Table B.1
Dice index
| Model | Dice index [0.2] | Dice index [0.4] | Dice index [0.6] | Dice index [0.8] |
|---|---|---|---|---|
| UNet[10–2] | 93.19 | 94.34 | 94.07 | 92.06 |
| UNet[10–3] | 93.13 | 94.37 | 94.16 | 92.43 |
| UNet[10–4] | 92.75 | 94.09 | 93.79 | 91.9 |
| UNet[10–5] | 82.4 | 89.6 | 91.3 | 87.66 |
| UNet++[10–3] | 93.46 | 94.61 | 94.28 | 92.36 |
| UNet++[10–4] | 93.04 | 94.07 | 93.6 | 91.57 |
| UNet++[10–5] | 90.24 | 91.88 | 91.02 | 87.67 |
Notes. Dice index evaluation on the test set for the schemes reported in Figure 14 at classification threshold values of 0.8, 0.6, 0.4, and 0.2. Blue (red) refers to the best (less performing) scheme in each column. The bold scores correspond to the absolute best (if in blue) or lowest (if in red) dice index. The closer the dice index to 1, the better. Dice index values in test were aligned with performances in training and validation steps as close performances are obtained for schemes with initial learning–rate values in [10–4, 10–3, 10–2] and lower performance are noted for schemes with a learning–rate value of 10–5. Moreover, UNet++[10–3] and UNet[10–5] result in the best (94.61% at 0.4) and least performing (82.4% at 0.2) schemes, respectively.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.