KCB-Net: A 3D knee cartilage and bone segmentation network via sparse annotation.
KCB-Net: A 3D knee cartilage and bone segmentation network via sparse annotation.
复制标题
DOI：
10.1016/j.media.2022.102574
复制
发表时间：
2022-11
影响因子：
10.9
通讯作者：
Chen, Danny Z.
中科院分区：
文献类型：
--
作者：
Peng, Yaopeng;Zheng, Hao;Liang, Peixian;Zhang, Lichun;Zaman, Fahim;Wu, Xiaodong;Sonka, Milan;Chen, Danny Z.
膝关节软骨和骨分割对于医生分析和诊断关节损伤和膝关节骨关节炎（OA）至关重要。用于医学图像分割的深度学习（DL）方法在很大程度上优于传统方法，但它们通常需要大量带注释的数据进行模型训练，这对于医学专家来说非常昂贵和耗时，特别是在3D图像上。在本文中，我们报告了一个新的膝关节软骨和骨分割框架，KCB-Net，基于稀疏注释的三维MR图像。KCB-Net从3D图像中选择一小部分切片进行注释，并寻求弥合稀疏注释和完整注释之间的性能差距。具体地说，它首先用无监督方案识别最有效和最具代表性的切片的子集;然后使用注释切片训练集成模型;接下来，它使用包含由集成方法生成的伪标签的3D图像自训练模型，并通过双向分层推土机距离（bi-HEMD）算法进行改进;最后利用原对偶内点法对分割结果进行微调。在四个3D MR膝关节数据集（SKI 10数据集，OAI ZIB数据集，爱荷华州数据集和iMorphics数据集）上的实验表明，我们的新框架在完整注释上优于最先进的方法，并且即使在低至10%的小注释率下也能产生高质量的结果。
Knee cartilage and bone segmentation is critical for physicians to analyze and diagnose articular damage and knee osteoarthritis (OA). Deep learning (DL) methods for medical image segmentation have largely outperformed traditional methods, but they often need large amounts of annotated data for model training, which is very costly and time-consuming for medical experts, especially on 3D images. In this paper, we report a new knee cartilage and bone segmentation framework, KCB-Net, for 3D MR images based on sparse annotation. KCB-Net selects a small subset of slices from 3D images for annotation, and seeks to bridge the performance gap between sparse annotation and full annotation. Specifically, it first identifies a subset of the most effective and representative slices with an unsupervised scheme; it then trains an ensemble model using the annotated slices; next, it self-trains the model using 3D images containing pseudo-labels generated by the ensemble method and improved by a bi-directional hierarchical earth mover’s distance (bi-HEMD) algorithm; finally, it fine-tunes the segmentation results using the primal-dual Internal Point Method (IPM). Experiments on four 3D MR knee joint datasets (the SKI10 dataset, OAI ZIB dataset, Iowa dataset, and iMorphics dataset) show that our new framework outperforms state-of-the-art methods on full annotation, and yields high quality results for small annotation ratios even as low as 10%.
登录
查看更多内容
影响因子：
10.9
作者：
Ambellan, Felix;Tack, Alexander;Zachow, Stefan
通讯作者：
Zachow, Stefan
影响因子：
5.5
作者：
Couteaux, V;Si-Mohamed, S.;Boussel, L.
通讯作者：
Boussel, L.
影响因子：
3.8
作者：
Xie，Hui;Pan，Zhe;Wu，Xiaodong
通讯作者：
Wu，Xiaodong
影响因子：
10.6
作者：
Yin Y;Zhang X;Williams R;Wu X;Anderson DD;Sonka M
通讯作者：
Sonka M
影响因子：
10.6
作者：
Liang, Peixian;Zhang, Yizhe;Chen, Danny Z.
通讯作者：
Chen, Danny Z.