Revert inference back to v0.4.8

docs/getting_started.md

@@ -40,7 +40,7 @@ naming convention for the files within your dataset. However, the MIST pipeline
 images in each patient directory is consistent or that each type of image is identifiable by a 
 list of identifier strings.
 
-MIST offers support for MSD and CSV formatted datasets. For more details, please see [MSD and CSV Formatted Data](#msd-and-csv-formatted-data).
+MIST offers support for MSD and CSV formatted datasets. For more details, please see [MSD and CSV Formatted Data](usage.md).
 
 Once your dataset is in the correct format, the final step is to prepare a small JSON 
 file containing the details of the dataset. We specifically ask for the following key-value 

mist/inference/main_inference.py

@@ -80,19 +80,19 @@ def back_to_original_space(pred, og_ants_img, config, fg_bbox):
         new_size = og_ants_img.shape
 
     # Bug fix for sitk resample
-    new_size = np.array(new_size, dtype='int').tolist()    
+    new_size = np.array(new_size, dtype='int').tolist()
+
     pred = resample_mask(pred,
                          labels=list(range(len(config["labels"]))),
                          target_spacing=og_ants_img.spacing,
                          new_size=new_size)
-    
+
     # Return prediction to original image space
-    # FIX: replace ants.reorient_image2 - there is a bug in there
-    pred = ants.from_numpy(pred.numpy())
-    pred.set_spacing(og_ants_img.spacing)
+    og_orientation = ants.get_orientation(og_ants_img)
+    pred = ants.reorient_image2(pred, og_orientation)
     pred.set_direction(og_ants_img.direction)
     pred.set_origin(og_ants_img.origin)
-    
+
     # Appropriately pad back to original size
     if fg_bbox is not None:
         pred = decrop_from_fg(pred, fg_bbox)