Pew-Thian Yap PhD, Presenter: Nothing to Disclose

Guorong Wu, Abstract Co-Author: Nothing to Disclose

Hongtu Zhu PhD, Abstract Co-Author: Nothing to Disclose

Weili Lin PhD, Abstract Co-Author: Nothing to Disclose

Dinggang Shen PhD, Abstract Co-Author: Nothing to Disclose

We propose a fast and accurate DTI registration algorithm, called F-TIMER, which leverages multiscale tensor regional distributions and local boundaries for hierarchically driving deformable matching of DT volumes.

Registration is achieved by aligning a set of automatically determined structural landmarks, via solving a soft correspondence detection problem. Based on the estimated correspondences, thin-plate splines are employed to generate a smooth, topology preserving, and dense transformation, and to avoid arbitrary mapping of non-landmark voxels. To avoid local minima, we employ a hierarchical strategy where a small subset of voxels with more distinctive attribute vectors are first deployed as landmarks to estimate a relatively robust low-degrees-of-freedom transformation. As registration progresses, an increasing number of voxels are permitted to participate in refining the correspondence matching. A scheme as such allows less conservative and faster progression of the correspondence matching towards the optimal solution. The dataset consists of DT images of 22 subjects, acquired using a 1.5T scanner. Each of the dataset consisted of 30 gradient directions with the diffusion weighting of b=700s/mm2. The imaging dimension was 256 x 256 with a rectangular FOV of 240 x 240 mm2 and image resolution of 0.9375 x 0.9375 x 2.5 mm3. All of the diffusion tensor data, as well as the derived scalar maps, were skull-stripped to extract the brain parenchyma before they were used in the experiments.

Registering using F-TIMER 21 of the subjects onto a randomly selected subject shows much improved alignment compared to pure affine registration. For further evaluation, the registration accuracy of F-TIMER was quantified by employing a set of images generated from simulated deformation fields, and the results were compared with state-of-the-art methods (Yang et al., MICCAI, 2008; Zhang et al., MedIA, 2006). Comparing the estimated deformation fields with the ground truths, F-TIMER yields 6% to 27% improvement over the other methods. Similar conclusion can be drawn from a fiber tracking experiment where F-TIMER gives 10% to 30% less error. All these are achieved at a speed 4 to 14 faster.

A novel fast and accurate DTI registration algorithm is proposed.

F-TIMER makes removing confounding morphological differences for drawing neurological related inference fast and effective.

Yap, P, Wu, G, Zhu, H, Lin, W, Shen, D, F-TIMER: Fast Tensor Image Morphing for Elastic Registration.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8009526.html