The Design And Implementation Of Multimodality Brain Registration Algorithm

In recent years, PET, as an important clinical examination imaging technology of radionuclide imaging, has become the indispensable tool of for cancer and neurological diseases diagnosis. Mulitmodality brain registration of PET via particular methods to create matching relationship between the origional brain image and standard atlas, then the original image is aligned to the reference brain atlas, so as to achieve the purpose achieve the information of different brain function area and diagnosis of neurological disease. Talairach atlas is an internationally recognized standard brain atlas in the field of brain image registration. Each brain area was labeled a different mark so that the clinical research could be implemented easily. Talairach atlas has been widely used in neurosurgery research. The information of functional metabolic in brain can be obtained by PET-Talairach atlas registration. It has important value in many areas such as diagnosis for stroke, epilepsy and other neurological disorders. However, due to the defect of poor resolution and brain structural features are not clear on PET image, so it is difficult to select the points accurately when PET brain registration. As CT has good spatial resolution and high density resolution, therefore, marking the feature points on the CT images, by PET-CT-standard brain atlas transform algorithm to achieve the registration of PET and standard brain atlas can compensate the defect such as the weakness description of the brain structure on PET. It is crucial for the diagnosis of neurological disorders. In this paper, we use the Talairach atlas as the standard brain image, design and implementation the algorithm of PET-Talairach atlas and PET-CT-Talairach atlas multimodality brain registration algorithms. The paper includes the following sections:First, we design and implement a PET-Talairach atlas registration algorithm. Based on affine transformation, the algorithm uses the cerebral cortex point to establish contradictory equation of PET-Talairach atlas to calculate its affine transformation matrix. Using the matrix to implement registration; Second, combined PET-CT registration techniques, we design a brain automatic extraction algorithm on CT, and then an oritend bounding box extraction algorithm is proposed after the brain segmentation. Finally a median sagittal plane extration algorithm based on mutual information is proposed. On the basis of those resrarch, the points of CT are calculated to implement the registration. By the transformation matrix of PET-CT based on mutual information, the points of CT are mapped to PET. And PET-Talairach atlas multimodality brain registration c achieved. Finally, experimental results show that the proposed algorithm works well in registration of PET images, and the registration algorithm by PET-CT received better registration accuracy and less calculation time.