Automatic Detection of Knee Joints and Quantification of Knee Osteoarthritis Severity using Convolutional Neural Networks

摘要

This paper introduces a new approach to automatically quantify the severityof knee OA using X-ray images. Automatically quantifying knee OA severityinvolves two steps: first, automatically localizing the knee joints; next,classifying the localized knee joint images. We introduce a new approach toautomatically detect the knee joints using a fully convolutional neural network(FCN). We train convolutional neural networks (CNN) from scratch toautomatically quantify the knee OA severity optimizing a weighted ratio of twoloss functions: categorical cross-entropy and mean-squared loss. This jointtraining further improves the overall quantification of knee OA severity, withthe added benefit of naturally producing simultaneous multi-classclassification and regression outputs. Two public datasets are used to evaluateour approach, the Osteoarthritis Initiative (OAI) and the MulticenterOsteoarthritis Study (MOST), with extremely promising results that outperformexisting approaches.