Meeting Banner
Abstract #1922

Estimating Registration Variance Using Deformation Field Perturbations

Jan Scholz1, Kaitlyn Easson2, and Jason P Lerch1,3

1Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada, 2Department of Biomedical and Molecular Sciences, Queen's University, Toronto, ON, Canada, 3Department of Medical Biophysics, Department of Medical Biophysics, Toronto, ON, Canada

Most image registration algorithms do not output any information about the variance of the transformation estimates. Here we show that by perturbing input files we can recover this information without modifying the underlying algorithms. We demonstrate that local brain volume estimates can be improved by using the determinant of the average across the distribution of transformations. Our methods will improve morphological analyses, registration-based label alignment, and help find optimal registration parameters.

This abstract and the presentation materials are available to members only; a login is required.

Join Here

Abstract