Computing Covariances for Mutual Information Coregistration
Paul A. Bromiley, Daniel Rueckert, Joseph V. Hajnal, Guang‐Zhong Yang · Research Explorer (The University of Manchester) · 2004
Mutual Information (MI) has become a popular similarity measure in multi-modal medical image registration since it was first applied to the problem in 1995. This paper describes a method for calculating the covariance matrix for MI coregistration. We derive an expression for the covariance matrix by identifying MI as a biased log-likelihood measure. The validity of this result is then demonstrated through comparison with the results of Monte-Carlo simulations of the coregistration of T1-weighted to T2-weighted synthetic MR scans of the brain. We conclude with some observations on the theoretical basis of MI as a log-likelihood.