Abstract Archives of the RSNA, 2013
Nadin Kokciyan, Abstract Co-Author: Nothing to Disclose
Rustu Turkay MD, Abstract Co-Author: Nothing to Disclose
Suzan Uskudarli, Abstract Co-Author: Nothing to Disclose
P?nar Yolum, Abstract Co-Author: Nothing to Disclose
Burak Acar PhD, Presenter: Nothing to Disclose
Baris Bakir, Abstract Co-Author: Nothing to Disclose
We present an ontology (ONLIRA - Ontology of the Liver for Radiology) developed for semantic expression of liver CT images. ONLIRA expands the common vocabulary offered by RadLex with domain knowledge for the liver. Specifically, anatomical properties of the liver, and the liver lesions are described. Thus, semantic analysis beyond pure image based analysis of liver CT images can be achieved.
The requirements for ONLIRA were determined via elicitation sessions with two expert radiologists. Each session consisted of three main threads: (1) questions and clarifications, (2) validation, and (3) tuning. During the first thread, we modeled the liver domain by interviewing the radiologists. Validation thread was used to assure the correctness of the developed liver model. The concepts represented by the ontology were refined in the third thread. The Protégé ontology editor was used for development of the ontology, consisting of 40 concepts, 12 relationships and 36 properties. In order to maintain the conformance with RadLex, all RadLex terms corresponding to an ONLIRA concept were referenced.
The radiologists validated the ONLIRA components (concepts, properties, relationships) by using 30 real liver patient cases. The validation was based on subjective assessment of the accuracy and the completeness of ONLIRA based representation of the cases.
75% of the 30 radiology reports (cases) were covered completely by ONLIRA. Remaining cases could not be completely expressed because current version of ONLIRA doesn’t cover concepts about gallbladder calculi (seen in 4 cases), hepatic steatosis (3 cases), hepatectomy (2 cases), lesion with heterogeneous density (1 case).
The novel ONLIRA ontology, that conforms with RadLex, was proposed for the semantic representation of liver CT images. It was shown to be capable of representing the information in radiology reports in a preliminary study over 30 cases. It can potentially be used for semantic analysis of radiology reports, comparison of cases based on semantic similarity as well as structured reporting. ONLIRA can be accessed via the link http://www.vavlab.ee.boun.edu.tr/pages.php?p=research/CARERA/carera2.html
Radiologists can use ONLIRA for structured reporting, and benefit from it in finding relevant patients similar to the image observations of a given patient.
ONLIRA—A Semantic Knowledge Representation of Liver CT Images. Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13044446.html