On instabilities of deep learning in image reconstruction and the potential costs of AI

@article{Antun2020OnIO,
  title={On instabilities of deep learning in image reconstruction and the potential costs of AI},
  author={Vegard Antun and Francesco Renna and C. Poon and B. Adcock and A. Hansen},
  journal={Proceedings of the National Academy of Sciences},
  year={2020},
  volume={117},
  pages={30088 - 30095}
}
  • Vegard Antun, Francesco Renna, +2 authors A. Hansen
  • Published 2020
  • Computer Science, Medicine
  • Proceedings of the National Academy of Sciences
  • Deep learning, due to its unprecedented success in tasks such as image classification, has emerged as a new tool in image reconstruction with potential to change the field. In this paper, we demonstrate a crucial phenomenon: Deep learning typically yields unstable methods for image reconstruction. The instabilities usually occur in several forms: 1) Certain tiny, almost undetectable perturbations, both in the image and sampling domain, may result in severe artefacts in the reconstruction; 2) a… CONTINUE READING
    73 Citations
    Interval Neural Networks as Instability Detectors for Image Reconstructions
    • Highly Influenced
    • PDF
    Adversarial Robust Training in MRI Reconstruction
    • 1
    • PDF
    Improving Robustness of Deep-Learning-Based Image Reconstruction
    • 7
    • Highly Influenced
    • PDF
    On the existence of stable and accurate neural networks for image reconstruction
    • 1
    • Highly Influenced
    • PDF
    Stabilizing Deep Tomographic Reconstruction Networks
    • 4
    • PDF
    Overcoming Measurement Inconsistency in Deep Learning for Linear Inverse Problems: Applications in Medical Imaging
    • PDF
    Artificial intelligence in image reconstruction: The change is here.
    • Ramandeep Singh, Weiwen Wu, Ge Wang, M. Kalra
    • Computer Science, Medicine
    • Physica medica : PM : an international journal devoted to the applications of physics to medicine and biology : official journal of the Italian Association of Biomedical Physics
    • 2020

    References

    SHOWING 1-10 OF 54 REFERENCES
    The Robustness of Deep Networks: A Geometrical Perspective
    • 88
    • PDF
    Neural Proximal Gradient Descent for Compressive Imaging
    • 61
    • PDF
    DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks
    • 2,054
    • PDF
    DAGAN: Deep De-Aliasing Generative Adversarial Networks for Fast Compressed Sensing MRI Reconstruction
    • 333
    • PDF
    One Network to Solve Them All — Solving Linear Inverse Problems Using Deep Projection Models
    • 203
    • PDF
    MR Image Reconstruction Using Deep Density Priors
    • 38
    • PDF
    Deep Convolutional Neural Network for Inverse Problems in Imaging
    • 853
    • PDF
    Geometric Robustness of Deep Networks: Analysis and Improvement
    • 64
    • PDF