rsip-vision / medio Goto Github PK

Currently, private fields in the header not read.

• For the itk backend, it should be simple as:

   reader.LoadPrivateTagsOn()

• For the pdcm, I need to check if it's read by default (I think it does, but not sure).
• Not sure if it's relevant for the nib backend.

from medio import read_img
read_img("path to image", backend='itk')

Exception:

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

If we can easily move between medio and ITK then we can use ITK functionality

It would be nice to be able to create an empty image with the same size and metadata as an existing one, without needing to read in the whole existing image, but probably just its header.

Probably by using nibabel's one.

Inspired by the hypermodern-python article series and modern python cookiecutters:

• https://github.com/fedejaure/cookiecutter-modern-pypackage
• https://github.com/cjolowicz/cookiecutter-hypermodern-python

Modernize the medio Python package.

Specifically:

• Use Poetry and the pyproject.toml to define the package (#7)
• Format the codebase with black (#9)
• Add typing to the code and static type analysis with mypy
• Add linting with flake8
• Enforce all the above in GitHub actions

Inspired by contrib-pydicom:

https://github.com/pydicom/contrib-pydicom/blob/cbd27f1ac523863aeba820edf796bf040d63993f/input-output/dicom_model/dicom_dir.py#L43-L50

Looks like the following should complement medio's is_dicom function:

def is_dcm(filename):
    with open(filename, 'rb') as f:
        x = f.read(132)
    return x[128:] == b"DICM"

It is particularly useful since many times dicom files lack any extension.
I checked the above is_dcm on some data and it looks reliable.

With that, we can change the default dicom reader backend to pydicom (single and series).

Edit:
Even better, use pydicom's built-in is_dicom function (equivalent to the above is_dcm function).