Knowledge Representation Research

Knowledge Represenation Research includes the components outlined below.

The Knowledge Representation Research Group

The importance of knowledge representation (KR) and ontology for software systems in support of medical applications is long-recognized. The KR research group in the computer science department specializes in KR languages based on default logic. The group studies and develops automated reasoning tools for knowledge representation languages, and investigates their applications in data and rule integration.

UK's Miroslaw Truszczynski, PhD, and Victor Marek, PhD, collaborated with Walton Sumner, MD (Washington University, St. Louis) to build a family medicine model used in simulations and automated generation of problems for certification exams. That work – which involved development of ontologies, rule systems, and a database design – resulted in a patent and two publications.

The integration of existing databases proposed for the CCTS will require the reconciliation of the existing standard dictionaries and nomenclature (e.g., SNOMED-CT, ICD-9). The tools developed by the UK KR group allow for modeling and reasoning of interactions among such terms.

Metadata Framework for Digitized Pathologic Images

BIC researcher, Sujin Kim, PhD, with the Markey Cancer Center and collaborators at the Massachusetts General Hospital, has received funding from the Institute of Museum and Library Sciences to create a standardized metadata framework for pathologic images to facilitate the storage, management, retrieval and sharing of well-described, integrated biomedical imaging information.

This three-year project addresses critical needs of the biomedical imaging community for metadata tools that support comprehensive biomedical image libraries. Digital imaging in pathology increases the capture and storage of visual findings related to diagnoses in digitized formats. The project will evaluate four existing sources of potential metadata identified in preliminary studies and collect relevant data elements.

However, since no single system currently provides all the data elements required to adequately describe pathologic images, her review will validate the strengths of disparate existing datasets, determine areas of overlap and duplication, and provide a foundation to collect, clean, and map potential candidate data elements.

Expected project results include significant translation of core concepts in information representation – cataloging, classification, authority and access control, subject analysis, arrangement and display, and vocabulary control, which have been developed, standardized and practiced – into new and emerging information management needs.