Is guessing mandatory metadata for ```dwi-preprocessing-*``` interesting for datasets containing DWIs ?

[Adam-Ismaili-92]

Description :

Some BIDS datasets containing DWI modalities — such as the IXI dataset — do not include certain metadata fields like PhaseEncodingDirection and TotalReadoutTime in their JSON sidecars.

These values are sometimes critical for preprocessing pipelines (e.g., dwi-preprocessing-*), but could potentially be estimated or guessed from the available raw imaging data, or be attributed a null value when not retrievable.

Question :

Would it make sense to implement a way to estimate/guess these values from the raw data so that DWI preprocessing pipelines can still run without manual metadata completion ?

[AliceJoubert]

That information is usually retrieved from dicoms and not niftis so it would be difficult to estimate at this point of the processing. Guessing or arbitrarily choosing the value would be exposing us to processing errors so it would be best not to do this. In the case of the IXI dataset the converter should be modified in order to include these values in the json files if they can be found in the metadata.

[Adam-Ismaili-92]

That information is usually retrieved from dicoms and not niftis so it would be difficult to estimate at this point of the processing. Guessing or arbitrarily choosing the value would be exposing us to processing errors so it would be best not to do this. In the case of the IXI dataset the converter should be modified in order to include these values in the json files if they can be found in the metadata.

Yes, I think you’re right. This information can be found in the DICOM metadata, and retrieving it from the raw NIFTI data seems not possible. It might be interesting to look for the information in the clinical data accompanying the imaging one, but it does not seem to be available for all datasets, especially IXI. Not to mention that estimating the information could introduce significant processing errors, as mentioned earlier. It may be more reasonable and safer to consider that these datasets simply cannot be processed by these particular pipelines.