Multimodal ChatGPT-4V for Electrocardiogram Interpretation: Promise and Limitations

Area of Science:

• Artificial Intelligence in Medicine
• Cardiology
• Medical Imaging Analysis

Background:

• Cardiovascular diseases require accurate interpretation of electrocardiograms (ECGs).
• Artificial intelligence (AI) models are increasingly explored for medical diagnostic support.
• Large language models (LLMs) with visual capabilities offer new avenues for medical data interpretation.

Purpose of the Study:

• To evaluate the performance of ChatGPT-4V in interpreting ECG waveforms.
• To assess its ability to answer multiple-choice questions related to ECGs.
• To determine its potential as a tool for cardiovascular care assistance.

Main Methods:

• The study utilized the ChatGPT-4V model, a multimodal large language model.
• The model was presented with various electrocardiogram (ECG) waveforms.
• Its interpretation accuracy was assessed against established diagnostic criteria and by evaluating its responses to ECG-related multiple-choice questions.

Main Results:

• ChatGPT-4V demonstrated notable capabilities in interpreting complex ECG waveforms.
• The model showed proficiency in answering multiple-choice questions pertaining to ECG analysis.
• Initial findings suggest a promising role for AI in supporting cardiovascular diagnostics.

Conclusions:

• ChatGPT-4V shows potential as a supplementary tool in cardiovascular care.
• Its visual interpretation abilities could aid clinicians in ECG analysis.
• Further research is warranted to validate its clinical utility and integration into healthcare workflows.