Attentive continuous generative self-training for unsupervised domain adaptive medical image translation.
Attentive continuous generative self-training for unsupervised domain adaptive medical image translation.
复制标题
DOI：
10.1016/j.media.2023.102851
复制
发表时间：
2023-08
影响因子：
10.9
通讯作者：
Woo, Jonghye
中科院分区：
文献类型：
--
作者：
Liu, Xiaofeng;Prince, Jerry L.;Xing, Fangxu;Zhuo, Jiachen;Reese, Timothy;Stone, Maureen;El Fakhri, Georges;Woo, Jonghye
关键词：
自训练是一类重要的无监督域适应（UDA）方法，用于在将从标记源域学到的知识应用于未标记和异构目标域时减轻域转移问题。虽然基于自训练的 UDA 通过基于最大 softmax 概率的可靠伪标签过滤，在判别任务（包括分类和分割）上显示出相当大的前景，但之前针对生成任务（包括图像模态翻译）的基于自训练的 UDA 的工作还很少。为了填补这一空白，在这项工作中，我们寻求开发一种生成式自我训练（GST）框架，用于具有连续值预测和回归目标的域自适应图像翻译。具体来说，我们使用变分贝叶斯学习来量化 GST 中的任意和认知不确定性，以衡量合成数据的可靠性。我们还引入了一种自我关注方案，该方案不再强调背景区域，以防止其主导训练过程。然后通过具有目标域监督的交替优化方案来进行适应，该方案将注意力集中在具有可靠伪标签的区域上。我们在两个跨扫描仪/中心、受试者间翻译任务上评估了我们的框架，包括标记到电影磁共振 (MR) 图像翻译和 T1 加权 MR 到分数各向异性翻译。对不配对目标域数据的广泛验证表明，与对抗性训练 UDA 方法相比，我们的 GST 产生了卓越的合成性能。
Self-training is an important class of unsupervised domain adaptation (UDA) approaches that are used to mitigate the problem of domain shift, when applying knowledge learned from a labeled source domain to unlabeled and heterogeneous target domains. While self-training-based UDA has shown considerable promise on discriminative tasks, including classification and segmentation, through reliable pseudo-label filtering based on the maximum softmax probability, there is a paucity of prior work on self-training-based UDA for generative tasks, including image modality translation. To fill this gap, in this work, we seek to develop a generative self-training (GST) framework for domain adaptive image translation with continuous value prediction and regression objectives. Specifically, we quantify both aleatoric and epistemic uncertainties within our GST using variational Bayes learning to measure the reliability of synthesized data. We also introduce a self-attention scheme that de-emphasizes the background region to prevent it from dominating the training process. The adaptation is then carried out by an alternating optimization scheme with target domain supervision that focuses attention on the regions with reliable pseudo-labels. We evaluated our framework on two cross-scanner/center, inter-subject translation tasks, including tagged-to-cine magnetic resonance (MR) image translation and T1-weighted MR-to-fractional anisotropy translation. Extensive validations with unpaired target domain data showed that our GST yielded superior synthesis performance in comparison to adversarial training UDA methods.
登录
查看更多内容
影响因子：
3.5
作者：
Garyfallidis E;Brett M;Amirbekian B;Rokem A;van der Walt S;Descoteaux M;Nimmo-Smith I;Dipy Contributors
通讯作者：
Dipy Contributors
影响因子：
10.9
作者：
He Y;Carass A;Zuo L;Dewey BE;Prince JL
通讯作者：
Prince JL
影响因子：
5.8
作者：
Kiureghian, Armen Der;Didevsen, Ove
通讯作者：
Didevsen, Ove
影响因子：
5.7
作者：
Avants BB;Tustison NJ;Song G;Cook PA;Klein A;Gee JC
通讯作者：
Gee JC
影响因子：
10.6
作者：
Chartsias A;Joyce T;Giuffrida MV;Tsaftaris SA
通讯作者：
Tsaftaris SA