Enhanced X-Ray Bone Fracture Detection Using MTBC and ResNet-50

Abstract

Early identification of X-ray picture bone fractures is crucial for effective patient treatment and care planning. This study presents an innovative automated system designed to identify bone fractures in radiographic images. The proposed approach involves multiple critical steps, beginning with image preprocessing to reduce noise and enhance the visibility of potential fracture lines. Next, a Multi-Trend Binary Code (MTBC) method is applied to capture detailed textural features that emphasize structural irregularities linked to fractures. These refined features are then fed into a customized ResNet-50 deep learning model to extract advanced, high-level patterns. A classification component then evaluates whether a fracture is present. The system’s effectiveness is thoroughly tested on an extensive bone X-ray dataset, with results showing notable enhancements in accuracy, sensitivity, and specificity over current methods. The combined use of MTBC for feature refinement and ResNet-50 for deep learning proves highly efficient in fracture detection, providing a useful tool for radiologists and healthcare professionals to enhance diagnostic accuracy and streamline workflows.