Imaging – based Detection of Bone Injuries using Hybrid Models

Bone fractures are among the most common physical injuries, affecting individuals across all age groups and requiring prompt and precise diagnosis. However, radiologists often face challenges in accurate fracture identification due to variations in image quality and complexity. The present work investigates a robust system for the automatic detection and classification of fractures in X-radiation (X-ray) images. The proposed model integrates Convolutional Neural Networks (CNN) and MobileNet with a hybrid approach that combines MobileNet and Random Forest to improve classification accuracy. The system is designed to be user-friendly, incorporating modules for registration, data processing, and real-time classification of fractures into 11 distinct types, including avulsion fracture, impacted fracture, longitudinal fracture, comminuted fracture, greenstick fracture, oblique fracture, pathological fracture, spiral fracture, and others. Experimental results demonstrate that the hybrid model significantly enhance diagnostic accuracy, underscoring its potential application in medical settings.