The shape of human cerebral cortex is influenced by factors such as heredity, development, plasticity, gender, neuropsychiatric disorders and so on. As a result, variances exist in different populations and among individuals. By analyzing the shape of cerebral cortex, we could explore the mechanisms of the factors affecting the cerebral cortex, and help diagnose neuropsychiatric disorders. The emergence of structural magnetic resonance imaging (MRI) provides us a better technique to investigate the shape of cerebral cortex. The shape analysis depends on various measures. In this paper, we described how to perform shape analysis of the cerebral cortex via three different meansures in a multiresolution manner: At the local level, we investigated the key neurodevelopmental factors that determine cortical thickness, namely synaptogenesis and regression, by analyzing the thickness of the visual cortex in humans with early and late onset blindness. The bilateral visual cortices of the early blind were significantly thicker than those of the late blind and the sighted controls, but the latter two groups did not differ significantly. This suggests reduced “pruning” of synapses in the visual cortex, which may be due to a lack of visual experience during a critical developmental period. These findings support the hypothesis that sensory experience is necessary for an appropriate regression and remodeling of neuronal processes and that synaptic regression might be a major determinant of macroscopic anatomical features like cortical thickness. At the intermediate level, we employed spherical wavelets to encode the shape information of the cerebral cortex. Based on this method, we proposed a classifier discriminating schizophrenia patients using the surface shape features of the cerebral cortex reconstructed from MRI. Leave-One-Out validation achieved high accuracy in classifying male schizophrenia patients. At the global level, we proposed a robust and accuracy algorithm for fractal dimension estimation. This algotrithm based on a cubic-triangle intersection checking method which was used in brain research for the first time, and the widely used box-counting method in fractal dimension estimation. The two features endowed our algorithm robustness, and its accuracy was validated using both artificial and real MR images.
修改评论