diff --git a/scripts/ImageReader.py b/scripts/ImageReader.py
index fffa310..5b03c0e 100644
--- a/scripts/ImageReader.py
+++ b/scripts/ImageReader.py
@@ -25,6 +25,14 @@ def get_image(self,imagefile):
         img = nibabel.load(os.path.join(self.sourcedir,imagefile))
         
         return img.get_data()
+        
+    def get_raw_image(self,imagefile):
+        """ Returns a data structure with the voxel contents """        
+        
+        # Load an image and convert it to an array
+        img = nibabel.load(os.path.join(self.sourcedir,imagefile))
+        
+        return img
     
     def mask_image(self,imagefile, maskfile):
         """ 
diff --git a/scripts/zscoring.py b/scripts/zscoring.py
index 72f8679..49f38ab 100644
--- a/scripts/zscoring.py
+++ b/scripts/zscoring.py
@@ -1 +1,34 @@
-#Hi.
+
+import os
+import numpy as np
+import nibabel as nb
+
+# TODO: we have to figure out how to make modules
+import ImageReader
+
+def zscoringNII(filename,sourcedir='../../brainimages'):
+    """ zscores a .nii.gz image only on non-zero values 
+        and saves the same file _Z.nii.gz    
+    """
+
+    # Read images
+    reader = ImageReader.ImageReader(sourcedir)
+    img = reader.get_raw_image(filename)
+    data = img.get_data()
+    affine = img.get_affine()
+    
+    # Compute non-zero mean
+    nonzdata = data[np.nonzero(data)]
+    meanz = np.mean(nonzdata)
+    
+    # z-scoring non-zero values
+    # TODO: Use a matrix formulation for speed
+    vect = np.vectorize(lambda x : (x-meanz)/meanz if x !=0.0 else 0.0)
+    x = vect(data)
+    
+    new_image = nb.Nifti1Image(x, affine)
+    nb.save(new_image, os.path.join(sourcedir, filename[:-7]+'_Z.nii.gz'))
+
+# This is the standard way of checking if the file is being executed    
+if __name__ == "__main__":
+    zscoringNII('111_GS_VBM_GM_reg2STD.nii.gz')
\ No newline at end of file