Deformation invariant attribute vector for 3D image registration: method and validation

Gang Li, Tianming Liu, Geoffrey S. Young, Lei Guo, Stephen T. C. Wong

Abstract: This paper proposes a novel method to define deformation invariant attribute vector for each voxel in 3D image for the purpose of anatomic correspondence detection. This is the extension of the work for 2D deformation invariant attribute using geodesic intensity histogram (GIH). Our original contribution is to extend this 2D technique to 3D image, and validate the method using synthesized deformation in 3D brain MRI image. Both theoretic analysis and initial validation result show that the proposed attribute vector is invariant to deformation. This deformation invariant attribute vector has wide applications in registration of 3D medical images