Improvement And Implementation Of Image Processing Performance On The X-ray Machine Workstation Based On DICOM

In recent years, incidence of mammary cancer of Chinese females is risingrapidly due to the change of diet and the improving of living standards. The breastcancer has become a kind of serious disease which threatens the health of females. Atpresent, the breast X-ray radiography is considered to be the most effective diagnosismethod. But because the imaging features of early mammary cancer is not obvious,the misdiagnosis of mammary cancer is often happened. Computer-aided detectionbased on mammary X-ray could effectively improve the diagnostic accuracy andreduce missed diagnosis. With the popularity of PACS system and promotion ofDICOM standard, modern digital X-ray machine are required to conform to theDICOM standard to blend in PACS system, to realize the sharing of medicalinformation.The image workstation in X-ray machine system is mostly just to process theDICOM image simply, such as filtering, enhanced and so on. In order to improve theimage processing performance of workstation, and make it play a more important rolein mammary cancer detection, this dissertation put forward to use CAD technology inthe detection of mammary mass on the workstation. Mass detection mainly dividedinto two parts, mass segmentation and mass classification. In the part of segmentation,this dissertation use Confidence Connection Filter method which is one of regiongrowing method to realize segmentation of suspected mass region; In the part ofclassification, this paper use the SVM classifier for mass training and testing. Itextracted the tumor area, regional average pixel, length-width ratio and contrast ascharacteristic parameters. In addition, the dissertation realized information sharingbetween medical equipment by introducing DCMTK library, including the upload,download, query/retrieval, etc. Its innovative is mainly shown in the following twoaspects:(1) First, due to the mass segmentation problems in mammary image,Confidence Connection segmentation method is proposed in this dissertation.Likewise, this dissertation used the SVM classifier for mass training and testing. They made the workstation has the function of auxiliary testing mass and promoteworkstation’s image processing performance.(2) Expand the X-ray machine workstation’s network communication functionby introducing the DCMTK library, including the upload, storage, query/retrieval, etc.That made the workstation to blend in the PACS system and realized the sharing ofmedical information.Experimental results show that the Confidence Connection segmentation methodwas accurate and efficient and could keep the edge information well. At the sametime, the mass characteristic parameters could become an important basis forclassification.The miss rate of classification is low and the result is satisfactory.