Automatic Grading of Individual Knee Osteoarthritis Features in Plain Radiographs Using Deep Convolutional Neural Networks

• Main Authors: Aleksei Tiulpin, Simo Saarakkala
• Format: Article
• Language: English
• Published: MDPI AG 2020-11-01
• Series: Diagnostics
• Subjects: multi-task learning; deep learning; transfer learning; knee Osteoarthritis; OARSI grading atlas
• Online Access: https://www.mdpi.com/2075-4418/10/11/932

Knee osteoarthritis (OA) is the most common musculoskeletal disease in the world. In<br />primary healthcare, knee OA is diagnosed using clinical examination and radiographic assessment.<br />Osteoarthritis Research Society International (OARSI) atlas of OA radiographic features allows<br />performing independent assessment of knee osteophytes, joint space narrowing and other knee<br />features. This provides a fine-grained OA severity assessment of the knee, compared to the gold<br />standard and most commonly used Kellgren–Lawrence (KL) composite score. In this study, we<br />developed an automatic method to predict KL and OARSI grades from knee radiographs. Our<br />method is based on Deep Learning and leverages an ensemble of residual networks with 50 layers.<br />We used transfer learning from ImageNet with a fine-tuning on the Osteoarthritis Initiative (OAI)<br />dataset. An independent testing of our model was performed on the Multicenter Osteoarthritis Study<br />(MOST) dataset. Our method yielded Cohen’s kappa coefficients of 0.82 for KL-grade and 0.79, 0.84,<br />0.94, 0.83, 0.84 and 0.90 for femoral osteophytes, tibial osteophytes and joint space narrowing for<br />lateral and medial compartments, respectively. Furthermore, our method yielded area under the<br />ROC curve of 0.98 and average precision of 0.98 for detecting the presence of radiographic OA, which<br />is better than the current state-of-the-art.