In this paper, we present an approach to obtain a beating heart digital phantom. The goal of this phantom is to obtain a dynamic model of the beating heart surrounded with its neighbor closest tissues and organs. In this phantom, thirteen different tissues were segmented from a dynamic sequence of 20 thoracic CT volumes. We use a semi-automatic method to describe this tissues of the first volume. The temporal evolution of these tissues has then been estimated by propagating this first segmentation to the others volumes by an elastic registration. This dynamic digital phantom can now be used to simulate a therapy, to model the cardiac electrical field, to evaluate registrations methods.