Table 2. Mean absolute distance (MAD) and rooted mean square error (RMSE) evaluated on 20 manually delineated scans for four of the tested methods. Smaller values are better and the best value for both MAD and RMSE are shown in bold font.
| MAD () |
RMSE () |
|||||||
|---|---|---|---|---|---|---|---|---|
| Boundary | AURA | SR-Net-T | ReLayNet | SR-Net | AURA | SR-Net-T | ReLayNet | SR-Net |
| ILM | 2.37 (0.36) | 2.96 (0.78) | 3.17 (0.61) | 2.38 (0.36) | 3.08 (0.51) | 4.22 (2.50) | 4.11 (1.06) | 3.20 (0.99) |
| RNFL-GCL | 3.09 (0.64) | 3.55 (0.55) | 3.75 (0.84) | 3.10 (0.55) | 4.58 (1.03) | 5.52 (2.07) | 5.65 (1.57) | 4.64 (1.19) |
| IPL-INL | 3.43 (0.53) | 3.71 (0.82) | 3.42 (0.45) | 2.89 (0.42) | 4.57 (0.79) | 5.08 (1.47) | 4.59 (0.75) | 3.91 (0.57) |
| INL-OPL | 3.25 (0.48) | 3.80 (0.69) | 3.65 (0.34) | 3.15 (0.56) | 4.29 (0.58) | 4.94 (0.96) | 4.75 (0.56) | 4.10 (0.71) |
| OPL-ONL | 2.96 (0.55) | 3.87 (0.78) | 3.28 (0.63) | 2.76 (0.59) | 4.36 (1.05) | 5.11 (1.06) | 4.57 (1.24) | 3.99 (1.12) |
| ELM | 2.69 (0.44) | 3.72 (1.26) | 3.04 (0.43) | 2.65 (0.66) | 3.38 (0.54) | 4.52 (1.34) | 3.78 (0.52) | 3.38 (0.81) |
| IS-OS | 2.07 (0.81) | 3.57 (1.28) | 2.73 (0.45) | 2.10 (0.75) | 2.62 (0.91) | 4.23 (1.27) | 3.31 (0.46) | 2.73 (0.89) |
| OS-RPE | 3.77 (0.94) | 4.72 (1.12) | 4.22 (1.48) | 3.81 (1.17) | 4.94 (1.14) | 5.80 (1.25) | 5.31 (1.66) | 4.89 (1.42) |
| BM | 2.89 (2.18) | 5.44 (2.98) | 3.09 (1.35) | 3.71 (2.27) | 3.61 (2.26) | 6.21 (2.96) | 3.87 (1.47) | 4.41 (2.34) |
| Overall | 2.95 (1.04) | 3.93 (1.47) | 3.37 (0.92) | 2.95 (1.10) | 3.94 (1.31) | 5.07 (1.86) | 4.44 (1.31) | 3.92 (1.37) |