Leveraging Domain Knowledge to Improve Microscopy Image Segmentation With Lifted Multicuts
Leveraging Domain Knowledge to Improve Microscopy Image Segmentation With Lifted Multicuts
复制标题
DOI：
10.3389/fcomp.2019.00006
复制
发表时间：
2019-10-24
影响因子：
2.6
通讯作者：
Kreshuk, Anna
中科院分区：
文献类型：
--
作者：
Pape, Constantin;Matskevych, Alex;Kreshuk, Anna
近年来，电子显微镜的通量显著增加，使得能够在相当大的组织体积中详细分析细胞形态和超微结构。以单突触分辨率分析神经回路仍然是这项技术的主要目标，但在细胞和发育生物学方面的应用也开始大规模出现。在光学显微镜方面，光片显微镜的不断发展导致了成像体积尺寸的快速增加，使得该领域的TB级采集成为常规。在这些研究中获取的数据量使得手动实例分割成为不可能，这是许多分析管道中的基本步骤。虽然由于采用了卷积神经网络，自动分割方法得到了显着改进，但其准确性仍然落后于人类注释，并且需要额外的手动校对。进一步改进的一个主要障碍是分割网络的有限视野，这阻止了它们学习利用预期的细胞形态或人类用来告知其分割决策的其他先验生物知识。在这篇文章中，我们展示了如何利用这些特定于域的信息，将其表达为图分区问题（称为提升多割问题）中的远程交互。使用这个公式，我们证明了显着提高分割精度的四个具有挑战性的基于边界的分割问题，从神经科学和发育生物学。
The throughput of electron microscopes has increased significantly in recent years, enabling detailed analysis of cell morphology and ultrastructure in fairly large tissue volumes. Analysis of neural circuits at single-synapse resolution remains the flagship target of this technique, but applications to cell and developmental biology are also starting to emerge at scale. On the light microscopy side, continuous development of light-sheet microscopes has led to a rapid increase in imaged volume dimensions, making Terabyte-scale acquisitions routine in the field. The amount of data acquired in such studies makes manual instance segmentation, a fundamental step in many analysis pipelines, impossible. While automatic segmentation approaches have improved significantly thanks to the adoption of convolutional neural networks, their accuracy still lags behind human annotations and requires additional manual proof-reading. A major hindrance to further improvements is the limited field of view of the segmentation networks preventing them from learning to exploit the expected cell morphology or other prior biological knowledge which humans use to inform their segmentation decisions. In this contribution, we show how such domain-specific information can be leveraged by expressing it as long-range interactions in a graph partitioning problem known as the lifted multicut problem. Using this formulation, we demonstrate significant improvement in segmentation accuracy for four challenging boundary-based segmentation problems from neuroscience and developmental biology.