UniECG: Understanding and Generating ECG in One Unified Model

The paper presents UniECG, a unified model designed for interactive ECG education that generates evidence-based explanations from ECG signals and creates ECG examples from textual learning objectives. It utilizes a two-stage architecture that learns from ECG signal-image-text data and employs specialized ECG generation tokens aligned with a pretrained text-conditioned ECG diffusion model for controllable signal generation. This approach enhances educational methodologies in ECG interpretation, providing a novel tool for case-based learning in medical education.