I'm experiencing a discrepancy in the display of DICOM images between MATLAB's imshow function and the Medical Image Labeling tool.

I am new on MATLAB. I am encountering a discrepancy in the display of DICOM images between two MATLAB environments: the standard MATLAB workspace using the imshow function and MATLAB's Medical Image Labeling tool. Both tools are used within the same MATLAB session to visualize the same DICOM images. However, the images appear differently in terms of contrast, brightness, and overall visibility. The preprocessing of these images includes steps like Hounsfield Unit conversion, windowing, noise reduction (Gaussian filtering), intensity normalization, and resampling. After preprocessing, the images displayed using imshow appear as intended. However, when these preprocessed images are loaded into the Medical Image Labeling tool, there is a noticeable difference in their appearance. This variation complicates the process of consistently analyzing and labeling the images, as the same image presents differently in two MATLAB-based tools. I am looking for insights or solutions to ensure uniformity in image display across both environments, which is critical for accurate image analysis and processing.

[Cris LaPierre]

Could you share your code? A sample image would also help. Otherwise, could you be more specific on what the differences are? For example, use the Image Viewer app to inspect the windowing parameters for an image where you see a difference. Consider sharing a screenshot of the contrast tab here (turn on Interactive Histogram).

One thing to consider - the medical imaging toolbox should automatically apply Hounsfield Unit conversion for you.

[Ozlem]

My code

filePath = 'C:\Users\ÖZLEM\Desktop\tez\fy pa\00000001\00000001\00000021.dcm';

if exist(filePath, 'file')
    dicomImage = dicomread(filePath);
    dicomInfo = dicominfo(filePath);
    
    huImage = double(dicomImage) * dicomInfo.RescaleSlope + dicomInfo.RescaleIntercept;
    
    windowCenter = 50;
    windowWidth = 350;
    
    lowerBound = windowCenter - (windowWidth / 2);
    upperBound = windowCenter + (windowWidth / 2);
    softTissueWindowedImage = huImage;
    softTissueWindowedImage(huImage < lowerBound) = lowerBound;
    softTissueWindowedImage(huImage > upperBound) = upperBound;
    
    sigma = 1;
    filteredImage = imgaussfilt(softTissueWindowedImage, sigma);
    
    normalizedImage = mat2gray(filteredImage);
    
    standardPixelSpacing = 0.6;
    
    currentPixelSpacing = dicomInfo.PixelSpacing(1);
    
    scalingFactor = currentPixelSpacing / standardPixelSpacing;
    
    resampledImage = imresize(normalizedImage, scalingFactor);
    
    resampledImage16 = uint16(resampledImage * 65535);
    
    newDicomPath = 'C:\Users\ÖZLEM\Desktop\tez\last\preprocessedImage.dcm';
    
    dicomwrite(resampledImage16, newDicomPath, dicomInfo, 'CreateMode', 'Copy');
    
    disp(['Preprocessed image saved as DICOM file at: ', newDicomPath]);
else
    disp(['File does not exist at ', filePath, '. Check the file path and ensure it includes the file extension.']);
end
then I wrote this
savedDicomImage = dicomread(newDicomPath);

figure;
imshow(savedDicomImage, []);
title('Saved DICOM Image');

figure;
imshow(resampledImage, []);
title('Original Preprocessed Image');

This is what it looks like when I open the saved DICOM data in Radiant, it looks the same in the Medical Image Labeling tool

[Cris LaPierre]

Can you zip and attach one of your images?

[Ozlem]

I attached unprocessed image. I am planning study with 2D roi therefore I use single slice only. Thank you very much for your time and help.

Answer 1

[Cris LaPierre]

Have you adjusted the Window Level and Window Width in the Medical Image Viewer App to match what you set in imshow?

When I open a medical image (uint16 xray) in a new app session, it seems to default to a level of 2048 and a width of 2048. Since a uint16 has pixel values ranging from 0 to 65535, this will make the contrast and brightness of the image appear different from a figure created using imshow(I,[]). The [] causes the figure to adjust the window width to the range of data in your image.

Since you have adjusted the image contrast, your resampledImage16 is likely using the full range of data (at least based on what I can see in the images you shared). Therefore, you will want to adjust the window level and window width settings in the Medical Image Labeler App to be 32768 and 65535 respectively to mimic what you see wtih imshow.

[Cris LaPierre]

Found something interesting. I needed to halve the window width in the Medical Image Labeler App to get it to appear the same as the Image Viewer App (screenshot below). I have passed that info along internally in case that is not the correct behavior.

For the image you shared, after running your code, I used a value of 32678 for both the level and width in the Medical Image Labeler App. When comparing preprocessedImage.dcm in the app with the figure created by imshow(resampledImage16,[]) the results looked the same.

[Cris LaPierre]

Final comment. Medical Image Labeler App will try to use the window center and width values in the metadata. However, the values in the metadata are for the original int16 image. Your preprocessing has convered it to a uint16 and applied histogram equalization. That means the original values are no longer valid for the new image.

I would suggest updating them in the metadata before writing your new dicom file. This will allow it to open with the same contrast and brightness as you specified in your code. Here is what the updated code might look like.

newDicomPath = 'preprocessedImageV2.dcm';
dicomInfo.WindowWidth = range(resampledImage16(:));
dicomInfo.WindowCenter = round(dicomInfo.WindowWidth/2);
dicomwrite(resampledImage16, newDicomPath, dicomInfo, 'CreateMode', 'Copy');