I Have to detect the nucleus of the cell...anyone provide me a code to detect the nucleus in the celll....

My criteria is to detect the contour of the cell and the contour of the nucleus.... i have done detecting the cell contour.... but i cannot detect the contour of the nucleus... i don't which segmentation method suits.... give me some idea...plz i have atttached the image of celll

thanks in advance.

Answer 1

[Image Analyst]

Just threshold and use bwboundaries().

[sou]

i have tried this sir but it fails to detect the nucleus....plz give me another choice

[Image Analyst]

Well what are you calling the nucleus? I don't think you need another choice if the nucleus is the big dark spot in the middle. Show us your code so we can see how it failed. My code is attached and it produced the following image:

[Image Analyst]

Yes, that code doesn't look good. Did you try my code on that image? Actually mine uses a fixed threshold. If you need to automatically find the threshold, then we'll have to develop something because the built in Otsu method that graythresh() uses is pretty bad, like it is in almost all cases.

[sou]

thanx for your code sir. it works good for some images.but in some cases it doesn,t detect the nucleus sir.I think automatic thresholdng is needed.

[Image Analyst]

So examine the various histograms of all the images and try to identify where the threshold is for the various images. You might also have to do some size and shape filtering after the thresholding to make sure you get oval/round things and not crazy shapes and sizes.

[sou]

ok sir. but i don't know how to get that value from histogram?in your program you mentioned binaryImage = grayImage < 110; how you used 110?i can't got it sir plz help me....

[Image Analyst]

I just looked at the histogram, and it worked for that image. You didn't say at first that you had any other images. So read my prior comment. You have to come up with something. I don't know the algorithm - you have to invent one. Maybe the triangle method will work well - it often does for me. Or else some ad hoc method of your own invention.

[sou]

ok sir i will try...I have extracted the cell boundary and nuclues boundary separately in binary form how should i overlay my extracted boundaries on my original gray image?

[Image Analyst]

Use plot(x,y, 'r-')

[sou]

okay thanx sir

[Image Analyst]

If you don't know how to format your code as code (highlight code then click {}Code button), then read this http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup or just attach the m-file with the paper clip.

[Image Analyst]

You're still using the lousy built-in auto thresholding algorithm, graythresh(). Aren't you going to write your own? One that actually works for your images? I've not seen 3.bmp or any other images than the one I worked on, so it's impossible for me to improve on the algorithm that I used. Post 5 images if you need help.

[sou]

sorry sir..i have attached 2 more images...<</matlabcentral/answers/uploaded_files/17602/3.bmp>>

[sou]

<<

>>

[sou]

i have attached images plz help me to get code