Hippocampus segmentation based on prior knowledge and GAC model

Abstract

This paper proposes a hippocampus segmentation method integrating prior knowledge with Geodesic active contour (GAC) model and applies it to the human brain sagittal MR images. GAC model is an effective image segmentation method. For specific objects, however, if some prior knowledge of objects is introduced, better segmentation results can be achieved. According to the form of external force in GAC model, the prior region and prior shape knowledge are both transformed into prior force, which can be embedded in the active contour model. Some important parameters of the model are also analyzed to find out their influence on the trend of contour evolution.