Table 4 Performance of artificial intelligence algorithms detecting implant failure in total joint arthroplasty
Author | AI Technique | DCNN | AUC | Accuracy | Sensitivity/Recall | Precision/PPV | Specificity |
|---|---|---|---|---|---|---|---|
Lau et al., 2022Â [37] | Pre-Trained on ImageNet and Tensor Flow | Xception Model | 0.935 | 96.3% | 96.1% | 92.4% | 90.9% |
Rouzrokh et al., 2022 [32] | U-Net Model | Efficient Net B0 | NR | Femur-97.0% Implant-98.0% Magnification Markers: 94.0% | NR | NR | NR |
Median (IQR) | NA | NA | 0.935 (NA) | 97.2% (96.7%–97.6%) | 96.1% (NA) | 92.4% (NA) | 90.9% (NA) |