Literature DB >> 31539791

Injecting and removing suspicious features in breast imaging with CycleGAN: A pilot study of automated adversarial attacks using neural networks on small images.

Anton S Becker1, Lukas Jendele2, Ondrej Skopek2, Nicole Berger3, Soleen Ghafoor4, Magda Marcon3, Ender Konukoglu5.   

Abstract

PURPOSE: To train a CycleGAN on downscaled versions of mammographic data to artificially inject or remove suspicious features, and to determine whether these AI-mediated attacks can be detected by radiologists.
MATERIAL AND METHODS: From two publicly available datasets, BCDR and INbreast, we selected 680 images with and without lesions as training data. An internal dataset (n = 302 cancers, n = 590 controls) served as test data. We ran two experiments (256 × 256 px and 512 × 408 px) and applied the trained model to the test data. Three radiologists read a set of images (modified and originals) and rated the presence of suspicious lesions on a scale from 1 to 5 and the likelihood of the image being manipulated. The readout was evaluated by multiple reader multiple case receiver operating characteristics (MRMC-ROC) analysis using the area under the curve (AUC).
RESULTS: At the lower resolution, the overall performance was not affected by the CycleGAN modifications (AUC 0.70 vs. 0.76, p = 0.67). However, one radiologist exhibited lower detection of cancer (0.85 vs 0.63, p = 0.06). The radiologists could not discriminate between original and modified images (0.55, p = 0.45). At the higher resolution, all radiologists showed significantly lower detection rate of cancer in the modified images (0.80 vs. 0.37, p < 0.001), however, they were able to detect modified images due to better visibility of artifacts (0.94, p < 0.0001).
CONCLUSION: Our proof-of-concept study shows that CycleGAN can implicitly learn suspicious features and artificially inject or remove them in existing images. The applicability of the method is currently limited by the small image size and introduction of artifacts.
Copyright © 2019 Elsevier B.V. All rights reserved.

Entities:  

Keywords:  Cancer; Cyber security; GAN; Mammography

Year:  2019        PMID: 31539791     DOI: 10.1016/j.ejrad.2019.108649

Source DB:  PubMed          Journal:  Eur J Radiol        ISSN: 0720-048X            Impact factor:   3.528


  3 in total

1.  A Cyber-Security Risk Assessment Methodology for Medical Imaging Devices: the Radiologists' Perspective.

Authors:  Tom Mahler; Erez Shalom; Arnon Makori; Yuval Elovici; Yuval Shahar
Journal:  J Digit Imaging       Date:  2022-02-17       Impact factor: 4.903

2.  Connected-UNets: a deep learning architecture for breast mass segmentation.

Authors:  Asma Baccouche; Begonya Garcia-Zapirain; Cristian Castillo Olea; Adel S Elmaghraby
Journal:  NPJ Breast Cancer       Date:  2021-12-02

Review 3.  Application of generative adversarial networks (GAN) for ophthalmology image domains: a survey.

Authors:  Aram You; Jin Kuk Kim; Ik Hee Ryu; Tae Keun Yoo
Journal:  Eye Vis (Lond)       Date:  2022-02-02
  3 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.