Table 7.

Dice score and Mean Absolute Error results of 4 class pixel segmentation with various distance maps for models with $3\times 3$ px kernel size and 32 initial features after removing anomalous data samples and expanding data set with 4 augmentation techniques.

Network	Distance	Dice Score				MAE [px]			TII [%]
	Map	Vitreous	Preretinal Space	Retina	Region b. Retina	PCV	ILM	RPE
UNet	—	0.9968	0.9591	0.9944	0.9984	1.33	0.50	0.57	15.1
	BasicOrient	0.9976	0.9606	0.9944	0.9987	1.42	0.47	0.56	4.8
	CumSum	0.9977	0.9609	0.9945	0.9987	1.25	0.47	0.55	6.8
	2NetR	0.9976	0.9601	0.9943	0.9986	1.48	0.48	0.59	9.4
	2NetPR	0.9975	0.9588	0.9943	0.9987	1.20	0.50	0.57	3.7
LFUNet	—	0.9973	0.9590	0.9942	0.9985	1.50	0.50	0.57	11.5
	BasicOrient	0.9975	0.9583	0.9942	0.9987	1.58	0.49	0.58	8.8
	CumSum	0.9977	0.9605	0.9944	0.9987	1.22	0.47	0.56	4.8
	2NetR	0.9974	0.9581	0.9944	0.9987	1.64	0.48	0.56	6.3
	2NetPR	0.9973	0.9571	0.9943	0.9987	1.26	0.51	0.56	5.0
AttUNet	—	0.9850	0.8802	0.9926	0.9956	4.02	0.60	0.80	50.5
	BasicOrient	0.9824	0.8551	0.9880	0.9962	7.31	0.87	2.05	52.9
	CumSum	0.9886	0.9025	0.9926	0.9981	4.50	0.59	0.79	38.7
	2NetR	0.9860	0.8675	0.9920	0.9980	4.74	0.72	0.98	44.1
	2NetPR	0.9963	0.9508	0.9926	0.9981	2.04	0.65	0.86	12.8
ReLayNet	—	0.9603	0.8015	0.9918	0.9859	12.63	0.71	0.96	88.3
	BasicOrient	0.9614	0.7528	0.9771	0.9930	20.00	0.59	0.79	92.0
	CumSum	0.9730	0.7937	0.9906	0.9976	15.06	0.79	1.00	79.5
	2NetR	0.9651	0.7594	0.9905	0.9974	18.55	0.70	1.11	88.1
	2NetPR	0.9968	0.9544	0.9927	0.9981	1.42	0.58	0.81	10.8