High-throughput screens of the gene function provide rapidly increasing amounts of data. In particular, the analysis of image data acquired in genome-wide cell phenotype screens constitutes a substantial bottleneck in the evaluation process and motivates the development of automated image analysis tools for large-scale experiments. Here we introduce a computational scheme to process multi-cell images as they are produced in high-throughput screens. We describe an approach to automatically segment and classify cell nuclei into different mitotic phenotypes. This enables automated identification of cell cultures that show an abnormal mitotic behavior. Our scheme proves a high classification accuracy, suggesting a promising future for automating the evaluation of high-throughput experiments