An efficient algorithm for registration of pre- and intra- operative brain MRI images to correct intensity inhomogeneity

Abstract

Spatially-varying intensity inhomogeneity is a severe artifact, which occurs in intra-operative MR images of the brain. This artifact causes implications in the accuracy of image guided navigation systems being performed during neurosurgery procedures. Therefore, it is highly desirable to correct intensity inhomogeneity along with registration process to achieve an efficient and adaptive method to the image guided surgery constraints. In this paper, a modified Residual Complexity similarity measure is used to correct this artifact with the registration of pre- and intra-operative images. The results show that this similarity measure outperforms some well-known similarity measures such as NMI and SSD in registration of pre- and intra-operative images of the brain, by 31.5%.