How to segment White Gray matter of Brain CT?

I tried to use the following thresholding 25–38HU of white matter, and 35–60HU of gray matter.

close all;
clear all;
I=imread(Normal00481.jpg');
I1=rgb2gray(I);

subplot(2,3,1);
imshow(I1,[]);

WhiteThresh= I1>= 25 & I1<=38;

I1(~WhiteThresh)=0;

subplot(2,3,2);
imshow(I1,[]);

BW1 = edge(I1,'canny');
subplot(2,3,3);
imshow(BW1);
GrayThrsh=I1>= 35 & I1<=60;
I1(~GrayThrsh)=0;
subplot(2,3,4);
imshow(I1,[]);
BW2 = edge(I1,'canny');
subplot(2,3,5);
imshow(BW2);

[Rik]

Segmenting white matter and grey matter is a nontrivial task. There is probably literature describing different methods to do this.

You should note as well that imread does not return the image in HU, even dicomread has to be scaled. SO the thresholding you're doing here is not actually connected to what you describe.

[Mohamad Ajam]

Thanks for advice. I tried this but did not know how to apply it correctly.

img = double(HU);
img = img - min(img(:));
maximum = max(img(:));
grayscale = (img / maximum) * 255;

[Rik]

You're approaching this problem from the wrong end. First try to think how you can get to your original dicom data. Then once you have access to that data you can start doing the more difficult things like segmentation. Pure guesses tend to work out badly.

Also, for correct markup of code, you can just select the code and then press the {}Code button.

[Mohamad Ajam]

Do you mean by dicom data as the following?

Preset1=Abdomen;45;250
Preset2=Bone;500;4000
Preset3=Crane;35;100
Preset4=Lung;-500;2000
Preset5=Pelvis;45;300

[ViewerLabels]
TL=[10:20]\n[10:10]\n[10:40]\n[10:1010]
TR=[8:103E]\n[8:80]\n[8:50]\n[8:20]-[8:32]\nSe:[20:11]\nIm:[20:13]\nLoc: [SLICELOCATION]
BL=ET: [18:91]\nTR: [18:80] ms\nTE: [18:81] ms\nThk: [18:50] mm\n[ZOOM]\n[LEVELANDWIDTH]

[Image Analyst]

This will throw a syntax error:

I=imread(Normal00481.jpg');

Anyway, attach 'Normal00481.jpg'. Of course you probably already know that you should never use lossy JPG images for image analysis. Who made this image in the incorrect format anyway?

[Mohamad Ajam]

Thanks for reply. I did unfortunately just to understand. I know that it is a mistake now. I attached the image.

[Rik]

It looks like there is some identifying information on that image. If you do not have permission from that person to publish it, you shouldn't upload it here.

[Mohamad Ajam]

Thanks for notification

[Mohamad Ajam]

I uploaded another image.

[Mohamad Ajam]

If I have the dicom data how can i can i identify the threshold for segmentation white/gray material from HU unit.

Answer 1

[Image Analyst]

Look here for algorithms: http://www.visionbib.com/bibliography/contentsmedical.html#Medical%20Applications,%20CAT,%20MRI,%20Ultrasound,%20Heart%20Models,%20Brain%20Models