Burned in annotations Dicom

Hi,

I have some manual burned in annotations in Dicom file of a lung HRCT scan. I need to extract the information of the annotations, thus, I need to know i.e. where exactly the annotations are located in order to identify that region in the raw Dicom file. I need to use this annotations to train a classifier, but I don't know how to do it. Can anyone help me?

Thank you in advance!

Answer 1

[sam0037]

I understand that you want to extract manual burned in annotations from a DICOM file and by annotations you mean the METADATA of the corresponding DICOM file. If you wish to extract the METADATA from a DICOM file, you can use DICOMINFO or DICOMDISP function from the Image Processing Toolbox.

Refer to the following MATLAB Documentation links to know more about 'Reading METADATA from DICOM Files', DICOMDISP function and DICOMINFO function respectively:

http://www.mathworks.com/help/releases/R2015b/images/read-metadata-from-dicom-files.html

http://www.mathworks.com/help/releases/R2015b/images/ref/dicomdisp.html

http://www.mathworks.com/help/releases/R2015b/images/ref/dicominfo.html

[Isabel]

Hi Abdus,

Thank you for you answer but no, by annotations I don't mean METADATA. I mean, that I have drawn contours in a CT scan and when it is saved the contours are burned into the DICOM. The question was if it is possible to get the information of the contours (coordinates/position) out of the scan.

Best regards

[Walter Roberson]

Did you draw the annotations in any particular color?

Did you save the annotated images as JPEG, or did you save as PNG or TIFF? (Please don't tell me you saved them as JPEG...)

[Isabel]

The annotations are in blue. The annotated images are saved as DICOM.

[Walter Roberson]

Is there anything else in the image that is the same color?

[Isabel]

no, the rest of the image is in black and white.

Answer 2

[Walter Roberson]

text_mask = Blue_channel > threshold1 & Red_channel < threshold2

For some small threshold2 (the black level) and larger threshold1 (the level of the blue)

The blue test finds the blue but also finds white. The red test rejects the white (white is near equal on all three channels)

You can now feed the mask itself to an ocr routine to extract the text.

[Isabel]

Thanks Walter, I will try that.