Support for labels in reader

[nathandecaux]

Hello,

Thanks for your work ! I've planned to used your plugin in my project in place on my own nifti reader/writer functions, however I'm missing a simple feature : loading layers as image or labels according to the nifti dtype.
There's no dev branch so I can't push it myself, but here's a sample of code that should do the trick :

def load_nifti(filename):
    image_data = sitk.ReadImage(filename)
    image = sitk.GetArrayFromImage(image_data)
    dtype = 'labels' if image.dtype in [np.uint8,np.uint16] else 'image'
    [...]
    return {"image": image, "affine": affine, "scale": scale, "dtype": dtype, "metadata": {"metadata": metadata, "origin": origin, "direction": direction, "spacing": spacing}}

and :

def reader_function(path):
    [...]
    layer_data = [(image_data["image"], {"affine": image_data["affine"], "metadata": image_data["metadata"]}, image_data["dtype"]) for image_data in image_data_list]
    return layer_data

[Karol-G]

Hey @nathandecaux,

Thank you for this proposition. However, many Nifti images (not segmentations) are also of dtype uint8 or uint16. Always loading Nifti files of these datatypes as labels will cause issues for this scenario. Nifti has sadly no internal flag for this case so the only solution is to load segmentations as images and convert them in Napari to a labels layer.

Best,
Karol