Automated analysis of zebrafish images for phenotypic changes in drug discovery

Abstract

Zebrafish has become one of the most popular and useful models in cell biology, development, and drug discovery. Because zebrafish embryo is transparent and can be observed under microscope without fixation, it is increasingly used in high-throughput screening. The small size of zebrafish embryos allows users to image them in a 96- or 384-well plate under various conditions, in turn, generating such a large amount of images that only automated analysis is feasible for processing and analyzing. We focus on developing an image processing algorithm to automatically quantify gene expression on zebrafish embryos that have been treated by various compounds. The challenge in this type of application includes aligning embryos of different orientations and automatically creating regions of interest (ROIs) to enclose specific areas in the head and torso of embryos. The image processing pipeline consists of alignment, segmentation, creation and quantification of ROIs. We test the algorithm using high-throughput images of zebrafish embryos obtained from an experiment screening compounds that may affect γ-secretase in Alzheimer's disease and results show that automated analysis can achieve satisfactory performance in a much shorter amount of time with a high level of objectivity.