Adversarial attacks and adversarial robustness in computational pathology

Nature Communications, vol. 13, pp. 5711

Abstract

Artificial Intelligence (AI) can support diagnostic workflows in oncology by aiding diagnosis and providing biomarkers directly from routine pathology slides. However, AI applications are vulnerable to adversarial attacks. Hence, it is essential to quantify and mitigate this risk before widespread clinical use. Here, we show that convolutional neural networks (CNNs) are highly susceptible to white- and black-box adversarial attacks in clinically relevant weakly-supervised classification tasks. Adversarially robust training and dual batch normalization (DBN) are possible mitigation strategies but require precise knowledge of the type of attack used in the inference. We demonstrate that vision transformers (ViTs) perform equally well compared to CNNs at baseline, but are orders of magnitude more robust to white- and black-box attacks. At a mechanistic level, we show that this is associated with a more robust latent representation of clinically relevant categories in ViTs compared to CNNs. Our results are in line with previous theoretical studies and provide empirical evidence that ViTs are robust learners in computational pathology. This implies that large-scale rollout of AI models in computational pathology should rely on ViTs rather than CNN-based classifiers to provide inherent protection against perturbation of the input data, especially adversarial attacks.

Authors 11

1. Narmin Ghaffari Laleh Aachen Department of Medicine III

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Department of Medicine III, University Hospital RWTH Aachen, RWTH Aachen university, Aachen, Germany

2. Daniel Truhn Aachen Department of Diagnostic and Interventional Radiology

   Universitätsklinikum Aachen

   Affiliation as printed

   Department of Diagnostic and Interventional Radiology, University Hospital Aachen, Aachen, Germany

3. Gregory Patrick Veldhuizen

   Else Kröner Fresenius Center for Digital Health · Technische Universität Dresden

   Affiliation as printed

   Else Kroener Fresenius Center for Digital Health, Medical Faculty Carl Gustav Carus, Technical University Dresden, Dresden, Germany

4. Tianyu Han Aachen Department of Physics of Molecular Imaging Systems Institute for Experimental Molecular Imaging

   RWTH Aachen University

   Affiliation as printed

   Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany

5. Marko van Treeck Aachen Department of Medicine III

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Department of Medicine III, University Hospital RWTH Aachen, RWTH Aachen university, Aachen, Germany

6. Roman David Buelow Aachen Institute of Pathology

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Institute of Pathology, University Hospital RWTH Aachen, Aachen, Germany

7. Rupert Langer

   University of Bern · Johannes Kepler University of Linz · Kepler Universitätsklinikum

   Affiliation as printed

   Institute of Pathology and Molecular Pathology, Kepler University Hospital, Johannes Kepler University Linz, Linz, Austria

   Institute of Pathology, University of Bern, Bern, Switzerland

8. Bastian Dislich

   University of Bern

   Affiliation as printed

   Institute of Pathology, University of Bern, Bern, Switzerland

9. Peter Boor Aachen Institute of Pathology

   Universitätsklinikum Aachen · RWTH Aachen University

   Affiliation as printed

   Institute of Pathology, University Hospital RWTH Aachen, Aachen, Germany

10. Volkmar Schulz Aachen Department of Physics of Molecular Imaging Systems Institute for Experimental Molecular Imaging Physics Institute III B

    Fraunhofer Institute for Digital Medicine · RWTH Aachen University

    Affiliation as printed

    Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany

    Fraunhofer Institute for Digital Medicine MEVIS, Aachen, Germany

    Hyperion Hybrid Imaging Systems GmbH, Aachen, Germany

    Physics Institute III B, RWTH Aachen University, Aachen, Germany

11. Jakob Nikolas Kather corresponding Aachen Department of Medicine III

    University of Leeds · Heidelberg University · University Hospital Heidelberg · National Center for Tumor Diseases · Universitätsklinikum Aachen · University Hospital Carl Gustav Carus · Else Kröner Fresenius Center for Digital Health · Technische Universität Dresden · RWTH Aachen University

    Affiliation as printed

    Department of Medicine 1, University Hospital and Faculty of Medicine Carl Gustav Carus, Technical University Dresden, Dresden, Germany. jakob_nikolas.kather@tu-dresden.de

    Department of Medicine III, University Hospital RWTH Aachen, RWTH Aachen university, Aachen, Germany. jakob_nikolas.kather@tu-dresden.de

    Division of Pathology and Data Analytics, Leeds Institute of Medical Research at St James's, University of Leeds, Leeds, UK. jakob_nikolas.kather@tu-dresden.de

    Else Kroener Fresenius Center for Digital Health, Medical Faculty Carl Gustav Carus, Technical University Dresden, Dresden, Germany. jakob_nikolas.kather@tu-dresden.de

    Medical Oncology, National Center for Tumor Diseases (NCT), University Hospital Heidelberg, Heidelberg, Germany. jakob_nikolas.kather@tu-dresden.de

    Else Kroener Fresenius Center for Digital Health, Medical Faculty Carl Gustav Carus, Technical University Dresden, Dresden, Germany