ACHARJYA]

Have you read TOM paper, regarding ROC space? I suggested you follow the paper, it is a great paper. Paper Click here

[chetna kaushal]

Respected Sir, when i run each image for one segmentation technique , it gives me different threshold value for each image. then when i run another segmentation technique, it also gives me different threshold values for each image. so then how to proceed? how can i just select one threshold and plot roc?

[KALYAN ACHARJYA]

• First, do the steps for 1 segmentation method, 1 ROC curve
• Second, do the same for another segmentation technique-2nd ROC curve
• So on..you can compare..all ROC curves from ROC space.

[arun kumar]

tq