Hi, could you provide the code you're running?

This is the code I am running:

import copy
from nipype.interfaces.ants import N4BiasFieldCorrection
n4 = N4BiasFieldCorrection()
n4.inputs.dimension = 3
n4.inputs.bspline_fitting_distance = 300
n4.inputs.shrink_factor = 3
n4.inputs.n_iterations = [50, 50, 30, 20]
n4.inputs.input_image = "/home/masoomeh/MR_Preprocess/data/Brats17_CBICA_AQA_1/Normalized_Brats17_CBICA_AQA_1_flair.nii"
n4.inputs.output_image ="/home/masoomeh/MR_Preprocess/data/Brats17_CBICA_AQA_1/corrected.nii "
n4.cmdline

Can you run:

res = n4.run()
print(res.outputs)

Great! It worked.
If I want to define an individual function of "biascorrection" using "N4BiasFieldCorrection" block, should I return "res.outputs" ?
Doing so, I got the following error:

AttributeError: 'NoneType' object has no attribute 'outputs'.

I also tried "res.output_image" as an output of N4BiasFieldCorrection.

Thanks for your help!

Sorry, I don't understand the question. What are you trying to do?

I want to create a node doing bias field correction like this:

bias = pe.Node(Function(input_names=["in_file"],
                           output_names=["bias_corrected"],
                           function=biasfield_correct), 
                           name="bias_correction")

While function biasfield_correct has been defined as follows:

def biasfield_correct(in_file):
   
    from nipype.interfaces.ants import N4BiasFieldCorrection
    n4 = N4BiasFieldCorrection()
    n4.inputs.dimension = 3
    n4.inputs.input_image = in_file
    n4.inputs.bspline_fitting_distance = 300
    n4.inputs.shrink_factor = 3
    n4.inputs.n_iterations = [50, 50, 30, 20]
    n4.inputs.convergence_threshold = 1e-6
    #n4.inputs.bspline_order = 5
    #n4.inputs.save_bias = True
    res = n4.run()

    return (res.outputs) 

putting this node in my workflow, I have the error: "res.output_image" as an output of N4BiasFieldCorrection!

You can wrap the interface in a Node without going through a Function interface:

from nipype.pipeline import engine as pe
from nipype.interfaces import ants

bias = pe.Node(
    N4BiasFieldCorrection(
        dimension=3,
        bspline_fitting_distance=300,
        shrink_factor=3,
        n_iterations=[50, 50, 30, 20],
        convergence_threshold=1e-6),
    name='bias_correction')

Then you would connect to input_image, rather than in_file, but otherwise it should be the same. And

res = bias.run()
res.outputs

should produce essentially the same results as running the interface directly, but all of the outputs should be contained in the Node's working directory.

I have the error: "res.output_image" as an output of N4BiasFieldCorrection!

I'm not sure how this would arise from what you've posted.

Hi,
Thank you for your help. Now it is working.
Sorry, I have posted the wrong error(I forgot to paste the error here) and I could not reproduce it!!

I have another problem with input stream, Selectfile node when I want to run the whole pipeline.

Following is my pipeline:

#1: import appropriate modules

import nipype.pipeline.engine as pe                        # the workflow and node wrappers
from nipype.interfaces.utility import IdentityInterface
from nipype.interfaces.ants import N4BiasFieldCorrection
from nipype.interfaces.io import SelectFiles, DataSink

import os

#2: Specify variables
base_dir = "/home/masoomeh/MR_Preprocess/"
output_dir = os.path.join(os.path.dirname(base_dir), "out")
in_file = "/home"   
subject_id = ["Brats17_CBICA_AQA_1"]

#3: Specify nodes
anat_input = pe.Node(IdentityInterface(fields=in_file, mandatory_inputs=True),name="anat_input")
biascor = pe.Node(N4BiasFieldCorrection(input_image = in_file,
                                        dimension=3,
        				bspline_fitting_distance=300,
       					shrink_factor=3,
        				n_iterations=[50, 50, 30, 20],
        				convergence_threshold=1e-6),
					name = "bias_correction")

#4: Create the workflow object, specify workflow 
wf = pe.Workflow(name='preproc')
wf.base_dir = '.'


#5: Input and output stream

templates = {'anat': 'data/{subject_id}/struct.nii'}
selectfiles = pe.Node(SelectFiles(templates,
                                  base_directory=base_dir), name="selectfiles")


datasink = pe.Node(DataSink(parameterization=False), name="datasink")
datasink.inputs.base_directory = "/home/masoomeh/MR_Preprocess/"
datasink.inputs.container = output_dir
substitutions = [('_subject_id_', '')]
datasink.inputs.substitutions = substitutions


infosource = pe.Node(IdentityInterface(fields=["subject_id"]), name="infosource")

wf.connect([
           # Connect SelectFiles and DataSink to the workflow
           (infosource, selectfiles, [("subject_id", "subject_id")]),
           (selectfiles, anat_input, [("anat", "in_file")]),
           (anat_input, biascor, [("in_file", "input_image")]),
           (anat_input, datasink, [("in_file", "output_anat")]),
            # input to datasink
           (biascor,   datasink    , [("output_image", "output_bias")]),
           ])
#6. run worflow
wf.write_graph()

wf.run()

Running the code, I have this error:

raise IOError(msg)
OSError: No files were found matching anat template: /home/masoomeh/MR_Preprocess/data/<undefined>/struct.nii

This is for only one data, while I want to extend it for more data.
I have not understood how to specify input stream!
I also need to initialize in_file (in_file = "/home" ), otherwise, I'll have this error:
NameError: name 'in_file' is not defined

Thanks in advance for your help!

anat_input = pe.Node(IdentityInterface(fields=in_file, mandatory_inputs=True),name="anat_input")

in_file == '/home', so I'm not really sure what this will do. I suspect what you want is fields=['in_file'].

biascor = pe.Node(N4BiasFieldCorrection(input_image = in_file,

Again, same issue, but I think it gets overridden by your connect statement below. You can just remove input_image=in_file here.

infosource = pe.Node(IdentityInterface(fields=["subject_id"]), name="infosource")

It doesn't appear you ever supply a subject ID to infosource. I'm guessing you actually want this to be an iterable.

I assume your datasink is okay, but it's really impossible to say without running.

Now I get the iterable idea and my pipeline is working quite well for whole subjects each includes different modalities.

infosource = pe.Node(IdentityInterface(fields=["subject_id","session_id"]), name="infosource")

infosource.iterables = [("subject_id", subject_id),
("session_id", session_id)]

templates = {'anat': 'data/{subject_id}/{session_id}.nii'}
selectfiles = pe.Node(SelectFiles(templates, base_directory=base_dir), name="selectfiles")

wf.connect([
(infosource, selectfiles, [("subject_id", "subject_id"),
("session_id","session_id")]),
...
])

Thank you very much for all your help!

No problem!

Hello , @effigies
In my application, I try "n4.run()" mentioned before, while getting an IOError: No command "N4BiasFieldCorrection" found. Is there any configuration problem or something wrong with my code?

def n4_corrction(im_input):
      n4 = N4BiasFieldCorrection()
       n4.inputs.dimension=3
       n4.inputs.input_image = im_input
       n4.inputs.bspline_fitting_distance = 300
       n4.inputs.shrink_factor = 3
       n4.inputs.n_iterations = [50, 50, 30, 20]
       n4.inputs.output_image = im_input.replace('.nii.gz', '_correcred.nii.gz')
       n4.run()

this usually happens if the command is not available in your path. try opening a terminal and see if N4BiasFieldCorrection is available in the shell.

also if you are using tools like spyder or pycharm, you will need to ensure that shell variables are exposed.

or if you are familiar with docker, you can create a docker image with ants and nipype (using neurodocker) and run the script you are trying to run within the docker instance.