From: Artificial intelligence in knee arthroplasty: current concept of the available clinical applications
Authors | Patients | Year | Type | Time | Assessment | Factors | Conclusion |
---|---|---|---|---|---|---|---|
Patient Decision Aid | |||||||
 Ramkumar et al. [16] | 175,042 | 2019 | Predictive of perioperative parameters (ANN) | Preop | Predict LOS, inpatient discharge Propose a risk-based plan for complex cases | Preop variables | Model can predict perioperative management |
 Bansback et al. [17] | 280 | 2019 | Patient Decision | Preop | Decision quality | PROMS, demographics | Predictive model with decision aid |
 Jayakumar et al. [18] | 150 | 2020 | Patient Decision | Preop | Decision quality, patient outcomes | PROMS, demographics | Presentation of RCT |
 Shah et al. [15] | 697 | 2020 | Patient decision | Preop | TKA loosening | Preop radiographs | Detection of implants loosening THA > TKA (Se 70%, Sp 96%) |
 Jayakumar et al. [19] | 129 | 2021 | Patient Decision | Preop | Decision quality, patient experience, functional outcomes | Education, preference assessment, PROMs | Better decision quality, satisfaction, improved PROMs |
 Yi et al. [14] | 237–274 | 2019 | TKA identification | Preop | Difference of TKA, UKA | Radiographs | Identification of TKA on X-ray and distinguish 2 models of TKA |
 Karnuta et al. [20] | 424 | 2020 | TKA identification | Preop | TKA models | Radiographs | Valid |
 Schwartz et al. [21] | 326 | 2020 | OA classification | Preop | OA stage | Preop radiographs | Convolutional neural network (CNN) and classify knee OA |
Surgical training | |||||||
 Aim et al. [22] | 330 | 2016 | VR training in arthroscopy | Preop | Review |  | Few assessments of VR training but promising |
 Goh et al. [23] |  | 2021 | VR and AR training in knee arthroplasty | Preop | Review |  | Few assessments of VR training but promising |
Preoperative planning | |||||||
 Wallace et al. [24] | 382 | 2020 | PM Implant Size | Preop | Component size prediction | Sex, height, weight, age, and ethnicity | More accurate than radiographic templating |
 Kunze et al. [25] | 17,283 | 2021 | PM Implant size | Preop | Component size prediction | Demographic variables (age, height, weight, BMI, sex) | Good to excellent performance for predicting TKA component Size. Main factor: sex Free app: https://orthopedics.shinyapps.io/TKASizing_Calculator/ |
 Li et al. [26] | 200 | 2021 | 3D reconstruction | Preop | AI-based 3D model construction | CT scan | As accurate as operator reconstruction. Faster than operator construction |
Surgery | |||||||
 Tsukada et al. [27] | 10 | 2019 | Augmented reality in surgery | Intraop | Tibial bone resection with AR | AR-KNEE system | Insufficient accuracy of bone cuts |
 Pokhrel et al. [28] | 15 | 2019 | Augmented reality in surgery | Intraop | Accuracy of bone cuts | Augmented reality system | Reliable accuracy |
 Verstraete et al. [29] | 479 | 2020 | ML PM | Intraop | Intraop planning (load) | Intraop alignment – tibiofemoral load | Validated ML algorithm |
Remote patient monitoring | |||||||
 Chiang et al. [30] | 18 | 2017 | Patient Monitoring | Postop | APDM sensors | Postop ROM | Continuous monitoring of ROM progress after TKA |
 Kang et al. [31] | 60 | 2018 | Patient Monitoring | Postop | Rehabilitation training instrument NEO-GAIT | VAS, ROM, HSS | NEO-GAIT plays more active and effective role in promoting rehabilitation after TKA |
 Ramkumar et al. [32] | 25 | 2019 | Remote Patient Monitoring | Postop | Feasibility – ROM – PROMs – exercise compliance | RPM mobile application | Pilot study — acquisition of continuous data |
 Mehta et al. [33] | 242 | 2020 | Remote Patient Monitoring | Postop | rate of discharge to home and clinical outcomes after hip or knee arthroplasty. | RPM mobile application | No significant difference in the rate of discharge to home. Significant reduction in rehospitalization rate with RPM |
 Bovonratwet et al. [34] | 319 | 2020 | NLP | Postop | Satisfaction | Patient narratives | Not efficient |
 Sagheb et al. [35] | 20,000 | 2020 | NLP | Postop | Identify data in OR report | OR report | NLP algorithms efficient |