Dataset Information

Prediction of Total Knee Replacement and Diagnosis of Osteoarthritis by Using Deep Learning on Knee Radiographs: Data from the Osteoarthritis Initiative.

ABSTRACT: Background The methods for assessing knee osteoarthritis (OA) do not provide enough comprehensive information to make robust and accurate outcome predictions. Purpose To develop a deep learning (DL) prediction model for risk of OA progression by using knee radiographs in patients who underwent total knee replacement (TKR) and matched control patients who did not undergo TKR. Materials and Methods In this retrospective analysis that used data from the OA Initiative, a DL model on knee radiographs was developed to predict both the likelihood of a patient undergoing TKR within 9 years and Kellgren-Lawrence (KL) grade. Study participants included a case-control matched subcohort between 45 and 79 years. Patients were matched to control patients according to age, sex, ethnicity, and body mass index. The proposed model used a transfer learning approach based on the ResNet34 architecture with sevenfold nested cross-validation. Receiver operating characteristic curve analysis and conditional logistic regression assessed model performance for predicting probability and risk of TKR compared with clinical observations and two binary outcome prediction models on the basis of radiographic readings: KL grade and OA Research Society International (OARSI) grade. Results Evaluated were 728 participants including 324 patients (mean age, 64 years ± 8 [standard deviation]; 222 women) and 324 control patients (mean age, 64 years ± 8; 222 women). The prediction model based on DL achieved an area under the receiver operating characteristic curve (AUC) of 0.87 (95% confidence interval [CI]: 0.85, 0.90), outperforming a baseline prediction model by using KL grade with an AUC of 0.74 (95% CI: 0.71, 0.77; P < .001). The risk for TKR increased with probability that a person will undergo TKR from the DL model (odds ratio [OR], 7.7; 95% CI: 2.3, 25; P < .001), KL grade (OR, 1.92; 95% CI: 1.17, 3.13; P = .009), and OARSI grade (OR, 1.20; 95% CI: 0.41, 3.50; P = .73). Conclusion The proposed deep learning model better predicted risk of total knee replacement in osteoarthritis than did binary outcome models by using standard grading systems. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Richardson in this issue.

SUBMITTER: Leung K

PROVIDER: S-EPMC7434649 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Objectives To evaluate the impact of total knee replacement on quality of life in people with knee osteoarthritis and to estimate associated differences in lifetime costs and quality adjusted life years (QALYs) according to use by level of symptoms.Design Marginal structural modeling and cost effectiveness analysis based on lifetime predictions for total knee replacement and death from population based cohort data.Setting Data from two studies-Osteoarthritis Initiative (OAI) and the Multicenter Osteoarthritis Study (MOST)-within the US health system.Participants 4498 participants with or at high risk for knee osteoarthritis aged 45-79 from the OAI with no previous knee replacement (confirmed by baseline radiography) followed up for nine years. Validation cohort comprised 2907 patients from MOST with two year follow-up.Intervention Scenarios ranging from current practice, defined as total knee replacement practice as performed in the OAI (with procedural rates estimated by a prediction model), to practice limited to patients with severe symptoms to no surgery.Main outcome measures Generic (SF-12) and osteoarthritis specific quality of life measured over 96 months, model based QALYs, costs, and incremental cost effectiveness ratios over a lifetime horizon.Results In the OAI, total knee replacement showed improvements in quality of life with small absolute changes when averaged across levels of confounding variables: 1.70 (95% uncertainty interval 0.26 to 3.57) for SF-12 physical component summary (PCS); -10.69 (-13.39 to -8.01) for Western Ontario and McMaster Universities arthritis index (WOMAC); and 9.16 (6.35 to 12.49) for knee injury and osteoarthritis outcome score (KOOS) quality of life subscale. These improvements became larger with decreasing functional status at baseline. Provision of total knee replacement to patients with SF-12 PCS scores <35 was the optimal scenario given a cost effectiveness threshold of $200?000/QALY, with cost savings of $6974 ($5789 to $8269) and a minimal loss of 0.008 (-0.056 to 0.043) QALYs compared with current practice. These findings were reproduced among patients with knee osteoarthritis from the MOST cohort and were robust against various scenarios including increased rates of total knee replacement and mortality and inclusion of non-healthcare costs but were sensitive to increased deterioration in quality of life without surgery. In a threshold analysis, total knee replacement would become cost effective in patients with SF-12 PCS scores ?40 if the associated hospital admission costs fell below $14?000 given a cost effectiveness threshold of $200?000/QALY.Conclusion Current practice of total knee replacement as performed in a recent US cohort of patients with knee osteoarthritis had minimal effects on quality of life and QALYs at the group level. If the procedure were restricted to more severely affected patients, its effectiveness would rise, with practice becoming economically more attractive than its current use.

Project description:Reduced external knee adduction moments in the second half of stance after total hip replacement have been reported in hip osteoarthritis patients. This reduction is thought to shift the load from the medial to the lateral knee compartment and as such increase the risk for knee osteoarthritis. The knee adduction moment is a surrogate for the load distribution between the medial and lateral compartments of the knee and not a valid measure for the tibiofemoral contact forces which are the result of externally applied forces and muscle forces. The purpose of this study was to investigate whether the distribution of the tibiofemoral contact forces over the knee compartments in unilateral hip osteoarthritis patients 1 year after receiving a primary total hip replacement differs from healthy controls. Musculoskeletal modeling on gait was performed in OpenSim using the detailed knee model of Lerner et al. (2015) for 19 patients as well as for 15 healthy controls of similar age. Knee adduction moments were calculated by the inverse dynamics analysis, medial and lateral tibiofemoral contact forces with the joint reaction force analysis. Moments and contact forces of patients and controls were compared using Statistical Parametric Mapping two-sample t-tests. Knee adduction moments and medial tibiofemoral contact forces of both the ipsi- and contralateral leg were not significantly different compared to healthy controls. The contralateral leg showed 14% higher medial tibiofemoral contact forces compared to the ipsilateral (operated) leg during the second half of stance. During the first half of stance, the lateral tibiofemoral contact force of the contralateral leg was 39% lower and the ratio 32% lower compared to healthy controls. In contrast, during the second half of stance the forces were significantly higher (39 and 26%, respectively) compared to healthy controls. The higher ratio indicates a changed distribution whereas the increased lateral tibiofemoral contact forces indicate a higher lateral knee joint loading in the contralateral leg in OA patients after total hip replacement (THR). Musculoskeletal modeling using a detailed knee model can be useful to detect differences in the load distribution between the medial and lateral knee compartment which cannot be verified with the knee adduction moment.

Dataset Information

Prediction of Total Knee Replacement and Diagnosis of Osteoarthritis by Using Deep Learning on Knee Radiographs: Data from the Osteoarthritis Initiative.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets