Mammogram - pectoral muscle removal

Hi all

I have a problem with removing pectoral muscle on mammogram. After identified pectoral muscle, what should I do next? I found on some paper, they are using loop function to travel up to First zero pixel and then replace all travelled pixels with zero and in some condition is violating for consecutive 5 times, then decrease pole position for next row by 1 and replace all pixels with zero.

Below is my sample code to identified pectoral muscle area:

pect = I;
[x y] = size(I);
thres=173;

for i=1:x
    for j=1:y
        if (pect(i,j)<thres)
            pect(i,j)=0;
        end
    end
end

*Note: I is the mammogram which I have removed "white text" area.

Thank you.

[krishnasai gudavalli]

How to identify the pectoral muscle?

[Image Analyst]

Post your image in a new question if my answer below about using thresholding does not work for you.

Answer 1

[Image Analyst]

I think you should just do

pect(pect <= thres) = 0;

instead of the loops.

[ws]

ok, thank you.

And for the removal part, I must using double for-loop?

[Image Analyst]

No. That did the removal. Just exactly what you did, except without loops. It merely sets the values below the threshold to zero. It is a grayscale image, like the original, except that pixels below the threshold are now pure black. It is not a binary image (logical image), which is often called a "mask" image. Isn't that how you were defining "removal"?

[ws]

Ok, after get the pectoral muscle area, how to remove that? Since there are different in size?

Thank you.

[Image Analyst]

I don't know what you're asking. As far as I can tell, it's removed. What you did was to make a copy of your image - "pect" is a copy of "I". Then you defined every pixel with a value less than "thres" as "pectoral muscle". Then we set those to zero, thus removing them. I don't know what the confusion is. You only have one image. WHAT is different in size? Sure the area of the pixels below "thres" is different than the brighter areas, but so what? It doesn't matter one bit. "pect" is the same as your original image except for the fact that the pectoral muscle (the pixels darker than "thres") are black (removed).

[Walter Roberson]

I wonder if wei siang is thinking of doing some cropping ??

[ws]

what i'm trying to do is find out the pectoral muscle, but threshold173 gave me the pecrtoral muscle area and useful breast area, so i need to find the method which will only crop out the pectoral muscle. after find out mask image(contain only pectoral muscle area), i wil use the orignal image to cut out pectoral muscle.

[Image Analyst]

If thresholding can't do it, as you first said but now deny, then I have no idea how to do it given the information you've provided. You'd have to provide an image. There are thousands of ways to segment things out of an image. You might use thresholding, region growing, texture, background subtraction, etc. but no one can tell unless they see your image and then try out a few things.

[ws]

Can you refer to the file I have upload?

http://www.2shared.com/file/T1M7ATyw/task.html

Thank you.

[ws]

Is my task feasible?

[Image Analyst]

Possibly, though it might take quite some development. It may be so complicated that it can't be done in a few lines given in the Answers forum. I never did see the image. If I don't see it in the first click, I don't continue. I don't download, save, unzip, create accounts, etc. I need to see it immediately or else I don't bother.

[Image Analyst]

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

[ws]

This is the image link: http://imageshack.us/photo/my-images/171/tasko.png/

My task:

1. Image 1 until image 5 are the steps I already done, now I want to find the way to get image 6 and 7.

2. I need image 6 to work with original image to get the final image (image 7).

3. After done this task I will use image7 for segmentation part.

Original image for testing purpose can get it from here: http://peipa.essex.ac.uk/ipa/pix/mias/mdb041.pgm

Thank you.

Answer 2

[Image Analyst]

You could threshold your image 5 to get a binary image. Then call imclearborder() to get the image of just the breast. Subtract that from the first binary image to get just the pect, which is touching the border. If you have cases where the breast also touches the border you'll have to do something like find the std dev of the regions and assume that the pect is the lower std dev.

[ws]

Thank you, I will try on it.

[ws]

This method is looking for the opposite, I want pectorial muscle, but it removed it.