IMAGE AUGMENTATION ON ECG IMAGE DATASET USING GAN

Abstract

Customized healthcare should organize an environment for the manipulation of reliable and secure private data. Using Generative Adversarial Networks (GANs), this paper proposes a method for constructing synthetic ECGs (electrocardiograms) (GANs). The objective is to develop data that may be used in educational and research contexts while reducing the risk of sensitive data leakage to the absolute minimum possible. For GANs to work, we recommend converting raw data to an image and then decoding it back to the original data domain so that GANs may operate on photos and video frames. Our transformation and processing theory' viability is basically shown. The primary disadvantages of each stage in the suggested process are then discussed for the specific situation of ECGs. As a result, a new study avenue into the use of GANs to anonymize health data is opened, and simple new advancements are anticipated.