added BACKGROUND_KEY in femoral_cartilage class; fixed bug in split r…

…egions method: background pixels are not take into account while computing center_of_mass; split_regions method returns an array with NaN values in the background

tissues/femoral_cartilage.py

@@ -39,6 +39,9 @@ class FemoralCartilage(Tissue):
     LATERAL_KEY = 1
     SAGGITAL_KEYS = [MEDIAL_KEY, LATERAL_KEY]
 
+    # Background Key
+    BACKGROUND_KEY = 3
+
     def __init__(self, weights_dir=None, medial_to_lateral=None):
         """
         :param weights_dir: Directory to weights files
@@ -171,20 +174,21 @@ def split_regions(self, unrolled_quantitative_map):
         """
 
         # create unrolled mask from unrolled map
+        unrolled_quantitative_map = np.nan_to_num(unrolled_quantitative_map)
         unrolled_mask_indexes = np.nonzero(unrolled_quantitative_map)
         unrolled_mask = np.zeros((unrolled_quantitative_map.shape[0], unrolled_quantitative_map.shape[1]))
         unrolled_mask[unrolled_mask_indexes] = 1
 
         # find the center of mass of the unrolled mask
         center_of_mass = sni.measurements.center_of_mass(unrolled_mask)
 
-        unrolled_mask[np.where(unrolled_mask < 1)] = 3
+        unrolled_mask[np.where(unrolled_mask < 1)] = self.BACKGROUND_KEY
 
         lateral_mask = np.copy(unrolled_mask)[:, 0:np.int(np.around(center_of_mass[1]))]
         medial_mask = np.copy(unrolled_mask)[:, np.int(np.around(center_of_mass[1])):]
 
-        lateral_mask[np.where(lateral_mask < 3)] = self.LATERAL_KEY
-        medial_mask[np.where(medial_mask < 3)] = self.MEDIAL_KEY
+        lateral_mask[np.where(lateral_mask < self.BACKGROUND_KEY)] = self.LATERAL_KEY
+        medial_mask[np.where(medial_mask < self.BACKGROUND_KEY)] = self.MEDIAL_KEY
 
         if self.medial_to_lateral:
             ml_mask = np.concatenate((medial_mask, lateral_mask), axis=1)
@@ -196,9 +200,9 @@ def split_regions(self, unrolled_quantitative_map):
         central_mask = np.copy(unrolled_mask)[np.int(center_of_mass[0]):np.int(center_of_mass[0]) + 10, :]
         posterior_mask = np.copy(unrolled_mask)[np.int(center_of_mass[0]) + 10:, :]
 
-        anterior_mask[np.where(anterior_mask < 3)] = self.ANTERIOR_KEY
-        posterior_mask[np.where(posterior_mask < 3)] = self.POSTERIOR_KEY
-        central_mask[np.where(central_mask < 3)] = self.CENTRAL_KEY
+        anterior_mask[np.where(anterior_mask < self.BACKGROUND_KEY)] = self.ANTERIOR_KEY
+        posterior_mask[np.where(posterior_mask < self.BACKGROUND_KEY)] = self.POSTERIOR_KEY
+        central_mask[np.where(central_mask < self.BACKGROUND_KEY)] = self.CENTRAL_KEY
 
         acp_mask = np.concatenate((anterior_mask, central_mask, posterior_mask), axis=0)
 
@@ -209,6 +213,9 @@ def split_regions(self, unrolled_quantitative_map):
 
         self.regions_mask = np.concatenate((ml_mask, acp_mask), axis=2)
 
+        # convert backgorund label to NaN
+        self.regions_mask[self.regions_mask == self.BACKGROUND_KEY] = np.nan
+
         assert (self.regions_mask[..., 0] == ml_mask[..., 0]).all()
         assert (self.regions_mask[..., 1] == acp_mask[..., 0]).all()