A brain nuclear magnetic resonance image segmentation algorithm based on non-rigid registration

Because different brain diseases, such as hippocampus, accumbens, and thalamus, are closely related with the shape changes of key brain structures. Nuclear Magnetic Resonance (NMR) Image segmentation plays a vital role in brain disease analysis. However, these structures often have no clear contrasted boundaries in brain NAIR images, a priori knowledge based on the atlas is required to better complete the segmentation. In this paper, a hybrid intensity and shape characteristic non-rigid registration algorithm, which is suitable for multiple objects segmentation problem, is proposed. Experiment results proved that better results can be obtained when the proposed algorithm is used to segment brain internal structures.