Identification of patterns in microscopy images of biological samples using evolution strategies

In this paper we describe an algorithm based on evolutionary algorithms for determining patterns in images of biological samples (especially living cells) generated using the micro-patterning assay approach. In order to identify these patterns it is necessary to identify symmetric grids in nanoscale microscopy images. The algorithm presented in this paper is based on evolution strategies (ES): After downsampling the image using a correlation based approach for estimating the optimal downsampling rate, initial grids are constructed which are repeatedly evaluated and mutated for creating new candidates from which the best ones are promoted to the next generation. In the experimental section of this paper we analyse the performance of several ES strategies for identifying optimal grids in several images of biological samples.