The IRMA database consists of 10000 annotated radiographs taken randomly frommedical routine at the RWTH Aachen University Hospital-Germany. The images are separated into 9000 training images and 1000 test images. The images are subdividedinto 57 classes. For CBMIR, the relevances are defined by the classes, given a query image from a certain class, all database images from the same class are considered relevant. The IRMA database was used in the ImageCLEF 2005 image retrieval evaluation forthe automatic annotation task.