Hi Chris,

first of all thanks for your tool and the great support you are giving here!

Please excuse my rather basic question:
I'm trying to convert a Siemens CT perfusion series from DICOM to NIfTI, using dcm2niix v1.0.20171215 under Linux Mint. The result are multiple NIfTIs with 100% identical .json, each containg a few slices of the head. Attached you'll find an examplatory .json

After working through the forum and the wiki,  I'm suspecting the problem lies within the DICOMs. Would you know of any way to confirm this? Or would you have an idea for a workaround within dcm2niix I could try out?

If any other information would be helpful please let me know!

Thanks in advance,
Adrian

My guess is that you are trying to stack slices of a CT where the X-Ray Exposure varies between the slices. If I am correct, these slices will be stacked if you do one of these two things:
 1. Update to a recent version of dcm2niix which will default to stacking CT scans where exposure varies (see release notes for v1.0.20190902). With modern versions you will need to explicitly run "-m n" if you want to separate slices based on exposure.
 2. Continue to use your old version, but provide "-m y" to force images to be merged despite varying exposure.

Thanks for the fast reply!
I've unfortunately already tried both (new version, -m y) without any effect. Would you know of other possible explanations for this problem?

Did the output provide any feedback regarding why the images were not stacked? My guess is these are not valid DICOM images, and the Series Instance UID is not shared across the images. Feel free to send me a private email with a link to the images from Box, DropBox or Google Drive.

Ah, could have thought of looking on the feedback myself. It says:

slices not stacked: Study Date/Time (0008,0020 / 0008,0030) varies 20200925131656.000000000000 ~= 20200925131659.000000000000

That means the DICOMs are carrying varying information on Study Date/Time, correct?

I'll have to check up whether I am allowed to share these images with you, many thanks for the offer anyway!

The [url=http://dicomlookup.com/lookup.asp?sw=Tnumber&q=(0008,0020)]Study Date and Study Time[/url] should be identical for all images in a single series (one study can contain multiple series, but all images in a series should be members of the same study). You might be able to rescue these but purging this, but be aware that tools may not be able to distinguish images from different sessions:

gdcmanon --dumb --remove 0008,0020 --remove 0008,0030 -i /path/to/input -o /path/to/output

Instead, I suggest you check the provenance of these images and get a copy before a poorly implemented anonymization or PACS corrupted them. Your images are not of archival quality.

The purging solution works! Thank you so much.

I'll certainly try to get access to the original images, but since they are from another university it's really helpful to have a work-around at hand.

Cheers to you!