DCAU-Net: dense convolutional attention U-Net for segmentation of intracranial aneurysm images

Visual Computing for Industry, Biomedicine, and Art

Table 6 Comparison of the segmentation performances of different methods, on the dataset of the Affiliated Hospital of Qingdao University

Model	< 3 mm			3–7 mm			> 7 mm
Model	Dice	Sensitivity	Specificity	Dice	Sensitivity	Specificity	Dice	Sensitivity	Specificity
HeadXNet [29]	0.5680	0.6029	0.9998	0.7207	0.7746	0.9996	0.8383	0.8449	0.9983
DeepMedic [32]	0.4296	0.4759	0.9897	0.5655	0.5541	0.9990	0.7280	0.7028	0.9938
GLIA-Net [33]	0.4015	0.4707	0.9993	0.7028	0.7059	0.9996	0.8253	0.7899	0.9986
DAResUNet [30]	0.5369	0.5412	0.9998	0.6791	0.8265	0.6584	0.8784	0.8929	0.9979
U-Net [36]	0.4286	0.4610	0.9995	0.6393	0.6585	0.9951	0.8191	0.8069	0.9975
Ours	0.6303	0.7103	0.9998	0.7607	0.7338	0.9997	0.8412	0.8267	0.9986