Artificial intelligence in neuroradiology for early stratification of critically ill neurological patients and optimisation of clinical decision-making: a scoping review.

Introduction: Critical neurological diseases constitute a major cause of mortality, disability, and healthcare resource utilisation worldwide. In this context, neuroradiology plays a central role in the early stratification of critically ill neurological patients by guiding time-sensitive diagnostic and therapeutic decisions. However, the increasing volume and complexity of neuroimaging have generated growing clinical pressure, driving interest in artificial intelligence (AI)–based tools to support diagnostic prioritisation, risk stratification, and clinical decision-making. Despite the rapid expansion of the literature, the available evidence remains fragmented and heterogeneous. Methodology: A scoping review will be conducted in accordance with the Joanna Briggs Institute recommendations and reported following the PRISMA-ScR guidelines. A systematic search will be performed in PubMed, Scopus, and Web of Science Core Collection, including studies published between 2015 and 2025. Original studies in living adult populations evaluating AI applications in neuroradiology for early clinical risk stratification, diagnostic prioritisation, or decision support will be considered. Study selection will be carried out in two phases through independent screening, with reference management performed using Rayyan. Data extraction will be conducted using a standardised Microsoft Excel spreadsheet, and findings will be synthesised using descriptive and narrative approaches. Expected results: We anticipate identifying a predominance of studies focused on neurovascular conditions and traumatic brain injury, with computed tomography and deep learning models being most frequently employed. The evidence is expected to be largely concentrated in retrospective studies with an emphasis on technical performance metrics, while prospective evaluations, external validation, and clinical impact studies are likely to be limited. Additionally, relevant gaps are expected to emerge regarding the integration of AI into neuroradiological workflows and its real-world impact on care timelines and clinical decision-making. This review will provide a structured mapping of the current state of the field and inform future research towards clinically meaningful and methodologically robust applications.