Generative Adversarial Networks for ECG generation, translation, imputation and denoising
Abstract
Artificial Intelligence is increasingly being used in medical applications. One challenge present in AI in medicine is not having high quality datasets available for training machine learning models. In this work, I explore two different methods of generating high quality medical data. In the first approach, I used a cycleGANs as novel method for ECG trans- lation, imputation and denoising. In the second method, I present a novel algorithm for generating high quality ECG data that uses a machine learning framework called Gener- ative Adversarial Networks and explanation AI systems. Through empirical evaluation, I show that both methods improve over state-of-the-art methods in their respective appli- cations. This thesis demonstrates that machine learning methods can be used to address the data scarcity problem in the medical setting.
Collections
Cite this version of the work
Alaina Mahalanabis
(2022).
Generative Adversarial Networks for ECG generation, translation, imputation and denoising. UWSpace.
http://hdl.handle.net/10012/18676
Other formats