A prospective randomised controlled trial of an AI-supported model of care for non-contrast Computed Tomography (CT) Brain interpretation in the Emergency Department

Patients will be triaged into the emergency department as per standard practice. Patients will be screened by the treating clinician or researcher for eligibility. Eligible patients who provide informed consent will be randomised in a 1:1 ratio. - Intervention Group (AI-Supported Care Model): Following scan acquisition, the images will be processed by the AI tool. A senior emergency department clinician will review the images with the assistance of the AI output, which provides a binary classification of 'remarkable' or 'unremarkable'. - If there is a consensus between the senior clinician and the AI that the scan is unremarkable, the patient can be considered for expedited discharge. - Discharged patients will be referred to the Royal Prince Alfred Virtual Hospital for a follow-up call from a physician within 24 hours, by which time the formal radiologist report will be available to confirm the unremarkable finding. - If the AI tool indicates a 'remarkable' scan, or if the clinician identifies a remarkable finding regardless of the AI output, the patient will follow the standard care pathway and remain in the emergency department to await the formal report. - If an AI deemed ‘unremarkable’ scan is later found to be remarkable by a radiologist this will be communicated to an emergency department doctor and the patient contacted and asked to return for in person assessment if necessary (as per standard practice for missed diagnoses). Participants undergo the same standard-of-care non-contrast Computed Tomography (CT) Brain scan that would ordinarily be ordered for clinical management. No additional imaging, radiation exposure, or scan acquisition procedures are required as part of the study. The intervention relates only to the interpretation workflow following image acquisition. Brief description of the AI tool - The AI tool is Annalise Enterprise V3, a commercially available, TGA-cleared (ARTG 513346) Class IIb software-as-a-medical-device. It functions as a computer-aided detection system that automatically analyses eligible non-contrast CT head scans and provides real-time outputs, visual overlays, and a study-level classification to assist clinician interpretation. The tool provides a binary classification of “remarkable” or “unremarkable” and is intended to support, rather than replace, clinical decision-making.