Home › Stroke & neuro › AI triage of ED CT brain scans

Stroke & neuro

AI triage of ED CT brain scans

Artificial Intelligence Triage of Urgent Versus Non‐Urgent CT Brain Findings to Support Expedited Emergency Department Disposition: A Retrospective Validation Study
Emergency Medicine Australasia · Added 29 September 2026

Retrospective validation of an AI model classifying 3,424 ED CT brain scans as urgent or not — 98.3% negative predictive value with no false negative needing urgent intervention at the index visit, putting ~64% of scans in the possibly-expedite bucket pending a prospective trial.

Study type

Cross-sectional

Design

Retrospective validation study of an AI model

Participants

3,424 adult non-contrast ED CT brain scans

Main finding

AI sensitivity 85.4% (95% CI 80.0–89.0%), specificity 69.1%, NPV 98.3%; post hoc re-adjudication gave sensitivity 94.3% and NPV 99.4%; ~64% of encounters potentially eligible for expedited disposition; no false negative required urgent intervention at index visit.

Caveat

Retrospective single-centre design with reference standard from radiologist reports; needs prospective validation.

Bottom line

Hypothesis-generating

Learning goal

Understand how AI-supported triage of CT brain findings could shape safe expedited emergency department disposition decisions.

More

• Topic: Stroke & neuro
• Daily scan: 29 September 2026
• Podcast: Paramedic Papers Daily, 29 September 2026

Related papers on Stroke & neuro

• Consensus on designing prehospital stroke trials

  Stroke · 6 Oct 2026 · Guideline / statement · Context only

  An expert consensus statement setting out operational, ethical and technological priorities for designing prehospital stroke trials in conventional EMS settings, covering consent models, randomisation, endpoints and statistical frameworks.

• Online stroke education boosts EMS-seeking

  Journal of Education and Health Promotion · 5 Oct 2026 · RCT · Hypothesis-generating

  RCT in 106 high-risk residents in Huanggang, China testing a 6-week online stroke education program against usual health education — it significantly improved stroke knowledge, warning-sign awareness and EMS-seeking intent.

• Two-thirds of stroke patients delay seeking care

  Cerebrovascular Diseases · 3 Oct 2026 · Systematic review · Context only

  Meta-analysis of 27 studies and 266,336 ischaemic stroke patients across 14 countries found 67.7% experienced prehospital delay before seeking care, with stroke knowledge and symptom recognition the key modifiable drivers.

• Stroke care gaps for First Nations patients

  The Medical Journal of Australia · 3 Oct 2026 · Cohort · Context only

  Victorian linked-data study of 32,163 EMS-attended stroke/TIA patients found First Nations patients had more than double the age-standardised incidence of EMS attendance, were less likely to receive a paramedic stroke assessment (58.2% vs 67.3%) and less often taken to thrombectomy-capable hospitals, though adjusted mortality was similar.