ADNI-to-BIDS TAU Conversion Issues

[DAVID AQUILUÉ LLORENS]

Hello clinica Team,

I have discovered some issues that do not seem to be documented in the ADNI-to-BIDS converter pipeline/documentation.

When converting some of the Co-registered and Averaged Tau PET files from DICOM to nifti using dcm2niix, there seems to be some problems in the conversion, that do not explicitly raise a strong enough warning but that results in corrupted nifti files. However, the issue does not happen for the step 4 PET scans (Coreg, Avg, Std Imgs and Vox Size, 6mm resolution). 

See the dcm2niix output for the ADNI3 baseline session of 041_S_6785:

(clinica) D:\ADNI Datasets\CHECKING_TAU\CS\raw\041_S_6785>dcm2niix -o .\nix AV1451_Coreg__Avg__Std_Img_and_Vox_Siz__Uniform_6mm_Res\2019-10-22_11_28_26.0\I1598320
Chris Rorden's dcm2niiX version v1.0.20230411  MSC1929  (64-bit Windows)
Found 96 DICOM file(s)
Warning: PatientOrient (0018,5100) not specified (issue 642).
Warning: Unable to determine manufacturer (0008,0070), so conversion is not tuned for vendor.
Convert 96 DICOM as .\nix\I1598320_AV1451_Coreg,_Avg,_Std_Img_and_Vox_Siz,_Uniform_6mm_Res_20191022142826_5 (160x160x96x1)
Saving as 32-bit float (slope, intercept or bits allocated varies).
Conversion required 0.236000 seconds.

(clinica) D:\ADNI Datasets\CHECKING_TAU\CS\raw\041_S_6785>dcm2niix -o .\nix AV1451_Co-registered__Averaged
Chris Rorden's dcm2niiX version v1.0.20230411  MSC1929  (64-bit Windows)
Found 63 DICOM file(s)
Warning: PatientOrient (0018,5100) not specified (issue 642).
Convert 63 DICOM as .\nix\AV1451_Co-registered_Averaged_AV1451_Co-registered,_Averaged_20191022142826_2 (128x128x63x1)
Conversion required 0.202000 seconds.

And see the visual comparison in comp_dcm2niix.png (top is Step4, bottom is Step2):

Using another conversion tool (SPM's spm_dicom_convert function), we can achieve correct conversion to nifti, see comp_dcm2niix_vs_spm.png (left is SPM-dicom and right is dcm2niix directly from clinica pipeline, with a dark band)

So, for some reason, the dcm2niix seems to have problems with some of these Step2 PET files. I can provide a broader list of subjects where I have observed this problem. However, I have been using only a subset of the available Tau scans, so it will not be comprehensive.

Here are some other subjects:

041_S_6136

041_S_6159

041_S_6354

They are mostly from sites 32, 41, 94, 127.

Let me know if I can do anything to help, I am currently working on fixing this manually as I believe that this may require some type of design change in the converter pipeline. As soon as I settle this issue I will focus on the OASIS PR.

Thanks,

David

[Alice Joubert]

Hello David,

Thank you very much for reaching out on this subject. We will investigate this issue and get back to you !

Best,

Alice

--
You received this message because you are subscribed to the Google Groups "Clinica" group.
To unsubscribe from this group and stop receiving emails from it, send an email to clinica-user...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/clinica-user/7c0c89f8-bb91-4210-8cbf-a807841f6e16n%40googlegroups.com.

[Alice Joubert]

Can you indicate which version of dcm2niix you are using for these tests ?

Thanks,

Alice

[AQUILUÉ LLORENS, DAVID]

Yes,

It is true that I am using a fairly outdated dcm2niix version, from when I first installed clinica. It's v1.0.20230411

The dataset was converted with clinica v0.9.4, python version 3.10.

Thanks,
David

Missatge de Alice Joubert <alice00...@gmail.com> del dia dj., 26 de març 2026 a les 16:11:

To view this discussion visit https://groups.google.com/d/msgid/clinica-user/CAP_epFu-ynkZgn%2BJd_KHsfCdf%2BgJeH7Jt2YccavY8E7YttZ9sA%40mail.gmail.com.

--

David Aquilué Llorens

PhD Student

Computational Neuroscience Group
https://www.upf.edu/web/cns

Center for Brain and Cognition 

Department of Information and Communication Technologies (DTIC-UPF)

Ramon Trias Fargas, 25-27 (Mercè Rodoreda building - Ciutadella campus)

08005 Barcelona                    

Office 24 335

[Alice Joubert]

I tested with a more recent version of dcm2niix (2025) and it still gives the same result, so this is not the cause (too bad). I will continue investigating !

Alice

To view this discussion visit https://groups.google.com/d/msgid/clinica-user/CACpJWuqjasioj7Y8bheiY8bXEHuAX%2BL5MLD4XD0Ypf4kxJeWLg%40mail.gmail.com.

[Alice Joubert]

Though, I think clinica is not really at fault here. The best way to go about it is probably to contact dcm2niix as they are quite reactive. Maybe they can fix this issue directly, which might interest other people using dcm2niix.

[AQUILUÉ LLORENS, DAVID]

Totally agree that clinica is not at fault here and that the most straightforward and overall better fix would be to directly fix it through dcm2niix.

I first contacted the ADNI data PET group who gave me the idea of checking with Step4 data. They also use dcm2niix on their data but taking the Step4 DICOM folders. Hope this helps for tracing back the error.

Best,

David

Missatge de Alice Joubert <alice00...@gmail.com> del dia dj., 26 de març 2026 a les 16:31:

To view this discussion visit https://groups.google.com/d/msgid/clinica-user/CAP_epFt9UfL3mFa3c0hZOhX9qzvw%2BtOUcbX%2Bz%2Bnpg1UupjrFZA%40mail.gmail.com.

[Alice Joubert]

Thanks for the additional info ! Do you know if they take step 4 whatever the PET tracer / subject ? We will discuss this possibility. If we don't use step 4 at least we could put a warning if dcm2niix outputs unusable images or choose not to convert these. 

Alice

To view this discussion visit https://groups.google.com/d/msgid/clinica-user/CACpJWurzmLZDUagCyjRuZioqg7GdmUt%3DYDa6UeLpYpMqVC_LOw%40mail.gmail.com.

[AQUILUÉ LLORENS, DAVID]

I could not say for sure, but I would assume that they do, as they underwent all the efforts to standardize vox size, bring to equal resolution of 6mm, etc. They applied the same approach for Tau (AV1451) and Abeta (AV45, FBB), but I am not sure about FDG.

They wondered (and I wondered as well in the beginning), why does the clinica pipeline not take the Step4 data directly? I see that the PET pipelines already take as input the FWHM of the PSF, but for ADNI maybe it would be easier to use Step4 with a common resolution of 6mm then decide if any other smoothing is needed. Just wondering, as I assume this was already decided previously for a justified reason.

Missatge de Alice Joubert <alice00...@gmail.com> del dia dj., 26 de març 2026 a les 16:43:

To view this discussion visit https://groups.google.com/d/msgid/clinica-user/CAP_epFspVjgYp-E2n%3Dw0x97xY84uz%3DcwRz_hffa-msHdL1MrZg%40mail.gmail.com.

[Alice Joubert]

STEP2 was taken to preserve the resolution of better slices that is lost for the resolution of worse slices with STEP4. Actually the converter treats pet modalities differently because PET FDG outputs both STEP 2 and STEP 4 so people choose what they want. I think the solution here for clinica will be to do the same for every PET modality involved, including PET TAU. I will patch this soon !

Best,

Alice

To view this discussion visit https://groups.google.com/d/msgid/clinica-user/CACpJWuo2u4eYEJVSZ7Oe1wVDXxqCF0RQJMTsvsTWq3WK-YoXYQ%40mail.gmail.com.