Table 5 Performance of the deep learning model for the presence of a meniscal tear
Model | Acc | Pre | Rec | Sen | Spe | AUC (95% CI) | Time (sec) | |
|---|---|---|---|---|---|---|---|---|
Medial tear | MobileNet | 64.11% | 62% | 54.87% | 54.87% | 71.85% | 0.675 (0.608–0.742) | 6.02 |
Ours | 85.08% | 83.93% | 83.19% | 83.19% | 86.67% | 0.889 (0.845–0.933) | 3.64 | |
Lateral tear | MobileNet | 64.32% | 40% | 64% | 64% | 64.44% | 0.674 (0.592–0.756) | 4.48 |
Ours | 80.54% | 62.96% | 68% | 68% | 85.19% | 0.817 (0.744–0.889) | 2.77 | |
Medial and lateral tear | MobileNet | 75.17% | 20.51% | 57.14% | 57.14% | 77.04% | 0.651 (0.476–0.825) | 3.15 |
Ours | 91.95% | 55% | 78.57% | 78.57% | 93.33% | 0.924 (0.863–0.985) | 1.88 |