Table 4 Comparing performance using only SoftBCEWithLogitsLoss with a fixed model architecture.

From: Multi-view hybrid encoder U-Net for 3D renal vascular medical image segmentation

Encoder Arch.

Lower encoder

Loss

FPR

Recall

DSC

MSD

CNNs + CNNs

Ese Vovnet39b32

0.0120

0.342

0.978

0.843

0.773

ConvNext Small34

0.0113

0.146

0.954

0.909

0.822

DM Nfnet F035

0.0128

0.104

0.956

0.928

0.834

CNNs + ViTNets

Sam2 Hiera Small38

0.0150

0.144

0.963

0.914

0.820

Swin Tiny39

0.0111

0.141

0.955

0.911

0.827

 

Proposed (CoaT Lite Small)41

0.0100

0.131

0.967

0.922

0.836

  1. Significant values are in bold.
Back to article page