Edge-Aware Contrast Enhancement Framework to Aid Automated Bone Fracture Detection

Bone fracture detection from X-ray images is a challenging task due to low contrast, noise, and poor visibility of fine structural details. Existing enhancement techniques often fail to preserve clinically important edges, resulting in reduced diagnostic accuracy in automated systems. To address this limitation, this study proposes an Edge-Aware Contrast Enhancement Framework (EACEF) that integrates laplacian sharpening, sobel gradient extraction, smoothing, unsharp masking, and adaptive gamma correction to highlight fracture lines while suppressing noise. Three deep learning models, CNN, VGG16, and ResNet50, were trained on two datasets that included both fractured and non-fractured bone images. Experimental results show that the proposed preprocessing technique significantly improves classification performance. For Dataset-I, ResNet50 achieved a highest accuracy of 96%, outperforming CNN and VGG16. For Dataset-II, ResNet50 achieved a further 99% accuracy. The findings indicate that the proposed EACEF method effectively improves bone structure and fracture boundaries, improving the reliability of the automatic fracture detection system. This work highlights the importance of edge-aware preprocessing in medical imaging and lays a strong foundation for future advances in computer-aided diagnosis.