Merge pull request #361 from chrisfilo/enh/similarity

Added similarity calculation interface

Author: satra

CHANGES

@@ -1,7 +1,7 @@
 Since last release
 ==================
 
-* ENH: Interface for WatershedBEM from the MNE software package
+* ENH: New interfaces: nipy.Similarity, WatershedBEM
 
 * FIX: Afni outputs should inherit from TraitedSpec
 

nipype/interfaces/nipy/__init__.py

@@ -1,2 +1,3 @@
 from .model import FitGLM, EstimateContrast
 from .preprocess import ComputeMask, FmriRealign4d
+from .utils import Similarity

nipype/interfaces/nipy/utils.py

@@ -0,0 +1,91 @@
+import warnings
+
+import nibabel as nb
+
+from ...utils.misc import package_check
+
+try:
+    package_check('nipy')
+except Exception, e:
+    warnings.warn('nipy not installed')
+else:
+    from nipy.algorithms.registration.histogram_registration import HistogramRegistration
+    from nipy.algorithms.registration.affine import Affine
+
+from ..base import (TraitedSpec, BaseInterface, traits,
+                    BaseInterfaceInputSpec, File, isdefined)
+
+
+class SimilarityInputSpec(BaseInterfaceInputSpec):
+
+    volume1 = File(exists=True, desc="3D volume", mandatory=True)
+    volume2 = File(exists=True, desc="3D volume", mandatory=True)
+    mask1 = File(exists=True, desc="3D volume", mandatory=True)
+    mask2 = File(exists=True, desc="3D volume", mandatory=True)
+    metric = traits.Either(traits.Enum('cc', 'cr', 'crl1', 'mi', 'nmi', 'slr'),
+                          traits.Callable(),
+                         desc="""str or callable
+Cost-function for assessing image similarity. If a string,
+one of 'cc': correlation coefficient, 'cr': correlation
+ratio, 'crl1': L1-norm based correlation ratio, 'mi': mutual
+information, 'nmi': normalized mutual information, 'slr':
+supervised log-likelihood ratio. If a callable, it should
+take a two-dimensional array representing the image joint
+histogram as an input and return a float.""", usedefault=True)
+
+
+class SimilarityOutputSpec(TraitedSpec):
+
+    similarity = traits.Float(desc="Similarity between volume 1 and 2")
+
+
+class Similarity(BaseInterface):
+    """Calculates similarity between two 3D volumes. Both volumes have to be in
+    the same coordinate system, same space within that coordinate system and
+    with the same voxel dimensions.
+
+    Example
+    -------
+    >>> from nipype.interfaces.nipy.utils import Similarity
+    >>> similarity = Similarity()
+    >>> similarity.inputs.volume1 = 'rc1s1.nii'
+    >>> similarity.inputs.volume2 = 'rc1s2.nii'
+    >>> similarity.inputs.metric = 'cr'
+    >>> res = similarity.run() # doctest: +SKIP
+    """
+
+    input_spec = SimilarityInputSpec
+    output_spec = SimilarityOutputSpec
+
+    def _run_interface(self, runtime):
+
+        vol1_nii = nb.load(self.inputs.volume1)
+        vol2_nii = nb.load(self.inputs.volume2)
+        
+        if isdefined(self.inputs.mask1):
+            mask1_nii = nb.load(self.inputs.mask1)
+            mask1_nii = nb.Nifti1Image(nb.load(self.inputs.mask1).get_data() == 1, mask1_nii.get_affine(),
+                                       mask1_nii.get_header())
+        else:
+            mask1_nii = None
+            
+        if isdefined(self.inputs.mask2):
+            mask2_nii = nb.load(self.inputs.mask2)
+            mask2_nii = nb.Nifti1Image(nb.load(self.inputs.mask2).get_data() == 1, mask2_nii.get_affine(),
+                                       mask2_nii.get_header())
+        else:
+            mask2_nii = None
+
+        histreg = HistogramRegistration(from_img = vol1_nii,
+                                        to_img = vol2_nii,
+                                        similarity=self.inputs.metric,
+                                        from_mask = mask1_nii,
+                                        to_mask = mask2_nii)
+        self._similarity = histreg.eval(Affine())
+
+        return runtime
+
+    def _list_outputs(self):
+        outputs = self._outputs().get()
+        outputs['similarity'] = self._similarity
+        return outputs