The Research Of Image Process Based On MRI Image Of Mild Cognitive Impairment

Computer vision, digital image processing and visualization became very hot inrecent years as the computers had been widely used in many fields such as medicalimage analysis. On the other hand, MRI which is short for Magnetic Resonance Imag-ing has been developed and applied very rapidly. It utilizes the magnetic resonance ofsome atomic nucleus in human organs to get signals. And then it processes these sig-nals to reconstruct the images of some slices. It could be a diagnostic tool for medicaltreatment especially for head and spine because it can shows the images of these partsvery clearly. At the same time, people can take advantage of computer to unleash thepotential power of MRI. For example, using digital image processing to increase thequality of MRI images; using image segmentation algorithms to get region of interestfor later processing; using 3D reconstruction to restore the original object in computer.Among these applications, 3D reconstruction is very important research and de-velopment areas. Its aim is to retrieve the object structure information from a series of2D images and visualize it.At the same time, in medical treatment there are many ambiguous images whichis hard or impossible for human beings to distinguish. For example, more and moreold people get Alzheimer's disease which is hard to diagnosis and cure. Mild cognitiveimpairment patients whose memory or cognitive have deceased but haven't reached tothe level of Alzheimer's disease, can very likely to get Alzheimer's disease. How-ever, there are errors in the methods used in medical treatments. If computer visionalgorithm could be combined with machine learning algorithm, efficient discriminatealgorithm could be raised according to form a auto diagnosis system based on MRIimages. This kind of system can decrease the diagnosis time while the accuracy rises.3D reconstruction and auto diagnosis could be the fundament for developing a good computer-aided diagnosis system.This paper start with the point of view of medical image processing. Firstly, usingray casting to reconstruct MRI data successfully, so that the researchers could lookthrough the skull and get the vision of infrastructure. In order to get more precise braindata for future diagnosis, further morphometric analysis and structural measurementis still needed. Different from voxel-based morphometry, cortical reconstruction isadopted for morphometric analysis. This process contains non-brain tissue stripping,segmentation, surface deformation and measures the thickness of gray matter to getthe thickness of different cortical regions.Since the number of cortical regions are much more than the number of subjectsand different cortical regions are of different capability of classification, feature selec-tion is used before classification in this paper. The method uses the idea of ensemblelearning, randomly selects the cortical regions to form some subsets, then assign a lin-ear classifier to each of these subsets, and finally use AdaBoost to combine all theseweak classifier to form a strong classifier.