MRT2-Based Localization Of The Cortical Landmarks

Using brain atlas to label practice T2-weighted MR images could supply maximum information about brain and is useful for the brain diseases diagnose and treatment program making. The key step of using brain atlas to label practice T2-weighted MR images is the foundation of the Talairach space, as the essential factor of the foundation, the precision of all the Talairach cortical landmarks which are located in T2-weighted MR images can affect the label result greatly. To found the Talairach space fast and accurately, this paper proposed a method to locate the Talairach cortical landmarks in T2-weighted MR images:Firstly, locate the planes all the landmarks are according to the definition of the Talairach cortical landmarks: the AP plane (the plane passes through the AC and PC and perpendicular to the MSP), the VPC plane (the plane perpendicular to both the MSP and the AP plane and passes through the PC), the VAC plane (the plane perpendicular to both the MSP and the AP plane and passes through the AC), this step is realized by using the method of minimizing the local symmetry index, location of the AC and PC and a kind of interpolation to rebuild the planes.Secondly, locate the cortical landmarks (A, P, R, L) in AP plane. This step is realized by using FCM algorithm to find the gray level scales of brain tissues and making a kind of projection which project the high gray level values of the CSF-fat-eyes to the lower gray level value scales of grey matter and white matter firstly, and then segment the projected images by using the Watershed algorithm, finally the cortical landmarks (A, P, R, L) is located in the segmented images.Thirdly, locate the cortical landmarks (S, I) in the VPC plane and VAC plane respectively. This step is realized by binarizing images according to a dynamic computed optimum threshold firstly, and then use the Mathematical Morphology opening-closing operation to segment the two-value images circularly, and finally locate the cortical landmarks (S, I) in the segmented images.According to the result, the method is fast, automatic, robust and repeatable.