Auto-extraction of brain region from MR image by using snake controlled by neural network

In this research, we developed a snake algorithm that was controlled by neural network for detecting the brain region from head images acquired from a MRI scanner. The way of movement of the snake algorithm is decided by several parameters. However, it is very difficult to decide the proper parameters for an image that is noisy or has its average intensity shifted between images like MRI. In our method, first, the neural network learns to output the proper parameters for the snake to converge to the right position. Second, the snake, controlled by neural network, was applied to the MRI image. Then the brain aria was detected automatically. In our study, it was shown that operator labor for detecting the brain region was dramatically reduced by using the proposed method.