Jundong Liu

Segmentation of external force field for automatic initialization and splitting of snakes

Active contours or snakes have been extensively utilized in handling image segmentation and classification problems. In traditional active contour models, snake initialization is performed manually by users, and topological changes, such as splitting of the snake, cannot be automatically handled. In this paper, we introduce a new method to solve the snake initialization and splitting problem, based on an area segmentation approach: the external force field is segmented first, and then the snake initialization and splitting can be automatically performed by using the segmented external force field. Such initialization and splitting produces multiple snakes, each of which is within the capture range associated to an object and will be evolved to the object boundary. The external force used in this paper is a gradient vector flow with an edge-preserving property (EPGVF), which can prevent the snakes from passing over weak boundaries. To segment the external force field, we represent it with a graph, and a graph-theory approach can be taken to determine the membership of each pixel. Experimental results establish the effectiveness of the proposed approach.

Segmentation of edge preserving gradient vector flow: an approach toward automatically initializing and splitting of snakes

Active contours or snakes have been extensively utilized in handling image segmentation and classification problems. In traditional active contour models, snake initialization is performed manually by users, and topological changes, such as splitting of the snake, can not be automatically handled. In this paper, we introduce a new method to solve the snake initialization and splitting problem, based on an area segmentation approach: the external force field is segmented first, and then the snake initialization and splitting can be automatically performed by using the segmented external force field. Such initialization and splitting produces multiple snakes, each of which is within the capture range associated to an object and evolved to the object boundary. The external force used in this paper is a gradient vector flow with an edge-preserving property (EPGVF), which can prevent the snakes from passing over weak boundaries. To segment the external force field, we represent it with a graph, and a graph-theory approach can be taken to determine the membership of each pixel. Experimental results establish the effectiveness of the proposed approach.

Local frequency representations for robust multimodal image registration

Automatic registration of multimodal images involves algorithmically estimating the coordinate transformation required to align the data sets. Most existing methods in the literature are unable to cope with registration of image pairs with large nonoverlapping field of view (FOV). We propose a robust algorithm, based on matching dominant local frequency image representations, which can cope with image pairs with large nonoverlapping FOV The local frequency representation naturally allows for processing the data at different scales/resolutions, a very desirable property from a computational efficiency view point. Our algorithm involves minimizing-over all rigid/affine transformations-the integral of the squared error (ISE or L/sub 2/E) between a Gaussian model of the residual and its true density function. The residual here refers to the difference between the local frequency representations of the transformed (by an unknown transformation) source and target data. We present implementation results for image data sets, which are misaligned magnetic resonance (MR) brain scans obtained using different image acquisition protocols as well as misaligned MR-computed tomography scans. We experimently show that our L/sub 2/E-based scheme yields better accuracy over the normalized mutual information.

Nonlinear feature transformation and deep fusion for Alzheimer's Disease staging analysis

Abstract In this study, we develop a novel nonlinear metric learning method to improve biomarker identification for Alzheimers Disease (AD) and Mild Cognitive Impairment (MCI). Formulated under a constrained optimization framework, the proposed method learns a smooth nonlinear feature space transformation that makes the mapped data more linearly separable for SVMs. The thin-plate spline (TPS) is chosen as the geometric model due to its remarkable versatility and representation power in generating sophisticated yet smooth deformations. In addition, a deep network based feature fusion strategy through stacked denoising sparse auto-encoder (DSAE) is adopted to integrate cross-sectional and longitudinal features estimated from MR brain images. Using the ADNI dataset, we evaluate the effectiveness of the proposed feature transformation and feature fusion strategies and demonstrate the improvements over the state-of-the-art solutions within the same category.

Robust Multimodal Image Registration Using Local Frequency Representations

Fusing of multi-modal data involves automatically estimating the coordinate transformation required to align the data sets. Most existing methods in literature are not robust and fast enough for practical use. We propose a robust algorithm, based on matching local-frequency image representations, which naturally allow for processing the data at different scales/resolutions, a very desirable property from a computational efficiency view point. This algorithm involves minimizing - over all affine transformations - the integral of the squared error (ISE or L2E) between a Gaussian model of the residual and its true density function. The residual here refers to the difference between the local frequency representations of the transformed (by an unknown transformation) source and target data. The primary advantage of our algorithm is its ability to cope with large non-overlapping fields of view of the two data sets being registered, a common occurrence in practise. We present implementation results for misalignments between CTan d MR brain scans.

Improved predictive model for pre-symptomatic mild cognitive impairment and Alzheimer's disease

Abstract Objective: Delineation of gray matter (GM) structures on brain MRI scans is termed segmentation. Accuracy of segmentation is a key factor in the valid comparison of GM density and volume between individuals and groups. Previously, it was demonstrated that a group of normal subjects who later developed mild cognitive impairment (MCI) had decreased GM volume in the medial temporal lobe compared to other normal subjects who remained normal an average 5.4 years after the scan. The objective of this study was to show whether accuracy of this predictive model was increased using an advanced segmentation technique. Methods: Structural MRI was performed on 74 longitudinally examined normal aged subjects. All subjects were cognitively normal at the time of their scan, but 18 later developed MCI, and six of these 18 went on from MCI to an AD diagnosis. We independently delineated GM using both a standard segmentation technique and a local Gaussian active contour (LGAC) technique. We compared the contribution of extracted volumes from each technique to a model predicting subjects who will eventually develop MCI. Results: Accuracy of the standard technique to distinguish pre-MCI from normal using imaging alone was 79% (sensitivity 78% and specificity 73%). Using LGAC, accuracy rose to 84% (sensitivity 78% and specificity 84%). Discussion: Structural brain changes precede MCI in longitudinally followed normal subjects. The LGAC technique improves the accuracy of a predictive model incorporating these structural changes by improving GM segmentation and the specificity of the model.

Distribution-based Level Set Segmentation for Brain MR Images

In this paper, we propose a distribution-based active contour model for brain MRI segmentation. As a generalization of the Chan-Vese piecewise-constant model, our solution uses Bayesian a posterior probabilities as the driving forces for curve evolution. Distribution prior, if available, can be seamlessly integrated into the level set evolution procedure. Unlike other region-based active contour models, our solution relaxes the global piecewise-constant assumption, and uses locally varying Gaussians to better account for intensity inhomogeneity and local variations existing in many MR images. More accurate and robust segmentations are therefore achieved. Experiments conducted on synthetic and real brain MRIs demonstrate the improvement made by our model.

Quad-mesh based radial distance biomarkers for Alzheimer's disease

Detailed analysis of brain structures is essential in identifying anatomical biomarkers in Alzheimers disease (AD). In this paper, we develop a new radial distance model to compare different hippocampal shapes and measure their atrophies over time. Using harmonic mappings, we project hippocampal surfaces onto cylinders to obtain evenly-spaced quadrilateral meshes. Surface radial distances estimated via the quad-meshes are invariant to global shifts in the surrounding tissues, leading to a powerful way to detect localized anatomical progressions. The novel quad-meshing method also provides an efficient means to align anatomical surfaces across subjects. Through regions of interest (ROI) analysis, we extract discriminative patches of radial distance and atrophy, and utilize them as anatomical features for patient classification. The effectiveness of the proposed surface modeling and feature extraction strategies in identifying shape biomarkers for AD/MCI is evaluated using the ADNI dataset.

Nonlinear Metric Learning for Alzheimer's Disease Diagnosis with Integration of Longitudinal Neuroimaging Features.

Identifying neuroimaging biomarkers of Alzheimer’s disease (AD) is of great importance for diagnosis and prognosis of the disease. In this study, we develop a novel nonlinear metric learning method to improve biomarker identification for Alzheimer’s disease and its early stage Mild Cognitive Impairment (MCI). Formulated under a constrained optimization framework, the proposed method learns a smooth nonlinear feature space transformation that pulls the samples of the same class closer to each other while pushing different classes further away. The thin-plate spline (TPS) is chosen as the geometric model due to its remarkable versatility and representation power in accounting for sophisticated deformations. In addition, a multi-resolution patch-based feature selection strategy is proposed to extract both cross-sectional and longitudinal features from MR brain images. Using the ADNI dataset, we evaluate the effectiveness of the proposed metric learning and feature extraction strategies and demonstrate the improvements over the state-of-the-art solutions within the same category.

Curve skeleton-based shape representation and classification

Computing the average anatomy and measuring the anatomical variability within a group of subjects are common practices in Computational Anatomy. In this paper, we propose a statistical analysis framework for 2D/3D shapes. At the core of the framework is a parametric shape representation formulated as a concatenation of skeleton points and the discs centered at the points. This shape representation possesses an excellent capability of capturing both global structures and local details. The constructed Riemannian manifold shape space provides a mathematically sound foundation for various groupwise operations, such as calculating the mean shape and conducting structure-specific normalization. Experiments with 2D shapes and 3D human brain structures show the effectiveness of our framework in calculating the distances among different shapes.