parse_ascconv errors

[bpinsard]

Are there examples of how to extract ascconv from a pydicom or dicomwrapper object?

wrapper= nibabel.nicom.dicomwrappers.wrapper_from_file('path/to/dicom')
csa = nibabel.nicom.csareader.get_csa_header(wrapper.dcm_data, 'series')
mrprot= csa['tags']['MrPhoenixProtocol']['items'][0]
asc=nibabel.nicom.ascconv.parse_ascconv(mrprot, '""') # 
--> 207     attrs, content = ASCCONV_RE.match(ascconv_str).groups()
    208     attrs = OrderedDict(tuple(x.split('=')) for x in attrs.split())
    209     # Normalize string start / end markers to something Python understands

AttributeError: 'NoneType' object has no attribute 'groups'

## hacking a bit for the regex to be smarter

nibabel.nicom.ascconv.ASCCONV_RE = re.compile(
     r'.*### ASCCONV BEGIN((?:\s*[^=\s]+=[^=\s]+)*) ###\n(.*?)\n### ASCCONV END ###.*',
     flags=re.M | re.S,
)

asc=nibabel.nicom.ascconv.parse_ascconv(mrprot, '""') #have to use double quote, as it seems to be the format when extracting the MrPhoenixProtocol the way above.

  Cell In[100], line 1
    asc=nibabel.nicom.ascconv.parse_ascconv(mrprot, '""')

  File ~/.virtualenvs/neuromod_data3.10/lib/python3.10/site-packages/nibabel/nicom/ascconv.py:212 in parse_ascconv
    tree = ast.parse(content)

  File /usr/lib/python3.10/ast.py:50 in parse
    return compile(source, filename, mode, flags,

  File <unknown>:1261
    sDiffusion.sFreeDiffusionData.sComment.0             = 0x55    # 'U'
                                                                    ^
SyntaxError: invalid syntax

Maybe ASCCONV format "evolved" and ast is no longer able to parse it when it contains new tag formats.

[moloney]

Could you check if this is fixed here: https://github.com/open-dicom/csa_header

[bpinsard]

Sorry for the late reply, and thanks for the pointer to that package.
I tested it and it worked without issue, and the parsed structure is really great.
Is the plan for nibabel to use that package in the near future to deprecate and replace the nibabel CSA parsers?

[moloney]

I believe that was the original idea. For now I created a branch of dcmstack that uses csa_header for parsing meta data to help test it out on more data: moloney/dcmstack#96