Home / Current Issue / Paper 1711963
Adversarially Robust Image Generation for Medical Imaging: Evaluating StyleGAN2 vs Diffusion Models under Noise and Domain Shift
Subject area: Biological & Medical Sciences · Area of research: Medical Image Analysis
DOI: https://doi.org/10.64388/IREV7I1-1711963
Abstract
Medical image generation using deep generative models has emerged as a promising tool for augmenting limited clinical datasets, enabling synthetic data generation for training and improving model generalization. However, adversarial robustness remains a major challenge, especially under noisy or domain-shifted conditions where models may produce artifacts that compromise diagnostic reliability. This study conducts a comprehensive comparison between two state-of-the-art frameworks?StyleGAN2 and Denoising Diffusion Probabilistic Models (DDPMs)?across noise, adversarial attacks, and modality shifts. Our experiments span over 120,000 medical images across chest X-rays and MRI scans. We observe that diffusion models exhibit a 22% higher structural fidelity and a 19% lower Fr?chet Inception Distance (FID) under Gaussian and adversarial perturbations. Additional robustness metrics show diffusion models outperform StyleGAN2 in both clean and noisy conditions, making them preferable for safety-critical applications.
How to cite this paper
@article{1711963,
author = {Rajat Yadav, Arihand Sinha, Saumya Tripathi, Arvind Goel},
title = {Adversarially Robust Image Generation for Medical Imaging: Evaluating StyleGAN2 vs Diffusion Models under Noise and Domain Shift},
journal = {Iconic Research And Engineering Journals},
year = {2023},
volume = {7},
number = {1},
pages = {758-762},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1711963.pdf},
abstract = {Medical image generation using deep generative models has emerged as a promising tool for augmenting limited clinical datasets, enabling synthetic data generation for training and improving model generalization. However, adversarial robustness remains a major challenge, especially under noisy or domain-shifted conditions where models may produce artifacts that compromise diagnostic reliability. This study conducts a comprehensive comparison between two state-of-the-art frameworks?StyleGAN2 and Denoising Diffusion Probabilistic Models (DDPMs)?across noise, adversarial attacks, and modality shifts. Our experiments span over 120,000 medical images across chest X-rays and MRI scans. We observe that diffusion models exhibit a 22% higher structural fidelity and a 19% lower Fr?chet Inception Distance (FID) under Gaussian and adversarial perturbations. Additional robustness metrics show diffusion models outperform StyleGAN2 in both clean and noisy conditions, making them preferable for safety-critical applications.},
month = {July},
doi = {https://doi.org/10.64388/IREV7I1-1711963}
}