Assessing the ability of generative adversarial networks to learn canonical medical image statistics

TL;DR

This paper evaluates whether modern GANs can reliably learn the statistical properties of canonical medical image models, revealing strengths in basic statistics but shortcomings in more complex, per-image metrics crucial for objective image quality assessment.

Contribution

It demonstrates that while GANs can learn some basic image statistics, they often fail to capture more complex, per-image statistics relevant for medical image quality evaluation.

Findings

GANs learned basic first- and second-order statistics

GANs generated high perceptual quality images

GANs failed to learn certain per-image statistics

Abstract

In recent years, generative adversarial networks (GANs) have gained tremendous popularity for potential applications in medical imaging, such as medical image synthesis, restoration, reconstruction, translation, as well as objective image quality assessment. Despite the impressive progress in generating high-resolution, perceptually realistic images, it is not clear if modern GANs reliably learn the statistics that are meaningful to a downstream medical imaging application. In this work, the ability of a state-of-the-art GAN to learn the statistics of canonical stochastic image models (SIMs) that are relevant to objective assessment of image quality is investigated. It is shown that although the employed GAN successfully learned several basic first- and second-order statistics of the specific medical SIMs under consideration and generated images with high perceptual quality, it failed to correctly learn several per-image statistics pertinent to the these SIMs, highlighting the urgent need to assess medical image GANs in terms of objective measures of image quality.