Page 19 - AIH-2-2

Artificial Intelligence in Health Machine learning in arthroplasty

Table 1. A summary of key studies on the use of artificial intelligence in total hip arthroplasties.
Author Study population/ Performance (AUROC, Input Function Model type
Training model size accuracy, sensitivity/
specificity, etc.)
Imaging
Karnuta et al. 10 2,954 AUROC: 0.991 Radiographic imaging Differentiate between DCNN InceptionV3
Accuracy: 97.9% eight implant types
Sensitivity: 88.6%
Specificity: 98.9%
Karnuta et al. 11 1,972 AUROC: 0.999 Radiographic imaging Discriminate between DCNN InceptionV3
Accuracy: 0.996 18 implant models
Sensitivity: 0.943
Specificity: 0.998
Borjali et al. 9 198 100% accuracy Radiographic imaging Differentiate between DCNN
three implant types
Urakawa et al. 15 1,173 Accuracy: 95.5% Radiographic imaging Identify fracture CNN
Sensitivity: 93.9%
Specificity: 97.4%
Cheng et al. 16 25,505 AUROC: 0.98 Radiographic imaging Identify fracture and DCNN
lesion
Krogue et al. 17 3,026 Accuracy: 93.7% Radiographic imaging Identify fracture DCNN
Sensitivity: 93.2%
Specificity: 94.2%
Sato et al. 18 10,484 AUROC: 0.99 Radiographic imaging Identify fractures and DCNN
Accuracy: 93.7% assist physicians in
Sensitivity: 93.2% diagnostic
Specificity: 94.2%
Rouzrokh 600 Mean differences of Radiographic imaging Determine acetabular CNN
et al. 19 1.35° for inclination and inclination and version
1.39° for anteversion
Rouzrokh 500 The mean absolute error Radiographic imaging Determine femoral CNN
et al. 20 of 0.6 mm component subsidence
Archer et al. 21 256 Good to excellent Radiographic Diagnose hip dysplasia Deep learning-based
compared to the human measurements software (vendor: HIPPO)
expert
Xue et al. 22 420 Sensitivity: 95.0%, Radiographic imaging Identify hip VGG-16 Layer CNN
Specificity: 90.7%, osteoarthritis
Accuracy: 92.8%
Üreten et al. 23 434 Accuracy: 90.2%, Radiographic imaging Identify hip VGG-16 Layer CNN
Sensitivity: 97.6%, osteoarthritis
Specificity: 83.0%
Nguyen et al. 24 510 Correlation of 0.8075 Radiographic imaging Determine bone CNN
(p<0.0001) to DXA mineral density
BMD values
PROMS
Fontana et al. 25 7,085 SF-36 PCS: 0.78 Demographic, medical, Ability to predict LASSO, RF, linear SVM
SF-36 MCS: 0.88 and outcome scores PROMs from patient
HOOS Jr: 0.77 characteristics
Kunze et al. 27 818 AUROC: 0.97 Demographic, medical, Ability to predict RF, stochastic gradient
and outcome scores PROMs from patient boosting, SVM,
characteristics neural network, elastic
net-penalized logistic
regression
(Cont'd...)

Volume 2 Issue 2 (2025) 13 doi: 10.36922/aih.3278