Unable to correctly stream bytes from DICOM-WADO servers

[kchawla-pi]

DICOM is a file format used by MRI scanners to store MRIs and metadata.
DICOM servers which serve DICOM images over HTTP are expected to adhere to the WADO Specification for DICOM retrieval.

When I use requests with requests-toolbelt's MultipartDecoder I can successfully retrieve DICOMs with a streaming response.
When I use httpx's streaming method, the parts received are invalid DICOMs.

Unfortunately, I could not find public access WADO server to help you test this, hopefully you can find one.

How do I fix this?

Here is the Python code to reproduce.
In the code, each part_ should be bytes for a single DICOM file.

import asyncio
import io
import logging
import httpx
import pydicom
import requests
from requests_toolbelt.multipart import decoder

logger = logging.getLogger(__name__)


async def async_wado(study_instance_uid, wado_request_url, wado_request_header):
    async with httpx.AsyncClient().stream(method="GET", url=wado_request_url, headers=wado_request_header) as response:
        async for part_ in response.aiter_bytes():
            dicom_slice_ = pydicom.dcmread(io.BytesIO(part_),)  
            # InvalidDicomError,Invalid DICOM header metadata or DICM Missing


def sync_wado(study_instance_uid, wado_request_url, wado_request_header):
    wado_response = requests.get(wado_request_url, headers=wado_request_header)    
    mp_data = decoder.MultipartDecoder.from_response(wado_response)
    for part_ in mp_data.parts: 
        if 'application/dicom' in part_.headers[b'Content-Type'].decode(encoding='utf-8'):
            dicom_slice_ = pydicom.read_file(io.BytesIO(part_.content))  
            # Works. .dcm file correctly read, can be saved to disk.


if __name__ == "__main__":
    logging.basicConfig(level=logging.DEBUG)
    study_instance_uid = "<StudyInstanceUIS>"
    wado_request_url = f"https://<hostname>:<port>/wado-rs/dicom_web/studies/{study_instance_uid}"
    wado_request_header = { 'Accept': 'multipart/related; type=application/dicom' }

    sync_wado(study_instance_uid, wado_request_url, wado_request_header)
    asyncio.run(async_wado(study_instance_uid, wado_request_url, wado_request_header))

This is the error:

pydicom.errors.InvalidDicomError: File is missing DICOM File Meta Information header or the 'DICM' prefix is missing from the header. Use force=True to force reading.

force=True gives an incomplete file and metadata.

Problems

1. I can't find a multipart decoder that works with httpx's async stream object. The MultipartDecoder() provided by requests-toolbelt works with requests.Response() object but not with httpx.Response() object. While each file is ultimately 50-100KB, the entire response is 1-2GB so reading the whole response content to use it with the requests' decoder is a slow process.
2. For async_wado(), it seems I need to decode the part_ which is in bytes, btu can't figure out how/with what to do it.