Time-lapse cell cycle quantitative data analysis using gaussian mixture models

Time-lapse fluorescence microscopy imaging provides an important method to study the cell cycle process under different conditions of perturbation. Existing methods, however, are rather limited in dealing with such time-lapse data sets, while manual analysis is unreasonably time-consuming. This chapter presents statistical data analysis issues and statistical pattern recognition to fill this gap. We propose to apply Gaussian mixture model (GMM) to study the classification problems. We first propose to model the time-lapse cell trace data by using autoregression (AR) model and to filter the cell features using this model. We then study whether there is significant difference in cell morphology between untreated and treated cases using Pearson correlation and GMM. Furthermore, we propose to study cell phase identification using GMM, and compare with other traditional classifiers. Once we identify the cell phase information, then we can answer questions such as when the cells are arrested. We employ the ordered Fisher clustering algorithm to study this problem. The GMM is shown to have a high accuracy to identify treated and untreated cell traces. From the cell morphologic similarity analysis, we found that there is no significant correlation between untreated and treated cases. For cell phase identification, the experiments show the GMM has the best recognition accuracy. Also, the experiments show the result from the Fisher clustering is consistent with biological observations as well as KS test.