Automatic Knee Osteoarthritis Diagnosis from Plain Radiographs: A Deep Learning-Based Approach

• DOI: 10.1038/s41598-018-20132-7
• Publication Date: 2018-01-23T10:04:03Z
• Authors: Aleksei Tiulpin; Jérôme Thevenot; Esa Rahtu; Petri Lehenkari; Simo Saarakkala
• Keywords: Kappa; Cohen's kappa; Grading (engineering)

Knee osteoarthritis (OA) is the most common musculoskeletal disorder. OA diagnosis currently conducted by assessing symptoms and evaluating plain radiographs, but this process suffers from subjectivity. In study, we present a new transparent computer-aided method based on Deep Siamese Convolutional Neural Network to automatically score knee severity according Kellgren-Lawrence grading scale. We trained our using data solely Multicenter Osteoarthritis Study validated it randomly selected 3,000 subjects (5,960 knees) Initiative dataset. Our yielded quadratic Kappa coefficient of 0.83 average multiclass accuracy 66.71% compared annotations given committee clinical experts. Here, also report radiological area under ROC curve 0.93. Besides this, attention maps highlighting features affecting network decision. Such information makes decision for practitioner, which builds better trust toward automatic methods. believe that model useful making research; therefore, openly release training codes set created in study.