Post

New research · Emergency Medicine
NPJ digital medicine · 1d
AI / informaticsNPJ digital medicine · 2026

Development and prospective shadow evaluation of a domain-specific large language model for emergency neurological diagnosis.

Yibing Guo, Xiangbin Meng, Erlan Yu … Chuanjie Wu
Read paper
Emergency MedicineAI / informatics

Large language model achieved 79.4% accuracy in emergency neurology, outperforming physicians.

Development and prospective shadow evaluation of a domain-specific large language model for emergency neurological diagnosis.

Yibing Guo et al. · NPJ digital medicine · 2026
Background

Large language models (LLMs) show promise in emergency medicine; however, their role in emergency neurology remains unclear.

Methods

We developed a customized LLM, Xuanwu-NeuroAid, and prospectively enrolled 433 patients.

n = 433 patients
Results
the model correctly identified the right diagnosis for brain and nerve emergencies
n = 433 patients
More results

The diagnostic outputs of the model and emergency physicians were compared with confirmed diagnoses, and an expert panel performed blinded evaluations of the recommendations generated by both the model and physicians.

More results

Blinded expert assessments indicated that the model's recommendations for examinations and treatments were significantly more comprehensive, accurate, and clinically applicable than those of physicians (p < 0.001).

“
Conclusion

Our findings highlight the potential of the LLM to enhance diagnostic precision and support decision-making in emergency neurology under simulated clinical conditions.

Read paper
0 comments

No comments yet. Be the first.

Related papers

LatestFoundational
AI / Informatics
odds ratio 2.37
Artificial intelligence assistance more than doubled the chances of correctly identifying a collapsed lung from injury
AI / Informatics
0.844
how well the new model predicted septic shock
Guideline
5 days
new guidelines recommend less antibiotic treatment than this for nonsevere lung infection
Cohort Study
76%
76% of patients offered virtual care were successfully diverted from the physical emergency department.
Cross-sectional
62%
most clinicians were interested in a research study for lower leg skin infection
Cohort Study
29.0 min
Patients during unscheduled drills had an average door-to-triage time of 29.0 minutes.
Cohort Study
IRR 0.21
Community Care Team Transitional Care Programme was associated with a lower 90-day death rate