Quantifying radiographic knee osteoarthritis severity using deep convolutional neural networks

By Joseph Antony, Kevin McGuinness, Noel Edward O'Connor, Kieran Moran. 2016

296 citations recorded by the source corpus.

This paper proposes a new approach to automatically quantify the severity of knee osteoarthritis (OA) from radiographs using deep convolutional neural networks (CNN). Clinically, knee OA severity is assessed using Kellgren & Lawrence (KL) grades, a five point scale. Previous work on automatically predicting KL grades from radiograph images were based on training shallow classifiers using a variety of hand engineered features. We demonstrate that classification accuracy can be significantly improved using deep convolutional neural network models pre-trained on ImageNet and fine-tuned on knee OA images. Furthermore, we argue that it is more appropriate to assess the accuracy of automatic knee OA severity predictions using a continuous distance-based evaluation metric like mean squared error than it is to use classification accuracy. This leads to the formulation of the prediction of KL grades as a regression problem and further improves accuracy. Results on a dataset of X-ray images and KL grades from the Osteoarthritis Initiative (OAI) show a sizable improvement over the current state-of-the-art.

Subfields: Rheumatology, Surgery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Rheumatology

1. Alexey A. Shvets
2. Alexander Rakhlin
3. Vladimir I. Iglovikov
4. Alexandr A Kalinin

Citations in the corpus

• A survey on deep learning in medical image analysis (2017, methodology): “(2015) MRI Anatomical landmark detection; uses CNN for slice classification for presence of landmark Antony et al. (2016) X-ray Osteoarthritis grading; pre-trained ImageNet CNN fine-tuned on knee X-rays Cai et al. (2016b) CT, MRI Vertebrae localization; RBM determines position, orientation and label”
• Automatic Knee Osteoarthritis Diagnosis from Plain Radiographs: A Deep Learning-Based Approach (2017, methodology): “To train the network by Antony et al. 16 , we implemented the architecture and the custom loss function proposed in the article.”
• Not‐so‐supervised: A survey of semi‐supervised, multi‐instance, and transfer learning in medical image analysis (2018, background): “[2015a] US standard plane localization different different feature, pretraining Antony et al. [2016] osteoarthritis quantification different different feature, pretraining Conjeti et al. [2016] tissue classification same different instance, align Elmahdy et al. [2017] skin lesion classification diff”
• Deep Convolutional Neural Networks Enable Discrimination of Heterogeneous Digital Pathology Images (2017, methodology): “For example, Antony et al. (2016) showed that fine-tuning clearly outperformed feature extraction, achieving 57.6% accuracy in multi-class grade assessment of knee osteoarthritis versus 53.4%.”
• Secure and Robust Machine Learning for Healthcare: A Survey (2020, result): “For instance, results obtained by transfer learning in [48] and [49] are contradictory.”
• Comparison of Deep Transfer Learning Strategies for Digital Pathology (2018, background): “Some recent publications in the biomedical field have shown that fine-tuning outperforms off-the-shelf features [1, 19, 31, 34].”