Applying Densely Connected Convolutional Neural Networks for Staging Osteoarthritis Severity from Plain Radiographs

Norman, Berk; Pedoia, Valentina; Noworolski, A.; Link, Thomas M.; Majumdar, Sharmila

doi:10.1007/s10278-018-0098-3

Cited by 130 publications

(114 citation statements)

References 18 publications

Supporting

Mentioning

108

Contrasting

Unclassified

Order By: Relevance

“…Radiographs were cropped to a 500 × 500 region centered around the knee joint. Briefly, 2D cross-correlation template matching was used to identify a 500 × 500 bounding box centered around the knee joint in 450 joints, and these cases were used to train a U-Net architecture that identified this region for all posteroanterior radiographs from the OAI study 32 . DESS MRIs were center-cropped to a 120 × 320 × 320 region, after which both sets of cropped images were normalized.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Predicts Total Knee Replacement from Magnetic Resonance Images

Tolpadi

Lee

Pedoia

et al. 2020

Sci Rep

Self Cite

View full text Add to dashboard Cite

Knee Osteoarthritis (OA) is a common musculoskeletal disorder in the United States. When diagnosed at early stages, lifestyle interventions such as exercise and weight loss can slow OA progression, but at later stages, only an invasive option is available: total knee replacement (TKR). Though a generally successful procedure, only 2/3 of patients who undergo the procedure report their knees feeling "normal" post-operation, and complications can arise that require revision. This necessitates a model to identify a population at higher risk of TKR, particularly at less advanced stages of OA, such that appropriate treatments can be implemented that slow OA progression and delay TKR. Here, we present a deep learning pipeline that leverages MRI images and clinical and demographic information to predict TKR with AUC 0.834 ± 0.036 (p < 0.05). Most notably, the pipeline predicts TKR with AUC 0.943 ± 0.057 (p < 0.05) for patients without OA. Furthermore, we develop occlusion maps for case-control pairs in test data and compare regions used by the model in both, thereby identifying TKR imaging biomarkers. As such, this work takes strides towards a pipeline with clinical utility, and the biomarkers identified further our understanding of OA progression and eventual TKR onset. otherwise would, thereby reducing time spent pursuing nonsurgical alternatives with minimal probability of success while dealing with serious pain. Beyond this, if the model were to draw from medical images of the knee, it could identify anatomic regions most correlated with a TKR prediction. To this point, few studies have been conducted in this space, and those that have primarily investigate the importance of cartilage volume loss, subchondral bone defects, and bone marrow lesions [17][18][19] . An identification of more such biomarkers for TKR, however, could greatly improve understanding of both OA and TKR, and ultimately guide treatment strategies.Predictive modeling of TKR, however, has a limited history, particularly with models that use medical images. A few studies have leveraged random forest regression, Cochran-Armitage tests for trend, and t-tests to identify demographic, general health, and physical examination measurements that most strongly correlate with TKR or total joint arthroplasty (TJA) 20, 21 . Others have taken these efforts further, using techniques such as multiple regression and multivariate risk prediction models to predict TKR outright 22, 23 . To our knowledge, only one group has developed a predictive model of TKR that accepts image inputs, attaining performance that surpasses that of models using only clinical and demographic information 24 . Notably, past TKR predictive models largely measure performance by evaluating the area under the receiver operating characteristic (ROC) curve, which plots true positive rate against false positive rate 25 . However, in most datasets used in this space, the number of patients who eventually undergo TKR is dramatically higher among those who have advanced OA as opposed to those with n...

show abstract

Section: Introductionmentioning

confidence: 99%

Deep Learning Predicts Total Knee Replacement from Magnetic Resonance Images

Tolpadi

Lee

Pedoia

et al. 2020

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…Figure 1B depicts two examples of saliency maps showing the importance of each pixel in the model decision making process. Performance of both these studies 21,22 are comparable to human reliability. A recent study aimed to determine the reliability of radiographic assessment of knee osteoarthritis (OA) by non-clinician readers compared to an experienced radiologist.…”

Section: Advances In Radiographic Assessment Of Knee Oamentioning

confidence: 74%

“…Tiulpln et al made use of the knee joint symmetry using state of the art Siamese network for the classification.Norman et. al . used a densely connected neural network (DenseNet) to make OA assessments.…”

Section: Advances In Radiographic Assessment Of Knee Oamentioning

confidence: 99%

“…al. 22 used a densely connected neural network (DenseNet) to make OA assessments. Figure 1A shows the agreement confusion matrix between DenseNet deep learning model and human radiologists in evaluating KL grades on a test set which was never seen in training or tuning phase of the automatic algorithm development.…”

Section: Advances In Radiographic Assessment Of Knee Oamentioning

confidence: 99%

See 1 more Smart Citation

Translation of morphological and functional musculoskeletal imaging

Pedoia

Majumdar

2018

Journal Orthopaedic Research

Self Cite

View full text Add to dashboard Cite

In an effort to develop quantitative biomarkers for degenerative joint disease and fill the void that exists for diagnosing, monitoring, and assessing the extent of whole joint degeneration, the past decade has been marked by a greatly increased role of noninvasive imaging. This coupled with recent advances in image processing and deep learning opens new possibilities for promising quantitative techniques. The clinical translation of quantitative imaging was previously hampered by tedious non‐scalable and subjective image analysis. Osteoarthritis (OA) diagnosis using X‐rays can be automated by the use of deep learning models and pilot studies showed feasibility of using similar techniques to reliably segment multiple musculoskeletal tissues and detect and stage the severity of morphological abnormalities in magnetic resonance imaging (MRI). Automation and more advanced feature extraction techniques have applications on larger more heterogeneous samples. Analyses based on voxel based relaxometry have shown local patterns in relaxation time elevations and local correlations with outcome variables. Bone cartilage interactions are also enhanced by the analysis of three‐dimensional bone morphology and the potential for the assessment of metabolic activity with simultaneous Positron Emission Tomography (PET)/MR systems. Novel techniques in image processing and deep learning are augmenting imaging to be a source of quantitative and reliable data and new multidimensional analytics allow us to exploit the interactions of data from various sources. In this review, we aim to summarize recent advances in quantitative imaging, the application of image processing and deep learning techniques to study knee and hip OA. ©2018 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res XX:XX–XX, 2018.

show abstract

“…When we used the test data (MOST) which is independent from the training data (OAI) in our validation experiments (Exp 3), we observed that classification performance was not affected, even slight improvements were obtained with LBP and HOG descriptors. This may be explained by different distribution and amount of of training samples (2915 vs 9012), and, potentially, imperfect annotations of KL grades 16,49 .…”

Section: /21mentioning

confidence: 99%

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

Bayramoglu

Tiulpin

Hirvasniemi

et al. 2020

Osteoarthritis and Cartilage

View full text Add to dashboard Cite

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) and Multicenter Osteoarthritis Study (MOST) datasets. A fully automatic method to locate the most informative region from subchondral bone using adaptive segmentation was developed. We used an oversegmentation strategy for partitioning knee images into the compact regions that follow natural texture boundaries. Local Binary Patterns (LBP), Fractal Dimension (FD), Haralick features, Shannon entropy, and Histogram of Oriented Gradients (HOG) methods were computed within the standard ROI and within the proposed adaptive ROIs. Subsequently, we built logistic regression models to identify and compare the performances of each texture descriptor and each ROI placement method using 5-fold cross validation setting. 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 (ROC) curve (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, 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.

show abstract

Applying Densely Connected Convolutional Neural Networks for Staging Osteoarthritis Severity from Plain Radiographs

Cited by 130 publications

References 18 publications

Deep Learning Predicts Total Knee Replacement from Magnetic Resonance Images

Deep Learning Predicts Total Knee Replacement from Magnetic Resonance Images

Translation of morphological and functional musculoskeletal imaging

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

Contact Info

Product

Resources

About