Diffusion-based conditional ECG generation with structured state space models

Juan Miguel Lopez Alcaraz, Nils Strodthoff

Abstract: Highlights•Synthetic ECG generation conditioned on 71 ECG statements.•State-of-the-art diffusion model backed by structured state-space models.•Proposed model outperforms GAN-based models according to different metrics.•Meaningful class interpolation in line with ECG domain knowledge.•Turing test with cardiologist confirms high quality of the generated samples.