Dicentric Chromosome Assay (DCA) has been used for estimating the absorbed radiation dose following occupational or incidental radiation exposure. The estimating procedure of DCA is labor intensive and time consuming, and therefore, it cannot be widely utilized for radiation mass casualty incidents. Automatic estimation systems of DCA has been studied to improve the efficiency of the estimation procedure in previous studies. These methods, however, cause significant problems on the accuracy of the chromosome detection. For improving the accuracy and efficiency of the DCA, an automatic estimation system of DCA using deep learning technique is under development in our research group.