Annu Int Conf IEEE Eng Med Biol Soc. 2021 Nov;2021:3509-3513. doi: 10.1109/EMBC46164.2021.9630424.
Toxoplasma gondii is a parasite that chronically infects about a third of the world's population. During chronic infection, the parasite resides within tissue cysts in the form of poorly understood bradyzoites which can number in the thousands. Our prior work showed that these bradyzoites are metabolically active exhibiting heterogeneous replication potential. The morphological plasticity of the mitochondrion potentially informs about parasite metabolic state. We developed an image processing based program to assist manual classification of mitochondrial morphologies by trained operators to collect data and statistics from the manual classification of shapes. We sought to determine whether certain morphologies were readily classifiable and the congruence among manual classifiers, i.e. the degree to which different operators would place the same objects within the same class. Results from three operators classifying mitochondrial morphologies from 5 tissue cyst images showed that among the four classes, one (Blobs) were the easiest to classify. There was remarkable congruence between 2 of the 3 operators in classifying the objects (96%), while the agreement among all 3 operators was somewhat modest (57%). Such information would be valuable for biologists studying these parasites as well as in development of fully automated methods of morphological classification.
刚地弓形虫是一种慢性感染全世界约三分之一人口的寄生虫。在慢性感染期间，寄生虫以未被充分了解的缓殖子形式存在于组织囊肿中，数量可能达到数千个。我们之前的工作表明，这些缓殖子具有代谢活性，表现出异质性的复制潜力。线粒体的形态可塑性可能反映了寄生虫的代谢状态。我们开发了一种基于图像处理的程序，以协助经过训练的操作人员对线粒体形态进行手动分类，从而从手动分类形状中收集数据和统计信息。我们试图确定某些形态是否易于分类，以及手动分类器之间的一致性，即不同操作人员将相同对象归入同一类别的程度。来自三名操作人员对 5 个组织囊肿图像中的线粒体形态进行分类的结果表明，在这四个类别中，有一个（Blob）是最容易分类的。在对物体进行分类时，有两名操作人员之间的一致性非常显著（96%），而所有三名操作人员之间的一致性则稍差（57%）。这种信息对于研究这些寄生虫的生物学家以及开发完全自动化的形态分类方法都非常有价值。
