Timezone: »

StyleGAN2-based Out-of-Distribution Detection for Medical Imaging
McKell Woodland · John Wood · Caleb O'Connor · Ankit Patel · Kristy Brock

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.

Author Information

McKell Woodland (MD Anderson Cancer Center; Rice University)

I am a computer science Ph.D. student at Rice University and I do my research through the Morfeus lab at MD Anderson Cancer Center.

John Wood (MD Anderson Cancer Center)
Caleb O'Connor (MD Anderson Cancer Center)
Ankit Patel (Rice University)
Kristy Brock (MD Anderson Cancer Centre)

More from the Same Authors