Video data modeling is the first step of building a video database. Meaningful representation of video data requires high-level knowledge of the underlying phenomenon and inter-relationships of objects embedded in the video. Video data generated by non-invasive medical diagnostic techniques,{}e.g.{} echocardiography, are used widely for monitoring of patients' condition as well as for research purpose. An appropriate video data model based on change of states of component objects present in echo video has been proposed in this paper. This Object State Transition (OST) model segments video based on objects' states and archives the necessary information in a database by suitable indexing method , so that retrieval of pertinent diagnostic information can be efficiently achieved.