Table 2 Performance comparison between the proposed method and other methods on the MOST dataset.
Methods | Top1 | |||||||
|---|---|---|---|---|---|---|---|---|
knee OA/non-knee OA | KL | FL | FM | TL | TM | JSN-L | JSN-M | |
Antony et al., 201719 | – | 49.48% | – | – | – | – | – | – |
Tiulpin et al., 201812 | – | 68.73% | – | – | – | – | – | – |
Chen et al., 201939 | – | 75.56% | – | – | – | – | – | – |
Mikhaylichenko et al., 202141 | – | 73.43% | – | – | – | – | – | – |
SE-ResNet-5014 | – | 74.39% | 79.32% | 78.33% | 76.87% | 71.19% | 93.20% | 84.00% |
SE-ResNext50-32x4d14 | – | 75.22% | 78.98% | 78.72% | 78.04% | 73.11% | 93.43% | 84.16% |
Ensemble14 | – | 76.13% | 80.31% | 79.23% | 78.42% | 73.56% | 93.69% | 84.83% |
Ours | 92.58% | 76.94% | 82.05% | 80.54% | 80.12% | 76.32% | 93.54% | 83.77% |
Ours (Ens.) | 92.84% | 77.86% | 80.88% | 79.91% | 79.52% | 76.71% | 93.31% | 84.19% |
Methods | Top±1 | |||||||
|---|---|---|---|---|---|---|---|---|
– | KL | FL | FM | TL | TM | JSN-L | JSN-M | |
Antony et al., 201719 | – | 73.06% | – | – | – | – | – | – |
Tiulpin et al., 201812 | – | 90.28% | – | – | – | – | – | – |
Chen et al., 201939 | – | 98.09% | – | – | – | – | – | – |
Mikhaylichenko et al., 202141 | – | 95.67% | – | – | – | – | – | – |
SE-ResNet-5014 | – | 98.26% | 93.72% | 92.16% | 94.55% | 97.73% | 98.17% | 98.18% |
SE-ResNext50-32x4d14 | – | 98.20% | 95.16% | 93.28% | 96.45% | 97.84% | 98.26% | 98.26% |
Ensemble14 | – | 98.43% | 94.64% | 93.20% | 95.80% | 97.97% | 98.30% | 98.28% |
Ours | – | 98.59% | 95.55% | 94.06% | 96.87% | 98.15% | 98.33% | 98.21% |
Ours (Ens.) | – | 98.12% | 94.86% | 93.72% | 96.59% | 98.10% | 98.20% | 98.31% |
