Deep Learning-Based Rib Segmentation and Fracture Classification in Chest CT Images

Abstract

Rib fractures are among the most common skeletal injuries in blunt thoracic trauma and are associated with serious respiratory complications, especially when multiple fractures are present. Accurate and rapid identification of fracture type and location is critical for treatment planning and prevention of potential complications. However, manual assessment of chest computed tomography (CT) images is time-consuming and prone to error. In this study, we propose a two-stage deep learning framework for automatic rib segmentation and fracture type classification on chest CT images. In the first stage, rib structures are segmented using an enhanced U-Net-based model with a multi-encoder design trained on an expert-annotated chest CT dataset. The results show that the model achieves high segmentation performance in terms of Dice Similarity Coefficient (DSC) and Intersection over Union (IoU). In the second stage, fracture-centered image patches extracted from the RibFrac dataset are classified into four fracture types using an ensemble classifier that exploits the outputs of multiple convolutional neural networks. Experimental results demonstrate that the proposed approach attains an accuracy above 99 percent in fracture classification, outperforming individual models and recent methods in the literature. Overall, the findings indicate that the proposed two-stage deep learning framework can reliably model rib anatomy and fracture types and has the potential to be integrated into future clinical workflows as a computer-aided decision-support system.