Pattern Recognition for High Throughput Zebrafish Imaging Using Genetic Algorithm Optimization

In this paper we present a novel approach for image based high–throughput analysis of zebrafish embryos. Zebrafish embryos can be made available in high numbers; specifically in groups that have been exposed to different treatments. Preferably, the embryos are processed in batches. However, this complicates an automated processing as individual embryos need to be recognized. We present an approach in which the individual embryos are recognized and counted in an image with multiple instances and in multiple orientations. The recognition results in a mask that is used in the analysis of the images; multichannel images with bright–field and fluorescence are used.

The pattern recognition is based on a genetic algorithm which is the base of an optimization procedure through which the pattern is found. The optimization is accomplished by a deformable template that is incorporated in the genetic algorithm. We show that this approach is very robust and produces result fast so that it becomes very useful in a high–throughput environment. The method is fully automated and does not require any human intervention. We have tested our approach on both synthetic and real life images (zebrafish embryos). The results indicate that the method can be applied to a broad range of pattern recognition problems that require a high–throughput approach.