Leukoaraiosis’ MR Image Registration And Venous Segmentation From White Matter

Leukoaraiosis (LA) is a clinical syndrome caused by different kinds of pathogenesis; its mainclinical manifestation is patient’s cognitive dysfunction. Besides the behaviour test scales, imagessuch as MR is focused on to learn LA at present. Clinical medical study indicates that there isrelevance between the progresses of the LA and the expansion of the venous of cerebral whitematter. Therefore, exploration the relationship between them will be helpful to understand thepathogenesis of LA. Now clinical diagnose to judge the expansion of venous is mainly byNeurologist’s observing and analyzing objectively the patients’MR image. This objective judgmentlacks of quantitative evaluation standard that leads erroneous diagnosis sometimes. So applyingimage processing method to analyze white matter venous change quantitatively is of meaningful fordiagnosising and prognosis of LA.The T1 flair of LA patients’ MR image can show clearly cerebral structure, both the grey matterand the white matter area, but it can’t show the detail information of white matter area venous; bycontraries, the phase image can show the detail information of venous, but lack of cerebral structure.Therefore, the two kinds of MR images show complementary information each other. Based on thiscomplementary characteristic of these two images, image processing works include imageregistration and venous segmentation from white matte was carred in this paper. The mainly workof this paper are as follows:1. The introduction of medical MR image processing knowledge: As the basic knowledge ofimage processing, the concept and methods of registration and segmentation of medical MR imageare introduced in this chapter.2. The registration of T1 image and phase image: First, linear filtering method was applied todenoise these images. Second , the principle of the point based method, the intensity basedcross-correlation method and the intensity based biggest mutual information method is gave out.Then T1 image and phase image was registrated by the three methods. Finally, the feasibility of theintensity based biggest mutual information method was verified.3. The segmentation of white matter: First, the morphological method was used to move thecranium. Second, the K-means method based on density was gave out to segment the area of whitematter. Finally, the advantage of the method was discussed based on the results.Finally, the white matter area of phase image was got on the base of the registraed images.4. The venous segmentation and quantification of white matter: First, the white matter area ofphase image was got on the base of the registraed images. Second, the top-bot method was used to do the pre-process to enhance the image contract. Second, wavelet transform was used to denoisethe equalized histogram. Then, multi-threshold method based on wavelet transform was used tosegment the venous. Finally, the venous information was quantified and the venous proportion ofthe white matter area was got. The basis was provided for further study.