Tutorial on

Generative adversarial networks and adversarial methods in biomedical image analysis

MICCAI 2019 - Sunday, October 13 - Shenzhen, China

Scope

Generative adversarial networks (GANs) are a powerful subclass of deep generative models that are currently receiving widespread attention from not only the computer vision and machine learning communities, but also the medical imaging community. The key idea behind GANs is that two neural networks are jointly optimized: one network tries to synthesize samples that resemble real data points while a second network assesses how well the result corresponds to a reference database of samples.

Using GANs for sample synthesis could potentially be used to address the shortage of large and diverse annotated databases. In addition, the concept of two networks that are optimized in an adversarial game has been exploited to provide an additional loss term to improve performance of existing image analysis methods. Adversarial methods have been successfully exploited in typical medical image analysis applications such as denoising, reconstruction, segmentation, and detection. Moreover, adversarial training has led to new applications in paradigms such as semi-supervised learning and abnormality detection.

In this tutorial, we will provide basic as well as advanced material on GANs and adversarial methods in medical image analysis in five sessions. We will focus on key state-of-the-art papers in the machine learning and computer vision literature and their relation to works in medical image analysis. To make these concepts tangible, we will provide examples of applications in medical imaging, both derived from our own work and that of others.

Topics

Generative adversarial networks

Anirban Mukhopadhyay, Technische Universität Darmstadt, Germany

Adversarial training methods

Jelmer Wolterink, Image Sciences Institute, University Medical Center Utrecht, The Netherlands

Adversarial methods for domain adaptation and semi-supervised learning

Konstantinos Kamnitsas, Imperial College London, United Kingdom

Adversarial methods for image segmentation, synthesis and quality enhancement

Jelmer Wolterink, Image Sciences Institute, University Medical Center Utrecht, The Netherlands

Organizers

Anirban Mukhopadhyay, Technische Universität Darmstadt, Darmstadt, Germany

Arjan Kuijper, Technische Universität Darmstadt & Fraunhofer IGD, Darmstadt, Germany

Ivana Išgum, Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands

Jelmer Wolterink, Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands

Bram van Ginneken, Diagnostic Image Analysis Group, Radboud University Medical Center, Nijmegen, The Netherlands