Automatic hippocampus segmentation based on convolutional neural networks and level set model

The hippocampus has a key role in a number of neurodegenerative diseases, such as Alzheimer's Disease. While many automatic segmentation methods exist, their performances are still challenged due to the poor image contrast of hippocampus in the MR images. In this paper, we present a fast and robust hippocampus segmentation method using convolutional neural networks and level set model. First, images are cropped which reserve the hippocampus region. The convolutional neural network (CNN) is then trained to segment the whole hippocampus region which can be taken as the initial level set contour. The final segmentation is achieved by the evolution of the level set contour. Our method performs better than other state-of-the-art methods. Results indicate the effectiveness of the proposed method, which yields mean Dice similarity coefficients of 0.864.