2.5D MFFAU-Net: a convolutional neural network for kidney segmentation

Table 1 Segmentation results in KiTS19

Algorithm or method	Input	Training time	Kidney Dice	Tumor Dice
2.5D U-Net	3 × 512 × 512	76 h	0.950	0.779
2.5D U-Net	5 × 512 × 512	87 h	0.950	0.788
2.5D ResU-Net	3 × 512 × 512	82 h	0.961	0.856
2.5D ResU-Net	5 × 512 × 512	94 h	0.965	0.855
2.5D DenseU-Net	3 × 512 × 512	83 h	0.963	0.852
2.5D DenseU-Net	5 × 512 × 512	96 h	0.962	0.850
2.5D V-Net	3 × 512 × 512	96 h	0.965	0.850
2.5D V-Net	5 × 512 × 512	103 h	0.970	0.855
2.5D MFFAU-Net (original conv)	3 × 512 × 512	92 h	0.972	0.868
2.5D MFFAU-Net (original conv)	5 × 512 × 512	105 h	0.972	0.865
2.5D MFFAU-Net (ResConv)	3 × 512 × 512	104 h	0.975	0.872
2.5D MFFAU-Net (ResConv)	5 × 512 × 512	133 h	0.970	0.860

ISSN: 1472-6947