Automatic classification of medical images for Content Based Image Retrieval Systems (CBIR)

This paper describes the results after using an automatic classification method to help improve the retrieval of medical images. Using a large dataset of medical images, we established links between low-level features from medical images and high-level features from textual codes of Image Retrieval for Medical Application (IRMA). This paper also explains the process and methods used to automatically classify these medical images, and the results from the classification process. Our best classification results were on image modality with an error-rate of 1%.