Konglerd Pirom, Reeb Catherine, Jansson Fredrik, Kaandorp Jaap A
Computational Science Lab, University of Amsterdam, Amsterdam, The Netherlands.
Institut de Systématique Évolution et Biodiversité UMR7205, UPMC-MNHN-CNRS-EPHE, 57 Rue Cuvier, BP39, 75005, Paris, France.
BMC Res Notes. 2017 Feb 20;10(1):103. doi: 10.1186/s13104-017-2424-0.
Many organisms such as plants can be characterized as complex-shaped branching forms. The morphological quantification of the forms is a support for a number of areas such as the effects of environmental factors and species discrimination. To date, there is no software package suitable for our dataset representing the forms. We therefore formulate methods for extracting morphological measurements from images of the forms.
As a case study we analyze two-dimensional images of samples from four groups belonging to three species of thalloid liverworts, genus Riccardia. The images are pre-processed and converted into binary images, then skeletonized to obtain a skeleton image, in which features such as junctions and terminals are detected. Morphological measurements known to characterize and discriminate the species in the samples such as junction thickness, branch thickness, terminal thickness, branch length, branch angle, and terminal spacing are then quantified. The measurements are used to distinguish among the four groups of Riccardia and also between the two groups of Riccardia amazonica collected in different locations, Africa and South America. Canonical discriminant analysis results show that those measurements are able to discriminate among the four groups. Additionally, it is able to discriminate R. amazonica collected in Africa from those collected in South America.
This paper presents general automated methods implemented in our software for quantifying two-dimensional images of complex branching forms. The methods are used to compute a series of morphological measurements. We found significant results to distinguish Riccardia species by using the measurements. The methods are also applicable for analyzing other branching organisms. Our software is freely available under the GNU GPL.
许多生物体，如植物，可被描述为具有复杂形状的分支形态。对这些形态进行形态学量化有助于多个领域，如环境因素的影响和物种鉴别。迄今为止，尚无适合我们所呈现的这些形态数据集的软件包。因此，我们制定了从形态图像中提取形态测量值的方法。
作为一个案例研究，我们分析了属于叶状苔类植物里卡多苔属三个物种的四组样本的二维图像。这些图像经过预处理并转换为二值图像，然后进行骨架化以获得骨架图像，在该图像中检测诸如节点和末端等特征。接着对已知用于表征和区分样本中物种的形态测量值进行量化，如节点厚度、分支厚度、末端厚度、分支长度、分支角度和末端间距。这些测量值用于区分里卡多苔属的四组样本，以及区分在非洲和南美洲不同地点采集的两组亚马逊里卡多苔。典型判别分析结果表明，这些测量值能够区分这四组样本。此外，它还能够区分在非洲采集的亚马逊里卡多苔和在南美洲采集的亚马逊里卡多苔。
本文介绍了我们软件中实现的用于量化复杂分支形态二维图像的通用自动化方法。这些方法用于计算一系列形态测量值。我们发现通过这些测量值能够显著区分里卡多苔属的物种。这些方法也适用于分析其他分支生物体。我们的软件可在GNU GPL协议下免费获取。
