pip install neuroworkfow
provide a way to format code for processing neural data, get researches rid of file finding, name jointing and other dirty works when writing data process workflow.
Offer tools for quick inspecting pipline such as:
- generate a pseudo dataset to simplify how to arrange a dataset to run a pipline or how a data set looks like after running
- generate a subpipline from a existing pipline, this is useful when only partially result is needed or using some intermediate result.
to write a data processing pipline using neuroworkflow, one need use three basic object: component, work and workflow. when run a processing pipline, one need additionally provide metadata which provide prossesed information such as subject, session and other setting.
component is the representation of a file.
when writing pipline, one need add propertis of a component to define a component. like
example_component = Component(suffix = 'bold', datatype = 'func', extension = 'nii')
some of the available properties are
| property | function | example |
|---|---|---|
| desc | describe | smooth6mm |
| datatype | data type | func |
| suffix | suffix | bold |
| extension | extension | nii.gz |
| task | task | eyesopen |
| space | space | MNI152 |
| echo | echo | 2 |
| data_place | folder to place | echo_2 |
| use_extension | force using extension | true |
when running, component will generate a file's name and directory according to its property, metadata and it's position in the whole pipline.
to simplify the initialization of a component, one can use init from to initialize a component
example_epi_json = Component.init_from(origin_epi_list, extension = 'json')
this will init a
a work is the representation of a processing step.
when writing pipline, one need add components to a work's input_components and output_components indicate what to process and what will be generated, add action as the processs.
copy_epis = Work(f'copy_epi',
[origin_epi_list],
[copied_origin_epi_list],
action = copy_file
)
and the copy_file is
def copy_file(input_file, output_file):
shutil.copyfile(input_file[0], output_file[0])
a action should accept two (input_file, output_file) or three (input_file, output_file, run_meta_data) parameters.