Artificial Intelligence for Procedural Guidance of Spinal Epidural Steroid Injections: A Scoping Review

Spinal epidural steroid injections (SESI) are commonly used for radiating low back and cervical pain but remain limited by target imprecision, operator variability, and inconsistent outcomes. Artificial intelligence (AI) has been increasingly investigated as a tool to augment image interpretation, trajectory planning, and decision support in interventional procedures, prompting interest in its potential role in procedural guidance for SESI. This scoping review aimed to summarize the current evidence on the use of AI to support procedural guidance for SESI. A search was conducted across three databases, PubMed (National Library of Medicine (NLM)), Web of Science, and Embase, for English-language primary studies published between 2015 and 2025. Nine studies met the inclusion criteria. The AI architectures included machine learning, deep learning, and neural network-based models. Evidence was clustered into three domains: pre-procedural planning  (e.g., anatomic segmentation and trajectory optimization using CT or MRI), intraprocedural guidance (e.g., real-time needle localization or plane identification using ultrasound or robotic systems), and post-procedural outcome prediction (e.g., forecasting pain reduction following SESI based on imaging data). Most studies demonstrated improvements in trajectory planning, needle localization, segmentation efficiency, or predictive performance. Potential benefits included reduced puncture time, decreased radiation exposure, and improved patient selection. However, AI performance was model-dependent, with some systems showing higher error rates than others. Additional limitations included small datasets, heterogeneity in AI model design, and underrepresentation of patients with spinal pathology or elevated body mass index. Overall, AI shows promise for enhancing SESI procedural guidance, but larger, more standardized studies are needed before its routine clinical use.