Multi-task learning for calcaneus fracture diagnosis of X-ray images.

Fracture diagnosis is critical in clinical settings, and imaging modalities like X-ray and CT scans are crucial for bone fracture diagnosis. Although X-ray scans are more convenient and affordable, they are prone to misdiagnosis and missed diagnosis due to the limited number of poses that can be observed and the uneven distribution of grayscales. Deep learning methods show great potential in improving the efficiency and accuracy of x-ray fracture diagnosis. This paper proposes MTL-DlinkNet, a multi-task learning model based on D-linkNet, that can perform both classification tasks for automatic identification of calcaneus fractures, and segmentation tasks for generation of regions of interest (RoI). The model achieves high accuracy (0.989 AUC) while annotating the data at the same time, making diagnosis more efficient and reducing the burden on doctors. The paper also presents an intelligent diagnostic tool for calcaneus fracture that assists doctors in clinical settings. The main contributions of this work include building a dataset of foot X-rays for fracture diagnosis, proposing a better-performing fracture diagnosis deep learning model, and developing an efficient diagnostic tool. Experimental results show that the MTL-DlinkNet model achieves better performance than the baseline models. Overall, this paper demonstrates the potential of deep learning methods in improving fracture diagnosis accuracy and efficiency, especially when multi-task learning is used, with practical applications in clinical settings. • We proposed a multi-task learning model for fracture diagnosis through X-ray images. • We developed a tool for calcaneus fracture diagnosis and data labeling. • We built a dataset of foot X-rays, with a total of 446 samples. [ABSTRACT FROM AUTHOR]