Efficient adversarial debiasing with concept activation vector - Medical image case-studies.
Efficient adversarial debiasing with concept activation vector - Medical image case-studies.
复制标题
使用概念激活向量进行有效的对抗性去偏 - 医学图像案例研究。
DOI：
10.1016/j.jbi.2023.104548
复制
发表时间：
2024
影响因子：
4.5
通讯作者：
Banerjee,Imon
中科院分区：
文献类型：
--
作者：
Correa,Ramon;Pahwa,Khushbu;Patel,Bhavik;Vachon,CelineM;Gichoya,JudyW;Banerjee,Imon
AI模型的真实的时间部署的一个主要障碍是确保这些模型对于看不见的人群的可信度。通常情况下，这些复杂的模型是黑盒子，在其中产生有希望的结果。然而，当仔细检查，这些模型开始，以揭示隐含的偏见在决策过程中，特别是少数subgroup.MethodWe开发了一个有效的对抗性去偏置的方法与部分学习，通过将现有的概念激活向量（CAV）的方法，以减少种族差异，同时保持目标任务的性能。CAV最初是一种模型可解释性技术，我们采用该技术来识别负责学习种族的卷积层，并且只微调到该层，而不是微调整个网络，限制性能下降。在胸部X射线用例的外部数据集上，去偏模型（平均AUC 0.87）优于基线卷积模型（平均AUC 0.57）以及使用流行的微调策略（平均AUC 0.81）训练的模型。此外，使用单个数据集对乳房X线照片模型进行去偏置（白色、黑人和亚洲人），并提高了外部数据集的性能（平均AUC 0.8 - 0.86），完全不同人群（主要是西班牙裔患者）。结论在这项研究中，我们证明了仅使用内部数据训练的对抗模型的表现与标准精细模型相当，甚至经常优于标准精细模型。从外部设置的数据调整策略。无论预测器的模型架构如何，只要使用基于梯度的方法训练卷积模型，就可以应用所描述的对抗训练方法。我们以学术开源许可证发布培训代码-https://github.com/ramon349/JBI2023_TCAV_debiasing。
BackgroundA major hurdle for the real time deployment of the AI models is ensuring trustworthiness of these models for the unseen population. More often than not, these complex models are black boxes in which promising results are generated. However, when scrutinized, these models begin to reveal implicit biases during the decision making, particularly for the minority subgroups.MethodWe develop an efficient adversarial de-biasing approach with partial learning by incorporating the existing concept activation vectors (CAV) methodology, to reduce racial disparities while preserving the performance of the targeted task. CAV is originally a model interpretability technique which we adopted to identify convolution layers responsible for learning race and only fine-tune up to that layer instead of fine-tuning the complete network, limiting the drop in performanceResultsThe methodology has been evaluated on two independent medical image case-studies - chest X-ray and mammograms, and we also performed external validation on a different racial population. On the external datasets for the chest X-ray use-case, debiased models (averaged AUC 0.87 ) outperformed the baseline convolution models (averaged AUC 0.57 ) as well as the models trained with the popular fine-tuning strategy (averaged AUC 0.81). Moreover, the mammogram models is debiased using a single dataset (white, black and Asian) and improved the performance on an external datasets (averaged AUC 0.8 to 0.86 ) with completely different population (primarily Hispanic patients).ConclusionIn this study, we demonstrated that the adversarial models trained only with internal data performed equally or often outperformed the standard fine-tuning strategy with data from an external setting. The adversarial training approach described can be applied regardless of predictor’s model architecture, as long as the convolution model is trained using a gradient-based method. We release the training code with academic open-source license - https://github.com/ramon349/JBI2023_TCAV_debiasing.