Deep learning-enabled multi-organ segmentation in whole-body mouse scans.
Deep learning-enabled multi-organ segmentation in whole-body mouse scans.
复制标题
DOI：
10.1038/s41467-020-19449-7
复制
发表时间：
2020-11-06
影响因子：
16.6
通讯作者：
Menze BH
中科院分区：
文献类型：
--
作者：
Schoppe O;Pan C;Coronel J;Mai H;Rong Z;Todorov MI;Müskes A;Navarro F;Li H;Ertürk A;Menze BH
小鼠的全身成像是研究的关键信息来源。器官分割是定量分析的先决条件，但如果手动完成，则是一项繁琐且容易出错的任务。在这里，我们提出了一种名为AIMOS的深度学习解决方案，它可以在不到一秒的时间内自动分割主要器官（大脑，肺，心脏，肝脏，肾脏，脾脏，膀胱，胃，肠）和骨骼，比以前的算法快了几个数量级。AIMOS匹配或超过最先进的方法和人类专家的分割质量。我们证明了生物医学研究定位癌症转移的直接适用性。此外，我们表明，专家注释受到人为错误和偏见。因此，我们表明，至少需要两个独立创建的注释来评估模型的性能。重要的是，AIMOS通过确定人类最有可能不同意的区域来解决人类偏见的问题，从而定位和量化这种不确定性，以改进下游分析。总之，AIMOS是一个强大的开源工具，可以在生物医学研究的许多领域提高可扩展性，减少偏差并促进可重复性。小鼠全身图像的器官分割对于定量分析是必不可少的，但繁琐且容易出错。在这里，作者开发了一个深度学习管道，可以在不到一秒的时间内分割体积全身扫描中的主要器官和骨骼，并提供概率图和不确定性估计。
Whole-body imaging of mice is a key source of information for research. Organ segmentation is a prerequisite for quantitative analysis but is a tedious and error-prone task if done manually. Here, we present a deep learning solution called AIMOS that automatically segments major organs (brain, lungs, heart, liver, kidneys, spleen, bladder, stomach, intestine) and the skeleton in less than a second, orders of magnitude faster than prior algorithms. AIMOS matches or exceeds the segmentation quality of state-of-the-art approaches and of human experts. We exemplify direct applicability for biomedical research for localizing cancer metastases. Furthermore, we show that expert annotations are subject to human error and bias. As a consequence, we show that at least two independently created annotations are needed to assess model performance. Importantly, AIMOS addresses the issue of human bias by identifying the regions where humans are most likely to disagree, and thereby localizes and quantifies this uncertainty for improved downstream analysis. In summary, AIMOS is a powerful open-source tool to increase scalability, reduce bias, and foster reproducibility in many areas of biomedical research. Organ segmentation of whole-body mouse images is essential for quantitative analysis, but is tedious and error-prone. Here the authors develop a deep learning pipeline to segment major organs and the skeleton in volumetric whole-body scans in less than a second, and present probability maps and uncertainty estimates.
登录
查看更多内容
影响因子：
3.1
作者：
Khmelinskii, Artem;Baiker, Martin;Kaijzel, Eric L.;Chen, Josette;Reiber, Johan H. C.;Lelieveldt, Boudewijn P. F.
通讯作者：
Lelieveldt, Boudewijn P. F.
影响因子：
10.9
作者：
Baiker, Martin;Milles, Julien;Lelieveldt, Boudewijn P. F.
通讯作者：
Lelieveldt, Boudewijn P. F.
影响因子：
10.9
作者：
Heinrich, Mattias P.;Oktay, Ozan;Bouteldja, Nassim
通讯作者：
Bouteldja, Nassim
影响因子：
11.5
作者：
Kaijzel, Eric L.;van der Pluijm, Gabri;Lowik, Clemens W. G. M.
通讯作者：
Lowik, Clemens W. G. M.
影响因子：
23.8
作者：
Davoudi, Neda;Dean-Ben, Xose Luis;Razansky, Daniel
通讯作者：
Razansky, Daniel