Content Based Medical Image Retrieval System

Abstract

In the field of medicine, digital images are produced in large quantities, but very few visual image retrieval systems have been used to diagnose diseases. Most of the image retrieval systems present todayare text-based, in which images are manually annotated by text-based keywords. For this human resources are required to annotate and index of medical images. Human intervention for indexing of images can cause errors due to the differences in human perception when describing images. Text based image retrieval methods are not making use of the visual content of images such as color, shape and texture. These methods are insufficient, time consuming and are expensive to implement. In Sri Lankan government hospitals there are no image retrieval systems for diagnostic purposes. The main objective of the proposed system is to provide a content-based medical image retrieval system for diagnostic purposes in the professorial medical ward, Colombo North teaching hospital.

The scope of the proposed system is targeted to implement for primary liver cancer or hepatocellular carcinoma (HCC). Retrieval method of this system is based on the visual content of images and automatic annotation for indexing of images. Extraction of images with regard to size and color is the basis of analyzing liver cancer. This is in accordance with the region based representation of the images. Graphical user web interface was developed to display the search results of similar disease category. The results show the visually most similar images in comparison with the relevant input of the cancer image. This system facilitates the medical professionals to diagnosethe relevant medical condition with greater accuracy which is a way forward in the decision making process. Due to the rapid progress in the technologies of computer science and image processing techniques, medical image retrieval systems have become an important area in many health related applications such as medical diagnostics. Proof-of-concept was conducted in the professorial ward at Colombo North teaching hospital with a hepatologist, in order to check the functionality of the system as per the requirements that were gathered at the beginning. During the POC, the system was tested with a set of images to verify the correctness of the retrieval process. The hepatologist who was involved with the development of the above system offered appropriate feedback regarding the accuracy, efficiency and use-friendliness from a medical perspective.