loading . . . Diagnostic accuracy of the ROSIER scale for detecting acute ischemic stroke: a comparison with FAST-ED, CPSS, and reference criteria - BMC Emergency Medicine Background Acute ischemic stroke (AIS) requires rapid diagnosis and treatment. While neuroimaging remains the gold standard, it is often unavailable in early emergency or prehospital settings. Clinical stroke scales may bridge this gap. Thus, this study aimed to compare the diagnostic accuracy of the ROSIER scale with the Cincinnati Prehospital Stroke Scale (CPSS), Field Assessment Stroke Triage for Emergency Destination (FAST-ED), and National Institutes of Health Stroke Scale (NIHSS) for the detection of AIS in patients with suspected stroke, using MRI as the imaging reference standard. Methods This prospective cross-sectional study was performed on adult patients admitted to the emergency department (ED) between June 2021 and 2023, with acute neurological symptoms and underwent brain MRI for a suspected stroke. After data gathering, receiver operating characteristic (ROC) curve analysis was performed, and sensitivity, specificity, positive predictive value, and negative predictive value of ROSIER for detecting of AIS were calculated and compared with CPSS, FAST-ED, and NIHSS. MRI was considered the gold standard for the diagnosis of AIS. Results A total of 655 patients were included, and 477 (72.8%) were diagnosed with AIS. The mean age was 68.43 ± 13.01 years, and 57.3% were male. The AUROC values for ROSIER, CPSS, FAST-ED, and NIHSS were 0.813, 0.860, 0.811, and 0.822, respectively. No significant differences were found between the tools. Sensitivities for ROSIER ≥ 2, CPSS ≥ 2, FAST-ED ≥ 3, and NIHSS ≥ 8 were 86.8%, 90.7%, 74.7%, and 73.5%, respectively. Their specificities were 58.1%, 67.7%, 77.4%, and 81.5%, respectively. Conclusion All four tools showed good diagnostic accuracy. CPSS was the most sensitive and suitable for prehospital detection, whereas NIHSS offered the best specificity for in-hospital confirmation. Selection of the tool should be context-dependent, based on resources, time, and examiner expertise. https://doi.org/10.1186/s12873-026-01654-0