Riemannian Shape Space for Subcortical Structure Analysis  
A novel framework to compare and classify subcortical brain structures. Three dimensional shapes are represented with their meridians, which are extracted from Fiedler vectors and possesses the merit of capturing the salient structure property along the direction of maximal shape variations. Projecting the 3D meridians onto a multi-dimensional sphere, similarity/dissimilarity between shapes can be computed based on a Riemannian spherical distance metric.

Robust Spectral Embeddings and Matching 
Shape matching in the spectral domain has gained great popularity in recent years. Most algorithms, however, rely on invariant global spectral embeddings of the shapes to find correspondence, where spatial neighborhood information is not explicitly incorporated into the matching procedure. Misalignments of global as well as local structures are often resulted due to the lack of spatial guidance. In this project, we identify a number of ambiguities existing in spectral embedding and matching, and subsequently propose a general framework to improve the matching coherence. At the center of the framework is a hybrid spatial-awareness spectral embedding (SASE), which allows various neighborhood and topological information, such as pair-wise distance, relative angles w.r.t. object centers, to be integrated into commute-time (CT) embeddings. A probabilistic expectation maximization (EM) algorithm with imposed regularity is employed to seek an optimal matching of the SASE embeddings.

Analysis of Viscreal and Subcutaneous Fat Tissue in MicroCT of Mice 
One of the common practices in obesity and diabetes studies is to measure the volumes and weights of various adipose tissues, among which, visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) play critical yet different physiological roles in mouse aging. In this project, a robust two-stage VAT/SAT separation framework for micro-CT mouse data is proposed. The first stage is to distinguish adipose from other tissue types, including background, soft tissue and bone, through a robust mixture of Gaussian model. Spatial recognition relevant to anatomical locations is carried out in the second step to determine whether the adipose is visceral or subcutaneous. We tackle this problem through a novel approach that relies on evolving the abdominal muscular wall to keep VAT/SAT separated. The VAT region of interest (ROI) is also automatically set up through an atlas based skeleton matching procedure.

Non-Twist Regularization for Image Registration 
Image registration is inherently ill-posed with infinitely many solutions for a pair of inputs, therefore appropriate regularizations are required to obtain legitimate and meaning transformation.We propose a mathematically sound formulation that explicitly controls the transformation to keep each grid in meaningful shape over the entire geometric matching procedure. The deformation regularity conditions are enforced by maintaining all the moving neighbors as Non-Twist grids. In contrast to similar work, we differentiate and formulate the convex and concave update cases under an efficient and straightforward point-toline/ surface orientation framework to prevent folding. The equality constrained optimization problem is solved by using the augmented Lagrangian multiplier method.

Brain structure segmentation for Parkinson\xa1\xafs Disease 
This project is in collaboration with the University of KentuckySanders-Brown Center of Aging. Our primary hypothesis is that depression in PD is linked to structural and functional abnormality in cortical and subcortical structures such as medial frontal cortex, caudate, maygdala etc. Brain structure segmentation algorithms and software packages are developed to measure the volume changes for these areas. This project is funded by University of Kentucky.

Brain atrophy measurement for Alzheimer\xa1\xafs Disease 
The primary goal of this project is to identify imaging and biomarker metrics that can potentially serve asreliable bases in AD diagnosis and prevention.Advanced image alignment tools are beingdeveloped to answer 1) for each individual AD patient, how does the atrophy change as the clinical stages progress? 2) What is the magnitude of the change? 3) For a group of AD patients, is there any common shrinkage pattern among them? This project is funded by the Sanders-Brown Center onAging at University of Kentucky.

CT/MR registration for neurosurgical planning 
Different modalities usually contain complementary info. For the information to be effectively combined, to align the images is the prerequisite.

Multiple Sclerosis lesion detection from multisequence (T2-weighted and FLAIR) images 
We have developed a fully automatic system for MS lesion that requires minimum user interaction. The separation of the lesion class from other normal tissue types is achieved by minimizing a statistically robust measure called L2E criterion. This project is funded by the Sanders-Brown Center on Aging at University of Kentucky.

T-AND: Toolbox for Analysis of Neurodegenerative Diseases 
Major componentsa. Whole brain tissue segmentation (tLGAC)b. Subcortical structure segmentation (sLGAC)c. Registration assisted cortex parcellation (CortexP)d. Individual Structure-Based Morphometry (iSBM-MS)

T-AND: MITK: Medical Imaging ToolKit 
Major functionalities include CT/MR rigid body registration, MRI/MRI non-rigid registration, etc.

Mice fat segmentation and volume measurement on OU-MicroCT 
This project is in collaboration with Edison Biotechnology Institute at Ohio University, and supported by Office of Vice President for Research (through the BMIT grant).