ONTOLOGY CREATION FOR WIRELESS CAPSULE ENDOSCOPY VIDEOS

In this paper we study multimedia ontology for Wireless Capsule Endoscopy (WCE) videos by enhancing its existing data structure.The 'wireless capsule' is a tiny disposable video camera that transmits 2 ~ 3 frames per second for a period of 8 ~ 11 hours.There are open problems in WCE, such as bleeding detection, as it is hard to identify accurately, using low-level features, i.e., color values.In addition, the physicians have to examine the videos continuously for two hours or more, which becomes restrictive.There have been research attempts to reduce this review time.However, they suffer from low accuracy and sensitivity, and do not process WCE videos with an efficient information data structure.To address this problem, we propose a new data structure named 'multimedia ontology for WCE videos' formed by combining medical and multimedia domain knowledge.Ontology represents a structure to describe the concepts and relationships in a specific domain with relevant data and its terminology.we define two types of ontology, i.e., generic and specific ontology.Generic ontology represents the broad concepts in WCE videos, such as medical terms, anatomic information, video format, etc., while specific ontology is a datadriven one including color, location, and region of images.The process of creating multimedia ontology consists of three steps: (1)collection of raw data from WCE videos, such as video data format, feature values, meta-data information and anomalies, (2) classification of the raw data into concepts including generic and specific ontology, and (3) identification of relationship between two concepts such as 'Is-A', 'Part-Of', and 'Has-A'.This WCE Ontology structure can be used to better address the open problems by providing 'relevant area focus' from the formed structure and can also be extended to other problems like detection of lesions and polyps.