infected white cell detection in image

Question

hello i'm a beginner of using matlab and as part of my Homework. i need to detect infected white cells in image and count them where the Nucleus of an infected white cell large and color Blue.

Original Image

Explain Image:

so i try to isolate the whit cells then detect the infected white cells but i'm stuck and do not what to do i will write my code and mention where i stuck and please if three is anther way to do this pleas help

1. convert the image from RGB to YcBcR space color to detect the whit sells

   OrgenalImg = imread('D:\Users\FADI\Desktop\cells\cells1.jpg');
   
   CopyOfOrgenalImg = OrgenalImg;
   
    YcbcrImage = rgb2ycbcr(CopyOfOrgenalImg);
    cb = YcbcrImage(:,:,2);
    cr = YcbcrImage(:,:,3);
   
   [r,c,v] = find(cb>=77 & cb<=127 & cr>=133 & cr<=173);
   
   index1 = size(r,1);
   
   %Mark the white cell pixel
   for i=1:index1
     CopyOfOrgenalImg(r(i),c(i),:) = 255;
   end
   
   figure, imshow(CopyOfOrgenalImg);title('White Cells');
   

   the White Cells image:

2.here i'm stuck i try to convert the White Cells image to gray grayscale to delete and remove the unwanted shape but i can't find way to do that so pleas help and how can i continue in my homework to detect and count the with cells and if there is anther ways to do this pleas let me know thanks in advance on any help.

Answer 1

You might attempt this by thresholding the image based on the blue channel to identify regions corresponding to each cell. For a non-research/homework exercise this may be sufficent given a carefully chosen threshold (0.66 in code below). Then you can iterate through each connected component and make a decision based on the area. In a real-world situation you would train up a system based on some kind of shape descriptor for this, but given you have only 2 example images some kind of hard-coded heuristic like this seems unavoidable.

Here is one solution:

INFECTED_AREA_THRESHOLD=3000; 

img=double(imread('\cell.jpg'))./255;
blueness=img(:,:,3)./(img(:,:,1)+img(:,:,2)+img(:,:,3));
blueness=blueness./max(max(blueness));
cellmask=blueness<0.66;

cellmask=bwfill(~cellmask,'holes');

map=bwlabel(cellmask);
labels=setdiff(unique(map),0)';

infectedcomponents=zeros(size(img,1),size(img,2));
areas=[];
infected=0;
for thislbl=labels
    connectedcomp=(map==thislbl);
    thisarea=sum(sum(connectedcomp));
    areas=[areas; thisarea];
    if (thisarea>INFECTED_AREA_THRESHOLD)
        infected=infected+1
        infectedcomponents=infectedcomponents | connectedcomp;
    end
end

imshow(rgb2gray(img).*infectedcomponents);
title(strcat(num2str(infected),' infected components detected'));
sort(areas)

The decision to threshold at 3000 was made by inspecting the list of component/cell areas in variable "areas". The blue threshold of 0.66 was by trial and error.

Sample output: