do not separate df for empty refs and preds

compute_metrics_reloaded.py

@@ -157,10 +157,10 @@ def compute_metrics_single_subject(prediction, reference, metrics):
     # append entry into the output_list to store the metrics for the current subject
     metrics_dict = {'reference': reference, 'prediction': prediction}
 
-    # NOTE: this is hacky fix to try to speed up metrics computation, tread very carefully
-    if len(unique_labels) == 2:
-        # compute metrics only for lesions
-        unique_labels = unique_labels[1:]
+    # # NOTE: this is hacky fix to try to speed up metrics computation, tread very carefully
+    # if len(unique_labels) == 2:
+    #     # compute metrics only for lesions
+    #     unique_labels = unique_labels[1:]
 
     # loop over all unique labels, e.g., voxels with values 1, 2, ...
     # by doing this, we can compute metrics for each label separately, e.g., separately for spinal cord and lesions
@@ -260,11 +260,11 @@ def main():
     # Convert JSON data to pandas DataFrame
     df = build_output_dataframe(output_list)
 
-    # create a separate dataframe for columns where EmptyRef and EmptyPred is True
-    df_empty_masks = df[(df['EmptyRef'] == True) & (df['EmptyPred'] == True)]
+    # # create a separate dataframe for columns where EmptyRef and EmptyPred is True
+    # df_empty_masks = df[(df['EmptyRef'] == True) & (df['EmptyPred'] == True)]
 
-    # keep only the rows where either pred or ref is non-empty or both are non-empty
-    df = df[(df['EmptyRef'] == False) | (df['EmptyPred'] == False)]
+    # # keep only the rows where either pred or ref is non-empty or both are non-empty
+    # df = df[(df['EmptyRef'] == False) | (df['EmptyPred'] == False)]
 
     # Compute mean and standard deviation of metrics across all subjects
     df_mean = (df.drop(columns=['reference', 'prediction', 'EmptyRef', 'EmptyPred']).groupby('label').