| Popis: |
This paper reviews three different methods for parameterisation of plant tissues which can be applied to images obtained with a confocal scanning laser microscope to create models for the simulation of the mechanical behaviour of biological cellular structures. Vectorisation, Voronoi tessellation and ellipse tessellation were tested. Potato tuber and carrot parenchyma were chosen as examples. For each method tested, five geometrical parameters were analysed: area, perimeter, orientation, elongation and a local indicator of spatial association of all individual regions which represented cells. The reconstruction accuracy of the original tissue microstructure by each parameterisation method was investigated by the comparison of the geometrical properties of the cells from the segmentation with their virtual equivalents. Linear regression models between the reference and modelled parameters were built. The model performance was expressed in terms of the coefficient of determination (R2), the slope of the regression line and the root mean square errors for prediction (RMSEP). Based on the results, Voronoi tessellation was considered to be inaccurate for tissue modelling. The vectorisation procedure only allowed for reproduction of the general shapes of cells, and the curvature of cell walls was neglected in this method. For both the Voronoi tessellation as well as vectorisation, created cells completely filled the space with no additional gaps and possessed sharp, angular shapes. The best overall reconstruction accuracy was obtained with ellipse tessellation. Models created with this method can be considered as representative equivalents of real tissues in terms of cell area, orientation, perimeter, shape and spatial arrangement. |
Full Text from ScienceDirect