ENH: Refactor NewSegmentInputSpec to support multichannel segmentation

Author: jhuguetn

nipype/interfaces/spm/preprocess.py

@@ -1628,22 +1628,34 @@ def _list_outputs(self):
 
 
 class NewSegmentInputSpec(SPMCommandInputSpec):
-    channel_files = InputMultiPath(
-        ImageFileSPM(exists=True),
-        mandatory=True,
-        desc="A list of files to be segmented",
-        field="channel",
-        copyfile=False,
-    )
-    channel_info = traits.Tuple(
-        traits.Float(),
-        traits.Float(),
-        traits.Tuple(traits.Bool, traits.Bool),
-        desc="""A tuple with the following fields:
-            - bias reguralisation (0-10)
-            - FWHM of Gaussian smoothness of bias
-            - which maps to save (Field, Corrected) - a tuple of two boolean values""",
-        field="channel",
+    channels = traits.List(
+        traits.Tuple(
+            InputMultiPath(
+                ImageFileSPM(exists=True),
+                mandatory=True,
+                desc="A list of files to be segmented",
+                field="channel",
+                copyfile=False,
+            ),
+            traits.Tuple(
+                traits.Float(),
+                traits.Float(),
+                traits.Tuple(traits.Bool, traits.Bool),
+                desc="""A tuple with the following fields:
+                        - bias reguralisation (0-10)
+                        - FWHM of Gaussian smoothness of bias
+                        - which maps to save (Field, Corrected) - a tuple of two boolean values""",
+                field="channel",
+            )
+        ),
+         desc="""A list of tuples (one per each channel)
+         with the following fields:
+            - a list of channel files (only 1rst channel files will be segmented)
+            - a tuple with the following channel-specific info fields:
+               - bias reguralisation (0-10)
+               - FWHM of Gaussian smoothness of bias
+               - which maps to save (Field, Corrected) - a tuple of two boolean values""",
+         field="channel",
     )
     tissues = traits.List(
         traits.Tuple(
@@ -1719,23 +1731,24 @@ class NewSegment(SPMCommand):
     """Use spm_preproc8 (New Segment) to separate structural images into
     different tissue classes. Supports multiple modalities.
 
-    NOTE: This interface currently supports single channel input only
+    NOTE: This interface supports multi-channel inputs
 
     http://www.fil.ion.ucl.ac.uk/spm/doc/manual.pdf#page=43
 
     Examples
     --------
     >>> import nipype.interfaces.spm as spm
     >>> seg = spm.NewSegment()
-    >>> seg.inputs.channel_files = 'structural.nii'
-    >>> seg.inputs.channel_info = (0.0001, 60, (True, True))
+    >>> seg.inputs.channels = [('structural.nii',(0.0001, 60, (True, True)))]
     >>> seg.run() # doctest: +SKIP
 
     For VBM pre-processing [http://www.fil.ion.ucl.ac.uk/~john/misc/VBMclass10.pdf],
     TPM.nii should be replaced by /path/to/spm8/toolbox/Seg/TPM.nii
 
     >>> seg = NewSegment()
-    >>> seg.inputs.channel_files = 'structural.nii'
+    >>> channel1= ('t1w.nii',(0.0001, 60, (True, True)))
+    >>> channel2= ('t2w.nii',(0.0001, 60, (True, True)))
+    >>> seg.inputs.channels = [channel1, channel2]
     >>> tissue1 = (('TPM.nii', 1), 2, (True,True), (False, False))
     >>> tissue2 = (('TPM.nii', 2), 2, (True,True), (False, False))
     >>> tissue3 = (('TPM.nii', 3), 2, (True,False), (False, False))
@@ -1764,16 +1777,20 @@ def _format_arg(self, opt, spec, val):
         """Convert input to appropriate format for spm
         """
 
-        if opt in ["channel_files", "channel_info"]:
+        #if opt in ["channel_files", "channel_info"]:
+        if opt == "channels":
             # structure have to be recreated because of some weird traits error
-            new_channel = {}
-            new_channel["vols"] = scans_for_fnames(self.inputs.channel_files)
-            if isdefined(self.inputs.channel_info):
-                info = self.inputs.channel_info
-                new_channel["biasreg"] = info[0]
-                new_channel["biasfwhm"] = info[1]
-                new_channel["write"] = [int(info[2][0]), int(info[2][1])]
-            return [new_channel]
+            new_channels = []
+            for channel in val:
+                new_channel = {}
+                new_channel["vols"] = scans_for_fnames(channel[0])
+                if isdefined(channel[1]):
+                    info = channel[1]
+                    new_channel["biasreg"] = info[0]
+                    new_channel["biasfwhm"] = info[1]
+                    new_channel["write"] = [int(info[2][0]), int(info[2][1])]
+                new_channels.append(new_channel)
+            return new_channels
         elif opt == "tissues":
             new_tissues = []
             for tissue in val:
@@ -1814,7 +1831,8 @@ def _list_outputs(self):
             outputs["normalized_class_images"].append([])
             outputs["modulated_class_images"].append([])
 
-        for filename in self.inputs.channel_files:
+        # main outputs are generated for the first channel images only
+        for filename in self.inputs.channels[0][0]:
             pth, base, ext = split_filename(filename)
             if isdefined(self.inputs.tissues):
                 for i, tissue in enumerate(self.inputs.tissues):
@@ -1853,15 +1871,19 @@ def _list_outputs(self):
                         os.path.join(pth, "y_%s.nii" % base)
                     )
 
-            if isdefined(self.inputs.channel_info):
-                if self.inputs.channel_info[2][0]:
-                    outputs["bias_field_images"].append(
-                        os.path.join(pth, "BiasField_%s.nii" % (base))
-                    )
-                if self.inputs.channel_info[2][1]:
-                    outputs["bias_corrected_images"].append(
-                        os.path.join(pth, "m%s.nii" % (base))
-                    )
+        # bias field related images are generated for images in all channels
+        for channel in self.inputs.channels:
+            for filename in channel[0]:
+                pth, base, ext = split_filename(filename)
+                if isdefined(channel[1]):
+                    if channel[1][2][0]:
+                        outputs["bias_field_images"].append(
+                            os.path.join(pth, "BiasField_%s.nii" % (base))
+                        )
+                    if channel[1][2][1]:
+                        outputs["bias_corrected_images"].append(
+                            os.path.join(pth, "m%s.nii" % (base))
+                        )
         return outputs