Uncertainty-aware domain alignment for anatomical structure segmentation

• Published in: Medical image analysis 2020-08, Vol.64, p.101732-101732, Article 101732
• Main Authors: Bian, Cheng, Yuan, Chenglang, Wang, Jiexiang, Li, Meng, Yang, Xin, Yu, Shuang, Ma, Kai, Yuan, Jin, Zheng, Yefeng
• Format: Article
• Language: English
• Subjects: Deep learning; Domain adaptation; Uncertainty; Unsupervised segmentation
• ISSN: 1361-8415; 1361-8423
• DOI: 10.1016/j.media.2020.101732

•An uncertainty estimation and segmentation module (UESM) is proposed, which can provide and speed up the uncertainty estimation for the UDA task.•An uncertainty-aware cross entropy loss is proposed to utilize the uncertainty maps to boost the segmentation performance on highly uncertain regions.•An uncertainty-aware self-training strategy is proposed to select the optimal target samples determined by uncertainty values.•An uncertainty feature recalibration module is proposed together with our adversarial learning block to minimize the cross-domain discrepancy.•The proposed method achieves the best performance on both cross-device and cross-modality datasets compared with the state-of-the-art methods. [Display omitted] Automatic and accurate segmentation of anatomical structures on medical images is crucial for detecting various potential diseases. However, the segmentation performance of established deep neural networks may degenerate on different modalities or devices owing to the significant difference across the domains, a problem known as domain shift. In this work, we propose an uncertainty-aware domain alignment framework to address the domain shift problem in the cross-domain Unsupervised Domain Adaptation (UDA) task. Specifically, we design an Uncertainty Estimation and Segmentation Module (UESM) to obtain the uncertainty map estimation. Then, a novel Uncertainty-aware Cross Entropy (UCE) loss is proposed to leverage the uncertainty information to boost the segmentation performance on highly uncertain regions. To further improve the performance in the UDA task, an Uncertainty-aware Self-Training (UST) strategy is developed to choose the optimal target samples by uncertainty guidance. In addition, the Uncertainty Feature Recalibration Module (UFRM) is applied to enforce the framework to minimize the cross-domain discrepancy. The proposed framework is evaluated on a private cross-device Optical Coherence Tomography (OCT) dataset and a public cross-modality cardiac dataset released by MMWHS 2017. Extensive experiments indicate that the proposed UESM is both efficient and effective for the uncertainty estimation in the UDA task, achieving state-of-the-art performance on both cross-modality and cross-device datasets.