Generating Realistic Multi-Beat ECG Signals
This work addresses a bottleneck in healthcare applications by enabling the generation of realistic multi-minute ECG sequences, which is incremental as it builds on existing diffusion models but extends them to longer sequences.
The paper tackled the problem of generating long-form synthetic ECG signals, which existing diffusion models struggle with, by proposing a three-layer synthesis framework that first generates single beats, then inter-beat features, and finally assembles them into coherent sequences, resulting in synthetic ECGs that maintain morphological fidelity and inter-beat relationships and significantly outperform end-to-end diffusion models in arrhythmia classification tasks.
Generating synthetic ECG data has numerous applications in healthcare, from educational purposes to simulating scenarios and forecasting trends. While recent diffusion models excel at generating short ECG segments, they struggle with longer sequences needed for many clinical applications. This paper proposes a novel three-layer synthesis framework for generating realistic long-form ECG signals. We first generate high-fidelity single beats using a diffusion model, then synthesize inter-beat features preserving critical temporal dependencies, and finally assemble beats into coherent long sequences using feature-guided matching. Our comprehensive evaluation demonstrates that the resulting synthetic ECGs maintain both beat-level morphological fidelity and clinically relevant inter-beat relationships. In arrhythmia classification tasks, our long-form synthetic ECGs significantly outperform end-to-end long-form ECG generation using the diffusion model, highlighting their potential for increasing utility for downstream applications. The approach enables generation of unprecedented multi-minute ECG sequences while preserving essential diagnostic characteristics.