oasis3-to-bids converter occurred FileNotFoundError: Clinical data not found for ADRC_ADRCCLINICALDATA ID."

[jianfeixu95]

Hi
I am currently working with the OASIS-3 dataset and following oasis3-to-bids documentation , which reference a clinical data named:

oasis3_adrc_clinical_data.csv

However, in the latest version of the OASIS-3 clinical data that I downloaded from the official website, this file does not appear to be present. Instead, the clinical data is provided as multiple CSV files, such as:

• OASIS3_UDSd1.csv
• OASIS3_UDSb4.csv
• OASIS3_demographics.csv
• OASIS3_diagnoses.csv

And when I ran the converter code, an error occurred:

"File "/home/cxxxxx/miniconda3/envs/clinica/lib/python3.11/site-packages/clinica/converters/oasis3_to_bids/_utils.py", line 44, in _get_df_based_on_index_name
raise FileNotFoundError(f"Clinical data not found for {index_name}.")
FileNotFoundError: Clinical data not found for ADRC_ADRCCLINICALDATA ID."

⸻

Questions

1. Has oasis3_adrc_clinical_data.csv been deprecated or removed in newer releases of OASIS-3?
2. If so, is there an official mapping or recommended way to reconstruct an equivalent table from the current clinical files (e.g., using OASISID and VISCODE)?
3. Are there any updated guidelines for oasis3-to-bids converter that still reference this file?

[AliceJoubert]

Hello @jianfeixu95 ,

Thanks for reaching out ! We have actually worked on a fix with @davidaquilue on branch oasis_new_modalities (see here). To answer your questions :

• It seems the file has not been available for quite some time on OASIS3 website
• I don't know of any official guideline from OASIS3 but we managed to reconstruct the information we needed in the branch I indicated
• The branch allows you to use the converter with the data currently available on OASIS3 website

Tell me if you need help retrieving the code from the branch. It should be part of next release but is still currently in development.

Best,

[jianfeixu95]

Many thanks, Alice @AliceJoubert! I will try to implement the new code from the oasis_new_modalities branch.

By the way, I noticed that the OASIS-3 dataset provides BIDS-format download scripts called "Downloading MR and PET Scan files in BIDS format".

Do you know the differences between the official BIDS data and the outputs processed by the Clinica converter?

Cheers!

[AliceJoubert]

From my understanding, their outputs follows and outdated version of BIDS specifications (1.0.1 vs 1.10.x) and does not include metadata tsv files (participants, sessions, scans). Depending on what you need (just the images for ex) it can be enough !

[jianfeixu95]

YES! They follow an outdated version of BIDS. Thanks again ~

Cheers!

[jianfeixu95]

Hi, @AliceJoubert, I occurred a new problem after process data using branch "oasis_new_modalities" converter.

Invalid NIfTI files generated by oasis3-to-bids causing ANTs failure in t1-linear pipeline

I first used the clinica oasis3-to-bids converter to transform the original OASIS-3 dataset into BIDS format. Then, I ran the t1-linear pipeline on the generated BIDS directory. However, the pipeline fails at the N4BiasFieldCorrection step with the following error:

file ...T1w.nii.gz is not recognized as a supported image format Segmentation fault (core dumped)

This suggests that ANTs cannot read some of the generated .nii.gz files.

From inspecting the generated BIDS directory, some NIfTI files have extremely small file sizes (~1.7 KB), which is not consistent with valid MRI data.

Could you please help look into this issue? Thanks ～

[AliceJoubert]

Hello @jianfeixu95,

I tried with this exact subject and did not have an issue. Normally, all the converter does is copy from the raw folder to the bids folder, at the right place, along with other files (tsv, json). From your screenshots :

• maybe something went wrong with the copy. can you tell me what os / python version are you using ?
• how did you download your files (with which options) ? (you are missing the json files and your path to the nifti seems long, so I suspect the options used to download the files might have been wrong)

Best,
Alice

[jianfeixu95]

Hi, @AliceJoubert

I am using Ubuntu 24.04.3 LTS with Python 3.12.

Since the XNAT download application does not work properly on Ubuntu 24.04 (it only shows a blank interface), I downloaded the OASIS-3 dataset using the official script: Downloading MR and PET Scan files, for example:

---

I was able to reproduce the issue and identified a potential root cause in the converter code:

File: git branch oasis3_new_modalities

clinica/clinica/converters/oasis3_to_bids/_utils.py

Line 424 in e5b1241

source_file = fs.open(fs.ls(sourcedata_dir)[0], mode="rb")

source_file = fs.open(fs.ls(sourcedata_dir)[0], mode="rb")

The converter assumes that there is only one file in sourcedata_dir, but in practice (from the official OASIS-3 download), each directory contains multiple files, typically:

sub-OAS30001_ses-d0757_run-01_T1w.json
sub-OAS30001_ses-d0757_run-01_T1w.nii.gz

Since fs.ls() does not guarantee ordering, the code may select json file instead of nii.gz file

As a result:

• The .json file is mistakenly opened and treated as an imaging file
• It is then renamed and saved as .nii.gz in the BIDS directory
• This produces invalid NIfTI files (~7 KB)
  
  CHEERS!

[AliceJoubert]

Hello @jianfeixu95, thanks for having investigated this. It's not very robust indeed.

When we use XNAT (or a zip folder) we get this :

.
└── OAS30001_MR_d0757
    ├── anat1
    │   ├── BIDS
    │   │   └── sub-OAS30001_ses-d0757_acq-TSE_T2w.json
    │   └── NIFTI
    │       └── sub-OAS30001_ses-d0757_acq-TSE_T2w.nii.gz

So I would not have been able to reproduce the error with XNAT.

I will see what I can do to adapt for both,

Best,

[jianfeixu95]

The dataset is typically organized in the following structure:

└── OAS30001_MR_d0129
    ├── anat1
    │   ├── sub-OAS30001_ses-d0129_acq-TSE_T2w.json
    │   └── sub-OAS30001_ses-d0129_acq-TSE_T2w.nii.gz
    ├── anat2
    │   ├── sub-OAS30001_ses-d0129_run-01_T1w.json
    │   └── sub-OAS30001_ses-d0129_run-01_T1w.nii.gz
    ├── anat3
    │   ├── sub-OAS30001_ses-d0129_run-02_T1w.json
    │   └── sub-OAS30001_ses-d0129_run-02_T1w.nii.gz
    ├── anat4
    │   ├── sub-OAS30001_ses-d0129_T2w.json
    │   └── sub-OAS30001_ses-d0129_T2w.nii.gz
    ├── func1
    │   ├── sub-OAS30001_ses-d0129_task-rest_run-01_bold.json
    │   └── sub-OAS30001_ses-d0129_task-rest_run-01_bold.nii.gz
    ├── func2
    │   ├── sub-OAS30001_ses-d0129_task-rest_run-02_bold.json
    │   └── sub-OAS30001_ses-d0129_task-rest_run-02_bold.nii.gz
    └── func3
        ├── sub-OAS30001_ses-d0129_task-rest_run-03_bold.json
        └── sub-OAS30001_ses-d0129_task-rest_run-03_bold.nii.gz

Best regards,