Artificial intelligence for lumbar spine anatomy and pathology detection: A scoping review

Background: Spinal imaging is vital for diagnosing spinal conditions. This scoping review consolidates existing research on the application of artificial intelligence (AI) in lumbar spine imaging, examining its role in normal and pathological conditions, with a focus on lumbar spinal stenosis (LSS). Methods: This study follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. A comprehensive search strategy was applied across multiple databases, with article selection and analysis performed by two reviewers. We evaluated methodologies, innovations, imaging protocols, AI model performance, and summarized findings. Results: Methodologies range from traditional machine learning to advanced deep learning techniques, showing diagnostic efficiency. AI was utilized to assess normal lumbar spine anatomy, as well as detect and quantify pathological changes within the spine, with high accuracy rates in classification, segmentation, and prediction tasks. Conclusion: AI algorithms are increasingly proficient in analyzing healthy and pathological lumbar spinal conditions. They provide faster analysis, greater accuracy in diagnosis, and valuable support for clinicians in diagnosis and treatment planning. AI in spinal imaging has great potential for enhancing patient care and clinical outcomes.