A comparison of two novel similarity measures based on mutual information in 2D/3D image registration

In image-guided intervention, 2D/3D medical image registration is crucial to assist the clinician to locate precisely the lesions of the patient, which involves many aspects, such as similarity measure. In this paper, two similarity measures, distance coefficient mutual information (DCMI) and distance weighted mutual information (DWMI), were compared for 2D/3D rigid registrations by applying them to porcine skull phantom datasets from the Medical University Vienna. The results showed that DWMI was more accurate than DCMI with lower by 5.25% mean of mTRE and higher by 5.06% success rate but greater 7.74% average iterations. This indicates that the two similarity measures, in which space distance information was used to restrict mutual information in different ways, have no dramatically difference in precision of 2D/3D registration.