Osteoarthritis Initiative Database Research Articles

Inadequate Classification of Poor Response After Total Knee Arthroplasty: A Comparative Analysis of 15 Definitions Using Data from the Dutch Arthroplasty Register and the Osteoarthritis Initiative Database

Variations in defining poor response to total knee arthroplasty (TKA) impede comparisons of response after TKA over time and across hospitals. This study aimed to compare the prevalence, overlap, and discriminative accuracy of 15 definitions of poor response after TKA using two databases. Data of patients one year after primary TKA from the Dutch Arthroplasty Register (LROI) (n = 12,275) and the Osteoarthritis Initiative (OAI) database (n = 204) were used to examine the prevalence, overlap (estimated by Cohen's kappa), and discriminative accuracy (sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and Youden index) of 15 different definitions of poor response after TKA. In the absence of a gold standard for measuring poor response to TKA, the numeric rating scale (NRS) satisfaction (≤ 6 'poor responder') and the global assessment of knee impact (dichotomized: ≥ 4 'poor responder') were used as anchors for assessing discriminative accuracy for the LROI and OAI dataset, respectively. These anchors were chosen based on a prior qualitative study that identified (dis)satisfaction as a central theme of poor responses to TKA by patients and knee specialists. The median (25th to 75th percentile) prevalence of poor responders in the examined definitions was 18.5% (14.0 to 25.5%), and the median Cohen's kappa for the overlap between pairs of definitions was 0.41 (0.32 to 0.59). Median (25th to 75th percentile) sensitivity was 0.45 (0.39 to 0.54), specificity was 0.86 (0.82 to 0.94), PPV was 0.45 (0.34 to 0.62), NPV was 0.89 (0.87 to 0.89), and the Youden index was 0.36 (0.20 to 0.43). This found a lack of overlap between different definitions of poor response to TKA. None of the examined definitions adequately classified poor responders to TKA. In contrast, the absence of a poor response could be classified with confidence.

Dimensional changes of Hoffa’s fat pad related to aging: evaluation by MRI

BackgroundHoffa's fat pad is an intra-capsular extra-sinovial structure of the knee joint that has a significant biomechanical and metabolic role, minimizing the influence of stresses created by loading and generating cytokines. Changes in its size can lead to variations in the homeostasis of the knee in elderly patients. This work intends to assess the dimensional variations of Hoffa's fat pad associated to aging in both sexes, using MRI sagittal sequences acquired from the OAI (Osteoarthritis Initiative) database.MethodsWe examined the Hoffa's fat pad sagittal thickness in 217 men and women with knee osteoarthritis who were grouped into four age groups for the study: 40–49; 50–59; 60–69; and 70–80. 3T sagittal IW 2D TSE Fat-suppressed MRI sequences, taken from the OAI (Osteoarthritis Initiative) database, were examined.ResultsHoffa’s fat pad thickness was shown to differ significantly between groups in both men and women, decreasing in the older individuals' groups (R = − 0.46; p 0.0001). By dividing the patients into ten-year age groups and by sex, the thickness of both the right knee and the left knee was examined. In fact, the average thickness of Hoffa's fat pad of the right knee was reported to be, in males, 33.6+/− 3 mm in subjects aged between 40 and 49 years, 31+/− 2.4 mm for patients aged between 50 and 59, 30.3 ± 1.8 mm in the group between 60 and 69 years and 28.7+/− 1.8 mm between 70 and 80 years. In women the values obtained were the following: 29+/− 1.6 mm between 40 and 49 years; 28.9+/− 2.6 mm in the group between 50 and 59 years, 25.3+/− 1.9 mm for patients aged 60 and 69 years and 26+/− 2 mm between 70 and 80 years. Similar results were obtained for the left knee.ConclusionsHoffa’s fat pad gradually thins with aging in both male and female patients with knee osteoarthritis, and this can be detected by evaluating the thickness of the fat pad on sagittal MRI sequences.

Association of Medial Meniscus Root Tears and Nonroot Tears With Worsening of Radiographic Knee Osteoarthritis.

Medial meniscus root tears (RTs) are associated with the development and worsening of knee osteoarthritis (OA), but little is known about their progression when compared with meniscal tears that spare the root (nonroot tears; NRTs). To compare radiographic worsening of OA in knees with RTs versus NRTs and to identify factors associated with radiographic worsening of OA in knees with RTs. Cohort study; Level of evidence, 3. Using the Osteoarthritis Initiative database, we included knees with medial meniscus RTs and NRTs present at the baseline visit (baseline tears) and new RTs and NRTs observed at 12- to 48-month annual follow-up visits (incident tears). Worsening of radiographic OA was defined for baseline tears as an increase in Kellgren-Lawrence grade (KLG) during the subsequent 12 months of follow-up; for incident tears, worsening was defined as either concurrent (increase in KLG over the 12 months preceding tear appearance on magnetic resonance imaging) or subsequent (increase in KLG during the 12 months after tear appearance). Odds ratios (ORs), adjusted for covariates, were calculated for the association of worsening by type of tear. Included were 39 knees with baseline RTs, 633 knees with baseline NRTs, 33 knees with incident RTs, and 234 knees with incident NRTs. Radiographic OA worsening subsequent to meniscal tear identification was no different for baseline RTs (15%) or baseline NRTs (14%; adjusted OR, 1.34; 95% CI, 0.52-3.47), nor did subsequent worsening differ for incident RTs (19%) versus incident NRTs (18%; adjusted OR, 0.52; 95% CI, 0.15-1.83). Concurrent radiographic OA worsening was seen at a significantly higher rate for incident RTs (64%) versus incident NRTs (21%; adjusted OR, 3.00; 95% CI, 1.21-7.47). Incident RTs in knees without radiographic OA (KLG 0-1) before the tear had a high rate of worsening (94%, n = 16) and were more likely to worsen than those in knees with radiographic OA (KLG ≥2) present before the tear. Compared with NRTs, incident RTs were associated with a significantly increased risk of worsening radiographic OA over a 12-month period concurrent with the appearance of the tear. Incident RTs in knees without radiographic OA at baseline had a high rate of worsening.

Association between Elevated Magnesium Intake and Reduced Risk of Recurrent Falls and Frailty in Osteoarthritis: Data from the Osteoarthritis Initiative.

The objective of this 8-year follow-up study was to investigate the relationship between magnesium intake and frailty, as well as recurrent falls, in individuals diagnosed with Osteoarthritis (OA) or those at a heightened risk for developing the condition. This study utilized data from the Osteoarthritis Initiative (OAI) database and conducted a prospective cohort study with a 8-year follow-up period. Total magnesium intake from both food sources and supplements was assessed using a food frequency questionnaire (FFQ), while frailty and recurrent falls were evaluated through established criteria and self-report, respectively. To account for potential confounding factors, various covariates were considered, and statistical analyses, including generalized additive mixed models (GAMMs), were employed to examine the associations. Among the 4,667 participants with OA, those with lower total magnesium intake were characterized by younger age, a higher proportion of African American individuals, higher body mass index (BMI), and lower dietary fiber intake (P<0.001). Notably, this group exhibited higher odds of experiencing recurrent falls and frailty (P = 0.034 and 0.006, respectively). Controlling for various factors, the GAMMs consistently revealed negative correlations between magnesium intake and the likelihood of frailty and recurrent falls, with each 1 mg/1000 kcal increase in magnesium intake associated with a 0.5% reduced frailty risk (p < 0.001) and a 0.2% decreased risk of recurrent falls (p = 0.001). Subgroup analyses suggested that increased total magnesium intake from both food sources and supplements may exert a more pronounced preventive effect on recurrent falls and frailty in men, older adults, individuals with normal BMI, and those with higher dietary fiber intake. Elevated total magnesium intake from both food sources and supplements was found to be associated with a decreased risk of recurrent falls and frailty in individuals diagnosed with OA or those at risk of developing the condition. These findings imply that increased total magnesium intake might be beneficial in managing the risk of these outcomes, particularly within specific subgroups, including men, older adults, those with a normal BMI, and those with higher dietary fiber intake.

Predicting hip-knee-ankle and femorotibial angles from knee radiographs with deep learning

BackgroundKnee alignment affects the development and surgical treatment of knee osteoarthritis. Automating femorotibial angle (FTA) and hip-knee-ankle angle (HKA) measurement from radiographs could improve reliability and save time. Further, if HKA could be predicted from knee-only radiographs then radiation exposure could be reduced and the need for specialist equipment and personnel avoided. The aim of this research was to assess if deep learning methods could predict FTA and HKA angle from posteroanterior (PA) knee radiographs. MethodsConvolutional neural networks with densely connected final layers were trained to analyse PA knee radiographs from the Osteoarthritis Initiative (OAI) database. The FTA dataset with 6149 radiographs and HKA dataset with 2351 radiographs were split into training, validation, and test datasets in a 70:15:15 ratio. Separate models were developed for the prediction of FTA and HKA and their accuracy was quantified using mean squared error as loss function. Heat maps were used to identify the anatomical features within each image that most contributed to the predicted angles. ResultsHigh accuracy was achieved for both FTA (mean absolute error 0.8°) and HKA (mean absolute error 1.7°). Heat maps for both models were concentrated on the knee anatomy and could prove a valuable tool for assessing prediction reliability in clinical application. ConclusionDeep learning techniques enable fast, reliable and accurate predictions of both FTA and HKA from plain knee radiographs and could lead to cost savings for healthcare providers and reduced radiation exposure for patients.

Longitudinal associations of body mass index and abdominal circumference with back pain among community-dwelling adults: data from the Osteoarthritis Initiative

BACKGROUND CONTEXTBack pain is the most common musculoskeletal problem in both developed and developing countries. The prevalence and burden of back pain increases with age, and the management of back pain becomes increasingly important in the context of global aging. There is increasing evidence that obesity is a modifiable risk factor for musculoskeletal pain in different locations. Understanding the role of obesity in back pain holds great potential for improving understanding of the mechanisms of back pain and for developing new preventive and therapeutic approaches. PURPOSETo evaluate the role of weight, body mass index (BMI) and abdominal circumference (AC) in risk of back pain over 96 months. DESIGNProspective cohort study. PATIENT SAMPLEThe sample was from 4,793 adults in the Osteoarthritis Initiative (OAI) database who had or were at increased risk for knee Osteoarthritis. OUTCOME MEASURESOutcome variables included the presence, severity, and frequency of back pain, using the past 30 days as the time frame. METHODSLongitudinal analysis of data from 4,793 participants enrolled in the Osteoarthritis Initiative, assessed every 12 or 24 months for weight, BMI (kg/m2), AC (cm), and presence, severity (none, mild, moderate, severe), and frequency (none, rarely, sometimes, often, always) of back pain. BMI and AC were decomposed into between-person and with-person components. Data analyses were performed using mixed-effects logistic (for presence of back pain) or ordered logistic regression (for severity and frequency of back pain) models. ResultsBack pain was reported in 58% of participants at baseline; 70% of those without back pain had incident back pain over 96 months. Both between-person (average value across a participant's all measurements) and within-person (deviations from the participant's average) effects of weight and BMI increased risk of presence, severity, and frequency of back pain (Odds radios (OR) per kg/m2: 1.010–1.046, p<.05) in females but not males, with statistically significant weight*sex and BMI*sex interactions. Similar findings were observed for between-person effects of AC on back pain, and the within-person effect of AC was only associated with back pain severity (OR per cm: 1.009, 95% confidence interval 1.002–1.017, p=.019) in females. CONCLUSIONSGreater average weight and BMI and increases in them increased odds of presence, severity, and frequency of back pain over 96 months in middle aged and older women but not men. Only average AC increased odds of back pain over time, in women. These findings suggest that preventing obesity and slowing weight gain is important for the management of back pain.

Factors Associated with Trajectories of Physical Activity Over 8 Years in Knee Osteoarthritis.

The aim of the study was to investigate (1) trajectories of physical activity (PA) over 96 months and (2) study to what extent knee pain, muscle strength, physical function, and radiographic disease were associated with PA trajectories in adults with or at risk of knee osteoarthritis (KOA). Using the Osteoarthritis Initiative (OAI) database, we described PA trajectories with the Physical Activity Scale for the Elderly (PASE) over 96 months. Knee pain was categorized into three groups: "no pain" [visual numeric pain rating scale (VAS=0)], "little to some pain" (VAS=1-3), or "moderate to severe pain" (VAS ≥ 4). Knee extensor strength was classified into high [>16.21 (men) and >10.82 (women) N/kg/m2 ] and low [12 seconds) and fast (<12 seconds) groups. Radiographic disease was classified as present [Kellgren-Lawrence (KL) ≥2] or absent (KL grade <2) of KOA. Among 3755 participants (age 61.0 ± 9.0 years, body mass index 28.5±4.8 kg/m2 , 58% female), we identified three trajectories: sedentary PA with slow decline (44.3%), low PA with slow decline (41.3%), and high PA with slow decline (14.4%). Poorer gait speed (OR: 2.32; 95% CI: 1.71-3.16), chair stand time (OR: 1.45; 95% CI: 1.07-1.96), and knee extensor strength (OR: 1.35; 95% CI: 1.03-1.76), but not pain or radiographic disease, were associated with PA trajectory of sedentary PA with slow decline. Physical function and strength, but not pain and radiographic disease, were associated with a trajectory of decline in PA among adults with or at risk of KOA.

The effect of interactions between BMI and sustained depressive symptoms on knee osteoarthritis over 4 years: data from the osteoarthritis initiative

BackgroundTo assess the compound effects of BMI and sustained depressive symptoms on changes in knee structure, cartilage composition, and knee pain over 4 years using statistical interaction analyses.MethodsOne thousand eight hundred forty-four individuals from the Osteoarthritis Initiative Database were analyzed at baseline and 4-year follow-up. Individuals were categorized according to their BMI and presence of depressive symptoms (based on the Center for Epidemiological Studies Depression Scale (threshold≥16)) at baseline and 4-year follow-up. 3 T MRI was used to quantify knee cartilage T2 over 4 years, while radiographs were used to assess joint space narrowing (JSN). Mixed effects models examined the effect of BMI-depressive symptoms interactions on outcomes of cartilage T2, JSN, and knee pain over 4-years.ResultsThe BMI-depressive symptoms interaction was significantly associated with knee pain (p < 0.001) changes over 4 years, but not with changes in cartilage T2 (p = 0.27). In women, the BMI-depressive symptoms interaction was significantly associated with JSN (p = 0.01). In a group-based analysis, participants with obesity and depression had significantly greater 4-year changes in knee pain (coeff.(obesity + depression vs. no_obesity + no_depression) = 4.09, 95%CI = 3.60–4.58, p < 0.001), JSN (coeff. = 0.60, 95%CI = 0.44–0.77, p < 0.001), and cartilage T2 (coeff. = 1.09, 95%CI = 0.68–1.49, p < 0.001) than participants without depression and normal BMI.ConclusionsThe compound effects of obesity and depression have greater impact on knee pain and JSN progression compared to what would be expected based on their individual effects.

Baseline Knee Pain Predicts Long-Term Response of Intra-Articular Steroid Injection in Symptomatic Knee Osteoarthritis: Data from OAI.

The current study aims to investigate the factors that could predict response to intra-articular corticosteroid injection (IACI) in patients with knee osteoarthritis (KOA). Data of participants were retrieved from the Osteoarthritis Initiative database. Participants with at least one IACI treatment on single or bilateral knees within the first 5 years of follow-up were retrospectively included. Demographic data, clinical and radiographic variables were collected at both baseline and the first follow-up after IACI treatment. Positive response to IACI treatment was defined as >20% reduction of Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain score from V0 to V1. All the variables with P < 0.2 after the comparison between the response and non-response groups were included in a multivariable logistic regression model to identify independent response predictive patient-specific valuables. Receiver operating characteristic curves were performed to establish the cutoff values of independent predictors. The current study included a total of 385 participants (473 knees), with 155 and 318 knees classified into the response group and non-response group, respectively. Those with satisfied responses to IACI treatment had significantly higher WOMAC pain score (P < 0.001), disability score (P = 0.002), and stiffness score (P = 0.015) at the baseline. Baseline WOMAC pain score showed significant association with positive response to IACI treatment in multivariate logistic analysis and the best cutoff value was 5 points. The rate of analgesics utilization was lower (P = 0.014) in the response group than the non-response group after the IACI treatment. KOA patients with a baseline WOMAC pain score ≥5 are more likely to benefit from IACI treatment.

Unsupervised machine-learning algorithms for the identification of clinical phenotypes in the osteoarthritis initiative database

ObjectivesOsteoarthritis (OA) is a complex disease comprising diverse underlying patho-mechanisms. To enable the development of effective therapies, segmentation of the heterogenous patient population is critical. This study aimed at identifying such patient clusters using two different machine learning algorithms. MethodsUsing the progression and incident cohorts of the Osteoarthritis Initiative (OAI) dataset, deep embedded clustering (DEC) and multiple factor analysis with clustering (MFAC) approaches, including 157 input-variables at baseline, were employed to differentiate specific patient profiles. ResultsDEC resulted in 5 and MFAC in 3 distinct patient phenotypes. Both identified a “comorbid” cluster with higher body mass index (BMI), relevant burden of comorbidity and low levels of physical activity. Both methods also identified a younger and physically more active cluster and an elderly cluster with functional limitations, but low disease impact. The additional two clusters identified with DEC were subgroups of the young/physically active and the elderly/physically inactive clusters. Overall pain trajectories over 9 years were stable, only the numeric rating scale (NRS) for pain showed distinct increase, while physical activity decreased in all clusters. Clusters showed different (though non-significant) trajectories of joint space changes over the follow-up period of 8 years. ConclusionTwo different clustering approaches yielded similar patient allocations primarily separating complex “comorbid” patients from healthier subjects, the latter divided in young/physically active vs elderly/physically inactive subjects. The observed association to clinical (pain/physical activity) and structural progression could be helpful for early trial design as strategy to enrich for patients who may specifically benefit from disease-modifying treatments.

Poster 242: Deep Learning for Identifying Patellofemoral Measurements Associated with Cartilage Lesions on MRI

Poster 242: Deep Learning for Identifying Patellofemoral Measurements Associated with Cartilage Lesions on MRI

Concept End Points Informing Design Considerations for Confirmatory Clinical Trials in Osteoarthritis.

There is an unmet need for therapies that target the underlying pathophysiology of osteoarthritis (OA). However, defining appropriate measures for clinical trials of such therapies is challenging. Our objective was to propose concept clinical end points that directly capture clinical benefit in this setting and evaluate the feasibility of their use. This analysis used the multicenter, longitudinal, observational Osteoarthritis Initiative (OAI) database. OAI participants primarily had knee OA, with follow-up of up to 9 years and assessments of joints, surgical interventions, performance outcomes, and patient-reported outcomes. We examined this data set to identify existing outcome measures of direct clinical benefit. We evaluated the feasibility of conducting trials using these candidate end points by estimating incidence rates and resulting required sample sizes and study durations in time-to-event analyses. We identified candidate end points based on total knee replacement (TKR) and composite end points defined by TKR and conservative thresholds of patient-reported outcomes of pain and function. Using time to TKR as an end point, a study with an average follow-up time of 3 years requires approximately 3,000 to 18,000 subjects, depending on effect size. Alternatively, for a composite end point, such as "time to TKR or severe pain or severely impaired functioning," the required sample sizes ranged from approximately 2,000 to 11,000 for a 3-year study. The proposed concept end points can reliably and feasibly evaluate the effectiveness of therapies for this unmet need. In particular, the composite end point approach can substantially reduce sample sizes (up to approximately 40%) compared to the use of TKR alone.

Predicting osteoarthritis onset and progression with 3D texture analysis of cartilage MRI DESS: 6-Year data from osteoarthritis initiative.

In this study, we developed a gray level co-occurrence matrix-based 3D texture analysis method for dual-echo steady-state (DESS) magnetic resonance (MR) images to be used for knee cartilage analysis in osteoarthritis (OA) studies and use it to study changes in articular cartilage between different subpopulations based on their rate of progression into radiographically confirmed OA. In total, 642 series of right knee DESS MR images at 3T were obtained from baseline, 36- and 72-month follow-ups from the OA Initiative database. At baseline, all 214 subjects included in the study had Kellgren-Lawrence (KL) grade <2. Three groups were defined, based on time of progression into radiographic OA (ROA) (KL grades ≥2): control (no progression), fast progressor (ROA at 36 months), and slow progressor (ROA at 72 months) groups. 3D texture analysis was used to extract textural features for femoral and tibial cartilages. All textural features, in both femur and tibia, showed significant longitudinal changes across all groups and tissue layers. Most of the longitudinal changes were observed in progressors, but significant changes were observed also in controls. Differences between groups were mostly seen at baseline and 72 months. The method is sensitive to cartilage changes before and after ROA. It was able to detect longitudinal changes in controls and progressors and to distinguish cartilage alterations due to OA and aging. Moreover, it was able to distinguish controls and different progressor groups before any radiographic signs of OA and during OA. Thus, texture analysis could be used as a marker for the onset and progression of OA.

Explainable machine learning for knee osteoarthritis diagnosis based on a novel fuzzy feature selection methodology.

Knee Osteoarthritis (ΚΟΑ) is a degenerative joint disease of the knee that results from the progressive loss of cartilage. Due to KOA’s multifactorial nature and the poor understanding of its pathophysiology, there is a need for reliable tools that will reduce diagnostic errors made by clinicians. The existence of public databases has facilitated the advent of advanced analytics in KOA research however the heterogeneity of the available data along with the observed high feature dimensionality make this diagnosis task difficult. The objective of the present study is to provide a robust Feature Selection (FS) methodology that could: (i) handle the multidimensional nature of the available datasets and (ii) alleviate the defectiveness of existing feature selection techniques towards the identification of important risk factors which contribute to KOA diagnosis. For this aim, we used multidimensional data obtained from the Osteoarthritis Initiative database for individuals without or with KOA. The proposed fuzzy ensemble feature selection methodology aggregates the results of several FS algorithms (filter, wrapper and embedded ones) based on fuzzy logic. The effectiveness of the proposed methodology was evaluated using an extensive experimental setup that involved multiple competing FS algorithms and several well-known ML models. A 73.55% classification accuracy was achieved by the best performing model (Random Forest classifier) on a group of twenty-one selected risk factors. Explainability analysis was finally performed to quantify the impact of the selected features on the model’s output thus enhancing our understanding of the rationale behind the decision-making mechanism of the best model.

A Multi-Task Deep Learning Method for Detection of Meniscal Tears in MRI Data from the Osteoarthritis Initiative Database.

We present a novel and computationally efficient method for the detection of meniscal tears in Magnetic Resonance Imaging (MRI) data. Our method is based on a Convolutional Neural Network (CNN) that operates on complete 3D MRI scans. Our approach detects the presence of meniscal tears in three anatomical sub-regions (anterior horn, body, posterior horn) for both the Medial Meniscus (MM) and the Lateral Meniscus (LM) individually. For optimal performance of our method, we investigate how to preprocess the MRI data and how to train the CNN such that only relevant information within a Region of Interest (RoI) of the data volume is taken into account for meniscal tear detection. We propose meniscal tear detection combined with a bounding box regressor in a multi-task deep learning framework to let the CNN implicitly consider the corresponding RoIs of the menisci. We evaluate the accuracy of our CNN-based meniscal tear detection approach on 2,399 Double Echo Steady-State (DESS) MRI scans from the Osteoarthritis Initiative database. In addition, to show that our method is capable of generalizing to other MRI sequences, we also adapt our model to Intermediate-Weighted Turbo Spin-Echo (IW TSE) MRI scans. To judge the quality of our approaches, Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) values are evaluated for both MRI sequences. For the detection of tears in DESS MRI, our method reaches AUC values of 0.94, 0.93, 0.93 (anterior horn, body, posterior horn) in MM and 0.96, 0.94, 0.91 in LM. For the detection of tears in IW TSE MRI data, our method yields AUC values of 0.84, 0.88, 0.86 in MM and 0.95, 0.91, 0.90 in LM. In conclusion, the presented method achieves high accuracy for detecting meniscal tears in both DESS and IW TSE MRI data. Furthermore, our method can be easily trained and applied to other MRI sequences.

A clinical model for predicting knee replacement in early-stage knee osteoarthritis: data from osteoarthritis initiative.

Knee osteoarthritis (OA) progresses in a heterogeneous way, as a majority of the patients gradually worsen over decades while some undergo rapid progression and require knee replacement. The aim of this study was to develop a predictive model that enables quantified risk prediction of future knee replacement in patients with early-stage knee OA. Patients with early-stage knee OA, intact MRI measurements, and a follow-up time larger than 108months were retrieved from the Osteoarthritis Initiative database. Twenty-five candidate predictors including demographic data, clinical outcomes, and radiographic parameters were selected. The presence or absence of knee replacement during the first 108months of the follow-up was regarded as the primary outcome. Patients were randomly divided into derivation and validation groups in the ratio of three to one. Nomograms were developed based on multivariable logistic regressions of derivation group via R language. Those models were further tested in the validation group for external validation. A total of 839 knees were enrolled, with 98 knees received knee replacement during the first 108months. Glucocorticoid injection history, knee OA in the contralateral side, extensor muscle strength, area of cartilage deficiency, bone marrow lesion, and meniscus extrusion were selected to develop the nomogram after multivariable logistic regression analysis. The bias-corrected C-index and AUC of our nomogram in the validation group were 0.804 and 0.822, respectively. Our predicting model provided simplified identification of patients with high risk of rapid progression in knee OA, which showed adequate predictive discrimination and calibration. • Knee OA progresses in a heterogeneous way and rises to a challenge when making treatment strategies. • Our predicting model provided simplified identification of patients with high risk of rapid progression in knee OA.

Machine learning to predict incident radiographic knee osteoarthritis over 8 Years using combined MR imaging features, demographics, and clinical factors: data from the Osteoarthritis Initiative

To develop a machine learning-based prediction model for incident radiographic osteoarthritis (OA) of the knee over 8 years using MRI-based cartilage biochemical composition and knee joint structure, demographics, and clinical predictors including muscle strength and symptoms. Individuals (n=1,044) with baseline Kellgren Lawrence (KL) grade 0-1 in the right knee from the Osteoarthritis Initiative database were analyzed. 3T MRI at baseline was used to quantify knee cartilage T2, and Whole-Organ Magnetic Resonance Imaging Scores (WORMS) were obtained for cartilage, meniscus, and bone marrow. The outcome was set as true if a subject developed KL grade 2-4 OA in the right knee over 8 years (n=183) and false if the subject remained at KL 0-1 over 8 years (n=861). We developed and compared three models: Model 1: 112 predictors based on OA risk factors; Model 2: top ten predictors based on feature importance score from Model 1 and clinical relevance; Model 3: Model 2 without the imaging predictors. We compared the models using the area under the ROC curve derived from hold-out data. The 10-predictor model (Model 2, that includes cartilage and meniscus WORMS scores and cartilage T2) had a slightly lower AUC (0.772) compared to the model with 112 predictors (Model 1: AUC=0.792, p=0.739); and had a significantly higher AUC compared to the model without MR imaging predictors (Model 3, AUC=0.669, p=0.011). A 10-predictor model including MRI parameters coupled with demographics, symptoms, muscle, and physical activity scores provides good prediction of incident radiographic OA over 8 years.

Towards novel osteoarthritis biomarkers: Multi-criteria evaluation of 46,996 segmented knee MRI data from the Osteoarthritis Initiative.

Convolutional neural networks (CNNs) are the state-of-the-art for automated assessment of knee osteoarthritis (KOA) from medical image data. However, these methods lack interpretability, mainly focus on image texture, and cannot completely grasp the analyzed anatomies’ shapes. In this study we assess the informative value of quantitative features derived from segmentations in order to assess their potential as an alternative or extension to CNN-based approaches regarding multiple aspects of KOA. Six anatomical structures around the knee (femoral and tibial bones, femoral and tibial cartilages, and both menisci) are segmented in 46,996 MRI scans. Based on these segmentations, quantitative features are computed, i.e., measurements such as cartilage volume, meniscal extrusion and tibial coverage, as well as geometric features based on a statistical shape encoding of the anatomies. The feature quality is assessed by investigating their association to the Kellgren-Lawrence grade (KLG), joint space narrowing (JSN), incident KOA, and total knee replacement (TKR). Using gold standard labels from the Osteoarthritis Initiative database the balanced accuracy (BA), the area under the Receiver Operating Characteristic curve (AUC), and weighted kappa statistics are evaluated. Features based on shape encodings of femur, tibia, and menisci plus the performed measurements showed most potential as KOA biomarkers. Differentiation between non-arthritic and severely arthritic knees yielded BAs of up to 99%, 84% were achieved for diagnosis of early KOA. Weighted kappa values of 0.73, 0.72, and 0.78 were achieved for classification of the grade of medial JSN, lateral JSN, and KLG, respectively. The AUC was 0.61 and 0.76 for prediction of incident KOA and TKR within one year, respectively. Quantitative features from automated segmentations provide novel biomarkers for KLG and JSN classification and show potential for incident KOA and TKR prediction. The validity of these features should be further evaluated, especially as extensions of CNN-based approaches. To foster such developments we make all segmentations publicly available together with this publication.

A deep learning method for predicting knee osteoarthritis radiographic progression from MRI

BackgroundThe identification of patients with knee osteoarthritis (OA) likely to progress rapidly in terms of structure is critical to facilitate the development of disease-modifying drugs.MethodsUsing 9280 knee magnetic resonance (MR) images (3268 patients) from the Osteoarthritis Initiative (OAI) database , we implemented a deep learning method to predict, from MR images and clinical variables including body mass index (BMI), further cartilage degradation measured by joint space narrowing at 12 months.ResultsUsing COR IW TSE images, our classification model achieved a ROC AUC score of 65%. On a similar task, trained radiologists obtained a ROC AUC score of 58.7% highlighting the difficulty of the classification task. Additional analyses conducted in parallel to predict pain grade evaluated by the WOMAC pain index achieved a ROC AUC score of 72%. Attention maps provided evidence for distinct specific areas as being relevant in those two predictive models, including the medial joint space for JSN progression and the intra-articular space for pain prediction.ConclusionsThis feasibility study demonstrates the interest of deep learning applied to OA, with a potential to support even trained radiologists in the challenging task of identifying patients with a high-risk of disease progression.

Identification of most important features based on a fuzzy ensemble technique: Evaluation on joint space narrowing progression in knee osteoarthritis patients

ObjectiveFeature selection (FS) is a crucial and at the same time challenging processing step that aims to reduce the dimensionality of complex classification or regression problems. Various techniques have been proposed in the literature to address this challenge with emphasis to medical applications. However, each one of the existing FS algorithms come with its own advantages and disadvantages introducing a certain level of bias. Materials and MethodsTo avoid bias and alleviate the defectiveness of single feature selection results, an ensemble FS methodology is proposed in this paper that aggregates the results of several FS algorithms (filter, wrapper and embedded ones). Fuzzy logic is employed to combine multiple feature importance scores thus leading to a more robust selection of informative features. The proposed fuzzy ensemble FS methodology was applied on the problem of knee osteoarthritis (KOA) prediction with special emphasis on the progression of joint space narrowing (JSN). The proposed FS methodology was integrated into an end-to-end machine learning pipeline and a thorough experimental evaluation was conducted using data from the Osteoarthritis Initiative (OAI) database. Several classifiers were investigated for their suitability in the task of JSN prediction and the best performing model was then post-hoc analyzed by using the SHAP method. ResultsThe results showed that the proposed method presented a better and more stable performance in contrast to other competitive feature selection methods, leading to an average accuracy of 78.14% using XG Boost at 31 selected features. The post-hoc explainability highlighted the important features that contribute to the classification of patients with JSN progression. ConclusionsThe proposed fuzzy feature selection approach improves the performance of the predictive models by selecting a small optimal subset of features compared to popular feature selection methods.