Author | AI Technique | DCNN | AUC | Accuracy | Sensitivity/Recall | Precision/PPV | Specificity |
---|---|---|---|---|---|---|---|
Belete et al., 2021 [22] | Hyperparameter, Manual Segmentation Pre-Processing, Data Augmentation | ResNet-18 | 1 | 100% | NR | NR | NR |
Bonnin et al., 2023 [23] | Exam Quality Control CNN Deep Learning | X-TKA | NR | 99.9% | 99.8% | 100% | 100% |
Ghose et al. 2020 [26] | Histogram Equalization Data Augmentation, Albumentations Deep Learning DCNN | MobileNetV2 | NR | 96.7% | NR | NR | NR |
DCNN | InceptionV3 | 0.992 | 98.9% | 94.6% | 94.6% | 99.4% | |
Klemt et al., 2022 [30] | CNN Preprocessing Hyp,erparameter Optimization, Class Activation Heat Maps | InceptionV3 | NR | Primary TKA: 97.4% Revision TKA: 96.3% | Primary TKA: 94.9% Revision TKA: 94.5% | NR | Primary TKA: 97.8% Revision TKA: 98.1% |
Patel et al., 2021 [31] | DCNN, Hyperparameter Optimization, Image Segmentation/Data Augmentation Ensembled Networks | EfficientNet & U-Net | NR | 98.9% Human: 76.1% | 98.9% | 99% | NR |
Sharma et al., 2021 [34] | BRISQUE Data Augmentation Fine-Tuning in Transfer Learning DCNN | ResNet-50v2, VGG16, MobileNetV2, DenseNet-201 | 0.9857 | 96.4% | 97.20% | NR | NR |
Tiwari et al., 2022 [20] | Transfer Machine Learning Models | ResNet-50, MobileNet, Efficient Net B7, InceptionV3, Nasnet, VGG16, Xception, Human | NR | ResNet-50-51.4% MobileNet -99.6% Efficient Net B7 -22.2% InceptionV3-96.2% Nasnet-94.6% VGG16-99.0% Xception-93.1% Human-78.2% | ResNet-50-42.0% MobileNet-99.6% Efficient Net B7-22.2% InceptionV3-96.2% Nasnet-94.6% VGG16-99.0% Xception-93.1% Human-50.0% | ResNet-50-62.0% MobileNet-99.6% Efficient Net B7-22.2% InceptionV3-96.2% Nasnet-94.6% VGG16-99.0% Xception-93.4% Human-80.1% | NR |
Yi et al., 2020 [35] | Data Augmentation DCNN | ResNet-18 | 1 | 100% | 100% | 100% | 100% |
Median (IQR) | NA | NA | 0.996 (0.990–1) | 98.9% (96.9%–99.8%) | 98.1% (94.8%–99.7%) | 99.6% (99.0%–100%) | 99.4% (98.1%–100%) |