arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2306.06503cs.LGcs.AIcs.CReess.IV
收藏

在医疗 AI 模型的领域迁移中保护隐私无需付出性能代价:差分隐私的关键作用

Preserving privacy in domain transfer of medical AI models comes at no performance costs: The integral role of differential privacy

University Hospital RWTH Aachen(亚琛工业大学医院);RWTH Aachen University(亚琛工业大学);Technical University of Munich(慕尼黑工业大学);Helmholtz-Zentrum Munich(慕尼黑亥姆霍兹中心)

Soroosh Tayebi Arasteh, Mahshad Lotfinia, Teresa Nolte, Marwin Saehn, Peter Isfort, Christiane Kuhl, Sven Nebelung, Georgios Kaissis, Daniel Truhn

AI总结:

本文利用五家机构逾59万张胸片证明,在医疗AI模型领域迁移中采用差分隐私,即使 epsilon 约为1,也能在域外诊断中保持与非DP模型相当的准确性和公平性。

摘要:

开发稳健且有效的医学人工智能(AI)模型需要访问大量患者数据。仅使用在大型多机构数据集上训练的 AI 模型可能有助于解决这一问题,但确保数据隐私的必要性仍然存在,尤其是成员推理可能危及患者保密性。作为一种拟议的补救措施,我们主张整合差分隐私(DP)。我们专门研究了在模型训练期间未曾见过的机构数据(即外部验证)上,使用 DP 训练的模型与未使用 DP 训练的模型相比的性能,这种情况反映了 AI 模型的临床使用。通过利用来自五家机构的超过 590,000 张胸部 X 光片,我们评估了 DP 增强领域迁移(DP-DT)在诊断心脏肥大、胸腔积液、肺炎、肺不张以及识别健康受试者方面的功效。我们将 DP-DT 与 non-DP-DT 进行对照,并以受试者工作特征曲线下面积(AUC)为主要指标,同时结合准确率、敏感性和特异性,检查诊断准确性和人口统计学公平性。结果表明,即使在极高隐私级别(epsilon 约为 1)下,DP-DT 的表现也与 non-DP-DT 相当(所有领域 P>0.119)。此外,相对于 non-DP-DT,DP-DT 在几乎所有亚组中仅导致不到 1% 的微小 AUC 差异。尽管一直有证据表明 DP 模型会在域内应用中造成显著性能下降,但我们证明域外性能几乎不受影响。因此,鉴于 DP 对性能影响极小,我们强烈主张在训练诊断性医疗 AI 模型时采用 DP。

英文摘要:

Developing robust and effective artificial intelligence (AI) models in medicine requires access to large amounts of patient data. The use of AI models solely trained on large multi-institutional datasets can help with this, yet the imperative to ensure data privacy remains, particularly as membership inference risks breaching patient confidentiality. As a proposed remedy, we advocate for the integration of differential privacy (DP). We specifically investigate the performance of models trained with DP as compared to models trained without DP on data from institutions that the model had not seen during its training (i.e., external validation) - the situation that is reflective of the clinical use of AI models. By leveraging more than 590,000 chest radiographs from five institutions, we evaluated the efficacy of DP-enhanced domain transfer (DP-DT) in diagnosing cardiomegaly, pleural effusion, pneumonia, atelectasis, and in identifying healthy subjects. We juxtaposed DP-DT with non-DP-DT and examined diagnostic accuracy and demographic fairness using the area under the receiver operating characteristic curve (AUC) as the main metric, as well as accuracy, sensitivity, and specificity. Our results show that DP-DT, even with exceptionally high privacy levels (epsilon around 1), performs comparably to non-DP-DT (P>0.119 across all domains). Furthermore, DP-DT led to marginal AUC differences - less than 1% - for nearly all subgroups, relative to non-DP-DT. Despite consistent evidence suggesting that DP models induce significant performance degradation for on-domain applications, we show that off-domain performance is almost not affected. Therefore, we ardently advocate for the adoption of DP in training diagnostic medical AI models, given its minimal impact on performance.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑