Automatic distal radius fracture detection and classification using deep convolutional neural network with radiological images

Distal radius fractures (DRF) are among the most common fractures and are often treated surgically. The accuracy and effectiveness of the surgical procedures greatly depend on the correct classification of distal radius fractures. Wrist fractures are the most commonly misclassified because of the wrist bone's complex anatomical structure, including several different bones. Thus, it is evident that models based on machine learning (ML) and artificial intelligence (AI) are required, with an emphasis on making them user-friendly for everyday clinical practice. Hence, this study proposes the Deep Convolutional Neural Network-based Distal Radius Fracture Classification Model (DCNN-DRFCM) to diagnose DRFs using anteroposterior and lateral wrist radiographs. The goal of this work is to develop an artificial intelligence system that can learn to utilize X-ray pictures to correctly diagnose distal radius fractures with a small amount of information. Labelling assessments with fractures and overlaying fracture masks generates images that may be used for testing and training segmentation and classification methods. The DCNN model analyzed DRF based on three views: lateral, anteroposterior, and lateral and anteroposterior views. The experimental outcomes demonstrate that the recommended model increases the classification accuracy rate of 99.3%, sensitivity rate of 96.5%, specificity rate of 97.8%, and F1-score rate of 95.6% and reduces the error rate of 11.2% compared to other popular approaches.