How to use matlab to count these particles?

So I have some neurospheres in a dish.

After I take these images, I run them through imagej. I sharpen the picture, "despeckle" the image, add a guassian blur, and then find the edges of the image. The end result of that process looks like this:

After this process, I take the image and run it through matlab's thresholding. The image then looks like:

Now I'd like to use some surface mapping and create peaks and valleys and count my neurospheres. The process I want to do is similar to the process found here.

I'd like to be able to count my neurospheres, come up with an area for each one, and then also have a diameter, and be able to discriminate each sphere by a minimum size diameter.

Would anyone be able to help?

Thanks,

Tushar The Ohio State University

[Bani Antonio Aguirre]

can you share your code for thresholding?

[Image Analyst]

@Bani Antonio Aguirre here is mine for interactive thresholding:

https://www.mathworks.com/matlabcentral/fileexchange/29372-thresholding-an-image?s_tid=srchtitle

I'm also attaching a triangle threshold which is good for skewed histograms.

Answer 1

[Image Analyst]

You might try adapthisteq() or imbothat() to get an image of the spheres. Then threshold to get your segmented image. Anyway, once you have an image you can threshold, just threshold it and call regionprops(). I don't think you have to do anything with "surface mapping and create peaks and valleys". Not sure what you were thinking there. So, basically

binaryImage = filteredImage < someThresholdValue;
% Measure areas and diameters.
props = regionprops(binaryImage, 'Area', 'EquivDiameter');
% Get measurements from structure into 1-D array
allAreas = [props.Area];
allDiameters = [props.EquivDiameter];
% Histogram them (will plot into current axes - create a new axes if you don't want that).
histogram(allAreas);
grid on;
% Sort from largest diameter to smallest, if you want (optional).
[sortedDiams, sortOrder] = sort(allDiameters, 'Descend');
% Sort areas the same way so they are kept matched up (together).
sortedAreas = allAreas(sortOrder);

See my File Exchange for a full demo in my Image Segmentation Tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862&sort=downloads_desc

I don't know what "discriminate each sphere by a minimum size diameter" means. The code above finds all blobs. If you want to throw out some and keep only some in a certain size range, call bwareafilt() or bwareaopen() before you call regionprops().

[tgohsu]

Thanks for the answer,

I don't want to count all blobs though. Part of the culture conditions create artifacts in the plate which aren't the particle of interest.

Further, I need to be able to quantify the size of my blobs with area and diameter.

[Image Analyst]

What blobs don't you want to count? My code DOES give you the area and diameter of every blob - not sure why you thought it didn't. My tutorial shows you how to filter based on virtually any attributes of the blobs that you want, that is, if you want something more sophisticated than what bwareafilt(), propfilt(), or bwareaopen() will do for you. But you need to specify what makes a good blob and what makes an unwanted blob.

[tgohsu]

fair point. Sorry for the confusion.

I just ran the code on the last image I posted, and I'm having some trouble understanding the histogram.

So what i'm looking at here is the area of blobs vs the number of blobs, right?

I'll look through your tutorials over the weekend and get back to you.

I really do appreciate your time.

[Image Analyst]

Yes, looks pretty typical. Quite often, nearly anything you measure in a particle size distribution, not matter what it is, will have a log-normal looking distribution. It looks exponential, but I bet if you added more histogram bins you'd see the familiar log-normal shape.

[Ham Man]

Hello evry one

I'm beginner and I have a picture as below and i want to count the particles in this image using matlab.

What is a simple script to run this.

Appreciate your help!

[Image Analyst]

Try imtophat(). If that doesn't work, start your own discussion thread.

Answer 2

[John BG]

Hi Tushar

this is John BG <mailto:[e-mail] [e-mail]>

1.- Acquiring petri image, using green layer only, manually binarising the whole image.

Play with threshold th1

clear all;clc;close all
A=imread('blood_cells_count.jpg');
% A2=A(:,:,1)+A(:,:,2)+A(:,:,3);    % a bit more contrast
A2=A(:,:,2);  % since green marker used no need for red blue layers
th1=190;  
A2(A2>th1)=255;
A2(A2<th1)=0;

figure(1);imshow(A);

figure(2);imshow(A2);

.

2.- Filling up holes in binarised cells

A20=~A2;
area_threshold=6;
A21 = bwareaopen(A20,area_threshold);
figure(10);imshow(A21);

A22=imfill(A21, 'holes');
figure(11); imshow(A22);

.

3.- It may not be really needed, but just in case, filling up and plotting cell contours

[B,L] = bwboundaries(A22,'noholes');
map=zeros(length(B),3);cmap=colormap(map);
figure(12);imshow(label2rgb(L,cmap, [.5 .5 .5]))
hold on
for k = 1:length(B)
   boundary = B{k};
   plot(boundary(:,2), boundary(:,1), 'r', 'LineWidth', 1)
end

.

.

4.- The asked cells count is:

length(B)

   ans =
      398

.

5.- Add a halo around each binarised blood cell, neurons are hairy aren't they?

PSF = fspecial('gaussian',5,5);
A3 = deconvlucy(uint8(255*(~A21)),PSF,5);
figure(3);imshow(A3);

.

Comment:

using area threshold may remove a few small cells. Sample arrowed small specks are removed but the circled small cell is also removed.

Perhaps next step would be to consider removing small AND no green marker (dark only) specks.

.

if you find this answer useful would you please be so kind to consider marking my answer as Accepted Answer?

To any other reader, if you find this answer useful please consider clicking on the thumbs-up vote link

thanks in advance

John BG

<mailto:[e-mail] [e-mail]>

[tgohsu]

John,

I appreciate the time you put forth in helping with my question.

I believe the next step would be to take out the little specks that you have annotated with the arrows in the last image.

What we're looking at here are neural stem cells before the agglomerate into neurosphere. This is a small concentration (20 cells/uL). The current schedule the lab is using is 2, 20, 100, 200 cell/uL. We're using a 1 time a week and 3 time a week feeding schedule.

The cell count your code provides, I'm assuming, uses the specks that you had pointed out with the arrows, correct? For neural stem cells, we generally don'e see cell counts above 300, so I think that's were the extra numbers are coming from.

Finally, could you advise on how to pull area and diameter data from the picture?

Thanks again for helping me as much as you have.

Tushar

[Image Analyst]

Tushar, I already gave you the code for that in my answer.

[John BG]

Tushar

the last image with the arrows and the circled cell is BEFORE removing the specs.

In the resulting image, black cells with red skin, the specks have already been removed.

Or in the resulting image with halos around cells, there are no small specs, they were removed.

The image was just a closing comment.

The next steps would be, from :

1. number the cells
2. measure the area of each cell using each boundary B{k}
3. colour the cells according respective areas

John BG <mailto:[e-mail] [e-mail]>