An automated method for cell phase identification in high throughput time-lapse screens

To better understand drug effects on cancer cells, it is important to measure cell cycle progression in individual cells as a function of time. Time-lapse fluorescence microscopy imaging provides an important method to study the dynamic cell cycle process under different conditions of perturbation. However, the assignment of a cell to a particular phase is done by visual inspection of images. To use timelapse fluorescence microscopy for high throughput cell cycle analysis and drug screens, improved approaches that are more automated and objective are needed. In this chapter, an automated method is proposed for cell phase identification in time-lapse microscopy data. A set of twelve shape and intensity features are first extracted to describe nuclei differences in different phases based on the experience of cell biologists. We then compare the performance of different pattern recognition techniques for cell phase identification. A k-nearest neighbor classifier with a set of seven features shows the best identification result. Final identification is performed by the k nearest neighbor classifier. The accuracy of the identification is further enhanced by applying knowledge rules of cell biology.