[ENH] Add copy_header option to N4BiasFieldCorrection

This N4 algorithm does not propagate the nifti headers after run.
This PR allows setting the option ``copy_header=True`` and that will
overwrite the N4 output with the corrected data and the original headers.

Author: oesteban

nipype/interfaces/ants/segmentation.py

@@ -259,6 +259,9 @@ class N4BiasFieldCorrectionInputSpec(ANTSCommandInputSpec):
                                   ' to file.'), xor=['bias_image'])
     bias_image = File(desc='Filename for the estimated bias.',
                       hash_files=False)
+    copy_header = traits.Bool(False, mandatory=True, usedefault=True,
+                              desc='copy headers of the original image into the '
+                                   'output (corrected) file')
 
 
 class N4BiasFieldCorrectionOutputSpec(TraitedSpec):
@@ -384,6 +387,23 @@ def _list_outputs(self):
                 self._gen_filename('bias_image'))
         return outputs
 
+    def _run_interface(self, runtime, correct_return_codes=(0,)):
+        runtime = super(N4BiasFieldCorrection, self)._run_interface(
+            runtime, correct_return_codes)
+
+        if self.inputs.copy_header:
+            import nibabel as nb
+            refnii = nb.load(self.inputs.input_image)
+            hdr = refnii.header.copy()
+            out_file = self._gen_filename('output_image')
+            nii = nb.load(out_file)
+            hdr.set_data_dtype(nii.header.get_data_dtype())
+            nb.Nifti1Image(nii.get_data(), refnii.affine, hdr).to_filename(
+                out_file)
+
+        return runtime
+
+
 
 class CorticalThicknessInputSpec(ANTSCommandInputSpec):
     dimension = traits.Enum(3, 2, argstr='-d %d', usedefault=True,

nipype/interfaces/ants/tests/test_auto_N4BiasFieldCorrection.py

@@ -14,6 +14,9 @@ def test_N4BiasFieldCorrection_inputs():
     ),
     convergence_threshold=dict(requires=['n_iterations'],
     ),
+    copy_header=dict(mandatory=True,
+    usedefault=True,
+    ),
     dimension=dict(argstr='-d %d',
     usedefault=True,
     ),