A novel mutual information-based similarity measure for 2D/3D registration in image guided intervention

In image-guided intervention, 2D/3D medical image registration is crucial to supply the clinician space and anatomy information. Digitally reconstructed radiographs (DRR) obtained from 3D volume data are usually compared iteratively with an x-ray image by selecting similarity measure until a match is achieved. In this paper, a new similarity measure based on mutual information (MI) was proposed for 2D/3D rigid registration by combining intensities with space coordinates. By applying the measure to porcine skull phantom datasets from the Medical University Vienna, it is shown that the mean iteration of the measure and mean target registration error (mTRE) is respectively lower by 49.51% and 27.29% than that of mutual information. The proposed similarity measure is more robust and convergent faster than MI in 2D/3D registration.