Using Adversarial Images to Assess the Robustness of Deep Learning Models Trained on Diagnostic Images in Oncology.

Journal: JCO clinical cancer informatics
Published Date:

Abstract

PURPOSE: Deep learning (DL) models have rapidly become a popular and cost-effective tool for image classification within oncology. A major limitation of DL models is their vulnerability to adversarial images, manipulated input images designed to cause misclassifications by DL models. The purpose of the study is to investigate the robustness of DL models trained on diagnostic images using adversarial images and explore the utility of an iterative adversarial training approach to improve the robustness of DL models against adversarial images.

Authors

  • Marina Z Joel
    Department of Therapeutic Radiology, Yale School of Medicine, New Haven, CT.
  • Sachin Umrao
    Department of Therapeutic Radiology, Yale School of Medicine, New Haven, CT.
  • Enoch Chang
    Yale Department of Therapeutic Radiology, New Haven, Connecticut, USA.
  • Rachel Choi
    Department of Therapeutic Radiology, Yale School of Medicine, New Haven, CT.
  • Daniel X Yang
    Department of Therapeutic Radiology, Yale School of Medicine, New Haven, CT.
  • James S Duncan
    Biomedical Engineering, Yale University, New Haven, CT 06511, USA.
  • Antonio Omuro
    Yale Brain Tumor Center at Yale Cancer Center and Smilow Cancer Hospital.
  • Roy Herbst
    Department of Medicine, Yale School of Medicine, New Haven, CT.
  • Harlan M Krumholz
    Section of Cardiovascular Medicine, Department of Internal Medicine, Yale School of Medicine, New Haven, CT, USA.
  • Sanjay Aneja
    Yale University, New Haven, Connecticut.