UB at CLEF2004 cross language medical image retrieval

Abstract

This paper presents the results of the State University of New York at Buffalo in the cross-language medical image retrieval task at CLEF 2004. Our work in image retrieval explores the combination of image and text retrieval using automatic query expansion. The system uses pseudo relevance feedback on the case descriptions associated with the top 10 images to improve ranking of images retrieved by a CBIR system. The results show significant improvements with respect to a base line that uses only image retrieval. © Springer-Verlag Berlin Heidelberg 2005.