Evaluating similarity measure for multimodal 3D to 2D registration

Usha Kiran, Roshan Ramakrishna Naik, Shyamasunder N. Bhat, Anitha H

Research output: Contribution to journalArticlepeer-review

Abstract

The 3D to 2D registration technique in spine surgery is vital to aid surgeons in avoiding the wrong site surgery by estimating the vertebral pose. The vertebral poses are estimated by generating the spatial correspondence relationship between pre-operative MR with intra-operative x-ray images, then evaluated using a similarity measure. Different similarity measures are used in 3D to 2D registration techniques to assess the spatial correspondence between the pre-operative and intra-operative images. However, to evaluate the registration performance of the similarity measures, the proposed framework employs three different similarity measures: Binary Image Matching, Dice Coefficients, and Normalized Cross-correlation technique to compare the images based on pixel positions. The registration accuracy of the proposed similarity measures is compared based on the mean Target Registration Error, mean Iteration Times, and success rate. In the absence of simulated test images, the experiment is conducted on the simulated AP and Lateral test images. The experiment conducted on the simulated test images shows that all three similarity measures work well for the feature based 3D to 2D registration in that BIM gives better results. The experiment also indicates high registration accuracy when the initial displacements are varied up to ±20 mm and ±100of the translational and rotational parameters, respectively, for three similarity measures.

Original languageEnglish
Article number055015
JournalBiomedical physics & engineering express
Volume9
Issue number5
DOIs
Publication statusPublished - 09-2023

All Science Journal Classification (ASJC) codes

  • Nursing(all)

Fingerprint

Dive into the research topics of 'Evaluating similarity measure for multimodal 3D to 2D registration'. Together they form a unique fingerprint.

Cite this