RADIOGRAPHIC OSTEOARTHRITIS PROGRESSION PREDICTION VIA MULTI-MODAL IMAGING DATA AND DEEP LEARNING

Abstract

Purpose: Prediction of radiographic osteoarthritis progression over short- and mid-terms may deepen understanding of the disease etiology and support the related clinical research. Prior imaging-oriented studies have focused mainly on individual imaging modalities. In particular, end-to-end deep learning methods have become increasingly common for knee X-rays. On the other hand, MR image analysis has been largely limited to manually designed imaging biomarkers. Such biomarkers are typically derived from rather small sample sizes and concern only the primarily affected tissues.