Deep Learning-Based Image Segmentation Techniques for Bone Fractures Using X-Ray Images: A Systematic Review

Abstract

Human bone fractures are common musculoskeletal disorders. The primary cause of fractures is often accidents or external pressure applied to the body, which can result in significant fractures. Medical image processing plays a crucial role in the segmentation and analysis of human bone fractures using X-ray images, assisting physicians in determining appropriate treatments. The use of artificial intelligence (AI) techniques, such as machine learning, deep learning (DL), and transfer learning, has garnered significant interest for medical diagnosis from X-ray image reports. The primary objective of this paper is to explore various deep learning-based methods for analyzing human bones using X-ray images. These methods include the evaluation of U-Net, ViT, TransUnet, Swin-Unet, and Swin-Unet++, with U-Net and SegNet being utilized for comparative analysis. The findings and discussion indicate that U-Net and ViT are among the most promising models for the MURA dataset, achieving high accuracy. A comparison chart is provided in the paper to highlight various fracture segmentation methods, dataset sizes, and evaluation metrics.