Knowledge-based image retrieval: A new generation design
Journal
Proceedings of SPIE - The International Society for Optical Engineering
Journal Volume
2165
Pages
137-148
Date Issued
1994
Author(s)
Abstract
A knowledge-based approach to predicting and pre-fetching prior images has been proposed to better meet radiologists' information and performance requirements during primary reading in a totally digitized radiology environment. The knowledge required to accurately predict the related images that a radiologist would wish to review for the diagnosis of a new examination includes classification/subclassification of reason-for-examination (based on examination requisition and patient information), and anatomical-portion/disease/abnormality-dependent cross-modality, cross-anatomical-portion and temporal examination relationships. Developing knowledge-based system to embody and process such knowledge involves a number of challenges. Specifically, because image retrieval knowledge infers many patient, examination and other pertinent data attributes, knowledge representation and processing need to be effective with knowledge and data relationships. Furthermore, because distributed and object- based architecture offer performance and easy-expansion advantages to database systems that house image-retrieval-related multimedia data, the design of a knowledge-based image retrieval system needs to handle distributed object-oriented data and knowledge interactions efficiently. A new generation Image Retrieval Expert System (IRES) has adopted a coupled knowledge-base/database architecture using an object-oriented knowledge and data representation. This paper will depict the architectural design as well as the detailed design/implementation results of IRES in this new generation.
SDGs
Publisher
SPIE
Type
conference paper