Description and classification of bivalve mollusks hemocytes: a computational approach

Abstract

The fractal formalism in combination with linear image analysis enables statistically significant description and classification of “irregular” (in terms of Euclidean geometry) shapes, such as, outlines ofin vitroflattened cells. We developed an optimal model for classifying bivalveSpisula sachalinensisandCallista brevisiphonataimmune cells, based on evaluating their linear and non-linear morphological features: size characteristics (area, perimeter), various parameters of cell bounding circle, convex hull, cell symmetry, roundness, and a number of fractal dimensions and lacunarities evaluating the spatial complexity of cells. Proposed classification model is based on Ward’s clustering method, loaded with highest multimodality index factors. This classification scheme groups cells into three morphological types, which can be distinguished both visually and by several linear and quasi-fractal parameters.