Integrating Human-AI Collaboration in ECG Analysis for Clinical Advancement.

Area of Science:

• Cardiology
• Artificial Intelligence in Healthcare
• Clinical Informatics

Background:

• Electrocardiogram (ECG) interpretation faces challenges in reliability and scalability due to human variability and increasing diagnostic demands.
• Existing artificial intelligence (AI) solutions in healthcare have shown potential but struggle with real-world clinical integration.
• Fuwai Hospital, a leading cardiovascular center, sought to improve ECG analysis efficiency and accuracy.

Purpose of the Study:

• To develop and implement the Synergistic Human-AI Partnership for ECG analysis (SHAPE) system.
• To establish a model for effective human-AI collaboration in clinical settings.
• To enhance ECG interpretation accuracy, efficiency, and clinician capacity.

Main Methods:

• Developed a closed-loop architecture with five interconnected platforms: Benchmark Labeling, Model Development, Clinical Collaboration, Education and Evaluation, and Visualization and Management.
• Implemented a rigorous two-stage expert adjudication process for generating gold-standard reference data.
• Integrated AI seamlessly into routine clinical workflows, fostering clinician engagement and feedback.

Main Results:

• Achieved a significant increase in diagnostic accuracy from 96.76% to 98.58% (P<0.001).
• Nearly doubled per-physician productivity, increasing daily interpretation volume from 272 to 504 reports, despite a 50% reduction in reporting staff.
• Increased clinical reliance on AI outputs to over 95% by mid-2025 (P<0.001).

Conclusions:

• The SHAPE system demonstrates substantial clinical and operational gains through human-AI collaboration.
• AI's greatest potential in healthcare lies in synergistic collaboration, not just automation.
• The SHAPE model offers a replicable blueprint for integrating AI to enhance care quality and clinician capacity in cardiovascular medicine.