Understanding tumor heterogeneity is an urgent and unmet need in cancer research. In this study, we used a morphology-based optical cell separation process to classify a heterogeneous cancer cell population into characteristic subpopulations. To classify the cell subpopulations, we assessed their morphology in hydrogel, a three-dimensional culture environment that induces morphological changes according to the characteristics of the cells (i.e., growth, migration, and invasion). We encapsulated the murine breast cancer cell line 4T1E, as a heterogeneous population that includes highly metastatic cells, in click-crosslinkable and photodegradable gelatin hydrogels, which we developed previously. We observed morphological changes within 3 days of encapsulating the cells in the hydrogel. We separated the 4T1E cell population into colony- and granular-type cells by optical separation, in which local UV-induced degradation of the photodegradable hydrogel around the target cells enabled us to collect those cells. The obtained colony- and granular-type cells were evaluated in vitro by using a spheroid assay and in vivo by means of a tumor growth and metastasis assay. The spheroid assay showed that the colony-type cells formed compact spheroids in 2 days, whereas the granular-type cells did not form spheroids. The tumor growth assay in mice revealed that the granular-type cells exhibited lower tumor growth and a different metastasis behavior compared with the colony-type cells. These results suggest that morphology-based optical cell separation is a useful technique to classify a heterogeneous cancer cell population according to its cellular characteristics.