使用CycleGAN在乳腺影像中注入和去除可疑特征:在小图像上使用神经网络进行自动对抗攻击的初步研究
研究利用两个公开数据集训练CycleGAN,在降采样乳腺影像中人为注入或去除可疑病灶,并由三名放射科医生对302例癌症和590例对照的图像进行判读。低分辨率下整体诊断性能未受显著影响(AUC 0.70 vs 0.76,p=0.67),但高分辨率下所有医生对修改后图像的癌症检出率显著降低(0.80 vs 0.37,p<0.001)。结论提示CycleGAN可隐式学习可疑特征并注入或去除,但受小图像尺寸和伪影引入的限制。
要点
- 高分辨率下CycleGAN修改图像使放射科医生癌症检出率从0.80降至0.37(p<0.001)。
- 高分辨率下医生可因伪影更明显而识别出修改图像(AUC 0.94,p<0.0001)。
- 低分辨率下整体诊断性能未受显著影响(AUC 0.70 vs 0.76,p=0.67)。
核心结论
- 概念验证研究显示CycleGAN能隐式学习可疑特征并在现有图像中人为注入或去除。
- 该方法目前受小图像尺寸和伪影引入的限制。
研究设计与样本
- 研究为概念验证研究,使用两个公开数据集BCD R和INbreast中680张有或无病灶的图像作为训练数据。
- 内部数据集(302例癌症、590例对照)作为测试数据,进行256×256像素和512×408像素两种实验。
- 三名放射科医生对修改和原始图像进行判读,采用多阅片者多病例ROC分析(MRMC-ROC)以曲线下面积(AUC)评估。
关键影像表现/征象
- 高分辨率下修改图像中伪影更明显,医生识别修改图像的AUC为0.94(p<0.0001)。
- 低分辨率下医生无法区分原始与修改图像(AUC 0.55,p=0.45)。
诊断标准与关键数值
- 放射科医生按1至5分评估可疑病灶存在可能性,并评估图像被操纵的可能性。
- 低分辨率下整体诊断性能AUC 0.70 vs 0.76(p=0.67),一名医生癌症检出率0.85 vs 0.63(p=0.06)。
- 高分辨率下所有医生对修改图像的癌症检出率0.80 vs 0.37(p<0.001)。
临床应用与阅片要点
- 原文未据此给出推荐。
局限性
- 方法的适用性目前受小图像尺寸和伪影引入的限制。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
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.
引用
- Becker AS, Jendele L, Skopek O, et al. Injecting and removing suspicious features in breast imaging with CycleGAN: A pilot study of automated adversarial attacks using neural networks on small images. Eur J Radiol. 2019. PMID: 31539791. doi:10.1016/j.ejrad.2019.108649
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。