Gradient descent-based automatic image registration algorithms typically fail when the initial misalignment between objects is large. This is a major limitation for routine clinical applications. The registration task is even more difficult for multi-modal images because of the nonlinear relationship between the pixel intensities in the images to be aligned. In this paper, we present a fast and accurate multi-modal image registration algorithm which successfully registers three-dimensional (3D) computed tomography to two-dimensional single-plane fluoroscopy data for large initial displacements between the images. Our experimental results show that the proposed approach can increase the range of initial displacements up to ± 20 mm for all translations and up to ± 20° for rotation while maintaining high precision and small bias in all six 3D rigid body transform parameters.
|Number of pages||25|
|Journal||Computer Methods in Biomechanics and Biomedical Engineering: Imagine and Visualisation|
|Publication status||Published - 2015|