ECGAN: Self-supervised generative adversarial network for electrocardiography
This work addresses the need for accessible and private ECG datasets, particularly for arrhythmias, by providing a method to generate synthetic data, though it appears incremental as it builds on existing GAN frameworks with domain-specific conditioning and assessment.
The authors tackled the problem of generating high-quality synthetic electrocardiography (ECG) data for biomedical applications, such as rare diseases and privacy constraints, by introducing ECGAN, a self-supervised generative adversarial network. The result was a substantial improvement in morphological plausibility, synchronization, and diversity compared to state-of-the-art models, with empirical results showing enhanced performance against existing generative models for sequences and audio synthesis.
High-quality synthetic data can support the development of effective predictive models for biomedical tasks, especially in rare diseases or when subject to compelling privacy constraints. These limitations, for instance, negatively impact open access to electrocardiography datasets about arrhythmias. This work introduces a self-supervised approach to the generation of synthetic electrocardiography time series which is shown to promote morphological plausibility. Our model (ECGAN) allows conditioning the generative process for specific rhythm abnormalities, enhancing synchronization and diversity across samples with respect to literature models. A dedicated sample quality assessment framework is also defined, leveraging arrhythmia classifiers. The empirical results highlight a substantial improvement against state-of-the-art generative models for sequences and audio synthesis.