If we can find or generate DICOM test data where we know exactly what each parsed attribute should be, that would be quite useful for testing purposes.
We can also create fixtures based on our current parser output if we assume that currently we are parsing things correctly (to prevent regressions in this case).