Knee Injury Diagnosis with Data And Feature Fusion-Enhanced Multi-Label Classification Network

Knee injuries are among the most common injuries people get despite of their age, gender, or lifestyle. They are often detected using a non-invasive diagnostic tool such as Magnetic Resonance Imaging (MRI) of the knee. With the increasing number of injuries yearly, radiologists are more prone to diagnostic errors. Therefore, several deep learning methods have been used for designing an automatic MRI interpreter. MRNet public dataset is an established knee injury dataset consisting of; multi-view MRI images with multi-label classification target. The designed deep learning solutions utilizing this dataset including the MRNet model, ignore the dependence and relations among the different injuries and views. Given the specific properties of this dataset and problem, we propose a tailored mutli-label classification network with enhanced data and feature fusion. The designed model outperforms the MRNet baseline model in terms of complexity and performance reaching an Area under the curve (AUC) of 0.925.