Classification of Medical X-ray Images for Automated Annotation

Of late, the amount of digital X-ray images that are produced in hospitals is increasing rapidly. Efficient storing, processing and classifying X-ray images have thus become an important research topic. Due to the increase in medical digital images, there is a rising need of managing this data properly and accessing them accurately. To overcome the difficulties of manual classification, the automated method is preferable and hence proposed in this work, wherein an effort has been made to automatically classify X-rays at the macro level (coarse level) using SVM classifier with six classes of X-ray images being taken, viz., chest, foot, spine, neck, head, and palm. Each class consists of 30 images collected from IRMA database. Initially, preprocessing is performed by using the M3 filter and its region-of-interest is found by applying connected component labeling (CCL) and both the shape and texture features were extracted. The fusion of shape and texture features gave a better performance of 96.56%.