Using statistical modelling and machine learning in detecting bone properties: A systematic review protocol.

Abstract

Introduction: Osteoporosis, a common condition characterised by decreased bone mass and microarchitectural deterioration, leading to increased fracture risk, is a significant health concern. Traditional diagnostic methods, such as Dual-energy X-ray Absorptiometry (DXA), have limitations in sensitivity and accessibility. However, the emergence of artificial intelligence (AI) and machine learning (ML) has brought promising tools capable of analysing complex medical data to enhance the detection and prediction of osteoporosis-related bone properties. This systematic review protocol outlines the methodology to evaluate the application and effectiveness of AI and ML methods in detecting bone properties and osteoporosis. It underscores their potential to revolutionise healthcare by providing more accurate and accessible osteoporosis detection and prediction tools.

Methods: This systematic review, which will follow the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) guidelines, will be comprehensive in its approach. A thorough search will be conducted across PubMed, Embase, IEEE Xplore, Scopus, Cochrane Library, and GitHub from their inception to March 2025. Studies involving adults aged 40 years and older that utilise AI/ML techniques to detect or predict bone density or other bone-related properties will be included. Two independent reviewers will perform screening, data extraction, and risk of bias assessments using appropriate tools such as RoB 2, ROBINS-I, QUADAS-2, PROBAST, and NOS. The comprehensive nature of this review ensures that no relevant study is overlooked. Data synthesis will involve narrative synthesis and, if applicable, meta-analysis using Review Manager (RevMan) and R software.

Discussion: This systematic review will comprehensively evaluate current AI and ML applications in detecting bone properties and osteoporosis. By identifying and analysing various AI/ML models and comparing them with traditional diagnostic methods, the review aims to highlight the effectiveness and potential of these technologies in clinical practice. The findings are expected to significantly impact healthcare professionals, researchers, and policymakers regarding advancements in AI/ML for bone health assessment and guide future research directions. Understanding the strengths and limitations of existing studies will be crucial in developing standardised protocols and facilitating the integration of AI/ML tools into routine osteoporosis screening and management.

Systematic review registration: This Systematic Review Protocol was registered in PROSPERO (CRD42024587326).