Discriminative Cue Integration for Medical Image Annotation

Automatic annotation of medical images is an increasingly important tool for physicians in their daily activity. Hospitals produce nowadays an increasing amount of data. Manual annotation is very costly and prone to human mistakes. This paper proposes a multi-cue approach to automatic medical image annotation. We represent images using global and local features. These cues are then combined together using three alternative approaches, a high-level, a mid-level and a low-level fusion scheme, all based on the Support Vector Machines (SVM) algorithm. We tested our methods on the IRMA database, and with the mid- and high-level integration scheme we did participate to the 2007 ImageCLEFmed benchmark evaluation, in the medical image annotation track. These algorithms ranked first and fifth respectively among all submission. Experiments using the low-level integration scheme also confirm the power of cue integration for this task.