DICOM Images show greyish

Hi there,

im curently working with CT and MRI Scans in DICOM Format. The MRI Images work completly fine, but the CT Images look this greyish. It is the same in the matlab Dicom Viewer but not in Online Viewers, there the Images are shown normal.

I already tried to rescale them, but it didn't help at all?

Has anybody an Idea where this comes from and how to fix it?

Thanks for your help :)

[Rik]

What exact code are you using to get to this image?

[Simon]

U mean 'dicomread'?

[DGM]

You might attach an example of the dcm image so that the native data range is known.

I'm assuming you intend to make the circular "background" region black, but that normalization techniques fail because the image corners are black. If the corners are indeed the global minima, then you'll have to decide how to otherwise identify the background color. How to do that would depend on what all the images look like.

For example, let's assume it's sufficient to sample the top row and find the most common color:

% load image
A = rgb2gray(imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/777493/image.jpeg')); 
% normalize based on background sample and global maximum
A = mat2gray(A,double([mode(A(1,:)) max(A(:))]));
imshow(A)

[Simon]

Thank you, its much better like that.

I think i can work on with this :)

But is there a possibility to change the contrast of dicom images in the code?

[Rik]

Dicomread will read the 16 bit CT numbers. You need to convert them to Hounsfield units yourself. You also need to set the window level (the second input to imshow) to avoid the default.

[DGM]

It depends what you mean by changing the contrast. The above example does increase the contrast.

A = rgb2gray(imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/777493/image.jpeg')); 
range(double(A(:)))/mean(double(A(:))) % example contrast metric

ans = 1.2804

A = mat2gray(A,double([mode(A(1,:)) max(A(:))]));
range(double(A(:)))/mean(double(A(:))) % check contrast again

ans = 12.5577

imadjust will let you adjust input and output black/white points (linear) and optionally adjust gamma. stretchlim() can be used for finding the values needed by imadjust(). I'm sure there are other things that can be done too, depending on what you need. Phrases like "contrast adjustment" or "contrast enhancement" are unfortunately kind of vague, since they can mean contradictory things in different contexts.

[Simon]

Thank you guys for all your help, so far im a a state, where I can continue in processing ;)

Answer 1

[Image Analyst]

Try using [] in imshow() to scale your min to 0 and max to 255:

imshow(yourGrayScaleImage, []);

[Simon]

already did this cause otherwise it only shows black

[Simon]

But i also got a bigger problem at the moment.

Is there any way to scale two object to the same size?

Cause i got now two object wich I want o compare, but they are in different size and location.

The spatial resolution is here very important.

[Simon]

P.S.: this Image is a result of glueing these two Object together, I pocessed them seperatlay.

[Image Analyst]

I'd use regionprops() to find the equivalent circular diameter. Then I'd use imresize() to resize the smaller one by the ratio of those diameters. Then I'd find the centroids with regionprops() and use imtranslate() to shift one of them so that the centroids align. Something like

props1 = regionprops(mask1, 'EquivDiameter');
props2 = regionprops(mask2, 'EquivDiameter');
d1 = props1.EquivDiameter 
d2 = props2.EquivDiameter
ratio = d1/d2;
mask2 = imresize(mask2, ratio) % Assuming mask2 is smaller.
props1 = regionprops(mask1, 'Centroid');
props2 = regionprops(mask2, 'Centroid');
imtranslate()
% etc.

[Rik]

You have the DICOM metadata of both? Then I would use that information to rescale. Only the translation should be done on the image, unless you have used fidutials on both scanners to tie the scanner coordinate grid to your object.

[Simon]

I got the original scandata, but i have no Idea hot to get to their metadata, neiter how to rescale the objects based on the metadata.

I have not used fidutials.

[Image Analyst]

Maybe try imfinfo().

[Simon]

Does this also works for .dcm format?

Cause it says "Unable to determine file format"

[Rik]

Try dicominfo instead.

[Simon]

Ok, i got now 174 fields of information, but how can I use them?

[Rik]

There are two fields that are important for you now: the RescaleSlope and the RescaleIntercept. You can convert the image data of your slice to HU with this line:

%convert to double to allow full data range
data=info.RescaleSlope*double(data)+info.RescaleIntercept;
%for CT:   genrally 1                generally -1024

That will give you the proper image data.

You should dig around in the other fields to figure out which fields might encode information about the voxel size. If the SliceIncrement doesn't exist, you need to use the ImagePositionPatient for two subsequent slices.

if ~isfield(info1,'SliceIncrement')
    if isempty(info2)
        %Single slice image, so increment doesn't exist.
        info=info1;
    else
        SliceIncrement=...
            (info2.ImagePositionPatient(3)-info1.ImagePositionPatient(3))/...
            (info2.InstanceNumber-info1.InstanceNumber);
        info=info1;info.SliceIncrement=abs(SliceIncrement);
    end
end

[Simon]

what do you mean with double(data)?

[Rik]

Your data is probably loaded as int16, so you need to convert it to the double datatype to allow decimal values as well. This might not be required, but it can prevent strange issues later on.

Have you done a basic Matlab tutorial? Dicom is complex enough on its own. You should first make sure you are familiar with Matlab itself.

[Simon]

I was just confused, cause it is already saved as double.

Iam kinda familiar with matlab, but am not the most experienced.

But I already learned a lot in the past month :p

But iam used to the basics already

[Image Analyst]

Attach what you see for dicominfo here.

[Image Analyst]

I don't see the field of view in real world units in the dicominfo you posted. Maybe I overlooked it. Do you see it? I know it's 512x512 pixels but do you see the real world field of view in cm there?

Otherwise, why do you think the spatial calibration is different between the two images? Are you sure the one disc object really isn't smaller than the other one?

[Simon]

Cause it is the same object, yes Iam sure that they are the same size in the real world.

[Simon]

This should be the real FOV

[Simon]

It is to be found in "Dicominfo2.PNG"

[Image Analyst]

OK since the reconstruction diameter is 500 for one of the images, and the image is 512x512 for both images, what is the reconstruction diameter for the other image? If the discs are the same size in the real world then the cm per pixel spatial calibration factor must be different for the two images because the discs shows up with a different pixel diameter in the two images.

[Simon]

Just for understanding, here you can see what these two discs originally looked like. (The right one is changed to binary already, cause of the original problem)

Also the left Image is saved as 256x256

Answer 2

[Simon]

Sorry for blacking out some,