In this thesis, we propose a multi-modal image registration method based on the a priori knowledge of the expected joint intensity distribution estimated from aligned training images. The goal of the registration is to find the optimal transformation such that the discrepancy between the expected and the observed joint intensity distributions is minimised. The difference between distributions is measured using the Kullback-Leibler distance (KLD) ....[ Read more ]

In this thesis, we propose a multi-modal image registration method based on the a priori knowledge of the expected joint intensity distribution estimated from aligned training images. The goal of the registration is to find the optimal transformation such that the discrepancy between the expected and the observed joint intensity distributions is minimised. The difference between distributions is measured using the Kullback-Leibler distance (KLD) .

The proposed method is conceptually different from the mutual information based registration method, which encourages the functional dependence between the two image random variables. The KLD based registration method guides the transformation based on the difference between the expected and observed joint intensity distributions, or, in other words, based on the expected outcomes learned from the training data.

Experimental results in T1-T2 (3D-3D) registration show that the KLD based registration algorithm is less dependent on the size of the sampling region than the Maximum log-Likelihood based registration method.

The proposed method has been applied to 3D-3D registration problems between time of flight magnetic resonance angiographic (TOFMRA) and phase contrast magnetic resonance angiographic (PCMRA) as well as 2D-3D registration problems between digital subtraction angiograms (DSAs) and magnetic resonance angiographic (MRA) image volumes.