The KNee OsteoArthritis Prediction (KNOAP2020) challenge: An image analysis challenge to predict incident symptomatic radiographic knee osteoarthritis from MRI and X-ray images.

OBJECTIVES: The KNee OsteoArthritis Prediction (KNOAP2020) challenge was organized to objectively compare methods for the prediction of incident symptomatic radiographic knee osteoarthritis within 78 months on a test set with blinded ground truth. DESIGN: The challenge participants were free to use any available data sources to train their models. A test set of 423 knees from the Prevention of Knee Osteoarthritis in Overweight Females (PROOF) study consisting of magnetic resonance imaging (MRI) and X-ray image data along with clinical risk factors at baseline was made available to all challenge participants. The ground truth outcomes, i.e., which knees developed incident symptomatic radiographic knee osteoarthritis (according to the combined ACR criteria) within 78 months, were not provided to the participants. To assess the performance of the submitted models, we used the area under the receiver operating characteristic curve (ROCAUC) and balanced accuracy (BACC). RESULTS: Seven teams submitted 23 entries in total. A majority of the algorithms were trained on data from the Osteoarthritis Initiative. The model with the highest ROCAUC (0.64 (95% confidence interval (CI): 0.57-0.70)) used deep learning to extract information from X-ray images combined with clinical variables. The model with the highest BACC (0.59 (95% CI: 0.52-0.65)) ensembled three different models that used automatically extracted X-ray and MRI features along with clinical variables. CONCLUSION: The KNOAP2020 challenge established a benchmark for predicting incident symptomatic radiographic knee osteoarthritis. Accurate prediction of incident symptomatic radiographic knee osteoarthritis is a complex and still unsolved problem requiring additional investigation.

Osteoarthritis year in review 2021: imaging.

PURPOSE: To provide a narrative review of original articles on imaging of osteoarthritis (OA) published between January 1, 2020 and March 31, 2021, with a special focus on imaging of inflammation, imaging of bone, cartilage and bone-cartilage interactions, imaging of peri-articular tissues, imaging scoring methods for OA, and artificial intelligence (AI) applied to OA imaging. METHODS: The Embase, Pubmed, Medline, Cochrane databases were searched for original research articles in the English language on human, in vivo, imaging of OA published between January 1, 2020 and March 31, 2021. Search terms related to osteoarthritis combined with all imaging modalities and artificial intelligence were applied. A selection of articles reporting on one of the focus topics was discussed further. RESULTS: The search resulted in 651 articles, of which 214 were deemed relevant to human OA imaging. Among the articles included, the knee joint (69%) and magnetic resonance imaging (MRI) (52%) were the predominant anatomical area and imaging modality studied. There were also a substantial number of papers (n = 46) reporting on AI applications in the field of OA imaging. CONCLUSION: Imaging continues to play an important role in the assessment of OA. Recent advances in OA imaging include quantitative, non-contrast, and hybrid imaging techniques for improved characterization of multiple tissue processes in OA. In addition, an increasing effort in AI techniques is undertaken to enhance OA imaging acquisition and analysis.

Detecting hip osteoarthritis on clinical CT: a deep learning application based on 2-D summation images derived from CT.

We developed and compared deep learning models to detect hip osteoarthritis on clinical CT. The CT-based summation images, CT-AP, that resemble X-ray radiographs can detect radiographic hip osteoarthritis and in the absence of large training data, a reliable deep learning model can be optimized by combining CT-AP and X-ray images. INTRODUCTION: In this study, we aimed to investigate the applicability of deep learning (DL) to assess radiographic hip osteoarthritis (rHOA) on computed tomography (CT). METHODS: The study data consisted of 94 abdominopelvic clinical CTs and 5659 hip X-ray images collected from Cohort Hip and Cohort Knee (CHECK). The CT slices were sequentially summed to create radiograph-like 2-D images named CT-AP. X-ray and CT-AP images were classified as rHOA if they had osteoarthritic changes corresponding to Kellgren-Lawrence grade 2 or higher. The study data was split into 55% training, 30% validation, and 15% test sets. A pretrained ResNet18 was optimized for a classification task of rHOA vs. no-rHOA. Five models were trained using (1) X-rays, (2) downsampled X-rays, (3) combination of CT-AP and X-ray images, (4) combination of CT-AP and downsampled X-ray images, and (5) CT-AP images. RESULTS: Amongst the five models, Model-3 and Model-5 performed best in detecting rHOA from the CT-AP images. Model-3 detected rHOA on the test set of CT-AP images with a balanced accuracy of 82.2% and was able to discriminate rHOA from no-rHOA with an area under the receiver operating characteristic curve (ROC AUC) of 0.93 [0.75-0.99]. Model-5 detected rHOA on the test set at a balanced accuracy of 82.2% and classified rHOA from no-rHOA with an ROC AUC of 0.89 [0.67-0.97]. CONCLUSION: CT-based summation images that resemble radiographs can be used to detect rHOA. In addition, in the absence of large training data, a reliable DL model can be optimized by combining CT-AP and X-ray images.

Adaptive segmentation of knee radiographs for selecting the optimal ROI in texture analysis.

OBJECTIVE: The purposes of this study were to investigate: 1) the effect of placement of region-of-interest (ROI) for texture analysis of subchondral bone in knee radiographs, and 2) the ability of several texture descriptors to distinguish between the knees with and without radiographic osteoarthritis (OA). DESIGN: Bilateral posterior-anterior knee radiographs were analyzed from the baseline of Osteoarthritis Initiative (OAI) (9012 knee radiographs) and Multicenter Osteoarthritis Study (MOST) (3,644 knee radiographs) datasets. A fully automatic method to locate the most informative region from subchondral bone using adaptive segmentation was developed. Subsequently, we built logistic regression models to identify and compare the performances of several texture descriptors and each ROI placement method using 5-fold cross validation. Importantly, we also investigated the generalizability of our approach by training the models on OAI and testing them on MOST dataset. We used area under the receiver operating characteristic curve (ROC AUC) and average precision (AP) obtained from the precision-recall (PR) curve to compare the results. RESULTS: We found that the adaptive ROI improves the classification performance (OA vs non-OA) over the commonly-used standard ROI (up to 9% percent increase in AUC). We also observed that, from all texture parameters, Local Binary Pattern (LBP) yielded the best performance in all settings with the best AUC of 0.840 [0.825, 0.852] and associated AP of 0.804 [0.786, 0.820]. CONCLUSION: Compared to the current state-of-the-art approaches, our results suggest that the proposed adaptive ROI approach in texture analysis of subchondral bone can increase the diagnostic performance for detecting the presence of radiographic OA.

Bone texture analysis for prediction of incident radiographic hip osteoarthritis using machine learning: data from the Cohort Hip and Cohort Knee (CHECK) study.

OBJECTIVE: To assess the ability of radiography-based bone texture variables in proximal femur and acetabulum to predict incident radiographic hip osteoarthritis (rHOA) over a 10 years period. DESIGN: Pelvic radiographs from CHECK at baseline (987 hips) were analyzed for bone texture using fractal signature analysis (FSA) in proximal femur and acetabulum. Elastic net (machine learning) was used to predict the incidence of rHOA (including Kellgren-Lawrence grade (KL) ≥ 2 or total hip replacement (THR)), joint space narrowing score (JSN, range 0-3), and osteophyte score (OST, range 0-3) after 10 years. Performance of prediction models was assessed using the area under the receiver operating characteristic curve (ROC AUC). RESULTS: Of the 987 hips without rHOA at baseline, 435 (44%) had rHOA at 10-year follow-up. Of the 667 hips with JSN grade 0 at baseline, 471 (71%) had JSN grade ≥ 1 at 10-year follow-up. Of the 613 hips with OST grade 0 at baseline, 526 (86%) had OST grade ≥ 1 at 10-year follow-up. AUCs for the models including age, gender, and body mass index (BMI) to predict incident rHOA, JSN, and OST were 0.59, 0.54, and 0.51, respectively. The inclusion of bone texture variables in the models improved the prediction of incident rHOA (ROC AUC 0.68 and 0.71 when baseline KL was also included in the model) and JSN (ROC AUC 0.62), but not incident OST (ROC AUC 0.52). CONCLUSION: Bone texture analysis provides additional information for predicting incident rHOA or THR over 10 years.

Association between radiography-based subchondral bone structure and MRI-based cartilage composition in postmenopausal women with mild osteoarthritis.

OBJECTIVE: Our aim was to investigate the relation between radiograph-based subchondral bone structure and cartilage composition assessed with delayed gadolinium enhanced magnetic resonance imaging of cartilage (dGEMRIC) and T2 relaxation time. DESIGN: Ninety-three postmenopausal women (Kellgren-Lawrence grade 0: n = 13, 1: n = 26, 2: n = 54) were included. Radiograph-based bone structure was assessed using entropy of the Laplacian-based image (ELap) and local binary patterns (ELBP), homogeneity indices of the local angles (HIAngles,mean, HIAngles,Perp, HIAngles,Paral), and horizontal (FDHor) and vertical fractal dimensions (FDVer). Mean dGEMRIC index and T2 relaxation time of tibial cartilage were calculated to estimate cartilage composition. RESULTS: HIAngles,mean (rs = -0.22) and HIAngles,Paral (rs = -0.24) in medial subchondral bone were related (P < 0.05) to dGEMRIC index of the medial tibial cartilage. ELap (rs = -0.23), FDHor,0.34 mm (r = 0.21) and FDVer,0.68 mm (r = 0.24) in medial subchondral bone were related (P < 0.05) to T2 relaxation time values of the medial tibial cartilage. FDHor at different scales in lateral subchondral bone were related (P < 0.01) to dGEMRIC index (r = 0.29-0.41) and T2 values of lateral tibial cartilage (r = -0.28 to -0.36). FDVer at larger scales were related (P < 0.05) to dGEMRIC index (r = 0.24-0.25) and T2 values of lateral tibial cartilage (r = -0.21). HIAngles,Paral (r = -0.25) and FDVer,0.68 mm (rs = 0.22) in the lateral tibial trabecular bone were related (P < 0.05) to dGEMRIC index of the lateral tibial cartilage. CONCLUSION: Our results support the presumption that several tissues are affected in the early osteoarthritis (OA). Furthermore, they indicate that the detailed analysis of radiographs may serve as a complementary imaging tool for OA studies.

Quantification of differences in bone texture from plain radiographs in knees with and without osteoarthritis.

OBJECTIVE: To quantify differences in bone texture between subjects with different stages of knee osteoarthritis (OA) and age- and gender-matched controls from plain radiographs using advanced image analysis methods. DESIGN: Altogether 203 knees were imaged using constant X-ray parameters and graded according to Kellgren-Lawrence (KL) grading scale (KL0: n = 110, KL1: n = 28, KL2: n = 27, KL3: n = 31, KL4: n = 7). Bone density-related and structure-related parameters were calculated from medial and lateral tibial subchondral bone plate and trabecular bone and from femur. Density-related parameters were derived from grayscale values and structure-related parameters from Laplacian- and local binary patterns (LBP)-based images. RESULTS: Reproducibilities of structure-related parameters were better than bone density-related parameters. Bone density-related parameters were significantly (P < 0.05) higher in KL2-4 groups than in control group (KL0) in medial tibial subchondral bone plate and trabecular bone. LBP-based structure parameters differed significantly between KL0 and KL2-4 groups in medial subchondral bone plate, between KL0 and KL1-4 groups in medial and lateral trabecular bone, and between KL0 and KL1-4/KL2-4 in medial and lateral femur. Laplacian-based parameters differed significantly between KL0 and KL2-4 groups in medial side regions-of-interest (ROIs). CONCLUSIONS: Our results indicate that the changes in bone texture in knee OA can be quantitatively evaluated from plain radiographs using advanced image analysis. Based on the results, increased bone density can be directly estimated if the X-ray imaging conditions are constant between patients. However, structural analysis of bone was more reproducible than direct evaluation of grayscale values, and is therefore better suited for quantitative analysis when imaging conditions are variable.

Arthroscopic ultrasound technique for simultaneous quantitative assessment of articular cartilage and subchondral bone: an in vitro and in vivo feasibility study.

Traditional arthroscopic examination is subjective and poorly reproducible. Recently, we introduced an arthroscopic ultrasound method for quantitative diagnostics of cartilage lesions. Here we describe our investigation of the feasibility of ultrasound arthroscopy for simultaneous measurements of articular cartilage and subchondral bone. Human osteochondral samples (n = 13) were imaged using a clinical 9-MHz ultrasound system. Ultrasound reflection coefficients (R, IRC), the ultrasound roughness index (URI) and the apparent integrated backscattering coefficient (AIB) were determined for both tissues. Mechanical testing, histologic analyses and micro-scale computed tomography imaging were the reference methods. Ultrasound arthroscopies were conducted on two patients. The ultrasound reflection coefficient correlated with the Mankin score and Young's modulus of cartilage (|r| > 0.56, p < 0.05). Ultrasound parameters (R, IRC, AIB) for subchondral bone correlated with the bone surface/volume ratio (|r| > 0.70, p < 0.05) and trabecular thickness (|r| > 0.59, p < 0.05). Furthermore, R and subchondral bone mineral density were significantly correlated (|r| > 0.65, p < 0.05). Arthroscopic ultrasound examination provided diagnostically valuable information on cartilage and subchondral bone in vivo.

In vivo comparison of delayed gadolinium-enhanced MRI of cartilage and delayed quantitative CT arthrography in imaging of articular cartilage.

OBJECTIVE: To compare delayed gadolinium-enhanced magnetic resonance imaging (MRI) of cartilage (dGEMRIC) and delayed quantitative computed tomography (CT) arthrography (dQCTA) to each other, and their association to arthroscopy. Additionally, the relationship between dGEMRIC with intravenous (dGEMRIC(IV)) and intra-articular contrast agent administration (dGEMRIC(IA)) was determined. DESIGN: Eleven patients with knee pain were scanned at 3 T MRI and 64-slice CT before arthroscopy. dQCTA was performed at 5 and 45 min after intra-articular injection of ioxaglate. Both dGEMRIC(IV) and dGEMRIC(IA) were performed at 90 min after gadopentetate injection. dGEMRIC indices and change in relaxation rates (ΔR(1)) were separately calculated for dGEMRIC(IV) and dGEMRIC(IA). dGEMRIC and dQCTA parameters were calculated for predetermined sites at the knee joint that were International Cartilage Repair Society (ICRS) graded in arthroscopy. RESULTS: dQCTA normalized with the contrast agent concentration in synovial fluid (SF) and dGEMRIC(IV) correlated significantly, whereas dGEMRIC(IA) correlated with the normalized dQCTA only when dGEMRIC(IA) was also normalized with the contrast agent concentration in SF. Correlation was strongest between normalized dQCTA at 45 min and ΔR(1,IV) (r(s) = 0.72 [95% CI 0.56-0.83], n = 49, P < 0.01) and ΔR(1,IA) normalized with ΔR(1) in SF (r(s) = 0.70 [0.53-0.82], n = 52, P < 0.01). Neither dGEMRIC nor dQCTA correlated with arthroscopic grading. dGEMRIC(IV) and non-normalized dGEMRIC(IA) were not related while ΔR(1,IV) correlated with normalized ΔR(1,IA) (r(s) = 0.52 [0.28-0.70], n = 50, P < 0.01). CONCLUSIONS: This study suggests that dQCTA is in best agreement with dGEMRIC(IV) at 45 min after CT contrast agent injection. dQCTA and dGEMRIC were not related to arthroscopy, probably because the remaining cartilage is analysed in dGEMRIC and dQCTA, whereas in arthroscopy the absence of cartilage defines the grading. The findings indicate the importance to take into account the contrast agent concentration in SF in dQCTA and dGEMRIC(IA).