Study of pattern understanding of medical X-ray images

点击翻译按钮获取中文摘要

In this project we studied basic mthodology of realizing computer aided diagnosis of X-ray images (CADX) in a highly intelligent level with applying image understanding techniques. We put emphasis on the following points :(1) CADX for screening, of Iring by 100mm Photofluorograns : We aim here that we develop the high level of mass screening system for detecting lung cancer in its early stage by improving the currently used screening system.(2) lntroduction of three dimensional (3D) images by using CT images : This is for developing quite new system of lung cancer detection which can find very early stages of cancer that can not be detected by any existing survey system. We eipect that this may be realized by using 3D lung images obtained by the recent helical scan X-ray CT equipment. Besides the above images we consider also the CADX of double contrast stomach X-ray images for diagnosing stomach cancer.Research results are classified into three categories : survey, basics of image pro … More cessing algorithms, and experiment of CADX using real X-ray images.(a) Survey : We performed the survey of recent trends and state of the arts in image understanding technology, 3D image analysis, computer graphics. and historical review and present state of CADX.(b) lmage processing algorithm : Fol-loving algorithms were developed and studied both theoretically and experimentally.・3D distance transformation and figure decomposition・region growing using both edge and region information・filter for detecting mass pattern・applications of neural net to image compression and feature extraction・display of 3D gray tone images(c) Applications to X-ray images : Basic experiments were performed concerning the following CADX.・detection of stomach cancer from double contrast stomach X-ray images・detection of lung cancer and diagnosis of pneumoconiosis. from chest X-ray images・diagnosis of lung cancer and pneumoconiosis from 3D chest CT images・extraction of heart and measurement of its functions from 3D chest MRI images Less