Skip to main content

Artificial intelligence in diagnosis of knee osteoarthritis and prediction of arthroplasty outcomes: a review

Abstract

Background

Artificial intelligence is an emerging technology with rapid growth and increasing applications in orthopaedics. This study aimed to summarize the existing evidence and recent developments of artificial intelligence in diagnosing knee osteoarthritis and predicting outcomes of total knee arthroplasty.

Methods

PubMed and EMBASE databases were searched for articles published in peer-reviewed journals between January 1, 2010 and May 31, 2021. The terms included: ‘artificial intelligence’, ‘machine learning’, ‘knee’, ‘osteoarthritis’, and ‘arthroplasty’. We selected studies focusing on the use of AI in diagnosis of knee osteoarthritis, prediction of the need for total knee arthroplasty, and prediction of outcomes of total knee arthroplasty. Non-English language articles and articles with no English translation were excluded. A reviewer screened the articles for the relevance to the research questions and strength of evidence.

Results

Machine learning models demonstrated promising results for automatic grading of knee radiographs and predicting the need for total knee arthroplasty. The artificial intelligence algorithms could predict postoperative outcomes regarding patient-reported outcome measures, patient satisfaction and short-term complications. Important weaknesses of current artificial intelligence algorithms included the lack of external validation, the limitations of inherent biases in clinical data, the requirement of large datasets in training, and significant research gaps in the literature.

Conclusions

Artificial intelligence offers a promising solution to improve detection and management of knee osteoarthritis. Further research to overcome the weaknesses of machine learning models may enhance reliability and allow for future use in routine healthcare settings.

Background

Total knee arthroplasty (TKA) is the only definitive surgical treatment for advanced knee osteoarthritis (KOA) [1,2,3]. Artificial intelligence (AI) and machine learning (ML) modeling is a new decision aid tool used in KOA diagnosis, patient selection, pre-TKA planning, prediction of disease progression, and estimation of treatment outcomes. The tool is improving with technological advancements and larger datasets but also requires extensive validation.

AI is a broad term referring to technologies that simulate human intelligence to automate tasks with high accuracy and precision. There are different methods to achieve this goal, such as designing algorithms with explicit rules and instructions or employing more “intelligent” algorithms such as those developed through machine learning. ML is a branch of AI involving algorithms that automatically “learn” from data, with incremental optimization and improvements in accuracy during the training process [2, 4]. Deep learning is a form of ML that does not require a labelled or structured dataset [4, 5]. For example, the use of artificial neural networks (utilizing the layers of increasing complexity and abstraction for information processing) to “learn” the important features of a model without human input [4].

AI can handle very large, complex datasets, and generate predictions to improve accuracy and efficiency of healthcare decisions, such as KOA and TKA [1]. ML algorithms have also been used to develop models to assist with pre-TKA planning and predict the value metrics of TKA, such as predicting implant size [6], reconstructing three-dimensional CT data of lower limb to facilitate robotic-assisted TKA [7], and assisting with component positioning and alignment [8]. ML potentially improves surgical precision and reduce the cost of manual labor. Regarding value metrics, ML methods have been used to predict the length of hospital stay, hospitalization charges, and discharge disposition. It impacts the economic burden of TKA and thus potentially affects decisions on payment models in healthcare settings [9,10,11].

This review aimed to summarize the existing evidence and highlight recent developments of AI and ML in diagnosis of KOA, prediction of the need for and outcomes of TKA.

Materials and methods

We searched PubMed and EMBASE databases for articles published in peer-reviewed journals between January 1, 2010 and May 31, 2021. We searched for the following terms: ‘AI’, ‘machine learning’, ‘knee’, ‘osteoarthritis’, and ‘arthroplasty’. We selected studies focusing on the use of AI in diagnosis of KOA, predicting the need for TKA, and predicting outcomes of TKA. We excluded non-English language articles and the articles with no English translation. A reviewer screened the articles for the relevance to the research questions and strength of evidence.

Results

The search produced 136 individual results, among which a total of 22 papers were included in the narrative synthesis following screening against inclusion/exclusion criteria (Table 1). Only one study was externally validated by testing the model using a dataset not used during model training to assess model performance and generalizability. The most commonly reported metric among the published articles was the area under the receiver operating characteristic curve (AUC), which evaluates the ability of an algorithm in discriminating between the individuals who experienced and those who did not experience the outcomes immediately after surgery and thereafter. AUC values ranged from 0.5 (indicating performance equal to a random predictor) to 1 (indicating a perfect predictor). Other reported metrics included sensitivity, specificity, Kappa coefficient (a measure of inter-rater reliability, where a value of 0 indicates no agreement while a value of 1 indicates perfect agreement), and positive and negative predictive values. The characteristics, performance, strengths, and weaknesses of AI algorithms are summarized in Table 2. AI algorithms used to predict the outcomes of TKA are shown in Table 3.

Table 1 Studies included in the scoping review
Table 2  A summary of reviewed studies on knee osteoarthritis diagnosis and knee arthroplasty prediction
Table 3  A summary of reviewed studies on predicting postoperative outcomes of total knee arthroplasty

Diagnosis and predicting the need for TKA

Multiple machine learning models have been developed for radiological diagnosis and severity grading of KOA (based on the most widely used the Kellgren-Lawrence Classification System) (Table 2). Tiulpin et al. [19] developed an automatic grading model based on the Deep Siamese Convolutional Neural Network. The model was first trained using 18,376 knee radiographs from the Multicenter Osteoarthritis Study (a longitudinal, prospective, observational study of KOA in older Americans), and further tuned for hyperparameters using 2,957 KOA radiographs from the Osteoarthritis Initiative (a multicenter, longitudinal, prospective observational study of knee osteoarthritis), and finally tested on 5,960 randomly selected KOA radiographs from the Osteoarthritis Initiative that are unseen during the training process. The model achieved a kappa coefficient of 0.83 and an average multiclass accuracy of 67%, indicating excellent agreement (comparable to intra- and inter-rater reliability by arthroplasty surgeons) [34, 35]. The key benefit of this model is the provision of probability distributions for each Kellgren-Lawrence grade prediction. In clinical practice, the model may be used to select the closest Kellgren-Lawrence grade in ambiguous cases. Similarly, Norman et al. [18] used DenseNet neural network architectures to develop an automatic Kellgren-Lawrence grading model. Saliency maps revealed important radiographic features in algorithm’s decision-making, such as osteophytes and joint space narrowing. For detecting Kellgren-Lawrence grades, the sensitivity and specificity of the model were 69–86% and 84–99%, respectively. The kappa coefficient was 0.83, which was the same as the model proposed by Tiulpin et al. [19]. Most existing algorithms focus on the radiographic diagnosis of KOA or rely heavily on radiographic information as candidate predictors of TKA. This may be due to substantially increased imaging data availability following the recent creation of public datasets such as the Osteoarthritis Initiative.

In a recent study, Leung et al. [15] developed a deep learning model that directly predicted the need for TKA based on knee radiographs. This model demonstrated superior performance in predicting TKA than the conventional binary outcome models based on the Kellgren-Lawrence or Osteoarthritis Research Society International grades. The deep learning model used additional image-based information that might not be captured by simple numerical grading systems [36].

The discrepancies between radiologic and clinical severity of KOA have been widely reported [37,38,39,40]. Clinical diagnosis is typically made according to American College of Rheumatology criteria, taking into account patient age, symptoms, physical examination, and radiographic assessments [41]. The decision for surgery is driven primarily by symptom severity instead of radiological findings. Thus, the ML algorithms (automate Kellgren-Lawrence grading or predict TKA using imaging data alone) are limited in clinical decision-making. Nevertheless, the ML-based studies mentioned above offer insight to the development of radiograph-based prediction models using different machine learning approaches and may serve as a stepping stone to future studies that include additional clinical parameters, which may be more suitable for clinical decision-making support.

In 2020, Heisinger et al. [13] first designed an ML prediction model by investigating knee symptomatology (e.g., pain, function, and quality of life), Kellgren-Lawrence grading, and socioeconomic and demographic factors four years before TKA. The longitudinal analyses showed that significant worsening in knee symptomatology before TKA was the most important factor in decision making for TKA, compared to the radiographic progression of KOA. The artificial neural network can predict patients who may undergo TKA in the next two years with an accuracy of 80%, with a positive predictive value of 84%, and a negative predictive value of 73%.

El-Galaly et al. [12] were the first to attempt to develop a clinical ML algorithm to predict early revision TKA using preoperative data. The models were trained on the Danish Knee Arthroplasty Registry. Patient age, post-fracture osteoarthritis, and weight were statistically significant preoperative factors. Nevertheless, the authors were unable to develop a clinically useful model based on preoperative information [12]. Hence, further study is needed to identify clinically useful predictors of revision TKA.

Predicting postoperative outcomes of TKA

The improvement following TKA is commonly assessed using the patient-reported outcome measures with or without accompanying “minimally clinically important improvement”, i.e., the minimum benefit assessed with the patient-reported outcome measures [42, 43]. Huber et al. [28] used ML algorithms to predict postoperative improvement in the patient-reported outcome measures. The models were trained and tested using the National Health Service data (130,945 observations), and the area under the receiver operating characteristic curve of the best performing models was approximately 0.86 (visual analogue scale) and 0.70 (Q score, i.e., sum of the Oxford Hip Score and Oxford Knee Score) for TKA. The results showed that preoperative visual analogue scale, Q score, and specific Q score dimensions were the most important predictors of postoperative patient-reported outcome measures [28]. Harris et al. [20] developed another model to predict post-TKA 1-year achievement of MCID and demonstrated fair discriminative ability for the prediction of some, but not all, PROMs included. Further development of similar machine learning algorithms for routine patient care could potentially assist postoperative outcome prediction.

AI can be used to predict post-TKA patient dissatisfaction. Kunze et al. [25] developed a random forest algorithm which demonstrated an AUC of 0.77 in identifying patients most likely to experience dissatisfaction. Farooq et al. [22] found that models built using ML achieved significantly higher AUC than using binary logistic regression on the same dataset (0.81 vs. 0.60). Given that a significant 20% of patients are dissatisfied following TKA and that existing statistical models cannot fully explain the reason for dissatisfaction [22], supervised machine learning models offer an alternative approach to automate the search for predictors of patient dissatisfaction.

The major complications of TKA are bleeding, thromboembolism, vascular injury, etc. [44] Many risk prediction calculators exist, such as the American College of Surgeons-National Surgical Quality Improvement Program universal surgical risk calculator and other arthroplasty-specific calculators [45, 46]. These conventional calculators have substantial weaknesses, such as poor accuracy, limited generalizability to external datasets, and preoperative use restrictions due to requiring intraoperative data as input variables [47, 48]. ML models offer an alternative approach to predict postoperative complications. Harris et al. [27] developed prediction models for 30-day mortality and major complications following elective arthroplasty. The models were trained on the American College of Surgeons National Surgical Quality Improvement data and externally validated using Veterans Affairs Surgical Quality Improvement Program data which had different patient demographics and clinical characteristics compared to the training data. The models showed acceptable performance in predicting mortality (AUC: 0.69) and cardiac complications (AUC: 0.72) (but not renal complications – AUC: 0.60) during external validation using the Veterans Affairs Surgical Quality Improvement Program data [27]. One important limitation of this study design is that the training dataset does not contain complete patient medical data (e.g., comorbidities) and only includes the patients from a small number of hospitals, limiting its generalizability [27]. Overall, ML has not been extensively applied in predicting post-TKA complications, and further efforts in model development with rigorous internal and external validation are warranted.

Discussion

We find AI and ML models improve automatic grading of knee radiographs, patient selection for TKA, and predictin of postoperative outcomes of patient-reported outcome measures, patient satisfaction, and short-term complications. The weaknesses of current AI algorithms include the lack of external validation, inherent biases of clinical data, the need for large datasets for training, and significant research and regulatory gaps.

Weaknesses of AI in arthroplasty

The current use of artificial intelligence algorithms has its limitations. First, accuracy and generalizability are key obstacles as very few models have been externally validated, and high AUC values do not necessarily translate to good clinical performance [26]. More rigorous external validation of prediction models is needed during algorithm development and testing, to ensure robustness and reliability before algorithms can be considered for routine clinical use. An important issue regarding generalizability lies in the fact that patient selection and postoperative outcomes are influenced by structure- and region-related confounders, such as institutional policies, hospital sites, and organizational culture [10]. For example, the threshold for booking TKA may differ between institutions depending on resource availability and hospital policy. Institutions may benefit from using region-specific machine learning algorithms for more accurate predictions.

Second, a practical disadvantage of machine learning models is the requirement of large datasets to train these models. These datasets often contain millions of unique data points and require hours or days of training, and additional datasets are needed to assess generalizability [49]. The increased availability of public datasets such as Multicenter Osteoarthritis Study and OAI could help overcome this obstacle and facilitate further research on machine learning in arthroplasty.

Third, a common concern surrounding the use of artificial intelligence is the “black-box” nature of machine learning models. Machine learning algorithms’ decision-making processes are opaque, using hidden layers and unknown connections between inputs and outputs, resulting in poor understanding and difficult scientific interpretation of how it generates predictions and recommendations [50]. Visualization of attention maps cannot directly provide information on these hidden relationships, and other efforts to increase the transparency of deep learning models are still ongoing [51]. Nevertheless, this poses more of a problem to scientific understanding rather than clinical application. By contrast, the reliance on data for model development is a key limitation of artificial intelligence in clinical use. Models developed are limited by the biases and limitations of current clinical data. Machine learning models are also “plastic”, i.e., changing when presented with new data [50], and the input parameters included in a machine learning algorithm, such as models predicting TKA need, may continuously change as new data becomes available to the model.

Finally, significant research and regulatory gaps exist, given the novel nature of this technology. There is a paucity of literature on the use of machine learning algorithms to predict the need for arthroplasty, and current machine learning models are unable to predict the long-term outcomes of TKA. ML models are limited by the biases of current clinical data, and future implementation of these algorithms into routine hospital care will also come with regulatory concerns of algorithm quality control, security issues and adversarial attacks.

Conclusions

KOA is an important public health problem worldwide. AI offers a promising solution to detect KOA and improve pre-TKA planning. Further research is needed to overcome the limitations of ML models and ensure reliability for future use in routine healthcare settings.

Availability of data and materials

All data generated or analysed during this study are included in this published article.

Abbreviations

ACS-NSQIP:

American College of Surgeons-National Surgical Quality Improvement Program

AUC:

Area under the receiver operating characteristic curve

KL:

Kellgren & Lawrence

LOS:

Length of stay

ML:

Machine learning

KOA:

Knee osteoarthritis

OAI:

Osteoarthritis Initiative

TKA:

Total knee arthroplasty

VASQIP:

Veterans Affairs Surgical Quality Improvement Program

References

  1. Myers TG, Ramkumar PN, Ricciardi BF, Urish KL, Kipper J, Ketonis C. Artificial Intelligence and Orthopaedics: An Introduction for Clinicians. J bone joint Surg Am volume. 2020;102(9):830–40.

    Google Scholar 

  2. Cabitza F, Locoro A, Banfi G. Machine Learning in Orthopedics: A Literature Review. Front Bioeng Biotechnol. 2018;6:75-.

    PubMed  PubMed Central  Google Scholar 

  3. Neogi T. The epidemiology and impact of pain in osteoarthritis. Osteoarthr Cartil. 2013;21(9):1145–53.

    CAS  Google Scholar 

  4. Bini SA. Artificial Intelligence M, Learning. Deep Learning, and Cognitive Computing: What Do These Terms Mean and How Will They Impact Health Care? J Arthroplasty. 2018;33(8):2358–61.

    PubMed  Google Scholar 

  5. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436–44.

    CAS  PubMed  Google Scholar 

  6. Lambrechts A, Ganapathi M, Wirix-Speetjens R. Clinical Evaluation of Artificial Intelligence based Preoperative Plans for Total Knee Arthroplasty. CAOS 2020 - The 20th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery: EasyChair; 2020. p. 169 – 73.

  7. Li Z, Zhang X, Ding L, Du K, Yan J, Chan MTV, Wu WKK, Li S. Deep learning approach for guiding three-dimensional computed tomography reconstruction of lower limbs for robotically-assisted total knee arthroplasty. Int J Med Robot. 2021;17(5):e2300. https://doi.org/10.1002/rcs.2300.

  8. Jacofsky DJ, Allen M. Robotics in Arthroplasty: A Comprehensive Review. J Arthroplasty. 2016 Oct;31(10):2353–63. doi:https://doi.org/10.1016/j.arth.2016.05.026. Epub 2016 May 18. PMID: 27325369.

    Article  PubMed  Google Scholar 

  9. Ramkumar PN, Karnuta JM, Navarro SM, Haeberle HS, Scuderi GR, Mont MA, et al. Deep Learning Preoperatively Predicts Value Metrics for Primary Total Knee Arthroplasty: Development and Validation of an Artificial Neural Network Model. J Arthroplasty. 2019;34(10):2220-7.e1.

  10. Li H, Jiao J, Zhang S, Tang H, Qu X, Yue B. Construction and Comparison of Predictive Models for Length of Stay after Total Knee Arthroplasty: Regression Model and Machine Learning Analysis Based on 1,826 Cases in a Single Singapore Center. J Knee Surg. 2022;35(1):7–14. https://doi.org/10.1055/s-0040-1710573.

  11. Karnuta JM, Navarro SM, Haeberle HS, Helm JM, Kamath AF, Schaffer JL, et al. Predicting Inpatient Payments Prior to Lower Extremity Arthroplasty Using Deep Learning: Which Model Architecture Is Best? J Arthroplasty. 2019;34(10):2235–41.e1.

    Google Scholar 

  12. El-Galaly A, Grazal C, Kappel A, Nielsen PT, Jensen SL, Forsberg JA. Can Machine-learning Algorithms Predict Early Revision TKA in the Danish Knee Arthroplasty Registry? Clin Orthop Relat Res. 2020;478(9):2088–101.

    PubMed  PubMed Central  Google Scholar 

  13. Heisinger S, Hitzl W, Hobusch GM, Windhager R, Cotofana S. Predicting Total Knee Replacement from Symptomology and Radiographic Structural Change Using Artificial Neural Networks-Data from the Osteoarthritis Initiative (OAI). J Clin Med. 2020;9(5):1298.

    PubMed Central  Google Scholar 

  14. Jafarzadeh S, Felson DT, Nevitt MC, Torner JC, Lewis CE, Roemer FW, et al. Use of clinical and imaging features of osteoarthritis to predict knee replacement in persons with and without radiographic osteoarthritis: the most study. Osteoarthr Cartil. 2020;28:308-S9.

    Google Scholar 

  15. Leung K, Zhang B, Tan J, Shen Y, Geras KJ, Babb JS, et al. Prediction of Total Knee Replacement and Diagnosis of Osteoarthritis by Using Deep Learning on Knee Radiographs: Data from the Osteoarthritis Initiative. Radiology. 2020;296(3):584–93.

    PubMed  Google Scholar 

  16. Tolpadi AA, Lee JJ, Pedoia V, Majumdar S. Deep Learning Predicts Total Knee Replacement from Magnetic Resonance Images. Sci Rep. 2020;10(1):6371.

    CAS  PubMed  PubMed Central  Google Scholar 

  17. Yi PH, Wei J, Kim TK, Sair HI, Hui FK, Hager GD, Fritz J, Oni JK. Automated detection & classification of knee arthroplasty using deep learning. Knee. 2020 Mar;27(2):535–42. doi: 10.1016/j.knee.2019.11.020. Epub 2019 Dec 26. PMID: 31883760.

    PubMed  Google Scholar 

  18. Norman B, Pedoia V, Noworolski A, Link TM, Majumdar S. Applying Densely Connected Convolutional Neural Networks for Staging Osteoarthritis Severity from Plain Radiographs. J Digit Imaging. 2019;32(3):471–7.

    PubMed  Google Scholar 

  19. Tiulpin A, Thevenot J, Rahtu E, Lehenkari P, Saarakkala S. Automatic Knee Osteoarthritis Diagnosis from Plain Radiographs: A Deep Learning-Based Approach. Sci Rep. 2018;8(1):1727-.

    PubMed  PubMed Central  Google Scholar 

  20. Harris AHS, Kuo AC, Bowe TR, Manfredi L, Lalani NF, Giori NJ. Can Machine Learning Methods Produce Accurate and Easy-to-Use Preoperative Prediction Models of One-Year Improvements in Pain and Functioning After Knee Arthroplasty? J Arthroplasty. 2021;36(1):112–7.e6.

    PubMed  Google Scholar 

  21. Bonakdari H, Pelletier JP, Martel-Pelletier J. A reliable time-series method for predicting arthritic disease outcomes: New step from regression toward a nonlinear artificial intelligence method. Comput Methods Programs Biomed. 2020 Jun;189:105315. doi:https://doi.org/10.1016/j.cmpb.2020.105315. Epub 2020 Jan 9. PMID: 31972347.

    Article  PubMed  Google Scholar 

  22. Farooq H, Deckard ER, Ziemba-Davis M, Madsen A, Meneghini RM. Predictors of Patient Satisfaction Following Primary Total Knee Arthroplasty: Results from a Traditional Statistical Model and a Machine Learning Algorithm. J Arthroplasty. 2020;35(11):3123–30.

    PubMed  Google Scholar 

  23. Hyer JM, White S, Cloyd J, Dillhoff M, Tsung A, Pawlik TM, Ejaz A. Can We Improve Prediction of Adverse Surgical Outcomes? Development of a Surgical Complexity Score Using a Novel Machine Learning Technique. J Am Coll Surg. 2020 Jan;230(1):43–52.e1. doi: 10.1016/j.jamcollsurg.2019.09.015. Epub 2019 Oct 28. PMID: 31672674.

    PubMed  Google Scholar 

  24. Ko S, Jo C, Chang CB, Lee YS, Moon YW, Youm JW, Han HS, Lee MC, Lee H, Ro DH. A web-based machine-learning algorithm predicting postoperative acute kidney injury after total knee arthroplasty. Knee Surg Sports Traumatol Arthrosc. 2020 Sep 3. doi: https://doi.org/10.1007/s00167-020-06258-0. Epub ahead of print. PMID: 32880677.

  25. Kunze KN, Polce EM, Sadauskas AJ, Levine BR. Development of Machine Learning Algorithms to Predict Patient Dissatisfaction After Primary Total Knee Arthroplasty. J Arthroplasty. 2020;35(11):3117–22.

    PubMed  Google Scholar 

  26. Fontana MA, Lyman S, Sarker GK, Padgett DE, MacLean CH. Can Machine Learning Algorithms Predict Which Patients Will Achieve Minimally Clinically Important Differences From Total Joint Arthroplasty? Clin Orthop Relat Res. 2019;477(6):1267–79.

    PubMed  PubMed Central  Google Scholar 

  27. Harris AHS, Kuo AC, Weng Y, Trickey AW, Bowe T, Giori NJ. Can Machine Learning Methods Produce Accurate and Easy-to-use Prediction Models of 30-day Complications and Mortality After Knee or Hip Arthroplasty? Clin Orthop Relat Res. 2019;477(2):452–60.

    PubMed  PubMed Central  Google Scholar 

  28. Huber M, Kurz C, Leidl R. Predicting patient-reported outcomes following hip and knee replacement surgery using supervised machine learning. BMC Med Inform Decis Mak. 2019;19(1):3-.

    PubMed  PubMed Central  Google Scholar 

  29. Lee HK, Jin R, Feng Y, Bain PA, Goffinet J, Baker C, Li J. An Analytical Framework for TJR Readmission Prediction and Cost-Effective Intervention. IEEE J Biomed Health Inform. 2019 Jul;23(4):1760–72. doi: 10.1109/JBHI.2018.2859581. Epub 2018 Jul 25. PMID: 30047916.

    PubMed  Google Scholar 

  30. Aram P, Trela-Larsen L, Sayers A, Hills AF, Blom AW, McCloskey EV, Kadirkamanathan V, Wilkinson JM. Estimating an Individual’s Probability of Revision Surgery After Knee Replacement: A Comparison of Modeling Approaches Using a National Data Set. Am J Epidemiol. 2018 Oct 1;187(10):2252–2262. doi: https://doi.org/10.1093/aje/kwy121. PMID: 29893799; PMCID: PMC6166214.

  31. Huang Z, Huang C, Xie J, Ma J, Cao G, Huang Q, Shen B, Byers Kraus V, Pei F. Analysis of a large data set to identify predictors of blood transfusion in primary total hip and knee arthroplasty. Transfusion. 2018 Aug;58(8):1855–1862. doi: https://doi.org/10.1111/trf.14783. Epub 2018 Aug 25. PMID: 30145838; PMCID: PMC6131039.

  32. Kluge F, Hannink J, Pasluosta C, Klucken J, Gaßner H, Gelse K, et al. Pre-operative sensor-based gait parameters predict functional outcome after total knee arthroplasty. Gait Posture. 2018;66:194–200.

    PubMed  Google Scholar 

  33. Van Onsem S, Van Der Straeten C, Arnout N, Deprez P, Van Damme G, Victor J. A New Prediction Model for Patient Satisfaction After Total Knee Arthroplasty. J Arthroplasty. 2016;31(12):2660-7.e1. https://doi.org/10.1016/j.arth.2016.06.004 Epub 2016 Jul 14. PMID: 27506723.

    Article  PubMed  Google Scholar 

  34. Gossec L, Jordan JM, Mazzuca SA, Lam MA, Suarez-Almazor ME, Renner JB, et al. Comparative evaluation of three semi-quantitative radiographic grading techniques for knee osteoarthritis in terms of validity and reproducibility in 1759 X-rays: report of the OARSI-OMERACT task force. Osteoarthritis Cartilage. 2008;16(7):742–8.

    CAS  PubMed  Google Scholar 

  35. Riddle DL, Jiranek WA, Hull JR. Validity and reliability of radiographic knee osteoarthritis measures by arthroplasty surgeons. Orthopedics. 2013;36(1):e25–32.

    PubMed  Google Scholar 

  36. Richardson ML. Deep Learning Improves Predictions of the Need for Total Knee Replacement. Radiology. 2020;296(3):594–5.

    PubMed  Google Scholar 

  37. Bedson J, Croft PR. The discordance between clinical and radiographic knee osteoarthritis: a systematic search and summary of the literature. BMC Musculoskelet Disord. 2008;9:116.

    PubMed  PubMed Central  Google Scholar 

  38. Bastick AN, Belo JN, Runhaar J, Bierma-Zeinstra SMA. What Are the Prognostic Factors for Radiographic Progression of Knee Osteoarthritis? A Meta-analysis. Clin Orthop Relat Res. 2015;473(9):2969–89.

    PubMed  PubMed Central  Google Scholar 

  39. Bastick AN, Runhaar J, Belo JN, Bierma-Zeinstra SMA. Prognostic factors for progression of clinical osteoarthritis of the knee: a systematic review of observational studies. Arthritis Res Therapy. 2015;17(1):152.

    Google Scholar 

  40. Hannan MT, Felson DT, Pincus T. Analysis of the discordance between radiographic changes and knee pain in osteoarthritis of the knee. J Rheumatol. 2000;27(6):1513–7.

    CAS  PubMed  Google Scholar 

  41. Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Development of criteria for the classification and reporting of osteoarthritis: Classification of osteoarthritis of the knee. Arthr Rhuem. 1986;29(8):1039–49.

    CAS  Google Scholar 

  42. Beaton DE, Boers M, Wells GA. Many faces of the minimal clinically important difference (MCID): a literature review and directions for future research. Curr Opin Rheumatol. 2002;14(2):109–14.

    PubMed  Google Scholar 

  43. Keurentjes JC, Van Tol FR, Fiocco M, Schoones JW, Nelissen RG. Minimal clinically important differences in health-related quality of life after total hip or knee replacement: A systematic review. Bone Joint Res. 2012;1(5):71–7.

    CAS  PubMed  PubMed Central  Google Scholar 

  44. Healy WL, Della Valle CJ, Iorio R, Berend KR, Cushner FD, Dalury DF, et al. Complications of total knee arthroplasty: standardized list and definitions of the Knee Society. Clin Orthop Relat Res. 2013;471(1):215–20.

    PubMed  Google Scholar 

  45. Romine LB, May RG, Taylor HD, Chimento GF. Accuracy and clinical utility of a peri-operative risk calculator for total knee arthroplasty. J Arthroplasty. 2013;28(3):445–8.

    PubMed  Google Scholar 

  46. Bozic KJ, Lau E, Kurtz S, Ong K, Rubash H, Vail TP, et al. Patient-related risk factors for periprosthetic joint infection and postoperative mortality following total hip arthroplasty in Medicare patients. J Bone Joint Surg Am. 2012;94(9):794–800.

    PubMed  Google Scholar 

  47. Manning DW, Edelstein AI, Alvi HM. Risk Prediction Tools for Hip and Knee Arthroplasty. J Am Acad Orthop Surg. 2016;24(1):19–27.

    PubMed  Google Scholar 

  48. Harris AHS, Kuo AC, Bozic KJ, Lau E, Bowe T, Gupta S, et al. American Joint Replacement Registry Risk Calculator Does Not Predict 90-day Mortality in Veterans Undergoing Total Joint Replacement. Clin Orthop Relat Res. 2018;476(9):1869–75.

    PubMed  PubMed Central  Google Scholar 

  49. Nichols JA, Herbert Chan HW, Baker MAB. Machine learning: applications of artificial intelligence to imaging and diagnosis. Biophys Rev. 2019;11(1):111–8.

    PubMed  Google Scholar 

  50. Price WN. Big data and black-box medical algorithms. Sci Transl Med. 2018;10(471):eaao5333.

    PubMed  PubMed Central  Google Scholar 

  51. Montavon G, Lapuschkin S, Binder A, Samek W, Müller K-R. Explaining nonlinear classification decisions with deep Taylor decomposition. Pattern Recogn. 2017;65:211–22.

    Google Scholar 

Download references

Acknowledgements

N/A.

Funding

Nil.

Author information

Authors and Affiliations

Authors

Contributions

L.S. Lee (study design, data acquisition and analysis, writing of manuscript); P.K. Chan (study design, analysis of data, providing revision comments); W.C. Fung (data acquisition and analysis); C. Wen, A. Cheung, V.W.K. Chan, M.H. Cheung, H. Fu, C.H. Yan and K.Y. Chiu (providing expert advices and revision comments). All authors read and approved the final manuscript.

Corresponding author

Correspondence to Ping Keung Chan.

Ethics declarations

Ethics approval and consent to participate

The need for approval was waived by Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (HKU/HA HKW IRB).

Consent for publication

Not applicable.

Competing interests

There are no competing interests to declare for any of the authors.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Lee, L.S., Chan, P.K., Wen, C. et al. Artificial intelligence in diagnosis of knee osteoarthritis and prediction of arthroplasty outcomes: a review. Arthroplasty 4, 16 (2022). https://doi.org/10.1186/s42836-022-00118-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s42836-022-00118-7

Keywords

  • Artificial intelligence
  • Machine learning
  • Arthroplasty
  • Replacement
  • Total knee arthroplasty
  • Osteoarthritis