Quantitative modeling of the phenotypic changes in the host cell during the bacterial infection makes it possible to explore an empirical relation between the infection stages and the quantifiable host-cell phenotype. A statistically reliable model of this relation can facilitate therapeutic defense against threats due to natural and genetically engineered bacterium. In the preliminary experiment, we have collected several thousand cell images over a period of 72 h of infection with a 2-h sampling frequency that covers various stages of infection by Francisella tularenesis (Ft). Segmentation of macrophages in images was accomplished using a fully automatic, parallel region growing technique. Over two thousand feature descriptors for the host cell were calculated. Multidimensional scaling, followed by hierarchical clustering, was used to group the cells. Preliminary results show that the host-cell phenotype, as defined by the set of measureable features, groups into different classes that can be mapped to the stages of infection.