Not writing JPEG encoded pixel data correctly

[davidssmith]

When I try to use DCMTK to dump a lossless-JPEG encoded DICOM that I wrote with this package, I get this error:

E: Found explicit length Pixel Data in top level dataset with transfer syntax JPEG Lossless, Non-hierarchical, 1st Order
Prediction: Only undefined length permitted
E: dcmdump: Pixel data in top level dataset in compressed Transfer Syntax uses explicit length: reading file: out.dcm

I think it is referring to the fact that this particular Transfer Syntax (1.2.840.10008.1.2.4.70) is encoded with sequence start/end elements instead of an explicit length value. So the pixel data is written with an explicit length (derived from the Julia arrays?), but also encoded with sequence start/end values, making DCMTK complain.

Is there an easy way to fix this?

[notZaki]

As far as I'm aware, this package doesn't support JPEG encoding.
I'd be surprised if it even reads JPEG encoded files correctly.

The fix would be to add in that support. Not sure how difficult that would be.
One way could be to rely on an external library, similar to how pydicom does it.

[davidssmith]

How hard would it be to fix so that it could read and write the raw bytes correctly without bothering to decode?

[notZaki]

I don't know---JPEG encoded DICOM files are foreign to me. Maybe someone else can chime in.

An alternative could be to use something like gdcm which might be able to convert an uncompressed out.dcm into a properly encoded file.

[davidssmith]

When the pixel data is encoded as a sequence instead of a fixed-length field, this package concatenates the sequence into a single, typed, pixel data array, so the fact that it was originally sequence encoded is lost. I guess the only clue to that after the read is complete is indirectly found by way of the Transfer Syntax.