Increasing computerization of the process of acquiring and storing diagnostic images requires more sophisticated approaches to image retrieval in order to optimally use this information for improved health-care delivery and research. This research will develop a new technique for image retrieval that will assist a physician in interpreting diagnostic imaging studies. The image retrieval system will give the physician the ability to query an image database of known cases and retrieve images (and associated textual information) that contain regions with features similar to what is in the image of interest. Through an interactive, menu-driven software interface, physicians will query the system by marking regions of interest in the query image. The system will then retrieve similar images ranked using a similarity criterion of the quantifiable properties of the abnormality. To facilitate development of the approach, the research will focus initially on content-based image retrieval of a specific category of disease and one form of imaging: the use of high resolution computed tomography (URCT) in patients with a variety of lung diseases. This area has been identified as one where inexperienced Radiologists would benefit greatly from the capability to query images by content to aid in making a diagnosis. The proposed research will address the following issues: the role of anatomical features in medical image retrieval, automatic selection of morphology delineation algorithms, optimal indexing of images for entry into and retrieval from a large database and automated improvement of the system based on physician feedback of query results.
| Aisen, Alex M; Broderick, Lynn S; Winer-Muram, Helen et al. (2003) Automated storage and retrieval of thin-section CT images to assist diagnosis: system description and preliminary assessment. Radiology 228:265-70 |
