How can I calculate the iou and dice for each slices in a CT volume to obtain the mean for all the volume?

조회 수: 3 (최근 30일)

Wan Faiz 2022년 5월 29일

0
링크

이 질문에 대한 바로 가기 링크

https://kr.mathworks.com/matlabcentral/answers/1729550-how-can-i-calculate-the-iou-and-dice-for-each-slices-in-a-ct-volume-to-obtain-the-mean-for-all-the

댓글: Image Analyst 2022년 5월 29일

This is my code:

load("25_1_50epochs.mat")
Error using load
Unable to find file or directory '25_1_50epochs.mat'.
destinationSkullTest11 = "C:\Users\User\Documents\SEM 7\FYP\Dataset\U-net\Image Test";
destinationMasksTest11 = "C:\Users\User\Documents\SEM 7\FYP\Dataset\U-net\Mask Test";
destinationPredictionFolder = "C:\Users\User\Documents\SEM 7\Deep Learning\Predicted Test";
pixelLabelID = [1 0];
classNames = ["fracture" "background"];
SkullFiles = dir(destinationSkullTest11);
SkullDirectoryNames = {SkullFiles([SkullFiles.isdir]).name};
SkullDirectoryNames = SkullDirectoryNames(~ismember(SkullDirectoryNames,{'.','..'}));
MaskFiles = dir(destinationMasksTest11);
MaskDirectoryNames = {MaskFiles([MaskFiles.isdir]).name};
MaskDirectoryNames = MaskDirectoryNames(~ismember(MaskDirectoryNames,{'.','..'}));
for z = 1:length(SkullDirectoryNames)
    fullfileSkull = fullfile(destinationSkullTest11,SkullDirectoryNames(z));
    fullfileMask = fullfile(destinationMasksTest11,MaskDirectoryNames(z));
    imdstest11 = imageDatastore(fullfileSkull,"FileExtensions",".png","LabelSource","foldernames");
    pxdstest11 = pixelLabelDatastore(fullfileMask,classNames,pixelLabelID,"FileExtensions",".png");
    mkdir(fullfile(destinationPredictionFolder, MaskDirectoryNames{z}));
    targetFolder = fullfile(destinationPredictionFolder, MaskDirectoryNames(z));
    pxdsResults11 = semanticseg(imdstest11,net,"WriteLocation",targetFolder);
    for t = 1:length(pxdsResults11.Files)
        gt = readimage(pxdstest11,t);
        pr = readimage(pxdsResults11,t);
        confmat = segmentationConfusionMatrix(pr,gt);
        % IoU = TP / (TP + FP + FN)
        numeiou(t,:) = confmat(1,1);
        denoiou(t,:) = confmat(1,1) + confmat(1,2) + confmat(2,1);
        numeiouall = sum(numeiou);
        denoiouall = sum(denoiou);
        % Dice = 2TP / (2TP + FP + FN)
        numedice(t,:) = 2*confmat(1,1);
        denodice(t,:) = 2*confmat(1,1) + confmat(1,2) + confmat(2,1);
        numediceall = sum(numedice);
        denodiceall = sum(denodice);    
    end
    miou(z,:) = sum(numeiouall)/sum(denoiouall);
    mdice(z,:) = sum(numediceall)/sum(denodiceall);
end
meaniou = mean(miou);
meandice = mean(mdice);