The impact of site-specific digital histology signatures on deep learning model accuracy and bias.
The impact of site-specific digital histology signatures on deep learning model accuracy and bias.
复制标题
DOI：
10.1038/s41467-021-24698-1
复制
发表时间：
2021-07-20
影响因子：
16.6
通讯作者：
Pearson AT
中科院分区：
文献类型：
--
作者：
Howard FM;Dolezal J;Kochanny S;Schulte J;Chen H;Heij L;Huo D;Nanda R;Olopade OI;Kather JN;Cipriani N;Grossman RL;Pearson AT
癌症基因组图谱 (TCGA) 是最大的数字组织学生物存储库之一。深度学习 (DL) 模型已在 TCGA 上进行训练，可直接根据组织学预测众多特征，包括生存、基因表达模式和驱动突变。然而，我们证明，对于 3,000 多名患有六种癌症亚型的患者，这些特征在 TCGA 中的组织提交站点之间存在很大差异。此外，我们还表明，通过 DL 可以轻松识别提交位点之间的组织学图像差异。尽管常用的颜色归一化和增强方法，位点检测仍然是可能的，并且我们量化了构成该位点特定数字组织学特征的图像特征。我们证明，这些位点特异性特征会导致生存、基因组突变和肿瘤分期等特征预测的准确性出现偏差。此外，种族也可以从特定地点的签名中推断出来，必须考虑到这一点以确保 DL 的公平应用。这些特定于站点的签名可能会导致对模型性能的过度乐观估计，我们提出了一种二次规划方法，该方法通过确保模型不在来自同一站点的样本上进行训练和验证来消除这种偏差。深度学习模型已经在癌症基因组图谱上进行了训练，可以直接根据组织学预测许多特征，包括生存、基因表达模式和驱动突变。在这里，作者证明了位点特异性的组织学特征可能导致对此类模型的准确性估计存在偏差，并提出了一种最小化这种偏差的方法。
The Cancer Genome Atlas (TCGA) is one of the largest biorepositories of digital histology. Deep learning (DL) models have been trained on TCGA to predict numerous features directly from histology, including survival, gene expression patterns, and driver mutations. However, we demonstrate that these features vary substantially across tissue submitting sites in TCGA for over 3,000 patients with six cancer subtypes. Additionally, we show that histologic image differences between submitting sites can easily be identified with DL. Site detection remains possible despite commonly used color normalization and augmentation methods, and we quantify the image characteristics constituting this site-specific digital histology signature. We demonstrate that these site-specific signatures lead to biased accuracy for prediction of features including survival, genomic mutations, and tumor stage. Furthermore, ethnicity can also be inferred from site-specific signatures, which must be accounted for to ensure equitable application of DL. These site-specific signatures can lead to overoptimistic estimates of model performance, and we propose a quadratic programming method that abrogates this bias by ensuring models are not trained and validated on samples from the same site. Deep learning models have been trained on The Cancer Genome Atlas to predict numerous features directly from histology, including survival, gene expression patterns, and driver mutations. Here, the authors demonstrate that site-specific histologic signatures can lead to biased estimates of accuracy for such models, and propose a method to minimize such bias.
登录
查看更多内容
影响因子：
50.3
作者：
Carrot-Zhang, Jian;Chambwe, Nyasha;Beroukhim, Rameen
通讯作者：
Beroukhim, Rameen
影响因子：
3.7
作者：
Badea L;Stănescu E
通讯作者：
Stănescu E
影响因子：
45.3
作者：
Bickell, NA;Wang, JJ;Guth, AA
通讯作者：
Guth, AA
影响因子：
8.8
作者：
Echle A;Rindtorff NT;Brinker TJ;Luedde T;Pearson AT;Kather JN
通讯作者：
Kather JN
影响因子：
64.8
作者：
通讯作者：
--