Automatic Analysis of Live Cell Image Sequences to determine Temporal Mitotic Phenotypes Nathalie Harder 1 , Felipe Mora-Berm´ udez 2 , William J. Godinez 1 , Annelie W¨ unsche 2 , Jan Ellenberg 2 , Roland Eils 1 , Karl Rohr 1 1 University of Heidelberg, BIOQUANT, IPMB, and DKFZ Heidelberg, Dept. Bioinformatics and Functional Genomics, Biomedical Computer Vision Group 2 European Molecular Biology Laboratory (EMBL) Heidelberg, Gene Expression and Cell Biology/Biophysics Programmes n.harder@dkfz-heidelberg.de Abstract. Automated screening platforms allow biologists to acquire large amounts of image data with high information content. However, reliable automatic methods for analyzing this data are often not avail- able. Here, we present an approach for detailed cell cycle analysis based on live cell fluorescence microscopy image sequences. Our approach com- prises segmentation and tracking of dividing cell nuclei, and classifies cells into seven cell cycle phases as well as five abnormal morphological phe- notypes. Moreover, we automatically quantify cell cycle phase durations and perform a statistical analysis to determine temporal phenotypes. Our approach was successfully applied to images from gene knockdown experiments and experiments treated with small molecule drugs. 1 Introduction Understanding gene regulation of the cell cycle is of high common interest since errors in this process may lead to serious diseases such as cancer. High-content image-based screening is a powerful technology for gene function studies, and comprises automated microscopy as well as computational analysis to automat- ically extract information in an unbiased way. In screening experiments for cell cycle analysis usually live cell images of multiple cells are acquired. Multi-cell im- age sequences can be either analyzed in a population-based manner, i.e. features are determined for all cells and changes are studied for the whole population over time. However, subtle effects, such as cell cycle phase prolongations of certain cells, cannot be detected in this way. Therefore, single cell-based analysis has to be performed, which requires to track the individual cells throughout an image sequence. Based on tracking, the temporal evolution of single cells can be in- vestigated, in particular, to study cell cycle phase progression. Previously, this has been done based on phase contrast ([1]) and fluorescence ([2, 3]) microscopy images. However, these studies distinguished only up to five cell cycle phases and did not consider abnormal cellular morphologies. Also, none of these stud- ies determined cell cycle phase durations which is an essential readout for gene