In order to reduce registration time, improve convergence ability and perform better in the presence of noise, a novel medical image registration measure is proposed, which is called generalized ...Jensen-Schur measure. According to Jensen-Schur measure, two kinds of special measures are constructed: Jensen-Schur- alpha and Jensen-Rényi-alpha. The characters of the two measures, mutual information and normalized mutual information are analyzed and compared by applying them to rigid registration of magnetic resonance (MR)/computed tomography (CT) and MR/positron emission tomography (PET) images. The results of tests show that Jensen-Schur-alpha (alpha = 2 or 3) outperforms other measures in the computational speed, convergence performance and noise immunity.
Mutual information has been proved an efficient measure for medical image registration. However it is confined in aligning two images and hard to be applied to mapping multiple images because of its ...large computational cost. A new measure for multiple medical image registration is proposed based on the theory of high dimensional mutual information and arithmetic geometric mean (AGM) divergence. The method first calculates the high dimensional arithmetic geometric mean matrix, and then calculates the entropy of the matrix. The maximal entropy corresponds to the optimal registration solution. The method is tested on brain images. The obtained results show that the proposed method can dramatically decrease registration time, which is a very important consideration in clinical use, with acceptable accuracy.
As a similarity measure of medical image registration, f-information is studied. Mutual information is considered a special type of f-information. In order to reduce sensitivity to changes in ...overlap, two novel normalized I-alpha-information measures are proposed. The function curves, computational time and convergence are studied by applying these measures to rigid registration of computed tomography (CT)/magnetic resonance (MR) and MR/positron emission tomography (PET) images. The results of tests show the effectiveness of normalized I-alpha-information.
A novel method for high-dimensional mutual information registration is proposed. This method first calculates high-dimensional mutual information matrix, and then calculates the entropy of that ...matrix. The maximal entropy corresponds to the optimal registration solution. The method was qualitatively and quantitatively evaluated on simulated and real brain images. The obtained results show that the proposed method can improve registration accuracy and decrease registration time.