Characterization of medical images using edge density

The use of medical images by medical practitioners has increased to an extent that computers have become a necessity in the image processing and analysis. These images along with their detail are crucial when practitioners are diagnosing medical problems in patients. This research investigates if the edge density of a medical image can be used to characterize it. The performance of the edge density feature is assessed by finding its accuracy to retrieve images of the same group from a database. The medical images used in this research are x-rays of the humane body from five different regions, namely; hands, breast, pelvis, skull and chest regions. The edge density feature has shown to produce considerably good results in both, classification of medical images and image retrieval. For the classification using the nearest neighbor and 5-nearest neighbor techniques yielded 82.5% and 85% classification success rates respectively and 75.75% for image retrieval. The edge density approach used in this research is comparable to approaches used in literature considering that other approaches used more than one feature to achieve a higher accuracy and the results obtained in this paper only uses the edge density feature.