Content Based Medical Image Retrieval System (CBMIRS) Using Patch Based Representation

This research work is to develop an efficient and powerful medical search engine to classify and search the radiographic medical images. It focuses on bag of visual words image representation and a similarity matching technique to represent match and retrieve the similar images. This work addresses the issues in content based image retrieval for medical images. In this system can handles different categories of medical images in organ level and the pathology level for chest X-ray images. This simple, efficient medical image categorization and retrieval system in large radiographic archives (IRMA database) is developed for a medicine physicians and researchers those who are interested in being able to retrieve medical images based on low level features. This would make these systems more helpful for radiologists in medical settings, researches in medical analysis and medical students as well as teachers in academic healthcare environments. The methodology presented is based on local patch representation of the image content using a bag of visual words approach with a kernel based SVM classifier. The system supports the classification of X-ray images and retrieval of similar medical images for given input query image. Key terms: CBIR, IRMA, Picture archiving and Communication System, Bag of Visual Words, Computer Aided Diagnosis, Chest Radiography, image Patches.