pyalfe.tasks.segmentation

pyalfe.tasks.segmentation#

class pyalfe.tasks.segmentation.Segmentation(inference_model: InferenceModel, image_processor: ImageProcessor)[source]#

Bases: Task

This is the parent class for segmentation tasks.

predict(images: list[str | Path] | tuple[str | Path], pred: str | Path)[source]#

Creates a prediction image given the input images.

Parameters:

  • images (list[str or Path] or tuple[str or Path]) – A list or tuple of input images.

  • pred (list[str or Path]) – Path to the output prediction segmentation image.

post_process(pred, seg, mask=None)[source]#

Post processes the prediction segmentation image by applying an optional mask to create the final segmentation. For example, in a lesion segmentation task you can remove any lesion outside the brain by applying a brain mask.

Parameters:

  • pred (str or Path) – Path to the prediction segmentation image.

  • mask (str or Path) – Path to the mask image.

  • seg (str or Path) – Path to the output post-processed segmentation image.

label_segmentation_components(seg, comp)[source]#

Breaks down a binary segmentation into individual connected components.

Parameters:

  • seg (str or Path) – Path to the segmentation image.

  • comp (str or Path) – Path to the output components image.

run(accession: str) → None[source]#

Runs the task for an accession.

class pyalfe.tasks.segmentation.MultiModalitySegmentation(inference_model: InferenceModel, image_processor: ImageProcessor, pipeline_dir: PipelineDataDir, modality_list: list[str], output_modality: str, image_type_input: str = 'skullstripped', image_type_output: str = 'abnormal_seg', image_type_mask: str | None = None, segmentation_dir: str = 'abnormalmap', components: bool = False, overwrite: bool = True)[source]#

Bases: Segmentation

This task generates a segmentation from several input modalities.

Parameters:

  • inference_model (InferenceModel) – The inference model object.

  • image_processor (ImageProcessor) – The image processor object.

  • pipeline_dir (PipelineDataDir) – The pipeline data directory object.

  • modality_list (list[Modality]) – The modalities used for segmentation. For example, [T1, T1Post].

  • output_modality (Modality) – The modality where the segmentation output will be saved in its directory.

  • image_type_input (str = 'skullstripped') – The type of image that should be used as input. Default is skullstripped.

  • image_type_output (str) – The type of output image that segmentation should be saved as. Default is abnormal_seg.

  • image_type_mask (str = None) – The type of output image modality that should be used as a mask. Default is None.

  • segmentation_dir (str = 'abnormalmap') – The directory where the segmentaiton should be saved into. Default is abnormalmap.

  • components (bool = False) – If True, creates a segmentation image that separately labels all the connected components of the original segmentation. Default is False.

  • overwrite (bool = True) – Whether to overwrite existing output segmentation images. Default is True.

logger = <Logger MultiModalitySegmentation (DEBUG)>#
run(accession)[source]#

Runs the task for an accession.

class pyalfe.tasks.segmentation.SingleModalitySegmentation(inference_model: InferenceModel, image_processor: ImageProcessor, pipeline_dir: PipelineDataDir, modality, image_type_input: str = 'skullstripped', image_type_output: str = 'abnormal_seg', image_type_mask: str | None = None, segmentation_dir: str = 'abnormalmap', components: bool = False, overwrite: bool = True)[source]#

Bases: MultiModalitySegmentation

This task generates a segmentation from a single input modality.

Parameters

image_processor: ImageProcessor

The image processor object.

pipeline_dir: PipelineDataDir

The pipeline data directory object.

modality: Modality

The modality used for segmentation. For example, [T1, T1Post].

image_type_input: str = ‘skullstripped’

The type of image that should be used as input. Default is skullstripped.

image_type_output: str

The type of output image that segmentation should be saved as. Default is abnormal_seg.

image_type_mask: str = None

The type of output image modality that should be used as a mask. Default is None.

segmentation_dir: str = ‘abnormalmap’

The directory where the segmentaiton should be saved into. Default is abnormalmap.

components: bool = False

If True, creates a segmentation image that separately labels all the connected components of the original segmentation. Default is False.

overwrite: bool = True

Whether to overwrite existing output segmentation images. Default is True.

logger = <Logger SingleModalitySegmentation (DEBUG)>#
class pyalfe.tasks.segmentation.TissueWithPriorSegementation(inference_model: InferenceModel, image_processor: ImageProcessor, pipeline_dir: PipelineDataDir, image_type_input: str = 'trim_upsampled', image_type_output: str = 'tissue_seg', template_name: str = 'Tissue', overwrite: bool = True)[source]#

Bases: Segmentation

This task generates tissue segmentation from T1 and a prior atlas based segementation.

Parameters:

  • inference_model (InferenceModel) – The inference model object.

  • image_processor (ImageProcessor) – The image processor object.

  • pipeline_dir (PipelineDataDir) – The pipeline data directory object.

  • image_type_input (str = 'skullstripped') – The type of image that should be used as input. Default is skullstripped.

  • image_type_output (str) – The type of output image that segmentation should be saved as. Default is tissue_seg.

  • overwrite (bool = True) – Whether to overwrite existing output segmentation images. Default is True.

logger = <Logger TissueWithPriorSegementation (DEBUG)>#
run(accession)[source]#

Runs the task for an accession.

class pyalfe.tasks.segmentation.SynthSegTissueSegmentation(inference_model: InferenceModel, image_processor: ImageProcessor, pipeline_dir: PipelineDataDir, image_type_input: str = 'trim_upsampled', image_type_output: str = 'tissue_seg', overwrite: bool = True)[source]#

Bases: Segmentation

This task generates tissue segmentation using freesurfer’s synthseg

Parameters:

  • inference_model (InferenceModel) – The inference model object.

  • image_processor (ImageProcessor) – The image processor object.

  • pipeline_dir (PipelineDataDir) – The pipeline data directory object.

  • image_type_input (str = 'skullstripped') – The type of image that should be used as input. Default is skullstripped.

  • image_type_output (str) – The type of output image that segmentation should be saved as. Default is tissue_seg.

  • overwrite (bool = True) – Whether to overwrite existing output segmentation images. Default is True.

logger = <Logger SynthSegTissueSegmentation (DEBUG)>#
post_process(pred, seg, mask=None)[source]#

Post processes the prediction segmentation image by applying an optional mask to create the final segmentation. For example, in a lesion segmentation task you can remove any lesion outside the brain by applying a brain mask.

Parameters:

  • pred (str or Path) – Path to the prediction segmentation image.

  • mask (str or Path) – Path to the mask image.

  • seg (str or Path) – Path to the output post-processed segmentation image.

run(accession)[source]#

Runs the task for an accession.