Rui Gan

Multi-dimensional mutual information based robust image registration using maximum distance-gradient-magnitude

In this paper, a novel spatial feature, namely maximum distance-gradient-magnitude (MDGM), is defined for registration tasks. For each voxel in an image, the MDGM feature encodes spatial information at a global level, including both edges and distances. We integrate the MDGM feature with intensity into a two-element attribute vector and adopt multi-dimensional mutual information as a similarity measure on the vector space. A multi-resolution registration method is then proposed for aligning multi-modal images. Experimental results show that, as compared with the conventional mutual information (MI)-based method, the proposed method has longer capture ranges at different image resolutions. This leads to more robust registrations. Around 1200 randomized registration experiments on clinical 3D MR-T1, MR-T2 and CT datasets demonstrate that the new method consistently gives higher success rates than the traditional MI-based method. Moreover, it is shown that the registration accuracy of our method obtains sub-voxel level and is acceptably high.

Statistical cerebrovascular segmentation in three-dimensional rotational angiography based on maximum intensity projections.

Segmentation of three-dimensional rotational angiography (3D-RA) can provide quantitative 3D morphological information of vasculature. The expectation maximization-(EM-) based segmentation techniques have been widely used in the medical image processing community, because of the implementation simplicity, and computational efficiency of the approach. In a brain 3D-RA, vascular regions usually occupy a very small proportion (around 1%) inside an entire image volume. This severe imbalance between the intensity distributions of vessels and background can lead to inaccurate statistical modeling in the EM-based segmentation methods, and thus adversely affect the segmentation quality for 3D-RA. In this paper we present a new method for the extraction of vasculature in 3D-RA images. The new method is fully automatic and computationally efficient. As compared with the original 3D-RA volume, there is a larger proportion (around 20%) of vessels in its corresponding maximum intensity projection (MIP) image. The proposed method exploits this property to increase the accuracy of statistical modeling with the EM algorithm. The algorithm takes an iterative approach to compiling the 3D vascular segmentation progressively with the segmentation of MIP images along the three principal axes, and use a winner-takes-all strategy to combine the results obtained along individual axes. Experimental results on 12 3D-RA clinical datasets indicate that the segmentations obtained by the new method exhibit a high degree of agreement to the ground truth segmentations and are comparable to those produced by the manual optimal global thresholding method.

Multiresolution Image Registration Based on Kullback-Leibler Distance

This paper extends our prior work on multi-modal image registration based on the a priori knowledge of the joint intensity dis- tribution that we expect to obtain, and Kullback-Leibler distance. This expected joint distribution can be estimated from pre-aligned training images. Experimental results show that, as compared with the Mutual Information and Approximate Maximum Likelihood based registration methods, the new method has longer capture range at different image resolutions, which can lead to a more robust image registration method. Moreover, with a simple interpolation algorithm based on non-grid point random sampling, the proposed method can avoid interpolation artifacts at the low resolution registration. Finally, it is experimentally demon- strated that our method is applicable to a variety of imaging modalities.

Maximum distance-gradient for robust image registration.

To make up for the lack of concern on the spatial information in the conventional mutual information based image registration framework, this paper designs a novel spatial feature field, namely the maximum distance-gradient (MDG) vector field, for registration tasks. It encodes both the local edge information and globally defined spatial information related to the intensity difference, the distance, and the direction of a voxel to a MDG source point. A novel similarity measure is proposed as the combination of the multi-dimensional mutual information and an angle measure on the MDG vector field. This measure integrates both the magnitude and orientation information of the MDG vector field into the image registration process. Experimental results on clinical 3D CT and T1-weighted MR image volumes show that, as compared with the conventional mutual information based method and two of its adaptations incorporating spatial information, the proposed method can give longer capture ranges at different image resolutions. This leads to more robust registrations. Around 2000 randomized rigid registration experiments demonstrate that our method consistently gives much higher success rates than the aforementioned three related methods. Moreover, it is shown that the registration accuracy of our method is high.

Vascular segmentation in three-dimensional rotational angiography based on maximum intensity projections

Three-dimensional rotational angiography (3D-RA) is a relatively new and promising technique for imaging blood vessels. In this paper, we propose a novel 3D-RA vascular segmentation algorithm, which is fully automatic and very computationally efficient, based on the maximum intensity projections (MIP) of 3D-RA images. Validation results on 13 clinical 3D-RA datasets reveal that, according to the agreement between the segmentation results and the ground truth, our method (a) outperforms both the maximum a posteriori-expectation maximization (MAP-EM)-based method and the MAP-Markov random field (MAP-MRF)-based segmentation method, and (b) works comparably to the optimal global thresholding method. Experimental results also show that our method can successfully segment major vascular structures in 3D-RA and produce a lower false positive rate than that of the MAP-EM-based and MAP-MRF-based methods.

Distance-Intensity for image registration

In this paper, a novel one-element voxel attribute, namely distance-intensity (DI), is defined for associating spatial information with image intensity for registration tasks. For each voxel in an image, the DI feature encodes spatial information at a global level, and is about the distance of the voxel to its closest object boundary, together with the original intensity information. Without the help of image segmentations, the computation of the DI map is carried out by applying a Poisson process on a vector field that combines both gradient and distance-gradient. Mutual information (MI) is adopted as a similarity measure on the DI feature space. A multi-resolution registration method is then used for aligning multi-modal images. Experimental results show that, as compared with the conventional MI-based method, the proposed method has longer capture ranges at different image resolutions. This leads to more robust registrations. Randomized registration experiments on clinical 3D CT, MR-T1 and MR-T2 datasets demonstrate that the new method gives higher success rates than the traditional MI-based method.

Robust optimization using disturbance for image registration

This paper exploits the different properties between the local neighborhood of global optimum and those of local optima in image registration optimization. Namely, a global optimum has a larger capture neighborhood, in which from any location a monotonic path exists to reach this optimum, than any other local optima. With these properties, we propose a simple and computationally efficient technique using transformation disturbance to assist an optimization algorithm to avoid local optima, and hence to achieve a robust optimization. We demonstrate our method on 3D rigid registrations by using mutual information as similarity measure, and we adopt quaternions to represent rotations for the purpose of the unique and order-independent expression. Randomized registration experiments on four clinical CT and MR-T1 datasets show that the proposed method consistently gives much higher success rates than the conventional multi-resolution mutual information based method. The accuracy of our method is also high.