Automated selection of trabecular bone regions in knee radiographs

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.

Early tibial subchondral bone texture changes after arthroscopic partial meniscectomy in knees without radiographic OA: A prospective cohort study

Arthroscopic partial meniscectomy (APM) may lead to changes in underlying trabecular bone (TB) structure potentially promoting the development of knee joint osteoarthritis. Our aim was to investigate if there are early changes occurring in tibial subchondral TB texture in the leg undergoing medial APM compared with the unoperated non-injured contra-lateral leg. The bone texture was measured as the medial-to-lateral ratio of fractal dimensions (FD) calculated for regions selected on weight-bearing anteroposterior tibiofemoral x-rays. Twenty-one subjects before and 12 months after APM were included from 374 patients scheduled for unilateral medial APM. The medial-to-lateral ratio was calculated for horizontal, vertical, and roughest FDs respectively. Higher FD means higher bone roughness. Each FD was calculated over a range of scales using a variance orientation transform method. Mean values of medial-to-lateral horizontal FD calculated for APM knees at follow-up were higher than those at baseline. For unoperated knees the values were lower. The difference in the horizontal FD change from baseline to follow-up between APM and contra-lateral legs was 0.028 (95% CI, 0.004-0.052). The bone roughness changes may reflect the increase in peak knee adduction moment (KAM) and KAM impulse during walking reported for the same cohort in a previous study. They may also reflect early signs of osteoarthritis development and thus, we speculate that individuals with increased bone texture roughness ratio after APM might be at higher risk of knee osteoarthritis development.

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.

Differences in tibial subchondral bone structure evaluated using plain radiographs between knees with and without cartilage damage or bone marrow lesions - the Oulu Knee Osteoarthritis study

Objectives

To investigate whether subchondral bone structure from plain radiographs is different between subjects with and without articular cartilage damage or bone marrow lesions (BMLs).

Methods

Radiography-based bone structure was assessed from 80 subjects with different stages of knee osteoarthritis using entropy of Laplacian-based image (E_Lap) and local binary patterns (E_LBP), homogeneity index of local angles (HI_Angles,mean), and horizontal (FD_Hor) and vertical fractal dimensions (FD_Ver). Medial tibial articular cartilage damage and BMLs were scored using the magnetic resonance imaging osteoarthritis knee score. Level of statistical significance was set to p < 0.05.

Results

Subjects with medial tibial cartilage damage had significantly higher FD_Ver and E_LBP as well as lower E_Lap and HI_Angles,mean in the medial tibial subchondral bone region than subjects without damage. FD_Hor, FD_Ver, and E_LBP were significantly higher, whereas E_Lap and HI_Angles,mean were lower in the medial trabecular bone region. Subjects with medial tibial BMLs had significantly higher FD_Ver and E_LBP as well as lower E_Lap and HI_Angles,mean in medial tibial subchondral bone. FD_Hor, FD_Ver, and E_LBP were higher, whereas E_Lap and HI_Angles,mean were lower in medial trabecular bone.

Conclusions

Our results support the use of bone structural analysis from radiographs when examining subjects with osteoarthritis or at risk of having it.

Key points

• Knee osteoarthritis causes changes in articular cartilage and subchondral bone

• Magnetic resonance imaging is a comprehensive imaging modality for knee osteoarthritis

• Radiography-based bone structure analysis can provide additional information of osteoarthritic subjects

Subchondral tibial bone texture analysis predicts knee osteoarthritis progression: data from the Osteoarthritis Initiative: Tibial bone texture & knee OA progression

OBJECTIVES

To examine whether trabecular bone texture (TBT) parameters assessed on computed radiographs could predict knee osteoarthritis (OA) progression.

METHODS

This study was performed using data from the Osteoarthritis Initiative (OAI). 1647 knees in 1124 patients had bilateral fixed flexion radiographs acquired 48 months apart. Images were semi-automatically segmented to extract a patchwork of regions of interest (ROI). A fractal texture analysis was performed using different methods. OA progression was defined as an increase in the joint space narrowing (JSN) over 48 months. The predictive ability of TBT was evaluated using logistic regression and receiver operating characteristic (ROC) curve. An optimization method for features selection was used to reduce the size of models and assess the impact of each ROI.

RESULTS

Fractal dimensions (FD's) were predictive of the JSN progression for each method tested with an area under the ROC curve (AUC) up to 0.71. Baseline JSN grade was not correlated with TBT parameters (R < 0.21) but had the same predictive capacity (AUC 0.71). The most predictive model included the clinical covariates (age, gender, body mass index (BMI)), JSN and TBT parameters (AUC 0.77). From a statistical point of view we found higher differences in TBT parameters computed in medial ROI between progressors and non-progressors. However, the integration of TBT results from the whole patchwork including the lateral ROIs in the model provided the best predictive model.

CONCLUSIONS

Our findings indicate that TBT parameters assessed in different locations in the joint provided a good predictive ability to detect knee OA progression.

Baseline trabecular bone and its relation to incident radiographic knee osteoarthritis and increase in joint space narrowing score: directional fractal signature analysis in the MOST study

PURPOSE

To explore the association of baseline trabecular bone structure with incident tibiofemoral (TF) osteoarthritis (OA) and with increase in joint space narrowing (JSN) score.

METHODS

The Multicenter Osteoarthritis Study (MOST) includes subjects with or at risk for knee OA. Knee radiographs were scored for Kellgren-Lawrence (KL) grade and JSN at baseline, 30, 60 and 84 months. Knees (KL ≤ 1) at baseline were assessed for incident OA (KL ≥ 2) and increases in JSN score. For each knee image at baseline, a variance orientation transform method (VOT) was applied to subchondral tibial bone regions of medial and lateral compartments. Seventeen fractal parameters were calculated per region. Associations of each parameter with OA incidence and with medial and lateral JSN increases were explored using logistic regression. Analyses were stratified by digitized film (DF) vs computer radiography (CR) and adjusted for confounders.

RESULTS

Of 894 knees with CR and 1158 knees with DF, 195 (22%) and 303 (26%) developed incident OA. Higher medial bone roughness was associated with increased odds of OA incidence at 60 and 84 months and also, medial and lateral JSN increases (primarily vertical). Lower medial and lateral anisotropy was associated with increased odds of medial and lateral JSN increase. Compared to DF, CR had more associations and also, similar results at overlapping scales.

CONCLUSION

Baseline trabecular bone texture was associated with incident radiographic OA and increase of JSN scores independently of risk factors for knee OA. Higher roughness and lower anisotropy were associated with increased odds for radiographic OA change.

Trabecular bone texture detected by plain radiography is associated with an increased risk of knee replacement in patients with osteoarthritis: a 6 year prospective follow up study

OBJECTIVE

To examine the association between trabecular bone texture and knee joint replacement (KJR) measured using a variance orientation transform (VOT) method.

METHODS

The association of trabecular bone texture and KJR was examined prospectively over 6 years in 123 subjects with symptomatic knee osteoarthritis (OA): data regarding KJR was available for 114 (93%). At baseline, weight-bearing anteroposterior tibio-femoral radiographs were acquired. Trabecular bone texture regions were selected from the medial and lateral subchondral tibia. The VOT method was applied to each region and five fractal bone texture parameters, i.e., mean fractal dimension (FDMEAN), fractal dimensions in the horizontal (FDH) and vertical (FDV) directions, and along the roughest part of trabecular bone (FD(Sta)), and texture aspect ratio (Str) were calculated. The association between groups with increasing baseline fractal parameters (defined using tertiles) with risk of JR was examined using logistic regression.

RESULTS

28 (25%) participants' study knees underwent KJR over 6 years. Participants with KJR had lower medial FD(MEAN) and FD(H) parameters (P = 0.02 for difference). With increasing FD(MEAN), adjusted for age, gender, body mass index (BMI), osteophyte grade, joint space narrowing (JSN) grade and WOMAC pain score, the odds of KJR was reduced (P = 0.04 for trend).

CONCLUSION

This study suggests that the texture of medial tibial trabecular bone measured from plain radiographs is related to the risk of KJR: with increasing FD(MEAN) (the overall measure of bone texture roughness) the risk of KJR was reduced, independent of other clinical predictors of joint replacement. Tibial trabecular bone texture may be a useful marker of disease progression and a target of therapy in OA.

Osteoarthritis classification using self organizing map based on gabor kernel and contrast-limited adaptive histogram equalization

Localization is the first step in osteoarthritis (OA) classification. Manual classification, however, is time-consuming, tedious, and expensive. The proposed system is designed as decision support system for medical doctors to classify the severity of knee OA. A method has been proposed here to localize a joint space area for OA and then classify it in 4 steps to classify OA into KL-Grade 0, KL-Grade 1, KL-Grade 2, KL-Grade 3 and KL-Grade 4, which are preprocessing, segmentation, feature extraction, and classification. In this proposed system, right and left knee detection was performed by employing the Contrast-Limited Adaptive Histogram Equalization (CLAHE) and the template matching. The Gabor kernel, row sum graph and moment methods were used to localize the junction space area of knee. CLAHE is used for preprocessing step, i.e.to normalize the varied intensities. The segmentation process was conducted using the Gabor kernel, template matching, row sum graph and gray level center of mass method. Here GLCM (contrast, correlation, energy, and homogeinity) features were employed as training data. Overall, 50 data were evaluated for training and 258 data for testing. Experimental results showed the best performance by using gabor kernel with parameters α=8, θ=0, Ψ=[0 π/2], γ=0,8, N=4 and with number of iterations being 5000, momentum value 0.5 and α0=0.6 for the classification process. The run gave classification accuracy rate of 93.8% for KL-Grade 0, 70% for KL-Grade 1, 4% for KL-Grade 2, 10% for KL-Grade 3 and 88.9% for KL-Grade 4.

A dissimilarity-based multiple classifier system for trabecular bone texture in detection and prediction of progression of knee osteoarthritis

There is a growing need for classification systems that can accurately detect and predict knee osteoarthritis (OA) from plain radiographs. For this purpose, a system based on a support vector machine (SVM) classifier and distances measured between trabecular bone (TB) texture images was developed and tested in previous work. Unlike other systems, it allows an image classification without the calculation and selection of numerous texture features, and it is invariant to a range of imaging conditions encountered in a routine X-ray screening of knees. Although the system exhibited 85.4% classification accuracy in OA detection, which was higher than those obtained from other systems, its performance could be further improved. To achieve this, a dissimilarity-based multiple classifier (DMC) system is developed in this study. The system measures distances between TB texture images and generates a diverse ensemble of classifiers using prototype selection, bootstrapping of training set and heterogeneous classifiers. A measure of competence is used to select accurate (i.e. better-than-random) classifiers from the ensemble, which are then combined through the majority voting rule. To evaluate the newly developed system in OA detection (prediction of OA progression), TB texture images selected on standardised radiographs of healthy and OA (non-progressive and progressive OA) knees were used. The results obtained showed that the DMC system has higher classification accuracies for the detection (90.51% with 87.65% specificity and 93.33% sensitivity) and prediction (80% with 82.00% specificity and 77.97% sensitivity) than other systems, indicating its potential as a decision-support tool for the assessment of radiographic knee OA.

Prediction of progression of radiographic knee osteoarthritis using tibial trabecular bone texture

OBJECTIVE

To develop a system for predicting the progression of radiographic knee osteoarthritis (OA) using tibial trabecular bone texture.

METHODS

We studied 203 knees with (n = 68) or without (n = 135) radiographic tibiofemoral OA in 105 subjects (90 men and 15 women with a mean age of 54 years) in whom 2 sets of knee radiographs were obtained 4 years apart. We determined medial and lateral compartment tibial trabecular bone texture using an automated region selection method. Three texture parameters were calculated: roughness, degree of anisotropy, and direction of anisotropy based on a signature dissimilarity measure method. We evaluated tibiofemoral OA progression using a radiographic semiquantitative outcome: an increase in the medial joint space narrowing (JSN) grade. We examined the predictive ability of trabecular bone texture in knees with and those without preexisting radiographic OA, with adjustment for age, sex, and body mass index, using logistic regression (generalized estimating equations) and receiver operating characteristic curves.

RESULTS

The prediction of increased medial JSN in knees with or without preexisting radiographic OA was the most accurate for medial trabecular bone texture; the area under the curve (AUC) was 0.77 and 0.75, respectively. For lateral trabecular bone texture, the AUC was 0.71 in knees with preexisting OA and 0.72 in knees without preexisting OA.

CONCLUSION

We have developed a system, based on analyzing tibial trabecular bone texture, which yields good prediction of loss of tibiofemoral joint space. The predictive ability of the system needs to be further validated.

A computer analysis method for correlating knee X-rays with continuous indicators

PURPOSE

To develop an image analysis method that can automatically find correlations between a set of plain radiographs and continuous clinical or physiological indicators.

METHODS

Knee X-rays taken from the Baltimore Longitudinal Study of Aging are used in this study. The computer analysis method is based on the WND-CHARM image feature set filtered by using the Pearson correlation of each feature with the continuous variable, and the estimated value is determined by a weighted nearest neighbor interpolation.

RESULTS

Experimental results using 300 radiographs show that the proposed method can correlate knee X-rays with physiological indicators such as sex, age, height, weight, and BMI. For instance, the Pearson correlation between the X-ray images and the height and weight were 0.59 and 0.62, respectively.

CONCLUSIONS

Using computer analysis, X-ray images can be correlated to continuous physiological variables that might not have a direct and straightforward link to the visual content of the radiograph. This approach of radiology image analysis can be used in population studies for detecting biomarkers and also in genome-wide association studies for studying the link between genes and anatomy.

Trabecular bone texture detected by plain radiography and variance orientation transform method is different between knees with and without cartilage defects

The objective of this work is to evaluate differences in trabecular bone (TB) texture between subjects with and without tibiofemoral cartilage defects using a variance orientation transform (VOT) method. A case-control study was performed in subjects without radiographic knee osteoarthritis (OA) (K&L grade <2) matched on sex, BMI, age, knee compartment, and meniscectomy where cases (n = 28) had cartilage defects (grade ≥2) and controls (n = 28) had no cartilage defects (grade <2). Cartilage defects were assessed from MRI using validated methods. The VOT was applied to TB regions selected on medial and lateral compartments in knee X-rays and fractal signatures (FS) in the horizontal (FS(H) ) and vertical (FS(V) ) directions, and along the roughest part of TB (FS(Sta) ) and texture aspect ratio signatures (StrS), at different trabecular image sizes (0.30-0.70 mm) were calculated. Compared with controls, FS(V) for cases were higher (p < 0.011) at image sizes 0.30-0.40 mm and 0.45-0.55 mm in the medial compartment. In the lateral compartment, FS(H) and FS(Sta) for cases were higher (p < 0.028) than those for controls at 0.30-0.40 mm and 0.45-0.55 mm, while FS(V) was higher (p < 0.02) at 0.30-0.40 mm. TB texture roughness was greater in subjects with cartilage defects than in subjects without, suggesting thinning and fenestration of TB occur early in OA and that the VOT identifies changes in TB in knees with early cartilage damage. No differences in StrS (p > 0.05) were found.

Progression analysis and stage discovery in continuous physiological processes using image computing

We propose an image computing-based method for quantitative analysis of continuous physiological processes that can be sensed by medical imaging and demonstrate its application to the analysis of morphological alterations of the bone structure, which correlate with the progression of osteoarthritis (OA). The purpose of the analysis is to quantitatively estimate OA progression in a fashion that can assist in understanding the pathophysiology of the disease. Ultimately, the texture analysis will be able to provide an alternative OA scoring method, which can potentially reflect the progression of the disease in a more direct fashion compared to the existing clinically utilized classification schemes based on radiology. This method can be useful not just for studying the nature of OA, but also for developing and testing the effect of drugs and treatments. While in this paper we demonstrate the application of the method to osteoarthritis, its generality makes it suitable for the analysis of other progressive clinical conditions that can be diagnosed and prognosed by using medical imaging.

A signature dissimilarity measure for trabecular bone texture in knee radiographs

PURPOSE

The purpose of this study is to develop a dissimilarity measure for the classification of trabecular bone (TB) texture in knee radiographs. Problems associated with the traditional extraction and selection of texture features and with the invariance to imaging conditions such as image size, anisotropy, noise, blur, exposure, magnification, and projection angle were addressed.

METHODS

In the method developed, called a signature dissimilarity measure (SDM), a sum of earth mover's distances calculated for roughness and orientation signatures is used to quantify dissimilarities between textures. Scale-space theory was used to ensure scale and rotation invariance. The effects of image size, anisotropy, noise, and blur on the SDM developed were studied using computer generated fractal texture images. The invariance of the measure to image exposure, magnification, and projection angle was studied using x-ray images of human tibia head. For the studies, Mann-Whitney tests with significance level of 0.01 were used. A comparison study between the performances of a SDM based classification system and other two systems in the classification of Brodatz textures and the detection of knee osteoarthritis (OA) were conducted. The other systems are based on weighted neighbor distance using compound hierarchy of algorithms representing morphology (WND-CHARM) and local binary patterns (LBP).

RESULTS

Results obtained indicate that the SDM developed is invariant to image exposure (2.5-30 mA s), magnification (x1.00 - x1.35), noise associated with film graininess and quantum mottle (< 25%), blur generated by a sharp film screen, and image size (> 64 x 64 pixels). However, the measure is sensitive to changes in projection angle (> 5 degrees), image anisotropy (> 30 degrees), and blur generated by a regular film screen. For the classification of Brodatz textures, the SDM based system produced comparable results to the LBP system. For the detection of knee OA, the SDM based system achieved 78.8% classification accuracy and outperformed the WND-CHARM system (64.2%).

CONCLUSIONS

The SDM is well suited for the classification of TB texture images in knee OA detection and may be useful for the texture classification of medical images in general.

Early detection of radiographic knee osteoarthritis using computer-aided analysis

OBJECTIVE

To determine whether computer-based analysis can detect features predictive of osteoarthritis (OA) development in radiographically normal knees.

METHOD

A systematic computer-aided image analysis method weighted neighbor distances using a compound hierarchy of algorithms representing morphology (WND-CHARM) was used to analyze pairs of weight-bearing knee X-rays. Initial X-rays were all scored as normal Kellgren-Lawrence (KL) grade 0, and on follow-up approximately 20 years later either developed OA (defined as KL grade=2) or remained normal.

RESULTS

The computer-aided method predicted whether a knee would change from KL grade 0 to grade 3 with 72% accuracy (P<0.00001), and to grade 2 with 62% accuracy (P<0.01). Although a large part of the predictive signal comes from the image tiles that contained the joint, the region adjacent to the tibial spines provided the strongest predictive signal.

CONCLUSION

Radiographic features detectable using a computer-aided image analysis method can predict the future development of radiographic knee OA.

Directional fractal signature analysis of trabecular bone: evaluation of different methods to detect early osteoarthritis in knee radiographs

There is ongoing research directed towards the development of cheap and reliable decision support systems for the detection and prediction of osteoarthritis (OA) in knee joints. Fractal analysis of trabecular bone texture X-ray images is one of the most promising approaches. It is cheap and non-invasive. However, difficulties arise when the fractal signature methods are used to quantify bone roughness and anisotropy on individual scales. This is because the fractal methods are able to quantify bone texture only in the vertical and horizontal directions, and previous studies showed that OA bone changes can occur in any direction. To address these difficulties, three directional fractal signature methods were developed in this study, i.e. a fractal signature Hurst orientation transform (FSHOT) method, a variance orientation transform (VOT) method, and a blanket with rotating-grid (BRG) method. These methods were tested and the best performing method was selected. Unlike other methods, the newly developed techniques are able to calculate fractal dimensions (FDs) on individual scales (i.e. fractal signature) in all possible directions. The accuracy of the methods developed in measuring texture roughness and anisotropy on individual scales was evaluated. The effects of imaging conditions such as image noise, blur, exposure, magnification, and projection angle and the effects of translation of the bone region of interest on texture parameters were also evaluated. Computer-generated fractal surface images with known FDs and X-ray images obtained for a human tibia head were used. Results obtained show that the VOT method performs better than the FSHOT and BRG methods.

Biometric identification using knee X-rays

Identification of people often makes use of unique features of the face, fingerprints and retina. Beyond this, a similar identifying process can be applied to internal parts of the body that are not visible to the unaided eye. Here we show that knee X-rays can be used for the identification of individual persons. The image analysis method is based on the wnd-charm algorithm, which has been found effective for the diagnosis of clinical conditions of knee joints. Experimental results show that the rank-10 identification accuracy using a dataset of 425 individuals is ~56%, and the rank-1 accuracy is ~34%. The dataset contained knee X-rays taken several years apart from each other, showing that the identifiable features correspond to specific persons, rather than the present clinical condition of the joint.