StyleGAN2-based Out-of-Distribution Detection for Medical Imaging

TL;DR

This paper presents a GAN-based method using StyleGAN2-ADA to detect out-of-distribution medical images, achieving over 90% AUROC in liver CT classification and effectively identifying artifacts.

Contribution

The study introduces a novel OOD detection approach for medical imaging using StyleGAN2-ADA and reconstruction-based metrics, tailored for liver CT scans.

Findings

Achieved >90% AUROC in liver vs. non-liver classification

Successfully identified liver artifacts like needles and ascites

Demonstrated effective OOD detection in medical images

Abstract

One barrier to the clinical deployment of deep learning-based models is the presence of images at runtime that lie far outside the training distribution of a given model. We aim to detect these out-of-distribution (OOD) images with a generative adversarial network (GAN). Our training dataset was comprised of 3,234 liver-containing computed tomography (CT) scans from 456 patients. Our OOD test data consisted of CT images of the brain, head and neck, lung, cervix, and abnormal livers. A StyleGAN2-ADA architecture was employed to model the training distribution. Images were reconstructed using backpropagation. Reconstructions were evaluated using the Wasserstein distance, mean squared error, and the structural similarity index measure. OOD detection was evaluated with the area under the receiver operating characteristic curve (AUROC). Our paradigm distinguished between liver and non-liver CT with greater than 90% AUROC. It was also completely unable to reconstruct liver artifacts, such as needles and ascites.