Perelman School of Medicine at the University of Pennsylvania

Section for Biomedical Image Analysis (SBIA)

participating with CBICA

DTI Warping

The growing clinical importance of DTI in disease investigation, concurrent with a significant improvement in the quality and speed of DTI acquisition, has enabled large population longitudinal and cross-sectional studies of WM changes in the brain, which can lead to early diagnosis of disease or to a more effective monitoring of treatment. This has generated the need for sophisticated techniques statistically analyzing DTI data, in order to identify and quantify complex patterns of structural changes associated with inter-individual variability and those induced by pathology. Group-based analysis requires spatial normalization of the DTI data, for which we are currently developing methods, followed by voxel-wise statistical analysis (Statistics of DTI) .

Diffusion tensor (DT) imaging is a relatively new magnetic resonance imaging method, which has emerged during the past few years as a potentially powerful way of understanding connectivity in the brain. DT imaging is based on measurements of microscopic diffusion of water molecules, which provides insight into homogeneous white matters and indicate the direction of nerve bundles. Since brain connectivity is important in studying brain development, aging, and disease processes, DTI is bound to play an important role in these scientific areas. Spatial normalization of tensor fields pose difficulties not previously considered in deformable registration methods of scalar images. In addition to the tensor's relocation to the template space, the orientation of each tensor has to be properly adjusted, which implies that the actual measurement on each voxel is both displaced and changed by the spatial transformation. Moreover, the reorientation of a tensor relies on the shape of the tensor in relation to the deformation field direction at that location. Therefore, the same deformation field prescribes a different re-orientation for different tensors. We are continuously in the process of developing new methods of feature-based spatial normalization with full tensor information being used to compute the features. We expect that these approaches to spatial normalization of DT images will consequently pave the way for the generation of DT statistical atlases for normal brains as well as brains with pathology, and form the base for large population studies.

Earlier Work

Spatial normalization was achieved by computing deformation fields from the deformable registration of Fractional Anisotropy maps followed by reorientation based on Procrustean Estimation. In order to pave the way for feature based matching, we investigated Gabor features.

Registration using Tissue Features

Scalar maps of the fractional anisotropy (FA) and the apparent diffusion coefficient (ADC) derived from DT images are combined together to obtain an estimate of the spatial distribution of white matter (WM), gray matter (GM), and cerebral-spinal fluid (CSF) directly from DTI. These estimates are then used to derive an attribute vector for each voxel, that serve as morphological signatures that assist in automatically determining anatomical correspondence in a hierarchical deformable registration algorithm. DTI warping is combined with tensor reorientation, based on a spatially adaptive procedure that estimates the underlying fiber orientation, so that properly oriented tensors are produced by tensor warping. Extensive experiments on simulated and real data of human and mouse brain images are utilized to assess this method quantitatively and qualitatively, and to evaluate the accuracy and robustness of the proposed registration method. Owing to the use of attribute vectors that facilitate finding anatomical correspondence, the method is also demonstrated to have the capability of handling large deformations, as in the case of large growth or aging related changes, as well as handle data of different resolutions.


  1. Jinzhong Yang, Dinggang Shen, Chandan Misra, Xiaoying Wu, Susan Resnick, Christos Davatzikos, Ragini Verma: "Spatial Normalization of Diffusion Tensor Images Based on Anisotropic Segmentation” International SPIE Medical Imaging, 2008
  2. Jinzhong Yang, Dinggang Shen, Christos Davatzikos, Ragini Verma: "Diffusion Tensor Image Registration Using Tensor Geometry and Orientation Features” International SMICCAI, 2008
  3. Ragini Verma, Feby Abraham, George Biros and Christos Davatzikos: "Landmark guided spatial normalization of Diffusion Tensor Images in the presence of large deformations. ISMRM, Seattle (2nd Prize in Best Poster Awards), May 2006.
  4. Ragini Verma, Feby Abraham, George Biros and Christos Davatzikos: "Correspondence Detection in Diffusion Tensor Images" International Symposium on Biomedical Imaging (ISBI) April 2006.
  5. Ragini Verma, Christos Davatzikos: "Matching and smoothing of diffusion tensor images using oriented Gabor morphological signatures” ISMRM Workshop on methods for quantitative diffusion of human brain, Canada, March 2005.
  6. Ragini Verma, Christos Davatzikos: "Matching of Diffusion Tensor Images Using Gabor Features",Proceedings of the IEEE International Symposium on Biomedical Imaging, p. 396-399, Arlington, Va., 15-18 April 2004.
  7. Ragini Verma and Christos Davatzikos: "Matching of Diffusion Tensor Images using Gabor Features” International Symposium on Biomedical Imaging (ISBI) Page: 396-399, April 2004.
  8. Dongrong Xu, Susumu Mori, Dinggang Shen, Peter C. M. van Zijl, Christos Davatzikos, "Spatial Normalization of Diffusion Tensor Fields", Magnetic Resonance in Medicine, 50(1):175-182 Jul 2003.
  9. Dongrong XU, Susumu Mori, Meiyappan Solaiyappan, Peter C. M. van Zijl, Christos Davatzikos, "A Framework for Callosal Fiber Distribution Analysis", Neuroimage, Vol.14, December 1, 2002, pp.1361-1369.
  10. Dongrong Xu, Susumu Mori,Dinggang Shen, Christos Davatzikos, "Statistically-based Reorientation of Diffusion Tensor Fields",IEEE International Symposium on Biomedical Imaging, Washington, D.C. 7-10 July 2002.