This paper presents the results of the University at Buffalo in the 2006 ImageCLEFmed task. Our approach for this task combines Content Based Image Retrieval (CBIR) and text retrieval to improve retrieval of medical images. Our results are comparable to other approaches presented in the task. Our results show that text retrieval performs well across the three different types of topics (visual, visual-semantic and semantic) and that the combination of CBIR and text retrieval yields moderate improvements.
展开▼
