Research On AD Brain MRI Segmetation Method Based On Edge Detection And Markov Random Filed

With the development of medical imaging technology and the rise of brain science, in recent years, magnetic resonance imaging (MRI) has been more and more widely applied in brain function and brain disease research because of its advantages, such as multi-angle, multi-plan, high resolution and without any harm to human bodies etc. And segmentation of brain MRI has become an important and hot topic to locate, research, diagnose the brain disease in currently. However, how to overcome the various factors on brain MRI such as non-uniformity artifacts, noise and partial volume effects, which caused by the uniformity of magnetic field and the hardware equipment, has become the focus of segmentation of brain MRI.In the existing segmentation methods on brain MRI, the method based-on Markov random field (MRF) model is attracting more and more experts and scholars of the world to conduct more thorough research to it, owing to its distinctive characteristics and stable and reliable results. In this thesis, we learnt the basic knowledge of MRI and based on its characteristics, and then combined the edge detection brain MRI segmentation and the improved MRF method to segment the Alzheimer’s disease data. In order to the subsequent analysis and comparison, we used the DARTEL (Diffeomorphic Anatomical Registration using Exponentiated Lie algebra) registration method to unify all brain tissue data to the same template.First of all, in order to eliminate the influence of non-tissue components in brain MRI, the edge detection method was used to remove non-tissue for all raw data which is convert into NIFTI format, moreover, morphological factors were used to optimize the results and then to increase the detection accuracy rate. Secondly, after extracting brain tissue structure, segmentation method based on optimized MRF was applied to segment the brain structure into gray matter and white matter. Finally, in order to compare the results, we used the DARTEL registration algorithm to register all data to the unified template. After all, we compared our results with the VBM-SPM results which is widely used and have best accuracy in brain MRI filed. Compared from two aspects:the statistics method and the machine learning classification, the statistics results showed that our method could locate the lesion area more precise, and the classification results is higher. These all can indicate that our method can improve the segmentation accuracy of brain MRI. In addition, considering the optimization and processing in the non-uniformity artifacts, noise and partial volume effects etc, our method is not sensitive to the various factors and has a good adaptability and accuracy.