Key result
Transformer-based GANs produce synthetic ECGs with superior morphological similarity and statistical parameters versus existing models.
Why the study?
Computer-assisted medical diagnosis requires substantial clinical ECG data, but data availability is limited by privacy concerns.
Does a Generative Adversarial Network with transformers improve the generation of synthetic ECG signals compared to existing models?
Does a Generative Adversarial Network with transformers improve the generation of synthetic ECG signals compared to existing models?
A novel GAN architecture using transformers generates synthetic ECG signals with better morphological similarity and statistical performance than existing models, potentially addressing data privacy challenges in computer-assisted medical diagnosis.
May facilitate privacy-preserving ECG dataset expansion; leaves open prospective clinical validation of diagnostic utility.
Heart diseases are a severe hazard to human health. By capturing the heart activity, electrocardiogram (ECG) tests were used to help diagnose cardiac illness. Nevertheless, the utilization of computer-assisted medical diagnosis necessitates a substantial volume of clinical data from the patient. This becomes an empirical challenge because of privacy concerns. Since the clinical data were unavailable, we present a Generative Adversarial Network (GAN), consisting of a generator composed of a transformer, a convolutional neural network (CNN), and a discriminator consisting of a transformer. Electrocardiogram information from the MIT-BIH dataset was used to train the generative adversarial network. The evaluation of the model was performed by calculating statistical parameters like RMSE, FD, and PRD and comparing them with existing published models. The results showed that the synthetic ECG produced by our generative adversarial network performed better than the existing models and had better morphological similarity to the actual ECG recordings
No takes yet. Share an insight, caveat, or question.
Kaleli et al. (2023) studied Heart diseases. Generative Adversarial Network (GAN) with transformers vs. Existing published models was evaluated on Statistical parameters (RMSE, FD, PRD) and morphological similarity. A Generative Adversarial Network with transformers produced synthetic ECGs with better morphological similarity to actual recordings and improved statistical parameters compared to existing models.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: