Improve Data validation: Fail early
We sometimes face issues, that within a workflow, some processes fail due to invalid DICOM data. This especially arises when working with DCMQI, which all our segmentation workflows do. From a User perspective it is frustrating that, e.g. in the nnUnet-predict case, the workflows fails only during the conversion of nrrd to dcmseg due to invalid dicoms, although this could have been seen earlier. Therefore, we should improve our strategy. One option might be, that we directly reject invalid DICOMs once they got sent to the platform. By this, we have the guarantee, that only valid DICOMs are on the platform.
Opinions?