Tailored Convolutional Neural Network Applied to Fragility Fracture Classification Using Ultrasonic Guided Wave Spectrum Images

Osteoporosis affects millions worldwide, significantly impairing quality of life and imposing substantial eco-nomic burdens on healthcare systems. Annually, osteoporosis leads to millions of fragility fractures, with hip fractures projected to double within the next few decades. Currently, dual-energy X-ray absorptiometry (DXA) is the standard diagnos-tic method, offering quick scans with low radiation exposure. However, its effectiveness is limited by several factors including the influence of bone size and patient-specific conditions, high costs, and accessibility issues. Alternative diagnostic methods such as trabecular bone score (TBS), computed tomography (CT), and magnetic resonance imaging (MRI) have been explored, yet each presents its own set of challenges. This study investigates the potential of advanced imaging techniques like quantitative ultrasound axial transmission technology to provide effective and safer alternatives to DXA. This research evaluates three CNN models, including ResNet and a BDAT-Net optimized and not optimized, to assess their effectiveness in medical image classification and their potential to best osteoporosis diagnostics.