Skip to main content
. Author manuscript; available in PMC: 2020 Apr 21.
Published in final edited form as: Acta Neurochir (Wien). 2020 Feb 1;162(3):553–566. doi: 10.1007/s00701-020-04234-8

Table 3.

Number of cases in test data with changes in predicted rupture status based on the two different models B with largest differences in predicted probabilities (left column) as well as low vs. high inflows for a fixed model B (right column)

Threshold Number of cases w/changed prediction (%)
Different models (flow fixed) Different flows (model fixed)
0.3 28 (11.24) 12 (4.82)
0.5 25 (10.04) 10 (4.02)
0.7 14 (5.62) 2 (0.8)