Related Experiment Video
Updated: Sep 17, 2025
08:22
Author Spotlight: Advancing the Study of Brain-Heart Interplay with a Comprehensive EEGLAB Plugin for Multimodal Signal Analysis
Published on: April 26, 2024
2.2K
Task-specific versus general-purpose AI models in ECG analysis: A comparative study with emergency medicine
Ertugrul Altinbilek1, Adem Az2, Ozgur Sogut2
1University of Health Sciences, Şişli Hamidiye Etfal Training and Research Hospital, Department of Emergency Medicine, Istanbul, Turkey.
The American Journal of Emergency Medicine
|July 3, 2025
Summary
A specialized AI model, ECG Reader-GPT, showed diagnostic accuracy comparable to emergency medicine specialists in interpreting electrocardiograms (ECGs). It outperformed general AI models, highlighting the value of domain-specific AI tools in clinical settings.
Area of Science:
- Cardiology
- Artificial Intelligence
- Medical Diagnostics
Background:
- Electrocardiogram (ECG) interpretation is crucial for diagnosing various cardiac conditions.
- The integration of Artificial Intelligence (AI) into medical diagnostics offers potential for enhanced accuracy and efficiency.
- General-purpose AI models may have limitations in specialized medical tasks like ECG analysis.
Purpose of the Study:
- To compare the diagnostic accuracy of three AI models (GPT-4o, Canva-GPT, ECG Reader-GPT) against emergency medicine specialists (EMSs).
- To evaluate AI model performance in interpreting electrocardiograms (ECGs) using a standardized test set.
- To assess the impact of domain specialization on AI diagnostic accuracy in cardiology.
Main Methods:
- A prospective diagnostic accuracy study involving 50 ECG questions from a validated test set.
- Thirty emergency medicine specialists (EMSs) completed the ECG test.
- Three AI models were evaluated daily over 30 days on the same test set.
- Diagnostic accuracy was compared across ECG subcategories and clinical case types.
Main Results:
- Emergency medicine specialists achieved the highest overall diagnostic accuracy (median: 41.5).
- ECG Reader-GPT demonstrated comparable accuracy to EMSs (median: 39.5), with no statistically significant difference (p=0.530).
- ECG Reader-GPT significantly outperformed GPT-4o and Canva-GPT across all categories (p < 0.001), particularly in rhythm disorders.
Conclusions:
- ECG Reader-GPT, a specialized AI model, exhibits diagnostic accuracy comparable to experienced emergency medicine specialists.
- Domain-specific AI models demonstrate superior performance over general-purpose models in ECG interpretation.
- These findings underscore the clinical utility of specialized AI tools for enhancing ECG analysis.
