Table 2.
Time to converge during training. The reported time is in hours and in parenthesis is the number of epochs that each model was trained for before application for inference.
| DeepMedic | 3D-U-Net | 3D-Res-U-Net | 2D-ResInc | FCN | |
|---|---|---|---|---|---|
| T1-T1 | 18 (35) | 6 (25) | 6 (25) | 150.5 (96) | 3 (25) |
| T2-T2 | 26 (35) | 6 (25) | 6 (25) | 71.5 (38) | 3 (25) |
| T1Gd-T1Gd | 24 (35) | 6 (25) | 6 (25) | 151.7 (100) | 3 (25) |
| Flair-Flair | 18 (35) | 6 (25) | 6 (25) | 147.7 (94) | 3 (25) |
| Multi-2 | 36 (35) | 8 (25) | 8 (25) | 157.3 (91) | 4 (25) |
| Multi-4 | 45 (35) | 8 (25) | 8 (25) | 113.6 (56) | 4 (25) |
| M-A | 62 (45) | 25 (25) | 25 (25) | 321.8 (54) | 17 (25) |