Types of application of artificial intelligence in the diagnosis and prognosis of osteoporosis; a narrative review

Abstract

Introduction: The rising impact of osteoporosis and fragility fractures highlights the need for advanced management strategies. Integrating digital health interventions, especially artificial intelligence (AI) algorithms, is essential. Osteoporosis, a major contributor to elderly disability, demands AI to minimize diagnostic errors. This review targets stakeholders interested in employing AI for osteoporosis management. Methods: We examined 16 articles from PubMed, Google Scholar, and Medline (January 1, 2015, to January 1, 2023) using keywords like AI, osteoporosis, fragility fracture, and machine learning. After excluding redundancies, 15 articles were selected, covering five key aspects of osteoporosis management: Bone mineral densitometry (BMD) predictive variables (n=1), diagnosis, screening, and classification of osteoporosis (n=5), diagnosis and screening of fractures (n=4), fracture risk forecast (n=2), and automated image segmentation (n=3). Results: Recent machine learning (ML) advances empower AI in assessing bone health beyond X-rays. Techniques, including AI-driven analysis with multi-detector computed tomography scans, extend beyond X-ray imaging. Convolutional neural networks (CNNs) excel in fracture diagnosis, surpassing medical professionals. Enhanced CNN performance is achieved through data augmentation and generative networks. Conclusion: Initial ML applications in osteoporosis research focus on the macroscopic scale, leaving a gap in microscale exploration. Establishing a robust system for bone micro-damage initiation detection is crucial for future applications in bone micromechanics. Ongoing development is essential to assess effectiveness and affordability through controlled studies.