Skip to main content
. Author manuscript; available in PMC: 2023 Nov 1.
Published in final edited form as: Med Image Anal. 2022 Sep 7;82:102574. doi: 10.1016/j.media.2022.102574

Table 4.

Comparison with state-of-the-art methods using full annotation on the OAI ZIB dataset with five-fold cross validation. Paired t-test values indicate the significance status of the improved performance of our method vs. the ensemble method (Zheng et al., 2020b).

Femoral bone Femoral cartilage Tibial bone Tibial cartilage
DSC (%) ASSD (mm) DSC (%) ASSD (mm) DSC (%) ASSD (mm) DSC (%) ASSD (mm)
Ensemble method (Zheng et al., 2020b) 98.49 ± 0.30 0.182 ± 0.046 89.26 ± 3.08 0.176 ± 0.067 98.60 ± 0.30 0.172 ± 0.049 86.36 ± 3.88 0.196 ± 0.081
UNet++ 3D (Zhou et al., 2018) 98.44 ± 0.31 0.190 ± 0.047 89.19 ± 2.74 0.174 ± 0.052 98.57 ± 0.31 0.222 ± 0.340 84.96 ± 4.55 0.226 ± 0.102
TransUNet 3D (Chen et al., 2021) 98.47 ± 0.29 0.188 ± 0.048 89.25 ± 2.94 0.173 ± 0.060 98.61 ± 0.30 0.171 ± 0.050 85.34 ± 4.24 0.240 ± 0.114
Attention UNet 3D (Oktay et al., 2018) 98.55 ± 0.30 0.174 ± 0.048 89.56 ± 2.64 0.169 ± 0.059 98.70 ± 0.31 0.162 ± 0.067 86.74 ± 4.01 0.196 ± 0.089
Our KCB-Net method 98.62 ± 0.26 0.164 ± 0.039 90.24 ± 2.76 0.153 ± 0.049 98.76 ± 0.30 0.149 ± 0.048 87.19 ± 3.96 0.185 ± 0.085
p-value ≪ 0.001 ≪ 0.001 ≪ 0.001 ≪ 0.001 ≪ 0.001 ≪ 0.001 ≪ 0.001 0.006