Automatic segmentation of lungs from CT image

I am trying to segment lungs from CT image, but while using Otsu method, some other regions (aroung lungs) also coming. Can some one help me to clear this problem? I can't use any manual technique. Need automatic lungs segmentation.

I=imread('NCP_39_1211_0023.png');
grayImage=I;
subplot(2, 3, 1);
imshow(grayImage, []);
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);
axis on;
title('Original Grayscale Image');

[pixelCount, grayLevels] = imhist(grayImage);
pixelCount(1) = 0;
pixelCount(end) = 0;
subplot(2, 3, 2); 
bar(grayLevels, pixelCount, 'BarWidth', 1, 'FaceColor', 'b');
grid on;
title('Histogram of Original Image');
%thresholdValue = 250;
%binaryImage = grayImage < thresholdValue;
level = graythresh(grayImage)
BW = imbinarize(I,level);
binaryImage = BW;
subplot(2, 3, 3);
imshow(binaryImage, []);
axis on;
title('Binary Image');

% Get rid of stuff touching the border
%binaryImage = imclearborder(binaryImage);
% Extract only the two largest blobs.
binaryImage = bwareafilt(binaryImage, 1);
% Fill holes in the blobs to make them solid.
binaryImage = imfill(binaryImage, 'holes');
% Display the binary image.
subplot(2, 3, 4);
imshow(binaryImage, []);
axis on;
title('Lungs-Only Binary Image');
drawnow;
% Mask image with lungs-only mask.  
% This will produce a gray scale image in the lungs and black everywhere else.
maskedImage = grayImage; % Initialize
maskedImage(~binaryImage) = 0;
% Display the masked gray scale image of only the lungs.
subplot(2, 3, 5);
imshow(maskedImage, []);
axis on;
title('Masked Lungs-Only Image');

Input image

Result image

Answer 1

[Image Analyst]

Looks like your automatic threshold was no good. And that's not the only thing. The image is total garbage. Way oversaturated. Try a different threshold, like around 150 or so.

thresholdValue = 150; % Automatic lungs segmentation at fixed threshold of 150
binaryImage = grayImage < thresholdValue;
% level = graythresh(grayImage)
% BW = imbinarize(I,level);
% binaryImage = BW;

[SILPA C S]

Sir, is there any other automatic segmentation methods

[SILPA C S]

Sir @Image Analyst, I am getting this result when I set 150 threshold value

[Rik]

You have barely any hope with an image that horrible. Why aren't you using the dicom image? You selected only 256 out of 4096 values. Any reliable workflow will require a better image. The image attached to the answer is also only 256 values, but at least those use a relevant window level.

[SILPA C S]

@Rik I'm working on a project having database in 'png' format. Also attached is one lung CT image among that, I can't alter this image. Could you suggest me any automatic lungs segmentation method or code for this kind of image.

[Image Analyst]

I did. What was wrong with using a threshold of 150 to threshold it automatically? No user-interaction is required so it's not manual. Why do you need a different automatic method?

[SILPA C S]

Sir @Image Analyst, in my program I'm accessing large number of CT images from a folder. For certain images 150 gives good segmentation. For some other 200, 250 like that. So each time I have to change values by seeing histogram. That's what I meant by manual. I would like to find a way where no such adjustment required each time. Sir, Is there any way to do this using histogram or any other way

[Image Analyst]

There are lots of ways. One way is to use a triangle threshold. Code is attached.

Another way might be to find the peak and just fall down the peak to a certain percentage of the peak. There are other possible ways one might think up.