Deformable Template Research Articles

Stochastic algorithm for Bayesian mixture effect template estimation

The estimation of probabilistic deformable template models in computer vision or of probabilistic atlases in Computational Anatomy are core issues in both fields. A first coherent statistical framework where the geometrical variability is modelled as a hidden random variable has been given by [S. Allassonnière et al., J. Roy. Stat. Soc. 69 (2007) 3–29]. They introduce a Bayesian approach and mixture of them to estimate deformable template models. A consistent stochastic algorithm has been introduced in [S. Allassonnière et al. (in revision)] to face the problem encountered in [S. Allassonnière et al., J. Roy. Stat. Soc. 69 (2007) 3–29] for the convergence of the estimation algorithm for the one component model in the presence of noise. We propose here to go on in this direction of using some “SAEM-like” algorithm to approximate the MAP estimator in the general Bayesian setting of mixture of deformable template models. We also prove the convergence of our algorithm toward a critical point of the penalised likelihood of the observations and illustrate this with handwritten digit images and medical images.

Learning Active Basis Model for Object Detection and Recognition

This article proposes an active basis model, a shared sketch algorithm, and a computational architecture of sum-max maps for representing, learning, and recognizing deformable templates. In our generative model, a deformable template is in the form of an active basis, which consists of a small number of Gabor wavelet elements at selected locations and orientations. These elements are allowed to slightly perturb their locations and orientations before they are linearly combined to generate the observed image. The active basis model, in particular, the locations and the orientations of the basis elements, can be learned from training images by the shared sketch algorithm. The algorithm selects the elements of the active basis sequentially from a dictionary of Gabor wavelets. When an element is selected at each step, the element is shared by all the training images, and the element is perturbed to encode or sketch a nearby edge segment in each training image. The recognition of the deformable template from an image can be accomplished by a computational architecture that alternates the sum maps and the max maps. The computation of the max maps deforms the active basis to match the image data, and the computation of the sum maps scores the template matching by the log-likelihood of the deformed active basis.

Multi-manifold diffeomorphic metric mapping for aligning cortical hemispheric surfaces

Cortical surface-based analysis has been widely used in anatomical and functional studies because it is geometrically appropriate for the cortex. One of the main challenges in the cortical surface-based analysis is to optimize the alignment of the cortical hemispheric surfaces across individuals. In this paper, we introduce a multi-manifold large deformation diffeomorphic metric mapping (MM-LDDMM) algorithm that allows simultaneously carrying the cortical hemispheric surface and its sulcal curves from one to the other through a flow of diffeomorphisms. We present an algorithm based on recent derivation of a law of momentum conservation for the geodesics of diffeomorphic flow. Once a template is fixed, the space of initial momentum becomes an appropriate space for studying shape via geodesic flow since the flow at any point on curves and surfaces along the geodesic is completely determined by the momentum at the origin. We solve for trajectories (geodesics) of the kinetic energy by computing its variation with respect to the initial momentum and by applying a gradient descent scheme. The MM-LDDMM algorithm optimizes the initial momenta encoding the anatomical variation of each individual relative to a common coordinate system in a linear space, which provides a natural scheme for shape deformation average and template (or atlas) generation. We applied the MM-LDDMM algorithm for constructing the templates for the cortical surface and 14 sulcal curves of each hemisphere using a group of 40 subjects. The estimated template shape reflects regions which are highly variable across these subjects. Compared with existing single-manifold LDDMM algorithms, such as the LDDMM-curve mapping and the LDDMM-surface mapping, the MM-LDDMM mapping provides better results in terms of surface to surface distances in five predefined regions.

Mesh Reconstruction Based on Deformable Template and Local ICP Algorithm

According to mesh reconstruction problem of point cloud models with similar topological structure, a mesh reconstruction algorithm of point cloud model is presented based on deformable template and local ICP algorithm. Firstly, point cloud model and template model are aligned in the least squares sense by human-computer interaction. Use the set of point cloud to construct k-d tree and search correspondence of vertices of template model in the point cloud model. Then, the set of correspondence Φ, between two models, is built. For any point Vi of template model, the transformation Ti of each point Vi is computed based on a local supporting neighborhood Vi : Nk(Vi) around Vi. For computation of Ti, we consider only the local correspondence that is inside the local supporting neighborhood Nk of Vi :Φ| Nk(Vi). In the iterative course of deformation, reduce local supporting neighborhood of each vertex Vi gradually to make it accurately approach point cloud model step by step. When the error between two of them is less than a given value, current template model can be taken as the reconstruction model of point cloud. It’s proved by example that the model we get can satisfy engineering request.

Evaluation of the Dynamic Deformable Elastic Template model for the segmentation of the heart in MRI sequences

We introduce a bio-inspired dynamic deformable (DET) model based on the equation of dynamics and including temporal smoothness constraints. The behaviour and characteristics of the dynamic DET model is studied in the context of the semi automatic spatio-temporal segmentation of the left ventricle myocardium in cine-MR images. The segmentation accuracy for endo/epicardium contours at end-diastole and end-systole, and as consequence the performance and limits of the current implementation, is evaluated in the context of the MICCAI LV Segmentation Challenge on a database of 15 multi-slice cine-MRI examinations.

Learning to Match: Deriving Optimal Template-Matching Algorithms from Probabilistic Image Models

Finding correspondences between images by template matching is a common problem in image understanding. Although a variety of solutions have been proposed, most of them rely on the arbitrary choice of a template and a matching function. Often, different cost functions lead to different results, and the choice of a good cost for a specific application remains an art. Statistical models on the other hand, allow us to derive optimal learning and matching algorithms from modeling assumptions using likelihood maximization principles. The key contribution of this paper is the development of a statistical framework for learning what function to optimize from training examples. We present a family of statistical models for grayscale images, which allow us to derive optimal template-matching algorithms. The intensity at each pixel is described by a random variable whose distribution is encoded by a deformable template. Firstly, we assume the intensity distribution to be Gaussian and derive an intensity-matching algorithm, which is a generalization of the classical sum-of-squared differences. Then, we introduce a hidden segmentation variable in the probabilistic model and derive a segmentation-matching algorithm that can handle photometric variations. Both models are exemplified on the automatic detection of anatomical landmarks in brain Magnetic Resonance Images.

REDUCING STRUCTURAL VARIATION TO DETERMINE THE GENETICS OF WHITE MATTER INTEGRITY ACROSS HEMISPHERES - A DTI STUDY OF 100 TWINS.

Studies of cerebral asymmetry can open doors to understanding the functional specialization of each brain hemisphere, and how this is altered in disease. Here we examined hemispheric asymmetries in fiber architecture using diffusion tensor imaging (DTI) in 100 subjects, using high-dimensional fluid warping to disentangle shape differences from measures sensitive to myelination. Confounding effects of purely structural asymmetries were reduced by using co-registered structural images to fluidly warp 3D maps of fiber characteristics (fractional and geodesic anisotropy) to a structurally symmetric minimal deformation template (MDT). We performed a quantitative genetic analysis on 100 subjects to determine whether the sources of the remaining signal asymmetries were primarily genetic or environmental. A twin design was used to identify the heritable features of fiber asymmetry in various regions of interest, to further assist in the discovery of genes influencing brain micro-architecture and brain lateralization. Genetic influences and left/right asymmetries were detected in the fiber architecture of the frontal lobes, with minor differences depending on the choice of registration template.

SU‐FF‐J‐174: Volume Rendered Cardiac Segmentation and Analysis for Breast Radiotherapy

Purpose: Development of a non‐rigid deformable template algorithm for the segmentation of cardiac anatomy allowing for quantification of dose delivered to the heart during breast irradiation. Method and Materials: A high resolution, contrast enhanced, cardiac CT scan was used for manual delineation of cardiac? anatomy. This reference scan consisted of 163 slices, 0.75 mm thick. In‐plane reconstruction voxel dimension of the scan was 0.373×0.373×0.75mm. Manual contours were generated to delineate cardiac anatomy, describing in detail position of relevant anatomy such as the left anterior descending coronary artery (LAD). This set of segmented structures, along with the corresponding cardiac CT image, define a reference set of cardiac data, to be used to match corresponding structures in our patient population. Plastimatch, a b‐spline, non‐rigid registration algorithm, was employed to determine the vector fields necessary to deform the reference CT image to match the designated patient volume. Using this vector field, the segmented model may then be deformed and registered to the new patient anatomy. Using the Fovia High Definition® Volume Rendering platform, the segmented volume may be analyzed volumetrically, visualizing anatomy, segmented volumes and delivered dose in near real‐time. Results: Retrospective application of this algorithm upon a selection of previously irradiated free breathing and breath hold left breast patients shows good agreement with expected geometry. However, alignment issues persist where significant thoracic and pulmonary motion obscures the registration image of the patients heart, resulting in poor registration of the treated anatomy with the cardiac template. Conclusion: Risk of late cardiopulmonary toxicity from left breast irradiation has motivated new treatment techniques to reduce cardiac exposure during radiotherapy. Implementation of a deformable segmentation model imaged volumetrically allows for careful analysis of delivered cardiac dose.

Image-driven population analysis through mixture modeling.

We present iCluster, a fast and efficient algorithm that clusters a set of images while co-registering them using a parameterized, nonlinear transformation model. The output of the algorithm is a small number of template images that represent different modes in a population. This is in contrast with traditional, hypothesis-driven computational anatomy approaches that assume a single template to construct an atlas. We derive the algorithm based on a generative model of an image population as a mixture of deformable template images. We validate and explore our method in four experiments. In the first experiment, we use synthetic data to explore the behavior of the algorithm and inform a design choice on parameter settings. In the second experiment, we demonstrate the utility of having multiple atlases for the application of localizing temporal lobe brain structures in a pool of subjects that contains healthy controls and schizophrenia patients. Next, we employ iCluster to partition a data set of 415 whole brain MR volumes of subjects aged 18 through 96 years into three anatomical subgroups. Our analysis suggests that these subgroups mainly correspond to age groups. The templates reveal significant structural differences across these age groups that confirm previous findings in aging research. In the final experiment, we run iCluster on a group of 15 patients with dementia and 15 age-matched healthy controls. The algorithm produces two modes, one of which contains dementia patients only. These results suggest that the algorithm can be used to discover subpopulations that correspond to interesting structural or functional "modes."

Figure-ground segmentation using factor graphs

Foreground–background segmentation has recently been applied [S.X. Yu, J. Shi, Object-specific figure-ground segregation, Computer Vision and Pattern Recognition (CVPR), 2003; S. Kumar, M. Hebert, Man-made structure detection in natural images using a causal multiscale random field, Computer Vision and Pattern Recognition (CVPR), 2003] to the detection and segmentation of specific objects or structures of interest from the background as an efficient alternative to techniques such as deformable templates [A.L. Yuille, Deformable templates for face recognition. Journal of Cognitive Neuroscience 3 (1) (1991)]. We introduce a graphical model (i.e. Markov random field)-based formulation of structure-specific figure-ground segmentation based on simple geometric features extracted from an image, such as local configurations of linear features, that are characteristic of the desired figure structure. Our formulation is novel in that it is based on factor graphs, which are graphical models that encode interactions among arbitrary numbers of random variables. The ability of factor graphs to express interactions higher than pairwise order (the highest order encountered in most graphical models used in computer vision) is useful for modeling a variety of pattern recognition problems. In particular, we show how this property makes factor graphs a natural framework for performing grouping and segmentation, and demonstrate that the factor graph framework emerges naturally from a simple maximum entropy model of figure-ground segmentation. We cast our approach in a learning framework, in which the contributions of multiple grouping cues are learned from training data, and apply our framework to the problem of finding printed text in natural scenes. Experimental results are described, including a performance analysis that demonstrates the feasibility of the approach.

An inverse association of cardiovascular risk and frontal lobe glucose metabolism

To investigate associations between vascular risk profile and cerebral glucose metabolism. Subjects ranged from normal to having dementia (age >55 years) and underwent neuropsychological testing, MRI, and FDG PET scanning (n = 58). The Framingham Cardiovascular Risk Profile (FCRP) and its individual components were used as covariates in regression analyses with each PET scan using SPM2. Analyses revealed broad areas of the frontal lobe in which higher FCRP was associated with lower normalized glucose metabolism including the superior medial frontal, superior frontal and superior orbital frontal cortex and the ventrolateral prefrontal cortex. Significant associations were predominately found in the left hemisphere. Independent component analyses revealed interesting regions but further confirm the relevance of the integrative measure of coronary risk. Although the mechanism of this association bears further investigation, this finding provides further evidence that vascular risk factors have malignant effects on the brain, particularly in the prefrontal cortex.

Construction of Bayesian deformable models via a stochastic approximation algorithm: A convergence study

The problem of the definition and estimation of generative models based on deformable templates from raw data is of particular importance for modeling non-aligned data affected by various types of geometric variability. This is especially true in shape modeling in the computer vision community or in probabilistic atlas building in computational anatomy. A first coherent statistical framework modeling geometric variability as hidden variables was described in Allassonnière, Amit and Trouvé [J. R. Stat. Soc. Ser. B Stat. Methodol. 69 (2007) 3–29]. The present paper gives a theoretical proof of convergence of effective stochastic approximation expectation strategies to estimate such models and shows the robustness of this approach against noise through numerical experiments in the context of handwritten digit modeling.

Foot Anatomical Landmarks Automatic Mark Method

In view of the disadvantage of existent landmark extraction methods, an efficient method is presented to automatically mark foot anatomical landmarks in human body. 3D data foot shape including anatomical landmarks is taken as deformable template. PCA and ICP methods are used to make target foot shape and template foot shape align and registrer in the least-squares sense. Then initial corresponding point sets between target foot shape and template foot shape are found. Sample point sets are obtained in the template foot shape by using method of combining uniform sampling and sampling based on distribution of curvature. Points in sample point sets corresponding to template foot shape are taken as position constrained condition. Discrete laplacian mesh deformation method is used to make template foot shape approach to target foot shape. Template foot shape is approached to target foot shape in error range via iteration. Therefore, the mapping relation between template foot shape points and the corresponding points on target foot shape is built. So the corresponding points of anatomical landmarks on template foot shape, on the target foot shape, can be treated as anatomical landmarks on target foot shape. The system can be applied to the healthy monitor of the flexible surface. The result of the experimentation indicates that foot anatomical landmarks mark method based on template has a high degree of automation and strong adaptability. It can be applied to difference analysis of 3D foot shape and shoe last individualized customization system based on foot shape.

Singular Perturbations For Heart Image Segmentation Tracking

In this note we give a result of convergence when time goes to infinity for a quasi static linear elastic model, the elastic tensor of which vanishes at infinity. This method is applied to segmentation of medical images, and improves the 'elastic deformable template' model introduced previously.

A Robust and Accurate Calibration Method by Computer Vision Techniques for Coordinate Transformation between Display Screens and Their Images

A robust and accurate calibration method for coordinate transformation between display screens and their images is proposed. The images of three specially designed calibration patterns, which are a white rectangle, a black rectangle, and a dot matrix, are captured and processed sequentially to assist achieving robust extraction of relevant fea- tures from the calibration pattern images, especially the border features in the image of the display screen. Image analysis techniques of deformable template matching, local thresholding, and inverse distance-weighted interpolation are integrated to perform fea- ture extraction and coordinate transformation to complete the calibration work. Good performance of the proposed method is demonstrated by comparing the experimental results of the method with those of a well-known calibration method.

IC-P2-114: Meta-analysis ROIs and minimal deformation templates improve PET-FDG as a candidate biomarker in Alzheimer's disease

Background: Research consistently finds hypometabolism of [18F]Fluorodeoxyglucose (FDG-PET) in patients diagnosed with Alzheimer’s Disease (AD). Identifying abnormalities in metabolism requires precise delineation of affected areas in order to maximize differences between groups, detect change over time and minimize variance. We generated meta-analysis ROIs (MetaROIs) from frequently cited coordinates in existing studies, and compared these regions with standard anatomical ROIs across three image preprocessing streams in order to identify potential bio-marker regions using data from the Alzheimer’s Disease Neuroimaging Initiative (AD 13), (Normal 12). Methods: Six terms relating to FDG-PET, AD, and Mild Cognitive Impairment (MCI) were used to search PubMed, resulting in 18 cross-sectional studies. Study coordinates were used to generate MetaROIs. A minimal deformation template (MDT) was generated for each group (AD, MCI, Normal), using high-resolution MR images, based on the method by Kochonuv et al. (2001). Preprocessing was done with three methods: (1) individual FDG-PET images were spatially normalized to the MNI FDG template, (2) individual MR images were spatially normalized to a group-specific template, then this transformation was applied to individual coregistered FDG-PET images, (3) same as (2) but additionally the FDG-PET image was masked with an individual gray-matter mask. Effect size and percent annual change delineated the most effective processing stream. Results: MetaROIs included frequently cited regions in bilateral posterior cingulate, bilateral angular gyri, and primarily left-sided frontal and temporal regions. MetaROIs exhibited larger mean group difference between AD and Normals compared to corresponding AAL MNI regions (effect size in posterior cingulate increased from 0.48 to 1.82). Normalization to MDT template further improved effect size by 7-15%. MetaROIs also increased twelve month percent change in AD patients (posterior cingulate increased from 8% to 11%). The same region in normals showed little change (MNI ROI 0.4% vs MetaROI 0.3%). Conclusions: The use of research-defined MetaROIs, along with a processing stream that reduced deformation and used individual gray matter masks, greatly improved the viability of FDG-PET as a candidate biomarker in AD.

Anatomical frame identification and reconstruction for repeatable lower limb joint kinematics estimates

The quantitative description of joint mechanics during movement requires the reconstruction of the position and orientation of selected anatomical axes with respect to a laboratory reference frame. These anatomical axes are identified through an ad hoc anatomical calibration procedure and their position and orientation are reconstructed relative to bone-embedded frames normally derived from photogrammetric marker positions and used to describe movement. The repeatability of anatomical calibration, both within and between subjects, is crucial for kinematic and kinetic end results. This paper illustrates an anatomical calibration approach, which does not require anatomical landmark manual palpation, described in the literature to be prone to great indeterminacy. This approach allows for the estimate of subject-specific bone morphology and automatic anatomical frame identification. The experimental procedure consists of digitization through photogrammetry of superficial points selected over the areas of the bone covered with a thin layer of soft tissue. Information concerning the location of internal anatomical landmarks, such as a joint center obtained using a functional approach, may also be added. The data thus acquired are matched with the digital model of a deformable template bone. Consequently, the repeatability of pelvis, knee and hip joint angles is determined. Five volunteers, each of whom performed five walking trials, and six operators, with no specific knowledge of anatomy, participated in the study. Descriptive statistics analysis was performed during upright posture, showing a limited dispersion of all angles (less than 3 deg) except for hip and knee internal–external rotation (6 deg and 9 deg, respectively). During level walking, the ratio of inter-operator and inter-trial error and an absolute subject-specific repeatability were assessed. For pelvic and hip angles, and knee flexion-extension the inter-operator error was equal to the inter-trial error—the absolute error ranging from 0.1 deg to 0.9 deg. Knee internal–external rotation and ab-adduction showed, on average, inter-operator errors, which were 8% and 28% greater than the relevant inter-trial errors, respectively. The absolute error was in the range 0.9–2.9 deg.

A multi-scale template method for shape detection with bio-medical applications

In this paper we present a novel methodology based on non-parametric deformable prototype templates for reconstructing the outline of a shape from a degraded image. Our method is versatile and fast and has the potential to provide an automatic procedure for classifying pathologies. We test our approach on synthetic and real data from a variety of medical and biological applications. In these studies it is important to reconstruct accurately the shape of the object under investigation from very noisy data. Here we assume that we have some prior knowledge about the object outline represented by a prototype shape. Our procedure deforms this shape by means of non-affine transformations and the contour is reconstructed by minimizing a newly developed objective function that depends on the transformation parameters. We introduce an iterative template deformation procedure in which the scale of the deformation decreases as the algorithm proceeds. We compare our results with those from a Gaussian Mixture Model segmentation and two state-of-the-art Level Set methods. This comparison shows that the proposed procedure performs consistently well on both real and simulated data. As a by-product we develop a new filter that recovers the connectivity of a shape.

Model-based analysis of flow-mediated dilation and intima-media thickness.

We present an end-to-end system for the automatic measurement of flow-mediated dilation (FMD) and intima-media thickness (IMT) for the assessment of the arterial function. The video sequences are acquired from a B-mode echographic scanner. A spline model (deformable template) is fitted to the data to detect the artery boundaries and track them all along the video sequence. The a priori knowledge about the image features and its content is exploited. Preprocessing is performed to improve both the visual quality of video frames for visual inspection and the performance of the segmentation algorithm without affecting the accuracy of the measurements. The system allows real-time processing as well as a high level of interactivity with the user. This is obtained by a graphical user interface (GUI) enabling the cardiologist to supervise the whole process and to eventually reset the contour extraction at any point in time. The system was validated and the accuracy, reproducibility, and repeatability of the measurements were assessed with extensive in vivo experiments. Jointly with the user friendliness, low cost, and robustness, this makes the system suitable for both research and daily clinical use.

Dynamic Shape and Appearance Modeling via Moving and Deforming Layers

We propose a model of the shape, motion and appearance of a scene, seen through a sequence of images, that captures occlusions, scene deformations, unconstrained viewpoint variations and changes in its radiance. This model is based on a collection of overlapping layers that can move and deform, each supporting an intensity function that can change over time. We discuss the generality and limitations of this model in relation to existing ones such as traditional optical flow or motion segmentation, layers, deformable templates and deformotion. We then illustrate how this model can be used for inference of shape, motion, deformation and appearance of the scene from a collection of images. The layering structure allows for automatic inpainting of partially occluded regions. We illustrate the model on synthetic and real sequences where existing schemes fail, and show how suitable choices of constants in the model yield existing schemes, from optical flow to motion segmentation and inpainting.