Determination of minimum pixel resolution for shape analysis: Proposal of a new data validation method for computerized images

Abstract

Abstract In many different research areas not only size, but also an exact description of particle shape is important in order to understand certain physical or chemical processes. Digital image analyses with specially developed software (DIP) have become increasingly popular in recent years, producing a huge amount of reproducible data. However, since image captures are pixelated, purely digitalisation problems/errors may occur and, hence, a minimum number of pixels for meaningful results has to be established. This may depend on different computer software or on calculation methods. Here we bring Elongation, Circularity and Sphericity in relation and calculate in a theoretical model maximum values of Circularity and Sphericity for specific Elongation values from 0 to 1. In these simple 2-dimensional plots, which can be applied to any DIP program, a line marking the upper limit of congruent shape analysis can be calculated. Points that fall far above the theoretical maximum curve are interpreted as digitalisation issues of the DIP programs: especially measuring the boundary length (perimeter) is not as simple. It can be shown that, with increasing particle size, the rate of obvious erroneous shape analysis data decreases, and thus a minimum pixel number can be established after “pixel size cleaning”. We tested our model with two commonly used plugins with two different shape calculation methods for particle analysis of the DIP program ImageJ: while an object has to be build-up of about 200 pixels using the preinstalled plugin, the threshold can be significantly reduced (50) using the Particles8_Plus plugin by Landini.