dicomread truecolor comes out pink and green R2015a trial

When I read in a dicom file I get a 720x960x3 uint8 output as I would expect. However, if I do an imshow() on this output the resulting image comes out with a green background (which should be black) and pink text (which should be white). All the other colors are off as well.

My code is pretty simple:

X = dicomread('KX000008');
imshow(X);

I've checked the headers and it is set to truecolor so I'm not sure what's going on. Opening the same dicom in other dicom viewer tools it comes out correct, black background and white text. Any help is appreciated.

[Image Analyst]

Probably won't get very much help until you attach the file. What other app did you open it with? ImageJ? Photoshop?

[Chris Fahim]

Syngo Fastview. It's jpeg lossless compression so imagej won't open

[Image Analyst]

Did you see the first sentence in my comment?

[Chris Fahim]

Yes, working on getting it deidentified.

Answer 1

[Chris Fahim]

Unfortunately, I think the only answer for this is a bug in Matlab dicomread for jpeg lossless transfer syntax 1.2.840.10008.1.2.4.70.

I used http://support.dcmtk.org/docs-dcmrt/dcmdjpeg.html to decompress the same image and matlab reads the decompressed image as expected.

[Eric Larson]

I have been experiencing the same problem. Pink and green images result from reading in DICOM with imagery encoded with 1.2.840.10008.1.2.4.70 transfer syntax. Same image file is unpacked/displayed correctly in other tools (e.g. ShowCase). (By the way, I am using 2014b and the MCR.) I'd drop an image sample here as an attachment, but unfortunately it is much larger than 5MB.

[Julio Chirinos]

Experiencing the exact same problem exactly on 2019b. it's dissapointing that the Mathworks has not fixed this bug after 4 years!

Answer 2

[Steven Lord]

http://www.mathworks.com/matlabcentral/newsreader/view_thread/341719#936401

You asked this on the newsgroup earlier this week. I responded.

[Chris Fahim]

It's truecolor so map is empty.

[Julio Chirinos]

Steven, your link above no longer works. So what is the workaround for this bug? Every other DICOM software I have tried reads the colors correctly. The problem is restricted to Matlab

[Image Analyst]

Julio, to solve your problem it's best if you just attach your dicom image, zipped up.

Answer 3

[Image Analyst]

Since we can't actually download and see your data and problem, all I can do is to assume what you're saying is right, and suggest the way to fix it is to just remap the colors into the colors you want. Are the colors two very specific colors with precise RGB values, or are they a range of alike colors?

[Chris Fahim]

It's a range of alike colors. It appears to be exactly as if the colormap is wrong. Is there a way I could force dicomread to use a different colormap ?

I've been using matlab to deid my images, once I find an alternative I'll post the image.

Thanks.

[Chris Fahim]

Quick update, I've tried deidentifing with Photoshop, GIMP, and MicroDicom. Everytime I edit the image and resave Matlab then reads the deided image correctly. I am at a loss right now.

[Image Analyst]

I don't know what De-ID means. You mean like removing the patient name from the image or something like that? Anyway, doesn't sound like a loss to me - sounds like a solution or at least a workaround. At least you know how to get it working.

[Chris Fahim]

Deid, meaning removing both names/info burned on the image and from the dicom tags. I'm beginning to think Matlab has a bug in dicomread, I just decompressed using http://support.dcmtk.org/docs-dcmrt/dcmdjpeg.html and the decompressed image shows up correctly in matlab.