The Spiking Cortical Model Based Structural Representations for Non-Rigid Multi-Modal Medical Image Registration

Non-rigid multi-modal medical image registration plays an important role in various clinical applications. The structural representation based registration methods have recently attached much attention due to the ability to address the influence of intensity differences of multi-modal images. However, these methods cannot represent the structural information of complicated medical images effectively, thereby leading to the unsatisfactory registration results. In this paper, we have proposed the spiking cortical model (SCM) based structural representations to determine the similarity metrics for the non-rigid image registration. The proposed method realizes image structural representations by means of the fractional order generalized entropy of the output pulse sequences generated by the SCM. The sum of squared differences (SSD) between structural representations of multi-modal images is used as the similarity metric. By using the free-from deformation as the transformation model, this similarity metric is optimized by the Limited memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) method to produce the registered image. Experiments on BrainWeb database and Atlas database show that the proposed method has higher registration accuracy than the methods based on the normalized mutual information and the SSD on the entropy images, the Weber local descriptor as well as the edge neighbourhood descriptor.