Page 23 - AIH-2-2

Artificial Intelligence in Health Machine learning in arthroplasty

Table 2. A summary of key studies on the use of artificial intelligence in total knee arthroplasty.
Author Study Performance (AUROC, Input Function Model type
population/ accuracy, sensitivity/
Training model specificity, etc.)
size
Imaging
Karnuta et al. 40 4,724/3568 AUROC 0.989, Accuracy Training epochs Identify nine different Deep learning system
97.4%, Sensitivity 89.2%, implant models
Specific 99.0%
Kunze et al. 42 11,777 Femoral 42.2%, 88.3%, 97.6% Demographics Identify implant size exactly, SVM
Tibial 43.8%, 90.0%, 97.7% (particularly height ±1, ±2 sizes
and sex)
Kim et al. 47 100 Model 1: 87.5% accuracy, 90% Radiographs Detect loosening of TKA DCNN
sensitivity 100% specific implants
Model 2: 97.5% accuracy,
100% sensitivity 97.5% specific
PROMS
Farooq et al. 52 1,325 AUROC 0.81 Demographic, medical, Predict patient satisfaction TreeNet
surgical variables
Kunze et al. 53 430 C-statistic: 0.77, calibration Demographic, medical, Predict patient dissatisfaction RF algorithm
intercept: 0.087, calibration PROMs
slope: 0.74, Brier score: 0.082
Huber et al. 28 130,945 AUROC VAS 0.86 Demographic, Predict VAS pain and Q score Extreme gradient
AUROC Q score 0.70 Subjective disease improvement boosting
burden
Fontana et al. 25 6,480 SF-36 PCS: 0.78 Demographic, medical Predict patients who will LASSO, RF, and
SF-36 MCS: 0.88 achieve minimum clinically linear SVM
KOOS Jr: 0.75 important difference (MCID)
at two years post-operation
Harris et al. 26 587 KOOS ADL: 0.76, Pain: Demographic, medical, Predict KOOS and KOOS Logistic regression,
0.72, Symptoms: 0.72, QoL pre-operative scores subscores LASSO, QDA, GBM
C-statistics: 0.71
Klemt et al. 7 2,389 AUROC: >88 Demographic, surgical Predict PROMs from patient ANN, SVM, RF, and
variables, outcome characteristics elastic-net penalized
scores logistic regression
Cost/LOS
Navarro et al. 55 141,446 LOS AUROC: 0.78 Demographic, medical Forecasting length of stay and Bayesian model
Payment AUROC: 0.74 patient-specific payment
Ramkumar 78,335 LOS AUROC: 0.820 Demographic, medical LOS, discharge disposition, ANN
et al. 56 Payment AUROC: 0.834 inpatient charges
Discharge disposition
AUROC: 0.794
Complications
Harris et al. 34 107,792 Renal complication AUROC: Demographic, medical Predict death and heart LASSO
0.78 complications
Cardiac complication AUROC:
0.73
Death AUROC: 0.73
Mohammadi 7,174 AUROC 0.82 Raw EHR Unplanned 30-day Neural network
et al. 54 documentation readmission
Ko et al. 57 5,757 AUROC: 0.78 Demographic, medical, Acute kidney injury GBM
labs
Jo et al. 58 1,686 AUROC: 0.84 Demographic, medical, Blood transfusion GBM
labs
(Cont'd...)

Volume 2 Issue 2 (2025) 17 doi: 10.36922/aih.3278