Saving a pre-processed (transformed) image

[philnovv]

Hi. I'd like to save the result of the following preprocessing of a batch of data.
Here's the preprocess and save transforms:

data = {
    'image: 'my_image.nii.gz',
}
FILE_KEYS = ['image']

preprocess = Compose([
    LoadImaged(FILE_KEYS),
    EnsureChannelFirstd(FILE_KEYS),
    monai.transforms.Orientationd(FILE_KEYS, axcodes="RAS"),
    monai.transforms.Spacingd(FILE_KEYS, pixdim=(2.0, 2.0, 2.0), mode=("bilinear")),
    monai.transforms.SpatialPadd(FILE_KEYS, spatial_size=[96,96,96]),
    monai.transforms.CenterSpatialCropd(FILE_KEYS, roi_size=[96,96,96])
])

save = SaveImaged(
    FILE_KEYS,
    output_dir='output',
    output_postfix="crop",
    resample=False,
    output_dtype=np.int16,
    separate_folder=False)

When I extract a batch of data as follows and then attempt to save, it fails:

train_ds = IXIDataset(
        root_dir='C:/Dev',
        section="training",
        transform=preprocess,
        download=True,
    )

train_loader = DataLoader(train_ds, batch_size=1, shuffle=True, num_workers=0)
b = next(train_loader)
save(b)

Here's the error it throws:

Traceback (most recent call last):
File "C:\Dev\plar.py", line 362, in
main()
File "C:\Dev\plar.py", line 309, in main
save(b)
File "C:\Users\15142\anaconda3\lib\site-packages\monai\transforms\io\dictionary.py", line 295, in call
self.saver(img=d[key], meta_data=meta_data)
File "C:\Users\15142\anaconda3\lib\site-packages\monai\transforms\io\array.py", line 458, in call
raise RuntimeError(
RuntimeError: SaveImage cannot find a suitable writer for output\IXI249-Guys-1072-T1.nii_crop.nii.gz.
Please install the writer libraries, see also the installation instructions:
https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies.
The current registered writers for .nii.gz: (<class 'monai.data.image_writer.NibabelWriter'>, <class 'monai.data.image_writer.ITKWriter'>).
Traceback (most recent call last):
File "C:\Users\15142\anaconda3\lib\site-packages\monai\transforms\io\array.py", line 445, in call
writer_obj.set_metadata(meta_dict=meta_data, **self.meta_kwargs)
File "C:\Users\15142\anaconda3\lib\site-packages\monai\data\image_writer.py", line 591, in set_metadata
self.data_obj, self.affine = self.resample_if_needed(
File "C:\Users\15142\anaconda3\lib\site-packages\monai\data\image_writer.py", line 270, in resample_if_needed
output_array = resampler(data_array[None], dst_affine=target_affine, spatial_size=output_spatial_shape)
File "C:\Users\15142\anaconda3\lib\site-packages\monai\utils\deprecate_utils.py", line 221, in wrapper
return func(*args, **kwargs)
File "C:\Users\15142\anaconda3\lib\site-packages\monai\transforms\spatial\array.py", line 251, in call
src_affine = to_affine_nd(spatial_rank, src_affine_).to(_dtype)
File "C:\Users\15142\anaconda3\lib\site-packages\monai\data\utils.py", line 929, in to_affine_nd
raise ValueError(f"affine must have 2 dimensions, got {affine_np.ndim}.")
ValueError: affine must have 2 dimensions, got 3.

[wyli]

After EnsureChannelFirstd, the data dimensionality should include a channel dimension even if it's a single channel image, for example, the shape should be (1, 300, 300, 300) for a single channel 300x300x300 volume... you can put monai.transforms.DataStatsd(FILE_KEYS), to print the debug info in the list of preprocessing transforms to check the shapes. EnsureChannelFirstd has a channel_dim parameter which you can adjust.