Key points are not available for this paper at this time.
INTRODUCTION: Cauda Equina Syndrome (CES) is a neurological emergency requiring rapid diagnosis. Traditional diagnostic methods face challenges due to variable presentations. This review evaluates the clinical performance, limitations, and validation needs of artificial intelligence and machine learning (AI/ML) compared to traditional CES diagnostic approaches. METHODS: A systematic PRISMA 2020 search of PubMed and ScienceDirect (2019-2025) using "cauda equina" yielded 520 articles. After removing 200 duplicates and 284 non-diagnostic papers, 36 diagnostic studies were identified. Ultimately, 9 studies evaluating distinct diagnostic techniques were included in the final analysis. RESULTS: Traditional approaches, such as GIRFT red flags, demonstrated limited specificity. Conversely, three AI/ML methodologies showed superior performance: Large Language Models (ChatGPT-3.5, Google Bard) successfully identified red flags in low back pain cases; machine learning algorithms improved predictive accuracy for surgical planning; and computer vision enhanced the detection of cauda equina compression on MRI scans. CONCLUSION: AI/ML applications offer promising diagnostic capabilities that could improve triage accuracy and optimise imaging utilisation for CES. However, safe translation into routine clinical practice requires prospective, multicenter validation to overcome existing limitations in sample size and diversity.
Aggarwal et al. (Mon,) studied this question.