Knowledge-Guided Multiview Deep Curriculum Learning for Elbow Fracture Classification

Elbow fracture diagnosis often requires patients to take both frontal and lateral views of elbow X-ray radiographs. In this paper, we propose a multiview deep learning method for an subtype classification task. Our strategy leverages transfer by first training two single-view models, one view the other view, then transferring weights corresponding layers in proposed network architecture. Meanwhile, quantitative medical knowledge was integrated into process through curriculum framework, which enables model learn from "easier" samples transition "harder" reach better performance. addition, our can work dual-view setting with single as input. We evaluate extensive experiments on task dataset 1,964 images. Results show that outperforms related methods bone study multiple settings, technique is able boost performance compared methods. The code available at https://github.com/ljaiverson/multiview-curriculum.