Matlab 自动核计数堆栈

Question

我正在使用这个借来的代码来自动计数 DAPI 染色的细胞核。

Hs.dirPath = pwd;
[Hs.dataFiles,Hs.dataNums] = getDataFiles(Hs.dirPath);
[Hs.foundChannels,Hs.fileNums,Hs.imgExts] = getImageFiles(Hs.dirPath);
for i = 1:length(Hs.fileNums)
    
    Hs.fileNum = Hs.fileNums(i);    
    fileName = ['dapi' sprintf('%03d',Hs.fileNum) '.tif'];    
    Hs.DI = readmm(fileName);
    Hs.DI = Hs.DI.imagedata;
    Hs.DI = scale(max(Hs.DI(:,:,round(linspace(10,size(Hs.DI,10),50))),[],3));
    binDapi = Hs.DI>adaptthresh(Hs.DI);

                masktmp = imclearborder(binDapi | ~Hs.DI);
                masktmp = bwareaopen(masktmp,50); 
                if any(masktmp(:))
                    binDapi = masktmp;
                end
                
    L = bwlabel(binDapi);
    numObjects = max(L,[],1:2);
    Hs.currObjs = [];
    Hs.allMasks = [];
    for j = 1:numObjects
        fnumStr = sprintf('%03d',Hs.fileNum);
        objMask = L == j;
        newObj = improc2.buildImageObject(objMask, fnumStr, Hs.dirPath);

        Hs.currObjs = [Hs.currObjs, newObj];
        Hs.allMasks = cat(3,Hs.allMasks,objMask);
    end

    objects = Hs.currObjs;
    save(sprintf('%s%sdata%03d.mat',Hs.dirPath,filesep,Hs.fileNum),'objects');
    [Hs.dataFiles,Hs.dataNums] = getDataFiles(Hs.dirPath);
    clear objects;
    Hs.currObjs = [];
end

它最终融合了不同平面上的原子核。在这里显示的图像中应该有 14 个原子核，但 1 和 6、2 和 5 以及 3 和 7 最终被视为相同。

我已尝试更改阈值等，但对堆栈没有帮助。

对修改代码有什么建议吗？谢谢。 Matlab output left, manually counted nuclei right

Here is the av intensity projection of the original image

Answer 1

我认为 MATLAB 的 imfindcircles 在这里可能有用。它高估了 #14，但似乎比你现在拥有的要好。

img = imread('Okqii.jpg');

mask = imbinarize(img,'adaptive');

cc = bwconncomp(mask);
lm = labelmatrix(cc);

imshow(img), hold on

for idx = 1:cc.NumObjects
  [c,r] = imfindcircles(lm == idx, [15,100], 'ObjectPolarity','bright');
  viscircles(c,r,'Color','b');
end