DICOM tag sequences management in Bio-Format

fiji
bio-formats
imagej
dicom

#1

Hi there,

I wanted to rise an issue that i’m facing with Fiji ecosystem and DICOM input.
This issue is problematic for our project in which we are trying to build medical image processing in nuclear medicine.

Originally ImageJ1 dosen’t support compressed DICOM, these last years we saw a significant growth of compressed DICOM as recent PACS system automatically compressed recieved DICOM to reduce by 50% the storage space.
The problem is that recent PACS do not enable by default decompression of DICOM at export so we saw a huge increase of compressed DICOM these last couple of years.

Bio-Format reader seems to be a very good alternative, I can successfully read compressed DICOM with it.

But there is another probleme, Bio-Format and IJ1 do not share the same Metadata presentation in the final ImagePlus.

The critical problem is in the management of DICOM tag sequences.
A DICOM tag sequence is a list of tags that belong to a specific tag. So it produce a sort of tree with parent tags that has child tags.

IJ1 has a clean way to organize child tags it is listed after the parent tag with an “>” marker.

BioFormat list all tags by its number and when the tag is available multiple time is has “#1” “#2” index.

Bio-Format presentation of tags has several problem :

  • The parent Tag itself is not listed in the metadata
  • Some child tags has a generic name like 0008,0100 Code Value, it apears several time in multiple different sequence that has no relationship. So BioFormat is listing a long list of “0008,0100” and it is impossible to know what is the parent tag of these code value. So it land with a loss of an important information, as all dicom sequences relationship are broken
  • Applications for ImageJ1 are not usable when DICOM opened with Bioformat, like IJ1 DicomTools to query a tag in the metadata. Also the applications we have developped can’t use BioFormat as reader because of this loss of informations.

I already raised the question in this forum without sucess : DICOM Tag presentation different Between IJ1 reader, SCIFIO and Bioformat - Issue with DicomTools

When I wrote this topic it was a theorical question but now it is a real issue for me to provide support of compressed DICOM.
I tried to imagine a way to rebuilt the metadata from Bio-Format to IJ1 style but because of the loss of Parent / Child hierachy i failed to build a clean “converter”.

So I have some questions about this issue :

  • Could Bio-Format implement an IJ1 metadata compatibility mode for DICOM ? (the original source code of IJ1 is here https://github.com/imagej/imagej1/blob/master/ij/plugin/DICOM.java)

  • If not is there a way to access a “raw header” in Bio-Format before the tags are grouped together and the parent/child tag hierachy are lost. If I get that input I will be able to write a method that will produce an IJ1 metadata style.

As I said in my previous topic this issue of DICOM Tag representation is also to affect SCIFIO reader as SCIFIO have another Sequence representation that are built in an array representation with also a break of Parent / Child dependency.

The DICOM header is a really complicated structure with a lot of hetereous tags, if we want medical imaging processing still possible in ImageJ ecosystem some standard in metadata presentation (or compatibility options) need to be shared accross readers.

Best regards,

Salim


#2

Hi Salim,

The handling of the tags is something which we can try and improve within Bio-Formats in order to at least allow you to access and maintain the parent child relationship between various tags. For other formats are current approach is to concatenate the key values in the hierarchy to create a value such as:

DisplaySetting|Channel|IlluminationType = Fluorescence

This is something which we can look to implement in the DICOM reader to improve the current handling. I have created a Bio-Formats Trello card to document the issue of the lost relationship between the tags: https://trello.com/c/SeNEOVDN/360-dicom-add-support-for-parent-child-relationship-in-metadata


#3

Thank you so much @dgault for you answer.

Very good, as long as I can have the parent /child relationship I will be able to fly.

In my group we read a lot of dicoms for our studies that can involve thousands of TEP/CT acquisitions coming for all over the world, so I would love to switch to Bio-Format to get support of compressed DICOM and test your DICOM reader in all these images (and provide bug identification if problems occurs)

Best regards,

Salim