Table 4.

Comparison of Various Methods with the Best Performing Semantic Segmentation Method

Method	ILM MAE (SD) [px]	RPE MAE (SD) [px]
FS-GS [Semantic ON 4 scSE] (ensemble)	0.23 (0.23)	1.12 (1.41)
Patch-based 64 × 32 (ensemble)	0.26 (0.17)	1.15 (1.13)
OCTExplorer^ [54]	1.32 (1.41)	4.95 (5.33)
AURA* [20]	1.26 (0.92)	7.17 (6.87)
Spectralis (automatic)	1.07 (1.63)	6.50 (3.27)

^*Some volume segmentations failed (18% of all testing volumes) and could not be included in error calculations.

^{^}Sections of some scans (1.7% of compared columns) returned undefined segmentations. px, pixels; scSE, squeeze + excitation blocks were incorporated.