HypVINN.utils.img_processing_utils¶

HypVINN.utils.img_processing_utils.get_clean_labels(segmentation)[source]¶

Get clean labels by removing non-connected components from a dilated mask and any connected component with size less than 3.

Parameters:

segmentationnp.ndarray: The segmentation mask.

Returns:

clean_seg: np.ndarray: The cleaned segmentation mask.
labels_cc: np.ndarray: The labels of the connected components in the segmentation mask.

HypVINN.utils.img_processing_utils.get_clean_mask(segmentation, optic=False)[source]¶

Get a clean mask by removing non-connected components from a dilated mask.

This function takes a segmentation mask and an optional boolean flag indicating whether to consider optic labels. It removes not connected components from the segmentation mask and returns the cleaned segmentation mask, the labels of the connected components, and a flag indicating whether to save the mask.

Parameters:

segmentationnp.ndarray: The input segmentation mask.
opticbool, default=False: A flag indicating whether to consider optic labels. Default is False.

Returns:

clean_segnp.ndarray: The cleaned segmentation mask.
labels_ccnp.ndarray: The labels of the connected components in the segmentation mask.
savemaskbool: A flag indicating whether to save the mask.

HypVINN.utils.img_processing_utils.img2axcodes(img)[source]¶

Convert the affine matrix of an image to axis codes.

This function takes an image as input and returns the axis codes corresponding to the affine matrix of the image.

Parameters:

imgnibabel image object: The input image.

Returns:

tuple: The axis codes corresponding to the affine matrix of the image.

HypVINN.utils.img_processing_utils.save_logits(logits, orig_path, ras_affine, ras_header, save_dir, mode)[source]¶

Save the logits (raw model outputs) as a NIfTI image.

This function takes the logits, reorients the image to match the original image’s orientation, and saves the results.

Parameters:

logitsnpt.NDArray[float]: The raw model outputs.
orig_pathPath: The path to the original image.
ras_affinenpt.NDArray[float]: The affine transformation of the RAS orientation.
ras_headernib.nifti1.Nifti1Header: The header of the RAS orientation.
save_dirPath: The directory where the logits will be saved.
modestr: The mode of operation.

Returns:

save_as: Path: The path where the logits were saved.

HypVINN.utils.img_processing_utils.save_segmentation(prediction, orig_path, ras_affine, ras_header, subject_dir, seg_file, mask_file, save_mask=False)[source]¶

Save the segmentation results.

This function takes the prediction results, cleans the labels, maps them to FreeSurfer Hypvinn Labels, and saves the results. It also reorients the mask and prediction images to match the original image’s orientation.

Parameters:

predictionnp.ndarray: The prediction results.
orig_pathPath: The path to the original image.
ras_affinenpt.NDArray[float]: The affine transformation of the RAS orientation.
ras_headernibabel header object: The header of the RAS orientation.
subject_dirPath: The directory where the subject’s data is stored.
seg_filePath: The file where the segmentation will be saved (relative to subject_dir/mri).
mask_filestr: The file where the mask will be saved (relative to subject_dir/mri).
save_maskbool, default=False: Whether to save the mask or not. Default is False.

Returns:

float: The time taken to save the segmentation.