Analysis Of MR Brain Image And Research Of Key Issues For Computer-aided Diagnosis Of Brain Atrophy

The incidence of brain atrophy is increasing. While, people mostly use manual and semi-auto method to diagonose brain atrophy disease at present. Along with the rapid development of computer technology, computer aided diagnosis method has become a major research field in medical image, radiation diagnosis and the computer science. Research shows that computer-aided diagnosis method can play a positive role for accurate diagnosis and reduce misdiagnosis. Doctors can use the measurement result to make quickly assess and diagnose the disease. While, cerebral organization segmentation and volume calculation is the basis of computer-aided diagnosis of brain atrophy.This thesis analyzed the related diagnosis methods of brain atrophy disease, including volume measurement of intracranial brain, brain white matter, gray matter, CSF and the hippocampus structure. While, the image segmentation is the basis of volume measurement. So, this thesis gave some experimental analysis about how to extract related brain regions.(1) Researched some methods of how to extract intracranial brain tissue and analyzed the influence of those threshold segmentation algorithms. And then, we chose the histogram threshold segmentation algorithm and used morphological methods to obtain intracranial brain area. At last, we used the overlap rate and image center to adjust the segmention result. Through the segmentation of all layers in different scanning direction, we finally obtained the whole image segmention result.(2) Researched the sub-segmentation methods of intracranial brain tissues. In this chapter, we used an improved markov random field algorithm with adaptive fuzzy similarity to segment the intracranial brain tissues, which added fuzzy neighborhood information to potential function groups in traditional markov. First, according to Bayes theorem, we transformed the segmentation issue to the Maximum A Posteriori (MAP), and then transformed it to the minimum energy function. Last, we used the Iterative Conditional Model (ICM) to find out the solution of the minium energy function.(3) Researched the segmentation algorithms about hippocampus structure. The hippocampus structure is related to specific brain atrophy. In this chapter, we researched a hippocampus segmentation method with an improved fast marching algorithm. The improved algorithm used the intensity of internal hippocampus structure to constrant the segmentation process, and finally according the time threshold value to complete the segmentation and obtain a very perfect segmentation result.Those experiment results showed that the proposed segmented algorithm of the intracraniale brain tissues can quickly and accurately extract brain tissues. And the sub-segmentation algorithms of intracraniale brain tissues and hippocampus structure had high segmentation accuracy. The mean correct segmentation rate of brain white matter, gray matter and cerebrospinal fluid was improved2.2%than traditional markov algorithm. The correct segmentation rate of improved fast marching algorithm was increased nearly10%than traditional segmentation algorithm. Accuracy segmentation has an important research value and meanings on subsequent calculation of brain volume and the diagnosis of brain atrophy.