Generating Synthetic Clinical Data that Capture Class Imbalanced Distributions with Generative Adversarial Networks: Example using Antiretroviral Therapy for HIV

TL;DR

This paper introduces an enhanced GAN framework with VAE and external memory to generate diverse, realistic synthetic clinical data that accurately reflects imbalanced class distributions, exemplified by HIV treatment data.

Contribution

The authors develop an extended GAN model incorporating VAE and external memory, effectively addressing mode collapse and producing high-quality synthetic healthcare data.

Findings

Overcomes mode collapse in GANs for clinical data

Generates synthetic data with accurate class imbalance representation

Maintains low patient disclosure risk and high utility

Abstract

Clinical data usually cannot be freely distributed due to their highly confidential nature and this hampers the development of machine learning in the healthcare domain. One way to mitigate this problem is by generating realistic synthetic datasets using generative adversarial networks (GANs). However, GANs are known to suffer from mode collapse thus creating outputs of low diversity. This lowers the quality of the synthetic healthcare data, and may cause it to omit patients of minority demographics or neglect less common clinical practices. In this paper, we extend the classic GAN setup with an additional variational autoencoder (VAE) and include an external memory to replay latent features observed from the real samples to the GAN generator. Using antiretroviral therapy for human immunodeficiency virus (ART for HIV) as a case study, we show that our extended setup overcomes mode collapse and generates a synthetic dataset that accurately describes severely imbalanced class distributions commonly found in real-world clinical variables. In addition, we demonstrate that our synthetic dataset is associated with a very low patient disclosure risk, and that it retains a high level of utility from the ground truth dataset to support the development of downstream machine learning algorithms.