研究院动态 当前所在位置:首页 > 信息公告 > 研究院动态
发布时间: 2023-10-11

 2023年108日,医学图像计算与计算机辅助介入领域的国际顶级会议MICCAIMedical Image Computing and Computer Assisted Intervention)在加拿大温哥华拉开帷幕。医疗机器人研究院郑国焱教授、顾运副教授等团队共7篇论文被录取。



CT-guided, Unsupervised Super-resolution Reconstruction of Single 3D Magnetic Resonance Image

CT引导的无监督的3D MRI超分辨率重建

Jiale Wang, Guoyan Zheng


Deep learning-based algorithms for single MR image (MRI) super-resolution have shown great potential in enhancing the resolution of low-quality images. However, many of these methods rely on supervised training with paired low resolution (LR) and high-resolution (HR) MR images, which can be difficult to obtain in clinical settings. This is because acquiring HR MR images in clinical settings requires a significant amount of time. In contrast, HR CT images are acquired in clinical routine. In this paper, we propose a CT-guided, unsupervised MRI super-resolution reconstruction method based on joint cross-modality image translation and super-resolution reconstruction, eliminating the requirement of high-resolution MRI for training. The proposed approach is validated on two datasets respectively acquired from two different clinical centers. Experimental results demonstrate the superior performance of the proposed approach over the state-of-the-art methods.




Partially Supervised Multi-Organ Segmentation via Affinity-aware Consistency Learning and Cross Site Feature Alignment


Qin Zhou, Peng Liu, Guoyan Zheng


Partially supervised segmentation (PSS) has attracted increasing attention in multi-organ segmentation (MOS). However, facing with challenges from lacking sufficiently labeled data and cross-site data discrepancy, PSS remains largely an unsolved problem. In this paper, to fully take advantage of the unlabeled data, we propose to model voxel-to-organ affinity in embedding space into consistency learning, ensuring consistency in both label space and latent feature space. Furthermore, to mitigate the cross-site data discrepancy, we propose to propagate the inter-organ affinity relationships across different sites, calibrating the multi-site feature distribution from a statistical perspective. Extensive experiments manifest that our method generates favorable results compared with other state-of-the-art methods, especially on hard organs with relatively smaller sizes.

部分监督分割在多器官分割任务中获得了越来越多的关注。然而, 由于缺乏足够的标注样本以及不同中心的数据差异部分监督分割在很大程度上仍然是一个未解决的问题本文中我们提出将体素-器官在特征空间的相似度关系引入一致性学习框架通过确保标签空间和特征空间的一致性来充分利用无标签数据此外为了减轻跨中心数据差异的影响我们提出在不同中心之间传播跨器官的相似度关系以此来从统计层面实现多中心数据的特征分布对齐大量实验表明与其他算法相比我们的算法取得了更好的性能尤其在尺寸相对较小的较难分割器官上表现更佳。



FedContrast-GPA: Heterogeneous Federated Optimization via Local Contrastive Learning and Global Process-aware Aggregation

FedContrast-GPA: 基于局部对比学习和全局过程感知聚合的异质联邦优化

Qin Zhou, Guoyan Zheng


Recently, federated learning has emerged as a promising strategy for performing privacy-preserving, distributed learning for medical image segmentation. However, the data heterogeneity as well as system heterogeneity makes it challenging to optimize. Existing methods try to tackle the heterogeneity issues in federated networks via proximal restriction or re-parameterization on local models, which may limit the convergence potential of local models. In this paper, we propose an end-to-end FedContrast-GPA framework to jointly address the data-level and system-level heterogeneity. In specific, for data heterogeneity, we aim to enhance the feature representations in local model training via an intra-client and inter-client local prototype based contrastive learning scheme. As for system heterogeneity, we further propose a simple process-aware aggregation scheme to achieve effective straggler mitigation. Experimental results on six prostate segmentation datasets demonstrate large performance boost over existing state-of-the-art federated optimization methods.




AirwayFormer: Structure-Aware Boundary-Adaptive Transformers for Airway Anatomical Labeling


Weihao Yu , Hao Zheng, Yun Gu, Fangfang Xie , Jiayuan Sun , Jie Yang


Pulmonary airway labeling identifies anatomical names for branches in bronchial trees. These fine-grained labels are critical for disease diagnosis and intra-operative navigation. Recently, various methods have been proposed for this task. However, accurate labeling of each bronchus is challenging due to the fine-grained categories and interindividual variations. On the one hand, training a network with limited data to recognize multitudinous classes sets an obstacle to the design of algorithms. We propose to maximize the use of latent relationships by a transformer-based network. Neighborhood information is properly integrated to capture the priors in the tree structure, while a U-shape layout is introduced to exploit the correspondence between different nomenclature levels. On the other hand, individual variations cause the distribution overlapping of adjacent classes in feature space. To resolve the confusion between sibling categories, we present a novel generator that predicts the weight matrix of the classifier to produce dynamic decision boundaries between subsegmental classes. Extensive experiments performed on publicly available datasets demonstrate that our method can perform better than state-of-the-art methods.




Pick the Best Pre-trained Model: Towards Transferability Estimation for Medical Image Segmentation.

挑选最佳预训练模型: 医疗图像分割的可迁移性评估

Yuncheng Yang, Meng Wei, Junjun He, Jie Yang, Jin Ye, Yun Gu


Transfer learning is a critical technique in training deep neural networks for the challenging medical image segmentation task that requires enormous resources. With the abundance of medical image data, many research institutions release models trained on various datasets that can form a huge pool of candidate source models to choose from. Hence, it’s vital to estimate the source models’ transferability (i.e., the ability to generalize across different downstream tasks) for proper and efficient model reuse. To make up for its deficiency when applying transfer learning to medical image segmentation, in this paper, we therefore propose a new  Transferability Estimation (TE) method. We first analyze the drawbacks of using the existing TE algorithms for medical image segmentation and then design a source-free TE framework that considers both class consistency and feature variety for better estimation. Extensive experiments show that our method surpasses all current algorithms for transferability estimation in medical image segmentation.




Triplet Consistent Learning for Odometry Estimation of Monocular Endoscope


Hao Yue,Yun Gu


The depth and pose estimations from monocular images are essential for computer-aided navigation. Since the ground truth of depth and pose are difficult to obtain, the unsupervised training method has a broad prospect in endoscopic scenes. However, endoscopic datasets lack sufficient diversity of visual variations, and appearance inconsistency is also frequently observed in image triplets. In this paper, we propose a triplet-consistency-learning framework (TCL) consisting of two modules: Geometric Consistency module(GC) and Appearance Inconsistency module(AiC). To enrich the diversity of endoscopic datasets, the GC module generates synthesis triplets and enforces geometric consistency via specific losses. To reduce the appearance inconsistency in the image triplets, the AiC module introduces a triplet-masking strategy to act on photometric loss. TCL can be easily embedded into various unsupervised methods without adding extra model parameters. Experiments on public datasets demonstrate that TCL effectively improves the accuracy of unsupervised methods even with limited number of training samples.




Semantic difference guidance for the uncertain boundary segmentation of CT left atrial appendage


Xin You, Ming Ding, Minghui Zhang, Yangqian Wu, Yi Yu, Yun Gu, and Jie Yang


Atrial fibrillation (AF) is one of the most common types of cardiac arrhythmia, which is closely relevant to anatomical structures including the left atrium (LA) and the left atrial appendage (LAA). Thus, a thorough understanding of the LA and LAA is essential for the AF treatment. In this paper, we have modeled relative relations between the LA and LAA via deep segmentation networks for the first time, and introduce a new LA & LAA CT dataset. To deal with uncertain boundaries between the LA and LAA, we propose the semantic difference module (SDM) based on diffusion theory to refine features with enhanced boundary information. Besides, disconnections between the LA and LAA are frequently observed in the segmentation results due to uncertain boundaries of the LAA region and CT imaging noise. To address this issue, we devise another connectivity-refined network with the connectivity loss. The loss function exerts a distance regularization on coarse predictions from the first-stage network. Experiments demonstrate that our proposed model can achieve state-of-the-art segmentation performance compared with classic convolutional-neural-networks (CNNs) and recent Transformer-based models on this new dataset. Specifically, SDM can also outperform existing methods on refining uncertain boundaries. Codes are available at https://github.com/AlexYouXin/LA-LAA-segmentation.




沪交ICP备20190057 版权所有 上海交通大学医疗机器人研究院   流量统计