{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/injecting-and-removing-malignant-features-in","title":"Injecting and removing malignant features in mammography with CycleGAN: Investigation of an automated adversarial attack using neural networks","arxiv_id":"1811.07767","date":"2018-11-19","proceeding":null,"authors":["Anton S. Becker","Lukas Jendele","Ondrej Skopek","Nicole Berger","Soleen Ghafoor","Magda Marcon","Ender Konukoglu"],"abstract":"$\\textbf{Purpose}$ To train a cycle-consistent generative adversarial network\n(CycleGAN) on mammographic data to inject or remove features of malignancy, and\nto determine whether these AI-mediated attacks can be detected by radiologists.\n$\\textbf{Material and Methods}$ From the two publicly available datasets, BCDR\nand INbreast, we selected images from cancer patients and healthy controls. An\ninternal dataset served as test data, withheld during training. We ran two\nexperiments training CycleGAN on low and higher resolution images ($256 \\times\n256$ px and $512 \\times 408$ px). Three radiologists read the images and rated\nthe likelihood of malignancy on a scale from 1-5 and the likelihood of the\nimage being manipulated. The readout was evaluated by ROC analysis (Area under\nthe ROC curve = AUC). $\\textbf{Results}$ At the lower resolution, only one\nradiologist exhibited markedly lower detection of cancer (AUC=0.85 vs 0.63,\np=0.06), while the other two were unaffected (0.67 vs. 0.69 and 0.75 vs. 0.77,\np=0.55). Only one radiologist could discriminate between original and modified\nimages slightly better than guessing/chance (0.66, p=0.008). At the higher\nresolution, all radiologists showed significantly lower detection rate of\ncancer in the modified images (0.77-0.84 vs. 0.59-0.69, p=0.008), however, they\nwere now able to reliably detect modified images due to better visibility of\nartifacts (0.92, 0.92 and 0.97). $\\textbf{Conclusion}$ A CycleGAN can\nimplicitly learn malignant features and inject or remove them so that a\nsubstantial proportion of small mammographic images would consequently be\nmisdiagnosed. At higher resolutions, however, the method is currently limited\nand has a clear trade-off between manipulation of images and introduction of\nartifacts.","url_abs":"http://arxiv.org/abs/1811.07767v1","url_pdf":"http://arxiv.org/pdf/1811.07767v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"injecting-and-removing-malignant-features-in","repo_url":"https://github.com/BreastGAN/experiment1","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"injecting-and-removing-malignant-features-in","repo_url":"https://github.com/BreastGAN/experiment2","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"adversarial-attack","task_name":"Adversarial Attack"},{"task_slug":null,"task_name":"Generative Adversarial Network"}],"methods":[{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"cycle-consistency-loss","method_name":"Cycle Consistency Loss"},{"method_slug":"gan-least-squares-loss","method_name":"GAN Least Squares Loss"},{"method_slug":"instance-normalization","method_name":"Instance Normalization"},{"method_slug":"patchgan","method_name":"PatchGAN"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}