Content and semantic context based image retrieval for medical image grid

Abstract

Physicians create Medical Image Libraries (MILs) to collect typical case images, and utilize CBIR (Content-based Image Retrieval) tools to search feature-similar samples within them to aid clinical intervention and diagnoses. This paper presents a CSBIR (Content and Semantic Context based Image Retrieval) scheme for MedlmGrid (Medical Image Grid) to tackle the sharing difficulties of special and heterogeneous MILs within wide areas. The scheme encapsulates distributed CBIR engines, MILs and their metadata as WS (Web Services), and links them in the grid environment. It combines CBIR and semantic context of images to automatically choose the optimal WS set to serve users. Our integrated-features based CSBIR engine for thorax CR (Computer Radiology Image) is related as one instance, which can merge the superiorities of randomly selected CBIR services. MedlmGrid CSBIR prototype is based on CGSP (ChinaGrid Supporting Platform) and its loosely coupled structure makes the integration of decentralized CBIR systems within or across hospitals more efficiently.