Images are an important part of medical diagnosis. They are produced in ever--increasing quantities and varieties in most modern hospitals. The growth of the volume of data associated with medical imaging generates a need for efficient and effective data access. CBIR (content--based image retrieval) is a technique that enables users to find information based on visual properties, which can be helpful to clinicians in data management and data exploration to complement text--based search for information. However, CBIR for large volume of data is known as to be computational intensive. In this thesis, Grid computing technology and CBIR based on local features are combined together to give solutions to data management and exploration in the medical image database of large volume.