ArticlePDF Available

Emergence of Groupwise Registration in MR Brain Study

Authors:
  • United Imaging Intelligence

Abstract and Figures

1. Background Modern medical imaging technologies such as Magnetic Resonance Imaging (MRI) [1] offer a safe and non-invasive means of performing clinical diagnosis and research, involving human brain development, aging, and disease-induced anomalies. For reliable estimation of disease related micro-structural difference, accurate deformable image registration plays a key fundamental role in dealing with confounding intra-subject variability in longitudinal studies and inter-subject variability in cross-sectional studies. Pairwise Registration: A plethora of pairwise deformable registration algorithms have flourished in the past two decades, comprehensive surveys of which can be found in [2-6]. In general, the goal of image registration is to estimate the deformation field for warping the subject image (or moving image) to the template image (or fixed image) by maximizing a certain similarity measure between the warped subject and the template. Upon successful registration, intra-or inter-subject differences are minimized, while at the same time disease-related changes and morphological variations are preserved. The majority of image registration algorithms fall into three categories: landmark-based [7-9], intensity-based [10-15], and feature-based [16-19]. Landmark-based algorithms take advantage of anatomical prior knowledge and are thus computationally fast, since only a few landmarks out of all voxels in an image volume need to be matched. It is however a challenging task even for the trained experts to accurately place a sufficient number of anatomical landmarks for achieving accurate registration. Moreover, inter-rater landmark-placement variability could be large, thus seriously undermining the performance of landmark-based registration algorithms. Intensity-based algorithms * Guorong Wu, Hongjun Jia, and Qian Wang contributed equally to this book chapter.
Content may be subject to copyright.
A preview of the PDF is not available
... In fact, performances of image registration can be greatly improved if information from other images in the population is well incorporated. For example, recent studies show very promising alignment of images in the groupwise manner [2]. Although it is difficult to directly warp a significantly different subject to the template in pairwise registration, the problem could become much easier when using other intermediate images as bridges [3][4][5]. ...
... Specifically, we convert (1) to an incremental refinement model by letting that reflects the tentatively deformed location of the template voxel x at time τ l−1 . Then, the prediction at time τ l , later than τ l−1 , is updated by (2) This incremental refinement is also illustrated in Fig. 1(c), where blue and red dashed arrows denote transformations at τ l−1 and τ l , respectively. ...
... only if restricted within a limited range such that is small. The latent intimacy between and y encourages us to select the training image with the field highly resembling according to (2). The importance in selecting can also be observed in Fig. 1(c). ...
Conference Paper
We propose a new approach to register the subject image with the template by leveraging a set of training images that are pre-aligned to the template. We argue that, if voxels in the subject and the training images share similar local appearances and transformations, they may have common correspondence in the template. In this way, we learn the sparse representation of certain subject voxel to reveal several similar candidate voxels in the training images. Each selected training candidate can bridge the correspondence from the subject voxel to the template space, thus predicting the transformation associated with the subject voxel at the confidence level that relates to the learned sparse coefficient. Following this strategy, we first predict transformations at selected key points, and retain multiple predictions on each key point (instead of allowing a single correspondence only). Then, by utilizing all key points and their predictions with varying confidences, we adaptively reconstruct the dense transformation field that warps the subject to the template. For robustness and computation speed, we embed the prediction-reconstruction protocol above into a multi-resolution hierarchy. In the final, we efficiently refine our estimated transformation field via existing registration method. We apply our method to registering brain MR images, and conclude that the proposed method is competent to improve registration performances in terms of time cost as well as accuracy.
... Iterating over this process and taking the average model of the previous iteration as the new reference produces an unbiased atlas. • Template-free approaches [33] on the other hand do not rely on an initial reference image thus avoiding the introduction of the bias in the first place. They can be subdivided into two categories: ...
Article
Online atlasing, i.e., incrementing an atlas with new images as they are acquired, is key when performing studies on very large, or still being gathered, databases. Regular approaches to atlasing however do not focus on this aspect and impose a complete reconstruction of the atlas when adding images. We propose instead a diffeomorphic online atlasing method that allows gradual updates to an atlas. In this iterative centroid approach, we integrate new subjects in the atlas in an iterative manner, gradually moving the centroid of the images towards its final position. This leads to a computationally cheap approach since it only necessitates one additional registration per new subject added. We validate our approach on several experiments with three main goals: 1- to evaluate atlas image quality of the obtained atlases with sharpness and overlap measures, 2- to assess the deviation in terms of transformations with respect to a conventional atlasing method and 3- to compare its computational time with regular approaches of the literature. We demonstrate that the transformations divergence with respect to a state-of-the-art atlas construction method is small and reaches a plateau, that the two construction methods have the same ability to map subject homologous regions onto a common space and produce images of equivalent quality. The computational time of our approach is also drastically reduced for regular updates. Finally, we also present a direct extension of our method to update spatio-temporal atlases, especially useful for developmental studies.
... Step 3: ihMT template and atlas construction. ihMT images of all control subjects and patients were realigned into a new common reference space by a 2D nonlinear registration using a symmetric group-wise normalization procedure 19,20 and further averaged to create a specific ihMT average template. Note that combining images of both patients and controls allowed optimization of the group-wise registration procedure by minimization of image deformations from all individual subjects, especially the patients, to the common average template. ...
Article
Background and purpose: Inhomogeneous magnetization transfer is a new endogenous MR imaging contrast mechanism that has demonstrated high specificity for myelin. Here, we tested the hypothesis that inhomogeneous magnetization transfer is sensitive to pathology in a population of patients with relapsing-remitting MS in a way that both differs from and complements conventional magnetization transfer. Materials and methods: Twenty-five patients with relapsing-remitting MS and 20 healthy volunteers were enrolled in a prospective MR imaging research study, whose protocol included anatomic imaging, standard magnetization transfer, and inhomogeneous magnetization transfer imaging. Magnetization transfer and inhomogeneous magnetization transfer ratios measured in normal-appearing brain tissue and in MS lesions of patients were compared with values measured in control subjects. The potential association of inhomogeneous magnetization transfer ratio variations with the clinical scores (Expanded Disability Status Scale) of patients was further evaluated. Results: The magnetization transfer ratio and inhomogeneous magnetization transfer ratio measured in the thalami and frontal, occipital, and temporal WM of patients with MS were lower compared with those of controls (P< .05). The mean inhomogeneous magnetization transfer ratio measured in lesions was lower than that in normal-appearing WM (P< .05). Significant (P< .05) negative correlations were found between the clinical scores and inhomogeneous magnetization transfer ratio measured in normal-appearing WM structures. Weaker nonsignificant correlation trends were found for the magnetization transfer ratio. Conclusions: The sensitivity of the inhomogeneous magnetization transfer technique for MS was highlighted by the reduction in the inhomogeneous magnetization transfer ratio in MS lesions and in normal-appearing WM of patients compared with controls. Stronger correlations with the Expanded Disability Status Scale score were obtained with the inhomogeneous magnetization transfer ratio compared with the standard magnetization transfer ratio, which may be explained by the higher specificity of inhomogeneous magnetization transfer for myelin.
Conference Paper
Understanding brain development involves studying the relationship between age as one of the explanatory variables and explained variables, observations of this organ, which can take many forms. Magnetic Resonance Imaging (MRI) gives the opportunity to extract such observations in a non-invasive and non-irradiating way. This powerful technique allows notably to gain insights about the functional activity of the brain or its internal diffusivity characteristics. Yet, it is rather on the purely morphological aspects that this thesis is focused on. The approach followed the study of the brain as a mathematical object, thus enabling the analysis of its shape and growth by the means of the geometric transformations connecting those objects. In the finding of those transformations, across structures of topological interest, lies the concept of registration. This opens the door to the statistical analysis of shapes and the creation of average anatomical models called atlases.
Thesis
Understanding brain development involves studying the relationship between age as one of the explanatory variables and explained variables, observations of this organ, which can take many forms. Magnetic Resonance Imaging (MRI) gives the opportunity to extract such observations in a non-invasive and non-irradiating way. This powerful technique allows notably to gain insights about the functional activity of the brain or its internal diffusivity characteristics. Yet, it is rather on the purely morphological aspects that this thesis is focused on. The approach followed the study of the brain as a mathematical object, thus enabling the analysis of its shape and growth by the means of the geometric transformations connecting those objects. In the finding of those transformations, across structures of topological interest, lies the concept of registration. This opens the door to the statistical analysis of shapes and the creation of average anatomical models called atlases.
Article
Full-text available
Listening effort may be reduced when hearing aids improve access to the acoustic signal. However, this possibility is difficult to evaluate because many neuroimaging methods used to measure listening effort are incompatible with hearing aid use. Functional near-infrared spectroscopy (fNIRS), which can be used to measure the concentration of oxygen in the prefrontal cortex (PFC), appears to be well-suited to this application. The first aim of this study was to establish whether fNIRS could measure cognitive effort during listening in older adults who use hearing aids. The second aim was to use fNIRS to determine if listening effort, a form of cognitive effort, differed depending on whether or not hearing aids were used when listening to sound presented at 35dB SL (flat gain). Sixteen older adults who were experienced hearing aid users completed an auditory n-back task and a visual n-back task; both tasks were completed with and without hearing aids. We found that PFC oxygenation increased with n-back working memory demand in both modalities, supporting the use of fNIRS to measure cognitive effort during listening in this population. PFC oxygenation was weakly and nonsignificantly correlated with self-reported listening effort and reaction time, respectively, suggesting that PFC oxygenation assesses a dimension of listening effort that differs from these other measures. Furthermore, the extent to which hearing aids reduced PFC oxygenation in the left lateral PFC was positively correlated with age and pure-tone average thresholds. The implications of these findings as well as future directions are discussed.
Article
Full-text available
The distance transform (DT) is a general operator forming the basis of many methods in computer vision and geometry, with great potential for practical applications. However, all the optimal algorithms for the computation of the exact Euclidean DT (EDT) were proposed only since the 1990s. In this work, state-of-the-art sequential 2D EDT algorithms are reviewed and compared, in an effort to reach more solid conclusions regarding their differences in speed and their exactness. Six of the best algorithms were fully implemented and compared in practice.
Article
A novel approach to correcting for intensity nonuniformity in magnetic resonance (MR) data is described that achieves high performance without requiring a model of the tissue classes present. The method has the advantage that it can be applied at an early stage in an automated data analysis, before a tissue model is available. Described as nonparametric nonuniform intensity normalization (N3), the method is independent of pulse sequence and insensitive to pathological data that might otherwise violate model assumptions. To eliminate the dependence of the field estimate on anatomy, an iterative approach is employed to estimate both the multiplicative bias field and the distribution of the true tissue intensities. The performance of this method is evaluated using both real and simulated MR data.
Article
In this paper, we present the concept of diffusing models to perform image-to-image matching. Having two images to match, the main idea is to consider the objects boundaries in one image as semi-permeable membranes and to let the other image, considered as a deformable grid model, diffuse through these interfaces, by the action of effectors situated within the membranes. We illustrate this concept by an analogy with Maxwell's demons. We show that this concept relates to more traditional ones, based on attraction, with an intermediate step being optical flow techniques. We use the concept of diffusing models to derive three different non-rigid matching algorithms, one using all the intensity levels in the static image, one using only contour points, and a last one operating on already segmented images. Finally, we present results with synthesized deformations and real medical images, with applications to heart motion tracking and three-dimensional inter-patients matching.
Article
We address the issue of low-level segmentation for real-valued images. The proposed approach relies on the formulation of the problem in terms of an energy partition of the image domain. In this framework, an energy is defined by measuring a pseudo-metric distance to a source point. Thus, the choice of an energy and a set of sources determines a tessellation of the domain. Each energy acts on the image at a different level of analysis; through the study of two types of energies, two stages of the segmentation process are addressed. The first energy considered, the path variation, belongs to the class of energies determined by minimal paths. Its application as a pre-segmentation method is proposed. In the second part, where the energy is induced by a ultrametric, the construction of hierarchical representations of the image is discussed.
Conference Paper
Groupwise registration has been recently introduced for simultaneous registration of a group of images with the goal of constructing an unbiased atlas. To this end, direct application of information-theoretic entropy measures on image intensity has achieved various successes. However, simplistic voxelwise utilization of image intensity often neglects important contextual information, which can be provided by more comprehensive geometric and statistical features. In this paper, we employ attribute vectors, instead of image intensities, to guide groupwise registration. In particular, for each voxel, the attribute vector is computed from its multiple-scale neighborhoods to capture geometric information at different scales. Moreover, the probability density function (PDF) of each attribute in the vector is then estimated from the local neighborhood, providing a statistical summary of the underlying anatomical structure. For the purpose of registration, Jensen-Shannon (JS) divergence is used to measure the PDF dissimilarity of each attribute at corresponding locations of different individual images. By minimizing the overall JS divergence in the whole image space and estimating the deformation field of each image simultaneously, we can eventually register all images and build an unbiased atlas. Experimental results indicate that our method yields better registration quality, compared with a popular groupwise registration method.
Article
The relatively good transparency of biological materials in the near infrared region of the spectrum permits sufficient photon transmission through organs in situ for the monitoring of cellular events. Observations by infrared transillumination in the exposed heart and in the brain in cephalo without surgical intervention show that oxygen sufficiency for cytochrome a,a3, function, changes in tissue blood volume, and the average hemoglobin-oxyhemoglobin equilibrium can be recorded effectively and in continuous fashion for research and clinical purposes. The copper atom associated with heme a3 did not respond to anoxia and may be reduced under normoxic conditions, whereas the heme-a copper was at least partially reducible.
Book
The theory of multidimensional scaling arose and grew within the field of the behavioral sciences and now covers several statistical techniques that are widely used in many disciplines. Intended for readers of varying backgrounds, this book comprehensively covers the area while serving as an introduction to the mathematical ideas behind the various techniques of multidimensional scaling.
Article
This work addresses the problem of approximating a manifold by a simplicial mesh, and the related problem of building triangulations for the purpose of piecewise-linear approximation of functions. It has long been understood that the vertices of such meshes or triangulations should be "well-distributed," or satisfy certain "sampling conditions." This work clarifies and extends some algorithms for finding such well-distributed vertices, by showing that they can be regarded as finding ε-nets or Delone sets in appropriate metric spaces. In some cases where such Delone properties were already understood, such as for meshes to approximate smooth manifolds that bound convex bodies, the upper and lower bound results are extended to more general manifolds; in particular, under some general conditions, the minimum Hausdorff distance for a mesh with n simplices to a d-manifold M is Θ((∫M√|κ(x)|/n)2/d) as n ⋺ ∞, where κ(x) is the Gaussian curvature at point x ∈ M. We also relate these constructions to Dudley's approximation scheme for convex bodies, which can be interpreted as involving an ε-net in a metric space whose distance function depends on surface normals.