The Research Of Automatic Feature Extraction For Medical Image Content-Based Retrieval

There is an enormous need for visual information organization, management, and retrieval in the growing field of digital archives or electronic patient recorders and by the increasing application domains of medical imaging and PACS (Picture Archiving and Communication Systems). In particular, techniques of content-based image retrieval (CBIR) have been major topics of research for medical image database queries instead of text-based searching techniques in recent years.Significant developments of CBIR techniques have been made in research and commercial applications since 1990's. However, there are some unique challenges to confront CBIR application with the medical image databases.There are still some medical constraints to image retrieval. Particularly, the number and kind of features to characterize medical images are subject to continuous evolution with the advancement of ages and deepness of understanding medical knowledge. Since the lack of effective indexing scheme that the general researchers can find to organize available features in medical images, the conventional CBIR-systems cannot guaramtee a meaningful query completion when used within the medical context.In the process of radiology medico-diagnosis, clinic determinations are usually based on the regional anatomic and physiologic information in medial images. In order to extract the regional features automatically, each of the medical images in database is segmented based on EM (Expectation-Maximization) algorithm. All of extracted regional features are mapped to the fuzzy feature space so that then uncertainty of image segmentation can becharacterized. Together with the features extracted from segmented regions, the features of regions of interest (ROI) are introduced to refine the query automatically. Matched with the integrated similarity measurement, the retrieved images are not only visually similar to the query image but also approximate in diagnosis meaning.The proposed CBIR method has been implemented in our experiments. Compared with the other method applied to retrieve images only based on the fuzzy regional feature, the proposed approach greatly improve the precise of medical image queries and preserve the robustness to segmentation-related uncertainties.In this paper, a general concept for integration is presented that relies on standard protocols and maintains the autonomy of both, PACS and CBIR. To enable medical CBIR system to guarantee the meaningful queries with priorly specified image class, a novel coding scheme so-called MOAB, which consists of four axes: Modality, Orientation, Anatomy, and Biology, is proposed. In our experiments, The MOAB classification coding scheme enables a unique classification of medical images so as to develop further content-based medical image retrieval. The code is flexible and easily to be extended. In this paper, we try to define general interfaces that are required for integration of CBIR to PACS, and maintain their autonomy as self-standing applications.