Subchondral tibial bone texture predicts the incidence of radiographic knee osteoarthritis: data from the Osteoarthritis Initiative

Diagnosing the Severity of Knee Osteoarthritis Using Regression Scores From Artificial Intelligence Convolution Neural Networks

BACKGROUND

This study focused on using deep learning neural networks to classify the severity of osteoarthritis in the knee. A continuous regression score of osteoarthritis severity has yet to be explored using artificial intelligence machine learning, which could offer a more nuanced assessment of osteoarthritis.

MATERIALS AND METHODS

This study used 8260 radiographic images from The Osteoarthritis Initiative to develop and assess four neural network models (VGG16, EfficientNetV2 small, ResNet34, and DenseNet196). Each model generated a regressor score of the osteoarthritis severity based on Kellgren-Lawrence grading scale criteria. Primary performance outcomes assessed were area under the curve (AUC), accuracy, and mean absolute error (MAE) for each model. Secondary outcomes evaluated were precision, recall, and F-1 score.

RESULTS

The EfficientNet model architecture yielded the strongest AUC (0.83), accuracy (71%), and MAE (0.42) compared with VGG16 (AUC: 0.74; accuracy: 57%; MAE: 0.54), ResNet34 (AUC: 0.76; accuracy: 60%; MAE: 0.53), and DenseNet196 (AUC: 0.78; accuracy: 62%; MAE: 0.49).

CONCLUSION

Convolutional neural networks offer an automated and accurate way to quickly assess and diagnose knee radiographs for osteoarthritis. The regression score models evaluated in this study demonstrated superior AUC, accuracy, and MAE compared with standard convolutional neural network models. The EfficientNet model exhibited the best overall performance, including the highest AUC (0.83) noted in the literature. The artificial intelligence-generated regressor exhibits a finer progression of knee osteoarthritis by quantifying severity of various hallmark features. Potential applications for this technology include its use as a screening tool in determining patient suitability for orthopedic referral. [Orthopedics. 202x;4x(x):xx-xx.].

Associating Knee Osteoarthritis Progression with Temporal-Regional Graph Convolutional Network Analysis on MR Images

BACKGROUND

Artificial intelligence shows promise in assessing knee osteoarthritis (OA) progression on MR images, but faces challenges in accuracy and interpretability.

PURPOSE

To introduce a temporal-regional graph convolutional network (TRGCN) on MR images to study the association between knee OA progression status and network outcome.

STUDY TYPE

Retrospective.

POPULATION

194 OA progressors (mean age, 62 ± 9 years) and 406 controls (mean age, 61 ± 9 years) from the OA Initiative were randomly divided into training (80%) and testing (20%) cohorts.

FIELD STRENGTH/SEQUENCE

Sagittal 2D IW-TSE-FS (IW) and 3D-DESS-WE (DESS) at 3T.

ASSESSMENT

Anatomical subregions of cartilage, subchondral bone, meniscus, and the infrapatellar fat pad at baseline, 12-month, and 24-month were automatically segmented and served as inputs to form compartment-based graphs for a TRGCN model, which containing both regional and temporal information. The performance of models based on (i) clinical variables alone, (ii) radiologist score alone, (iii) combined features (containing i and ii), (iv) composite TRGCN (combining TRGCN, i and ii), (v) radiomics features, (vi) convolutional neural network based on Densenet-169 were compared.

STATISTICAL TESTS

DeLong test was performed to compare the areas under the ROC curve (AUC) of all models. Additionally, interpretability analysis was done to evaluate the contributions of individual regions. A P value <0.05 was considered significant.

RESULTS

The composite TRGCN outperformed all other models with AUCs of 0.841 (DESS) and 0.856 (IW) in the testing cohort (all P < 0.05). Interpretability analysis highlighted cartilage's importance over other structures (42%-45%), tibiofemoral joint's (TFJ) dominance over patellofemoral joint (PFJ) (58%-67% vs. 12%-37%), and importance scores changes in compartments over time (TFJ vs. PFJ: baseline: 44% vs. 43%, 12-month: 52% vs. 39%, 24-month: 31% vs. 48%).

DATA CONCLUSION

The composite TRGCN, capturing temporal and regional information, demonstrated superior discriminative ability compared with other methods, providing interpretable insights for identifying knee OA progression.

LEVEL OF EVIDENCE: 4

TECHNICAL EFFICACY

Stage 2.

A Comprehensive Review on MRI-based Knee Joint Segmentation and Analysis Techniques

Using magnetic resonance imaging (MRI) in osteoarthritis pathogenesis research has proven extremely beneficial. However, it is always challenging for both clinicians and researchers to detect morphological changes in knee joints from magnetic resonance (MR) imaging since the surrounding tissues produce identical signals in MR studies, making it difficult to distinguish between them. Segmenting the knee bone, articular cartilage and menisci from the MR images allows one to examine the complete volume of the bone, articular cartilage, and menisci. It can also be used to assess certain characteristics quantitatively. However, segmentation is a laborious and time-consuming operation that requires sufficient training to complete correctly. With the advancement of MRI technology and computational methods, researchers have developed several algorithms to automate the task of individual knee bone, articular cartilage and meniscus segmentation during the last two decades. This systematic review aims to present available fully and semi-automatic segmentation methods for knee bone, cartilage, and meniscus published in different scientific articles. This review provides a vivid description of the scientific advancements to clinicians and researchers in this field of image analysis and segmentation, which helps the development of novel automated methods for clinical applications. The review also contains the recently developed fully automated deep learning-based methods for segmentation, which not only provides better results compared to the conventional techniques but also open a new field of research in Medical Imaging.

Machine learning and deep neural network-based learning in osteoarthritis knee

Osteoarthritis (OA) of the knee joint is considered the commonest musculoskeletal condition leading to marked disability for patients residing in various regions around the globe. Application of machine learning (ML) in doing research regarding OA has brought about various clinical advances viz, OA being diagnosed at preliminary stages, prediction of chances of development of OA among the population, discovering various phenotypes of OA, calculating the severity in OA structure and also discovering people with slow and fast progression of disease pathology, etc. Various publications are available regarding machine learning methods for the early detection of osteoarthritis. The key features are detected by morphology, molecular architecture, and electrical and mechanical functions. In addition, this particular technique was utilized to assess non-interfering, non-ionizing, and in-vivo techniques using magnetic resonance imaging. ML is being utilized in OA, chiefly with the formulation of large cohorts viz, the OA Initiative, a cohort observational study, the Multi-centre Osteoarthritis Study, an observational, prospective longitudinal study and the Cohort Hip & Cohort Knee, an observational cohort prospective study of both hip and knee OA. Though ML has various contributions and enhancing applications, it remains an imminent field with high potential, also with its limitations. Many more studies are to be carried out to find more about the link between machine learning and knee osteoarthritis, which would help in the improvement of making decisions clinically, and expedite the necessary interventions.

Short-term variations in trabecular bone texture parameters associated to radio-clinical biomarkers improve the prediction of radiographic knee osteoarthritis progression

The present study aims to examine whether the short-term variations in trabecular bone texture (TBT) parameters, combined with a targeted set of clinical and radiographic data, would improve the prediction of long-term radiographic knee osteoarthritis (KOA) progression. Longitudinal (baseline, 24 and 48-month) data, obtained from the Osteoarthritis Initiative cohort, were available for 1352 individuals, with preexisting OA (1 < Kellgren-Lawrence < 4) at baseline. KOA progression was defined as an increase in the medial joint space narrowing score from the 24-months to the 48-months control point. 16 regions of interest were automatically selected from each radiographic knee and analyzed using fractal dimension. Variations from baseline to 24 months in TBT descriptors as well as selected radiographic and clinical readings were calculated. Different logistic regression models were developed to evaluate the progression prediction performance when associating TBT variations with the selected clinical and radiographic readings. The most predictive model was mainly determined using the area under the receiver operating characteristic curve (AUC). The proposed prediction model including short-term variations in TBT parameters, associated with clinical covariates and radiographic scores, improved the capacity of predicting long-term radiographic KOA progression (AUC of 0.739), compared to models based solely on baseline values (AUC of 0.676, p-value < 0.008).

A discriminative shape-texture convolutional neural network for early diagnosis of knee osteoarthritis from X-ray images

Knee Osteoarthritis (OA) is one of the most common causes of physical disability worldwide associated with a significant personal and socioeconomic burden. Deep Learning approaches based on Convolutional Neural Networks (CNNs) achieved remarkable improvements in knee OA detection. Despite this success, the problem of early knee OA diagnosis from plain radiographs remains a challenging task. This is due to the high similarity between the X-ray images of OA and non-OA subjects and the disappearance of texture information regarding bone microarchitecture changes in the top layers during the learning process of the CNN models. To address these issues, we propose a Discriminative Shape-Texture Convolutional Neural Network (DST-CNN), which automatically diagnoses early knee OA from X-ray images. The proposed model incorporates a discriminative loss to improve class separability and deal with high inter-class similarities. In addition, a new Gram Matrix Descriptor (GMD) block is embedded in the CNN architecture to compute texture features from several intermediate layers and combine them with the shape features in the top layers. We show that merging texture features with deep ones leads to better prediction of the early stages of OA. Comprehensive experimental results on two large public databases, Osteoarthritis Initiative (OAI) and Multicenter Osteoarthritis Study (MOST) demonstrate the potential of the proposed network. Ablation studies and visualizations are provided for a detailed understanding of our proposed approach.

Radiographic Biomarkers for Knee Osteoarthritis: A Narrative Review

Conventional radiography remains the most widely available imaging modality in clinical practice in knee osteoarthritis. Recent research has been carried out to develop novel radiographic biomarkers to establish the diagnosis and to monitor the progression of the disease. The growing number of publications on this topic over time highlights the necessity of a renewed review. Herein, we propose a narrative review of a selection of original full-text articles describing human studies on radiographic imaging biomarkers used for the prediction of knee osteoarthritis-related outcomes. To achieve this, a PubMed database search was used. A total of 24 studies were obtained and then classified based on three outcomes: (1) prediction of radiographic knee osteoarthritis incidence, (2) knee osteoarthritis progression and (3) knee arthroplasty risk. Results showed that numerous studies have reported the relevance of joint space narrowing score, Kellgren-Lawrence score and trabecular bone texture features as potential bioimaging markers in the prediction of the three outcomes. Performance results of reviewed prediction models were presented in terms of the area under the receiver operating characteristic curves. However, fair and valid comparisons of the models' performance were not possible due to the lack of a unique definition of each of the three outcomes.

Prediction Models to Estimate the Future Risk of Osteoarthritis in the General Population: A Systematic Review

OBJECTIVE

To evaluate the performance and applicability of multivariable prediction models for osteoarthritis (OA).

METHODS

This was a systematic review and narrative synthesis using 3 databases (EMBASE, PubMed, and Web of Science) from inception to December 2021. We included general population longitudinal studies reporting derivation, comparison, or validation of multivariable models to predict individual risk of OA incidence, defined by recognized clinical or imaging criteria. We excluded studies reporting prevalent OA and joint arthroplasty outcome. Paired reviewers independently performed article selection, data extraction, and risk-of-bias assessment. Model performance, calibration, and retained predictors were summarized.

RESULTS

A total of 26 studies were included, reporting 31 final multivariable prediction models for incident knee (23), hip (4), hand (3) and any-site OA (1), with a median of 121.5 (range 27-12,803) outcome events, a median prediction horizon of 8 years (range 2-41), and a median of 6 predictors (range 3-24). Age, body mass index, previous injury, and occupational exposures were among the most commonly included predictors. Model discrimination after validation was generally acceptable to excellent (area under the curve = 0.70-0.85). Either internal or external validation processes were used in most models, although the risk of bias was often judged to be high with limited applicability to mass application in diverse populations.

CONCLUSION

Despite growing interest in multivariable prediction models for incident OA, focus remains predominantly on the knee, with reliance on data from a small pool of appropriate cohort data sets, and concerns over general population applicability.

MRI-based Texture Analysis of Infrapatellar Fat Pad to Predict Knee Osteoarthritis Incidence

Background Infrapatellar fat pad (IPFP) quality has been implicated as a marker for predicting knee osteoarthritis (KOA); however, no valid quantification for subtle IPFP abnormalities has been established. Purpose To investigate whether MRI-based three-dimensional texture analysis of IPFP abnormalities could help predict incident radiographic KOA. Materials and Methods In this prospective nested case-control study, 690 participants whose knees were at risk for KOA were included from the Pivotal Osteoarthritis Initiative MRI Analyses incident osteoarthritis cohort. All knees had a Kellgren-Lawrence grade of 1 or less at baseline. During the 4-year follow-up, case participants were matched 1:1 to control participants, with incident radiographic KOA as the outcome. MRI scans were segmented at the incident time point of KOA (hereafter, P0), 1 year before P0 (hereafter, P-1), and baseline. MRI-based three-dimensional texture analysis was performed to extract IPFP texture features. Least absolute shrinkage and selection operator and multivariable logistic regressions were applied in the development cohort and evaluated in the test cohort. The area under the receiver operating characteristic curve (AUC) was used to evaluate the discriminative value of the clinical score, IPFP texture score, and MRI Osteoarthritis Knee Score. Results Participants were allocated to development (n = 500, 340 women; mean age, 60 years) and test (n = 190, 120 women; mean age, 61 years) cohorts. In both cohorts, IPFP texture scores (AUC ≥0.75 for all) showed greater discrimination than clinical scores (AUC ≤0.69 for all) at baseline, P-1, and P0, with significant differences in pairwise comparisons (P ≤ .002 for all). Greater predictive and concurrent validities of IPFP texture scores (AUC ≥0.75 for all) compared with MRI Osteoarthritis Knee Scores (AUC ≤0.66 for all) were also demonstrated (P < .001 for all). Conclusion MRI-based three-dimensional texture of the infrapatellar fat pad was associated with future development of knee osteoarthritis. ClinicalTrials.gov registration no.: NCT00080171 © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Fischer in this issue.

Subchondral tibial bone texture of conventional X-rays predicts total knee arthroplasty

Lacking disease-modifying osteoarthritis drugs (DMOADs) for knee osteoarthritis (KOA), Total Knee Arthroplasty (TKA) is often considered an important clinical outcome. Thus, it is important to determine the most relevant factors that are associated with the risk of TKA. The present study aims to develop a model based on a combination of X-ray trabecular bone texture (TBT) analysis, and clinical and radiological information to predict TKA risk in patients with or at risk of developing KOA. This study involved 4382 radiographs, obtained from the OsteoArthritis Initiative (OAI) cohort. Cases were defined as patients with TKA on at least one knee prior to the 108-month follow-up time point and controls were defined as patients who had never undergone TKA. The proposed TKA-risk prediction model, combining TBT parameters and Kellgren–Lawrence (KL) grades, was performed using logistic regression. The proposed model achieved an AUC of 0.92 (95% Confidence Interval [CI] 0.90, 0.93), while the KL model achieved an AUC of 0.86 (95% CI 0.84, 0.86; p < 0.001). This study presents a new TKA prediction model with a good performance permitting the identification of at risk patient with a good sensitivy and specificity, with a 60% increase in TKA case prediction as reflected by the recall values.

Study of the Efficacy of Artificial Intelligence Algorithm-Based Analysis of the Functional and Anatomical Improvement in Polynucleotide Treatment in Knee Osteoarthritis Patients: A Prospective Case Series

Knee osteoarthritis (OA) is one of the most common degenerative diseases in old age. Recent studies have suggested new treatment approaches dealing with subchondral remodeling, which is a typical feature of OA progression. However, diagnostic tools or therapeutic approaches related to such a process are still being researched. The automated artificial intelligence (AI) algorithm-based texture analysis is a new method used for OA-progression detection. We designed a prospective case series study to examine the efficacy of the AI algorithm-based texture analysis in detecting the restoration of the subchondral remodeling process, which is expected to follow therapeutic intervention. In this study, we used polynucleotide (PN) filler injections as the therapeutic modality and the treatment outcome was verified by symptom improvement, as well as by the induction of subchondral microstructural changes. We used AI algorithm-based texture analysis to observe these changes in the subchondral bone with the bone structure value (BSV). A total of 51 participants diagnosed with knee OA were enrolled in this study. Intra-articular PN filler (HP cell Vitaran J) injections were administered once a week and five times in total. Knee X-rays and texture analyses with BSVs were performed during the screening visit and the last visit three months after screening. The Visual Analogue Scale (VAS) and Korean-Western Ontario MacMaster (K-WOMAC) measurements were used at the screening visit, the fifth intra-articular injection visit, and the last visit. The VAS and K-WOMAC scores decreased after PN treatment and lasted for three months after the final injection. The BSV changed in the middle and deep layers of tibial bone after PN injection. This result could imply that there were microstructural changes in the subchondral bone after PN treatment, and that this change could be detected using the AI algorithm-based texture analysis. In conclusion, the AI- algorithm-based texture analysis could be a promising tool for detecting and assessing the therapeutic outcome in knee OA.

Prediction of knee osteoarthritis progression using radiological descriptors obtained from bone texture analysis and Siamese neural networks: data from OAI and MOST cohorts

BACKGROUND

Trabecular bone texture (TBT) analysis has been identified as an imaging biomarker that provides information on trabecular bone changes due to knee osteoarthritis (KOA). In parallel with the improvement in medical imaging technologies, machine learning methods have received growing interest in the scientific osteoarthritis community to potentially provide clinicians with prognostic data from conventional knee X-ray datasets, in particular from the Osteoarthritis Initiative (OAI) and the Multicenter Osteoarthritis Study (MOST) cohorts.

PATIENTS AND METHODS

This study included 1888 patients from OAI and 683 patients from MOST cohorts. Radiographs were automatically segmented to determine 16 regions of interest. Patients with an early stage of OA risk, with Kellgren and Lawrence (KL) grade of 1 < KL < 4, were selected. The definition of OA progression was an increase in the OARSI medial joint space narrowing (mJSN) grades over 48 months in OAI and 60 months in MOST. The performance of the TBT-CNN model was evaluated and compared to well-known prediction models using logistic regression.

RESULTS

The TBT-CNN model was predictive of the JSN progression with an area under the curve (AUC) up to 0.75 in OAI and 0.81 in MOST. The predictive ability of the TBT-CNN model was invariant with respect to the acquisition modality or image quality. The prediction models performed significantly better with estimated KL (KLprob) grades than those provided by radiologists. TBT-based models significantly outperformed KLprob-based models in MOST and provided similar performances in OAI. In addition, the combined model, when trained in one cohort, was able to predict OA progression in the other cohort.

CONCLUSION

The proposed combined model provides a good performance in the prediction of mJSN over 4 to 6 years in patients with relevant KOA. Furthermore, the current study presents an important contribution in showing that TBT-based OA prediction models can work with different databases.

Machine learning based texture analysis of patella from X-rays for detecting patellofemoral osteoarthritis

OBJECTIVE

To assess the ability of texture features for detecting radiographic patellofemoral osteoarthritis (PFOA) from knee lateral view radiographs.

DESIGN

We used lateral view knee radiographs from The Multicenter Osteoarthritis Study (MOST) public use datasets (n  = 5507 knees). Patellar region-of-interest (ROI) was automatically detected using landmark detection tool (BoneFinder), and subsequently, these anatomical landmarks were used to extract three different texture ROIs. Hand-crafted features, based on Local Binary Patterns (LBP), were then extracted to describe the patellar texture. First, a machine learning model (Gradient Boosting Machine) was trained to detect radiographic PFOA from the LBP features. Furthermore, we used end-to-end trained deep convolutional neural networks (CNNs) directly on the texture patches for detecting the PFOA. The proposed classification models were eventually compared with more conventional reference models that use clinical assessments and participant characteristics such as age, sex, body mass index (BMI), the total Western Ontario and McMaster Universities Arthritis Index (WOMAC) score, and tibiofemoral Kellgren-Lawrence (KL) grade. Atlas-guided visual assessment of PFOA status by expert readers provided in the MOST public use datasets was used as a classification outcome for the models. Performance of prediction models was assessed using the area under the receiver operating characteristic curve (ROC AUC), the area under the precision-recall (PR) curve -average precision (AP)-, and Brier score in the stratified 5-fold cross validation setting.

RESULTS

Of the 5507 knees, 953 (17.3%) had PFOA. AUC and AP for the strongest reference model including age, sex, BMI, WOMAC score, and tibiofemoral KL grade to predict PFOA were 0.817 and 0.487, respectively. Textural ROI classification using CNN significantly improved the prediction performance (ROC AUC = 0.889, AP = 0.714).

CONCLUSION

We present the first study that analyses patellar bone texture for diagnosing PFOA. Our results demonstrates the potential of using texture features of patella to predict PFOA.

Trabecular bone texture analysis of conventional radiographs in the assessment of knee osteoarthritis: review and viewpoint

BACKGROUND

Trabecular bone texture analysis (TBTA) has been identified as an imaging biomarker that provides information on trabecular bone changes due to knee osteoarthritis (KOA). Consequently, it is important to conduct a comprehensive review that would permit a better understanding of this unfamiliar image analysis technique in the area of KOA research. We examined how TBTA, conducted on knee radiographs, is associated to (i) KOA incidence and progression, (ii) total knee arthroplasty, and (iii) KOA treatment responses. The primary aims of this study are twofold: to provide (i) a narrative review of the studies conducted on radiographic KOA using TBTA, and (ii) a viewpoint on future research priorities.

METHOD

Literature searches were performed in the PubMed electronic database. Studies published between June 1991 and March 2020 and related to traditional and fractal image analysis of trabecular bone texture (TBT) on knee radiographs were identified.

RESULTS

The search resulted in 219 papers. After title and abstract scanning, 39 studies were found eligible and then classified in accordance to six criteria: cross-sectional evaluation of osteoarthritis and non-osteoarthritis knees, understanding of bone microarchitecture, prediction of KOA progression, KOA incidence, and total knee arthroplasty and association with treatment response. Numerous studies have reported the relevance of TBTA as a potential bioimaging marker in the prediction of KOA incidence and progression. However, only a few studies have focused on the association of TBTA with both OA treatment responses and the prediction of knee joint replacement.

CONCLUSION

Clear evidence of biological plausibility for TBTA in KOA is already established. The review confirms the consistent association between TBT and important KOA endpoints such as KOA radiographic incidence and progression. TBTA could provide markers for enrichment of clinical trials enhancing the screening of KOA progressors. Major advances were made towards a fully automated assessment of KOA.

A machine learning approach to distinguish between knees without and with osteoarthritis using MRI-based radiomic features from tibial bone

Objectives

Our aim was to assess the ability of semi-automatically extracted magnetic resonance imaging (MRI)–based radiomic features from tibial subchondral bone to distinguish between knees without and with osteoarthritis.

Methods

The right knees of 665 females from the population-based Rotterdam Study scanned with 1.5T MRI were analyzed. A fast imaging employing steady-state acquisition sequence was used for the quantitative bone analyses. Tibial bone was segmented using a method that combines multi-atlas and appearance models. Radiomic features related to the shape and texture were calculated from six volumes of interests (VOIs) in the proximal tibia. Machine learning–based Elastic Net models with 10-fold cross-validation were used to distinguish between knees without and with MRI Osteoarthritis Knee Score (MOAKS)–based tibiofemoral osteoarthritis. Performance of the covariate (age and body mass index), image features, and combined covariate + image features models were assessed using the area under the receiver operating characteristic curve (ROC AUC).

Results

Of 665 analyzed knees, 76 (11.4%) had osteoarthritis. An ROC AUC of 0.68 (95% confidence interval (CI): 0.60–0.75) was obtained using the covariate model. The image features model yielded an ROC AUC of 0.80 (CI: 0.73–0.87). The model that combined image features from all VOIs and covariates yielded an ROC AUC of 0.80 (CI: 0.73–0.87).

Conclusion

Our results suggest that radiomic features are useful imaging biomarkers of subchondral bone for the diagnosis of osteoarthritis. An advantage of assessing bone on MRI instead of on radiographs is that other tissues can be assessed simultaneously.

Key Points

• Subchondral bone plays a role in the osteoarthritis disease processes.

• MRI radiomics is a potential method for quantifying changes in subchondral bone.

• Semi-automatically extracted radiomic features of tibia differ between subjects without and with osteoarthritis.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00330-021-07951-5.

Analyzing Femorotibial Cartilage Thickness Using Anatomically Standardized Maps: Reproducibility and Reference Data

Alterations in cartilage thickness (CTh) are a hallmark of knee osteoarthritis, which remain difficult to characterize at high resolution, even with modern magnetic resonance imaging (MRI), due to a paucity of standardization tools. This study aimed to assess a computational anatomy method producing standardized two-dimensional femorotibial CTh maps. The method was assessed with twenty knees, processed following three common experimental scenarios. Cartilage thickness maps were obtained for the femorotibial cartilages by reconstructing bone and cartilage mesh models in tree-dimension, calculating three-dimensional CTh maps, and anatomically standardizing the maps. The intra-operator accuracy (median (interquartile range, IQR) of -0.006 (0.045) mm), precision (0.152 (0.070) mm), entropy (7.02 (0.71) and agreement (0.975 (0.020))) results suggested that the method is adequate to capture the spatial variations in CTh and compare knees at varying osteoarthritis stages. The lower inter-operator precision (0.496 (0.132) mm) and agreement (0.808 (0.108)) indicate a possible loss of sensitivity to detect differences in a setting with multiple operators. The results confirmed the promising potential of anatomically standardized maps, with the lower inter-operator reproducibility stressing the need to coordinate operators. This study also provided essential reference data and indications for future research using CTh maps.

Bone Structure Analysis of the Radius Using Ultrahigh Field (7T) MRI: Relevance of Technical Parameters and Comparison with 3T MRI and Radiography

Bone fractal signature analysis (FSA-also termed bone texture analysis) is a tool that assesses structural changes that may relate to clinical outcomes and functions. Our aim was to compare bone texture analysis of the distal radius in patients and volunteers using radiography and 3T and 7T magnetic resonance imaging (MRI)-a patient group (n = 25) and a volunteer group (n = 25) were included. Participants in the patient group had a history of chronic wrist pain with suspected or confirmed osteoarthritis and/or ligament instability. All participants had 3T and 7T MRI including T1-weighted turbo spin echo (TSE) sequences. The 7T MRI examination included an additional high-resolution (HR) T1 TSE sequence. Radiographs of the wrist were acquired for the patient group. When comparing patients and volunteers (unadjusted for gender and age), we found a statistically significant difference of horizontal and vertical fractal dimensions (FDs) using 7T T1 TSE-HR images in low-resolution mode (horizontal: p = 0.04, vertical: p = 0.01). When comparing radiography to the different MRI sequences, we found a statistically significant difference for low- and high-resolution horizontal FDs between radiography and 3T T1 TSE and 7T T1 TSE-HR. Vertical FDs were significantly different only between radiographs and 3T T1 TSE in the high-resolution mode; FSA measures obtained from 3T and 7T MRI are highly dependent on the sequence and reconstruction resolution used, and thus are not easily comparable between MRI systems and applied sequences.

Patients with lateral compartment knee osteoarthritis during arthroscopy are at highest risk of subsequent knee arthroplasty

BACKGROUND

Arthroscopic treatment of knee osteoarthritis has declined, in part due to concerns with conversion to arthroplasty. Some studies have investigated the demographic predictors for conversion to arthroplasty, few have assessed the risk factors within the knee itself. Our aim was to analyse the demographics and anatomical wear features of a large cohort of patients undergoing knee arthroscopy.

METHODS

A retrospective analysis of 1760 cases spanning over 17 years undergoing knee arthroscopy was performed. Patients were 36 years or older at time of the index arthroscopy. Each patient received the International Cartilage Regeneration and Joint Preservation Society (ICRS) grade of all regions as well an estimate of the remaining meniscal percentage. Demographic factors as well as intraoperatively collected data were analysed as predictive variables for subsequent conversion to arthroplasty using a multi-step Cox regression analysis.

RESULTS

A total of 102 patients (6.2%) were converted to arthroplasty. Age at arthroscopy (hazard ratio (HR) 1.073; 95% confidence interval (CI) 1.058-1.088) and ICRS grade of the lateral tibial plateau (HR 1.166; 95% CI 1.066-1.276) were statistically significant predicting variables for conversion to arthroplasty.

CONCLUSIONS

The results of this study indicate that higher ICRS grade of the lateral tibial plateau at arthroscopy is the most significant predictor for conversion to knee arthroplasty, with a hazard equal to an increase in age. The absence of these factors does not justify arthroscopic treatment of patients with knee osteoarthritis.

LEVEL OF EVIDENCE

III.

Discriminative Regularized Auto-Encoder for Early Detection of Knee OsteoArthritis: Data from the Osteoarthritis Initiative

OsteoArthritis (OA) is the most common disorder of the musculoskeletal system and the major cause of reduced mobility among seniors. The visual evaluation of OA still suffers from subjectivity. Recently, Computer-Aided Diagnosis (CAD) systems based on learning methods showed potential for improving knee OA diagnostic accuracy. However, learning discriminative properties can be a challenging task, particularly when dealing with complex data such as X-ray images, typically used for knee OA diagnosis. In this paper, we introduce a Discriminative Regularized Auto Encoder (DRAE) that allows to learn both relevant and discriminative properties that improve the classification performance. More specifically, a penalty term, called discriminative loss is combined with the standard Auto-Encoder training criterion. This additional term aims to force the learned representation to contain discriminative information. Our experimental results on data from the public multicenter OsteoArthritis Initiative (OAI) show that the developed method presents potential results for early knee OA detection.

A novel method for determining the Femoral-Tibial Angle of Knee Osteoarthritis on X-ray radiographs: data from the Osteoarthritis Initiative

Femoral-tibial alignment is a prominent risk factor for Knee Osteoarthritis (KOA) incidence and progression. One way of assessing alignment is by determining the Femoral-Tibial Angle (FTA). Several studies have investigated FTA determination; however, methods of assessment of FTA still present challenges. This paper introduces a new method for semi-automatic measurement of FTA as part of KOA research. Our novel approach combines preprocessing of X-ray images and the use of Active Shape Model (ASM) as the femoral and tibial segmentation method, followed by a thinning process. The result of the thinning process is used to predict FTA automatically by measuring the angle between the intersection of the two vectors of branching points on the femoral and tibial areas. The proposed method is trained on 10 x-ray images and tested on 50 different x-ray images of the Osteoarthritis Initiative (OAI) dataset. The outcomes of this approach were compared with manually obtained FTA measurements from the OAI dataset as the ground truth. Based on experiments, the difference in measurement results between the FTA of the OAI and the FTA obtained using our method is quite small, i.e., below 0.81° for the right FTA and below 0.77° for the left FTA with minimal average errors. This result indicates that this method is clinically suitable for semi-automatic measurement of the FTA.

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.

Deep learning risk assessment models for predicting progression of radiographic medial joint space loss over a 48-MONTH follow-up period

OBJECTIVE

To develop and evaluate deep learning (DL) risk assessment models for predicting the progression of radiographic medial joint space loss using baseline knee X-rays.

METHODS

Knees from the Osteoarthritis Initiative without and with progression of radiographic joint space loss (defined as ≥ 0.7 mm decrease in medial joint space width measurement between baseline and 48-month follow-up X-rays) were randomly stratified into training (1400 knees) and hold-out testing (400 knees) datasets. A DL network was trained to predict the progression of radiographic joint space loss using the baseline knee X-rays. An artificial neural network was used to develop a traditional model for predicting progression utilizing demographic and radiographic risk factors. A combined joint training model was developed using a DL network to extract information from baseline knee X-rays as a feature vector, which was further concatenated with the risk factor data vector. Area under the curve (AUC) analysis was performed using the hold-out test dataset to evaluate model performance.

RESULTS

The traditional model had an AUC of 0.660 (61.5% sensitivity and 64.0% specificity) for predicting progression. The DL model had an AUC of 0.799 (78.0% sensitivity and 75.5% specificity), which was significantly higher (P < 0.001) than the traditional model. The combined model had an AUC of 0.863 (80.5% sensitivity and specificity), which was significantly higher than the DL (P = 0.015) and traditional (P < 0.001) models.

CONCLUSION

DL models using baseline knee X-rays had higher diagnostic performance for predicting the progression of radiographic joint space loss than the traditional model using demographic and radiographic risk factors.

Multimodal Machine Learning-based Knee Osteoarthritis Progression Prediction from Plain Radiographs and Clinical Data

Knee osteoarthritis (OA) is the most common musculoskeletal disease without a cure, and current treatment options are limited to symptomatic relief. Prediction of OA progression is a very challenging and timely issue, and it could, if resolved, accelerate the disease modifying drug development and ultimately help to prevent millions of total joint replacement surgeries performed annually. Here, we present a multi-modal machine learning-based OA progression prediction model that utilises raw radiographic data, clinical examination results and previous medical history of the patient. We validated this approach on an independent test set of 3,918 knee images from 2,129 subjects. Our method yielded area under the ROC curve (AUC) of 0.79 (0.78–0.81) and Average Precision (AP) of 0.68 (0.66–0.70). In contrast, a reference approach, based on logistic regression, yielded AUC of 0.75 (0.74–0.77) and AP of 0.62 (0.60–0.64). The proposed method could significantly improve the subject selection process for OA drug-development trials and help the development of personalised therapeutic plans.

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.

Osteoarthritis year in review 2018: imaging

PURPOSE

To provide a narrative review of the most relevant original research published in 2017/2018 on osteoarthritis imaging.

METHODS

The PubMed database was used to recover all relevant articles pertaining to osteoarthritis and medical imaging published between 1 April 2017 and 31 March 2018. Review articles, case studies and in vitro or animal studies were excluded. The original publications were subjectively sorted based on relevance, novelty and impact.

RESULTS AND CONCLUSIONS

The publication search yielded 1,155 references. In the assessed publications, the most common imaging modalities were radiography (N = 708) and magnetic resonance imaging (MRI) (355), followed by computed tomography (CT) (220), ultrasound (85) and nuclear medicine (17). An overview of the most important publications to the osteoarthritis (OA) research community is presented in this narrative review. Imaging studies play an increasingly important role in OA research, and have helped us to understand better the pathophysiology of OA. Radiography and MRI continue to be the most applied imaging modalities, while quantitative MRI methods and texture analysis are becoming more popular. The value of ultrasound in OA research has been demonstrated. Several multi-modality predictive models have been developed. Deep learning has potential for more automatic and standardized analyses in future OA imaging research.

Bone Density and Texture from Minimally Post-Processed Knee Radiographs in Subjects with Knee Osteoarthritis

Plain radiography is the most common modality to assess the stage of osteoarthritis. Our aims were to assess the relationship of radiography-based bone density and texture between radiographs with minimal and clinical post-processing, and to compare the differences in bone characteristics between controls and subjects with knee osteoarthritis or medial tibial bone marrow lesions (BMLs). Tibial bone density and texture was evaluated from radiographs with both minimal and clinical post-processing in 109 subjects with and without osteoarthritis. Bone texture was evaluated using fractal signature analysis. Significant correlations (p < 0.001) were found in all regions (between 0.94 and 0.97) for calibrated bone density between radiographs with minimal and clinical post-processing. Correlations varied between 0.51 and 0.97 (p < 0.001) for FD_Ver texture parameter and between − 0.10 and 0.97 for FD_Hor. Bone density and texture were different (p < 0.05) between controls and subjects with osteoarthritis or BMLs mainly in medial tibial regions. When classifying healthy and osteoarthritic subjects using a machine learning-based elastic net model with bone characteristics, area under the receiver operating characteristics (ROCAUC) curve was 0.77. For classifying controls and subjects with BMLs, ROCAUC was 0.85. In conclusion, differences in bone density and texture can be assessed from knee radiographs when using minimal post-processing.