Deep Learning for Rheumatoid Arthritis: Joint Detection and Damage Scoring in X-rays
This work addresses the need for automated, objective diagnosis of rheumatoid arthritis, which currently relies on manual inspection by doctors, but it is incremental as it builds on existing deep learning methods for medical imaging.
The authors tackled the problem of automating diagnosis of rheumatoid arthritis from X-ray images by developing a multi-task deep learning model for joint detection and damage scoring, achieving 4th and 5th place in a global challenge for joint space narrowing and erosion, respectively, with a 5% relative error reduction from a novel loss function.
Recent advancements in computer vision promise to automate medical image analysis. Rheumatoid arthritis is an autoimmune disease that would profit from computer-based diagnosis, as there are no direct markers known, and doctors have to rely on manual inspection of X-ray images. In this work, we present a multi-task deep learning model that simultaneously learns to localize joints on X-ray images and diagnose two kinds of joint damage: narrowing and erosion. Additionally, we propose a modification of label smoothing, which combines classification and regression cues into a single loss and achieves 5% relative error reduction compared to standard loss functions. Our final model obtained 4th place in joint space narrowing and 5th place in joint erosion in the global RA2 DREAM challenge.