Table 3 Segmentation results of an ablation study of the proposed model on the four validation datasets.

From: Deep 3D attention CLSTM U-Net based automated liver segmentation and volumetry for the liver transplantation in abdominal CT volumes

Dataset

Model

Recall

Precision

DSC

IOU

HD (mm)

Left lobe

3D U-Net

0.838 ± 0.124

0.803 ± 0.081

0.771 ± 0.082

0.783 ± 0.127

6.895 ± 1.794

AU-Net

0.838 ± 0.122

0.808 ± 0.086

0.785 ± 0.081

0.776 ± 0.128

6.691 ± 1.785

AU-Net w/ DS

0.847 ± 0.121

0.812 ± 0.083

0.787 ± 0.082

0.779 ± 0.128

6.630 ± 1.732

DALU-Net

0.851 ± 0.121

0.806 ± 0.083

0.789 ± 0.084

0.787 ± 0.128

6.629 ± 1.696

Right lobe

3D U-Net

0.966 ± 0.069

0.786 ± 0.067

0.855 ± 0.065

0.821 ± 0.138

5.144 ± 1.635

AU-Net

0.966 ± 0.067

0.807 ± 0.072

0.867 ± 0.037

0.856 ± 0.115

5.125 ± 1.714

AU-Net w/ DS

0.972 ± 0.063

0.804 ± 0.071

0.868 ± 0.066

0.827 ± 0.121

5.087 ± 1.731

DALU-Net

0.975 ± 0.058

0.805 ± 0.071

0.869 ± 0.067

0.896 ± 0.141

5.074 ± 1.705

Caudate lobe

3D U-Net

0.825 ± 0.326

0.793 ± 0.316

0.849 ± 0.307

0.819 ± 0.316

2.133 ± 1.476

AU-Net

0.905 ± 0.241

0.817 ± 0.209

0.922 ± 0.236

0.896 ± 0.236

1.557 ± 1.273

AU-Net w/ DS

0.914 ± 0.229

0.822 ± 0.217

0.929 ± 0.215

0.907 ± 0.222

1.583 ± 1.221

DALU-Net

0.943 ± 0.183

0.877 ± 0.174

0.955 ± 0.167

0.936 ± 0.181

1.238 ± 1.131

Whole liver

3D U-Net

0.899 ± 0.155

0.850 ± 0.287

0.822 ± 0.295

0.770 ± 0.288

3.236 ± 1.256

AU-Net

0.891 ± 0.162

0.893 ± 0.241

0.852 ± 0.258

0.800 ± 0.259

3.280 ± 0.982

AU-Net w/ DS

0.922 ± 0.110

0.907 ± 0.213

0.887 ± 0.218

0.840 ± 0.219

2.980 ± 0.784

DALU-Net

0.923 ± 0.120

0.924 ± 0.188

0.899 ± 0.201

0.855 ± 0.205

2.762 ± 0.728