Abstract

The combination of a fractal formalism with linear methods of image analysis is the basis of a statistically reliable description of the morphology and classification of forms that are irregular from the point of view of Euclidean geometry, such as in vitro flattened amoeboid cells. We have developed a classification algorithm that is based on the morphological characteristics of immune cells of four species of marine invertebrates: the bivalves Spisula sachalinensis (Schrenck, 1862) and Callista brevisiphonata (Carpenter, 1864) and the echinoderms Aphelasterias japonica (Bell, 1881) and Asterias amurensis (Lütken, 1871). The morphological signs of the cells included dimensional characteristics (area and perimeter), features of the circle and the convex hull that circumscribes the cell, characteristics of the symmetry and roundness of the cell, as well as fractal dimension and lacunarity, which were determined in several ways while assessing the spatial complexity of cells. It was shown that the optimal classification structure of immune cells of all studied species is based on a similar universal algorithm, including hierarchical classification by the Ward method, using variables with the highest multimodality index that load different factors in the factor analysis.

中文翻译：

---

海洋无脊椎动物免疫细胞的比较形态学分类

摘要

分形形式主义与图像分析的线性方法的结合是统计学上可靠的形态学描述的基础，从欧几里得几何学的角度来看，这些形态是不规则的，例如体外扁平的变形虫细胞。我们已经开发了一种分类算法，该算法基于四种无脊椎动物的免疫细胞的形态学特征：双壳类长毛棘轮虫（Schrenck，1862）和短壳ista（Callista brevisiphonata）（Carpenter，1864）和棘皮棘皮鱼（Bell，1881）和天竺葵（吕特肯，1871年）。细胞的形态学标志包括尺寸特征（面积和周长），围绕细胞的圆形和凸包的特征，细胞的对称性和圆度以及分形维数和空隙度，这些特征在评估细胞空间复杂性的几种方法。结果表明，所有研究物种的免疫细胞的最佳分类结构均基于相似的通用算法，包括通过Ward方法进行的分级分类，其中使用了具有最高多模态指数的变量，这些变量在因子分析中具有不同的因子。