Abstract
Currently, the images of brain and minute biotissue of the body are obtained by MRI, and it is used for the diagnosis of cancer and injury. Moreover, for diagnosis of diseases of brain and research for brain physiology, it gives us important evidence. So far, the extraction of brain region in MR image has been processed by medical doctors manually. It requires a large amount of labor. Then several algorithms were developed for extracting brain region. Region growing, snake and density histogram are popular algorithms. It was, however, very difficult to extract brain region automatically by these algorithms. Region growing easily induces leakage from the brain region. A snake is not appropriate to apply when detecting an objective region that is surrounded by other objects like brain and scalp. The density histogram is heavily affected by the distribution of density of each MR image, which is changed by MR protocols. In this study, we developed an algorithm for automatic extraction of the brain region from head images acquired from an MRI scanner. This algorithm uses a snake and multi-seed region growing. In this algorithm, the scalp region was deleted before extracting the brain region. Moreover, the leak from brain is checked at the points of scalp detected by the snake. By this method, leakage of the region growing from the brain region was well suppressed. In our research, it was revealed that the brain region was extracted more easily and accurately by using our proposed method compared with conventional methods.
