Ordering of 4D DICOM images while writing

Keyn34 · August 18, 2022, 4:36pm

Hello everyone!

I am currently trying to write a 4D series of DICOM images (with Python and SimpleITK). Can anyone tell me which tag identifies the location of a volume in the fourth/time dimension (if there even is one)? I couldn’t find much information regarding this…

I really appreciate any help you can provide.

zivy · August 18, 2022, 5:29pm

Hello @Keyn34,

A 4D image is usually saved as a set (series is a reserved word in DICOM speak) of 3D images.

Each 3D image has its own series number and time information (content time tag, 0008|0033) and they all share the same study number. See this example showing how to save a single 3D DICOM image and this discussion on 4D DICOM images.

lassoan · August 19, 2022, 10:38am

A 4D image is a usually stored as a single DICOM series (Series Instance UID is the same for all frames) and the series is split based on some other “grouping” tag. Unfortunately, the choice of grouping tag is not controlled by the DICOM standard but it is up to imaging device manufacturers. We have identified a dozen or so tags that are commonly used, you can find them here:

github.com

fedorov/MultiVolumeImporter/blob/master/MultiVolumeImporterPlugin.py#L37


      
          self.tags['seriesDescription'] = "0008,103E"
          self.tags['instanceUID'] = "0008,0018"
          self.tags['position'] = "0020,0032"
          self.tags['studyDescription'] = "0008,1030"
          self.tags['seriesNumber'] = "0020,0011"
          self.tags['instanceNumber'] = "0020,0013"
          self.tags['repetitionTime'] = "0018,0080"
          
          
# tags used to identify multivolumes
          self.multiVolumeTags = {}
          self.multiVolumeTags['TriggerTime'] = "0018,1060"
          self.multiVolumeTags['EchoTime'] = "0018,0081"
          self.multiVolumeTags['FlipAngle'] = "0018,1314"
          self.multiVolumeTags['RepetitionTime'] = "0018,0080"
          self.multiVolumeTags['AcquisitionTime'] = "0008,0032"
          self.multiVolumeTags['SeriesTime'] = "0008,0031"
          self.multiVolumeTags['ContentTime'] = "0008,0033"
          # Siemens Somatom Cardiac CT 'ScanOptions' tag contains info on cardiac cycle
          self.multiVolumeTags['CardiacCycle'] = "0018,0022"
          # GE Revolution CT uses 'NominalPercentageOfCardiacPhase' tag to identify cardiac cycle
          self.multiVolumeTags['NominalPercentageOfCardiacPhase'] = "0020,9241"

What kind of 4D image would you like to write - CT, MRI, US?

Keyn34 · August 19, 2022, 11:11am

Hello @zivy, Thank you for the clarification. I tried it with your mentioned tags, but no luck for now. My goal would be to load a DICOM series into, for example, ITK-Snap and to have the data displayed as one 4D volume where I can scroll through. I’ll look further. Maybe I can find a configuration where it works.

Thank you, @lassoan, for that helpful information. I want to write either PET, CT or MRI. I thought there was maybe a universal tag for all modalities that specifies the location in a sequence of multiple 3D volumes. I will give it a try with your provided tags!

Thank you both again for your already constructive input!

lassoan · August 19, 2022, 11:21am

If you want to store 4D images for use in research software, such as 3D Slicer or ITK-Snap then I would recommend to save them in NRRD file format. Image loading and saving in NRRD is much simpler and faster than in DICOM.

Keyn34 · August 19, 2022, 11:31am

Thank you, and I very much agree with you (we prefer nifti, for example, for all sorts of cases), but a few analysis tools we have sometimes apparently demand DICOM.

I’m just trying to develop a small utility solution if needed.