StrokeSegLab.postprocessing.postprocessor.Postprocessor Class Reference

Public Member Functions
None	__init__ (self, gui=None)
None	check_viewer (self, str viewer)
None	run (self, np.ndarray data, np.ndarray affine, str input_path, list[tuple[int, int]] bbox, tuple[int, int, int] original_shape, str temp_dir, str trsf_path, tuple[float, float, float] old_spacing, list[tuple[int, int]] padding, str bet, str MNI_base_image, float threshold, bool open_viewer=False)

Public Attributes
	option = Option()
	logger = logging.getLogger()
	wrapper = AnimaWrapper()
	resampler = Resampler()
	config = Config()
	viewer = Viewer()
	gui = gui

Protected Member Functions
str	_save_img (self, str temp_dir, np.ndarray data, str base_name, np.ndarray affine, str name)
tuple[np.ndarray, np.ndarray]	_convert_to_segmentation (self, np.ndarray data, float threshold)
None	_register_seg_to_reference (self, str img_path, str trsf_path, str ref)
None	_binarize_seg (self, str seg_path, float threshold=0.5)
None	_print_action (self, str action_name)
np.ndarray	_remove_padding (self, np.ndarray data, list[tuple[int, int]] padding)
np.ndarray	_uncrop_from_bbox (self, np.ndarray data, tuple[slice, slice, slice] slicer, tuple[int, int, int] original_shape)

Detailed Description

This class performs postprocessing on 3D images

Constructor & Destructor Documentation

__init__()

None StrokeSegLab.postprocessing.postprocessor.Postprocessor.__init__	(		self,
			gui = None )

Initialize the main postprocessing class with optional GUI integration.

Args:
    gui (GUIMain, optional): Instance of the gui class to enable the display of messages and status updates. Defaults to None.

Member Function Documentation

_binarize_seg()

None StrokeSegLab.postprocessing.postprocessor.Postprocessor._binarize_seg ( self, str seg_path, float threshold = 0.5 )

protected

@public
Convert float segmentation into a binary mask (0 / 1).

_convert_to_segmentation()

tuple[np.ndarray,np.ndarray] StrokeSegLab.postprocessing.postprocessor.Postprocessor._convert_to_segmentation ( self, np.ndarray data, float threshold )

protected

@public
Convert the output of the model into a segmentation based on the threshold given : 
- Take the first element of the batch axis
- Apply a softmax on the channel dimensions (2 channels : background and lesion)
- Exctract the probabilities of the lesion class (1)
- Save it if the option 'pmap' is specified
- Apply a threshold to create a binary mask

Args:
    data (np.ndarray): Model output with shape (batch_size, channels, x, y, z)
    threshold (float): Threshold to create the binary mask

Returns:
    tuple[np.ndarray,np.ndarray]: seg : binary mask, pmap : probability map if requested

_print_action()

None StrokeSegLab.postprocessing.postprocessor.Postprocessor._print_action ( self, str action_name )

protected

@public
Log the current action
Args:
    action_name (str): Name of the action

_register_seg_to_reference()

None StrokeSegLab.postprocessing.postprocessor.Postprocessor._register_seg_to_reference ( self, str img_path, str trsf_path, str ref )

protected

@public
Register a segmentation to a reference using Anima.
Applies a binarization step afterward to remove float interpolation artifacts.

_remove_padding()

np.ndarray StrokeSegLab.postprocessing.postprocessor.Postprocessor._remove_padding ( self, np.ndarray data, list[tuple[int, int]] padding )

protected

@public
Remove padding from a numpy array based on a list of tuples specifying the intervals to keep for each dimension

Args:
    data (np.ndarray): Data numpy array to process
    padding (list[tuple[int, int]]): List of tuples that specify the padding to remove (start, end)

Returns:
    np.ndarray: Array with padding removed

_save_img()

str StrokeSegLab.postprocessing.postprocessor.Postprocessor._save_img ( self, str temp_dir, np.ndarray data, str base_name, np.ndarray affine, str name )

protected

@public
Save a NIFTI image with a dynamically constructed name

Args:
    temp_dir (str): Directory where the image will be save
    data (np.ndarray): Data of the image 
    base_name (str): Basename of the image, without extension
    affine (np.ndarray): Affine transformation matrix of the image
    name (str): Identifier to determine filename suffix

Returns:
    str: Full path of the saved file

_uncrop_from_bbox()

np.ndarray StrokeSegLab.postprocessing.postprocessor.Postprocessor._uncrop_from_bbox ( self, np.ndarray data, tuple[slice,slice,slice] slicer, tuple[int,int,int] original_shape )

protected

@public
Place the cropped data back into a full-size volume at the location specified by slicer. Then transpose the volume axes from (x, y, z) to (z, y, x)

Args:
    data (np.ndarray): Cropped data
    slicer (tuple[slice,slice,slice]): Tuple of slices that specify where to place data
    original_shape (tuple[int,int,int]): Shape of full original volume

Returns:
    np.ndarray: Uncropped data

check_viewer()

None StrokeSegLab.postprocessing.postprocessor.Postprocessor.check_viewer	(		self,
		str	viewer )

Call for check_viewer method from the viewer class to check if the viewer specify exist

Args:
    viewer (str): Name of the viewer specify

run()

None StrokeSegLab.postprocessing.postprocessor.Postprocessor.run	(		self,
		np.ndarray	data,
		np.ndarray	affine,
		str	input_path,
		list[tuple[int, int]]	bbox,
		tuple[int,int,int]	original_shape,
		str	temp_dir,
		str	trsf_path,
		tuple[float,float,float]	old_spacing,
		list[tuple[int, int]]	padding,
		str	bet,
		str	MNI_base_image,
		float	threshold,
		bool	open_viewer = False )

Run the entire postprocessing pipeline on the data produced by the inference step : 
- Convert to segmentation, return a pmap if the option is specified by the user
- Remove padding
- Uncrop
- Resampling to the original spacing
- Saving image
- Register to reference only if the inverse transformation was applied during preprocessing
- Open viewer with input image and binary mask if the option was specified by the user


Args:
    data (np.ndarray): Data prostprocessed
    affine (np.ndarray): affine matrix
    input_path (str): Input path
    bbox (list[tuple[int, int]]): Bounding box coordinates used for cropping
    original_shape (tuple[int,int,int]): Original shape of the image before preprocessing
    temp_dir (str): Path to the temporary directory
    trsf_path (str): Path to the transformation text file
    old_spacing (tuple[float,float,float]): Voxel spacing of the input image
    padding (list[tuple[int, int]]): List of padding tuples applied to each dimension
    bet (str): Path to the brain-extracted T1 image
    MNI_base_image (str): Reference MNI image array or None
    threshold (float): Segmentation threshold value
    open_viewer (bool, optional): If True open a viewer with input image and segmentation. Defaults to False.

---

The documentation for this class was generated from the following file:

• /home/ymahe/Desktop/StrokeSeg/StrokeSegLab/postprocessing/postprocessor.py