Change 'label = 0' to 'label = 1' when both the reference and predict…

…ion images are empty.

The original 'label = 0' caused that the metrics were corresponding to background and were not easy to aggregate across subjects.
Now, with 'label = 1', even cases with both empty reference and prediction are considered when computing group mean and std.

compute_metrics/compute_metrics_reloaded.py

@@ -163,7 +163,7 @@ def compute_metrics_single_subject(prediction, reference, metrics):
             break       # break to loop to avoid processing the background label ("else" block)
     # Special case when both the reference and prediction images are empty
     else:
-        label = 0
+        label = 1
         print(f'Processing label {label} -- both the reference and prediction are empty')
         bpm = BPM(prediction_data, reference_data, measures=metrics)
         dict_seg = bpm.to_dict_meas()