Application of Large Language Models in Emergency Neurology

Application of Multimodal Large Language Models in Emergency Neurology Diagnosis

Emergency neurology covers a wide range of conditions, often involving urgent situations such as acute cerebrovascular diseases, seizures, central nervous system infections, and consciousness disorders. However, due to the time constraints in emergency care and limited patient information collection, misdiagnosis and missed diagnoses are common issues. Large language models (LLMs) possess powerful natural language processing and knowledge reasoning capabilities, enabling them to directly handle and understand complex, unstructured medical data such as patient medical records, dialogue notes, and laboratory test results. LLMs show broad potential for application in complex medical scenarios. This study aims to evaluate the application value of LLMs in emergency neurology, specifically examining their diagnostic accuracy in emergency neurology conditions, analyzing the feasibility of treatment plans and further examination recommendations proposed by the model, and exploring their potential in improving diagnostic efficiency and aiding decision-making.

Study Overview

• Status: Completed
• Conditions: Emergency; Neurology
• Intervention / Treatment: Diagnostic test: Large Language Model Diagnosis

• Study Type: Observational
• Enrollment (Actual): 433

Contacts and Locations

Study Locations

• China
  • Beijing
    • Beijing, Beijing, China, 100053
      • Xuanwu Hospital, Capital Medical University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Probability Sample

Study Population

Patients in the emergency neurology department

Description

Inclusion Criteria:

• Age ≥18-80 years, male or female.
• Patients seeking emergency neurology care.
• Patients who can provide complete medical records (including consultation recordings, physical examination, test results, etc.).
• Voluntary participation and signing of informed consent.

Exclusion Criteria:

• Patients who directly enter the resuscitation process due to the severity of their condition(e.g., patients who are immediately placed in the ICU).
• Patients with unstable vital signs.
• Patients who are unable to communicate effectively (e.g., severe consciousness impairment or severe cognitive disorders).
• Patients who are currently participating in other clinical trials.

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Patients presenting to the emergency neurology department.	Diagnostic test: Large Language Model Diagnosis; Using the large language model for diagnosing emergency neurology conditions.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
dignostic accuracy	To evaluate the consistency between the diagnosis made by large language models for emergency patients and the confirmed diagnosis after inpatient or outpatient visits.	1 month

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Feasibility of treatment plans	Experts use the Emergency Treatment Recommendation Scoring Scale to evaluate the treatment suggestions from conventional methods and large language models. The maximum score is 5 and the minimum score is 1, with 5 representing strong agreement with the recommendation.	1 month
dignostic specificity	A comparison of dianostic specificity between large language model diagnosis and emergency department physicians diagnosis	1 month
Diagnostic Sensitivity	A comparison of dianostic sensitivity between large language model diagnosis and emergency department physicians diagnosis.	1 month
False Discovery Rate	A comparison of the false discovery rate between large language model diagnosis and emergency department physicians diagnosis.	1 month

Collaborators and Investigators

• Sponsor: Capital Medical University

Study record dates

Study Major Dates

• Study Start (Actual): February 1, 2025
• Primary Completion (Actual): April 7, 2025
• Study Completion (Actual): April 7, 2025

Study Registration Dates

• First Submitted: January 6, 2025
• First Submitted That Met QC Criteria: January 15, 2025
• First Posted (Actual): January 16, 2025

Study Record Updates

• Last Update Posted (Actual): April 15, 2025
• Last Update Submitted That Met QC Criteria: April 14, 2025
• Last Verified: April 1, 2025

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Pathologic Processes; Disease Attributes; Emergencies
• Other Study ID Numbers: ALEGN

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No