Human Body Bone Fracture Identification using Improved Deep Learning Model

Accidental human bone fractures require repair. Fractured bones include geometric information. The resolved body fractures show better conformance along the sensitive substrate, notably in the frontal face. Computer vision can repair and reconstruct bones. CAD tools assist diagnose and repair fractured bones. A difficult research topic is developing an autonomous and accurate human bone fracture detection CAD framework using bone scans. Deep Learning (DL) inspire us to propose the Automatic Detection of Human Bone Fractures (ADHBF) using input bone pictures. The ADHBF model includes pre-processing, bone segmentation, feature extraction, and classification. Pre-processing the bone picture with general filtering and contrast modification improves its quality. To accurately locate the ROI, the ADHBF framework segmented the bone using a dynamic threshold-based technique. ADHBF extracts bone features using the suggested CNN model. Using alternative activation functions with dropout layer, we improve CNN layers for classification. We adjusted the pre-trained model’s CNN layers to minimize training time and overfitting/underfitting. For human bone fracture prediction, the softmax classifier classifies automated CNN features. Experimental results utilizing the public dataset show the model’s efficiency. The ADHBF boosts anticipation accuracy by 3.56 % and processing efficiency by 6.27%.