Aggregation-and-Attention Network for brain tumor segmentation

BMC Medical Imaging

Table 2 The ablation experiments of basic modules

EDS	UPL	DAF	MSC	L_down	L_up	Dice↑			Precision↑			Sensitivity↑			Hausdorff↓
EDS	UPL	DAF	MSC	L_down	L_up	WT	CT	ET	WT	CT	ET	WT	CT	ET	WT	CT	ET
Baseline						0.771	0.839	0.698	0.781	0.910	0.677	0.921	0.880	0.887	1.904	1.293	2.117
+						0.800	0.883	0.741	0.814	0.937	0.743	0.946	0.913	0.926	1.488	0.981	1.650
+	+					0.808	0.880	0.751	0.828	0.935	0.760	0.943	0.913	0.922	1.423	0.986	1.569
+	+	+				0.818	0.874	0.758	0.836	0.932	0.760	0.944	0.902	0.926	1.401	1.033	1.574
+	+	+	+			0.829	0.868	0.770	0.841	0.924	0.770	0.946	0.904	0.925	1.397	1.064	1.560
+	+	+	+	+		0.849	0.891	0.849	0.860	0.945	0.792	0.952	0.913	0.934	1.360	0.929	1.507
+	+	+	+	+	+	0.869	0.896	0.814	0.886	0.952	0.818	0.950	0.912	0.934	1.308	0.934	1.456

The best results are marked with bold
^**↑ indicates that the greater the index value, the better the network segmentation performance.↓ indicates that the smaller the index value, the better
The network segmentation performance. The L_dual in the DAF module is considered as an inherent attribute

ISSN: 1471-2342