Catálogo de publicaciones - libros

This paper investigates and characterizes sources of variability in MEG signals in multi-site, multi-subject studies. Understanding these sources will help to develop efficient strategies for comparing and pooling data across repetitions of an experiment, across subjects, and across sites. In this work, we investigated somatosensory MEG data collected at three different sites and applied variance component analysis and nonparametric KL divergence analysis in order to characterize the sources of variability. Our analysis showed that inter-subject differences are the biggest factor in the signal variability. We demonstrated that the timing of the deflections is very consistent in the early somatosensory response, which justifies a direct comparison of deflection peak times acquired from different visits, subjects, and systems. Compared with deflection peak times, deflection magnitudes have larger variation across sites; modeling of this variability is necessary for data pooling.

A novel method is presented for creating a probability map from histologically defined cytoarchitectonic data, customised for the anatomy of individual fMRI volunteers. Postmortem structural and cytoarchitectonic information from a published dataset is combined with high resolution structural MR images using deformable registration of a region of interest. In this paper, we have targeted the three sub-areas of the primary auditory cortex (located on Heschl’s gyrus); however, the method could be applied to any other cytoarchitectonic region. The resulting probability maps show a significantly higher overlap than previously generated maps using the same cytoarchitectonic data, and more accurately span the macroanatomical structure of the auditory cortex. This improvement indicates a high potential for spatially accurate fMRI analysis, allowing more reliable correlation between anatomical structure and function. We validate the approach using fMRI data from nine individuals, taken from a published dataset. We compare activation for stimuli evoking a pitch percept to activation for acoustically matched noise, and demonstrate that the primary auditory cortex (Te1.0) and the lateral region Te1.2 are sensitive to pitch, whereas Te1.1 is not.

In this article, we develop a new method to segment Q-Ball imaging (QBI) data. We first estimate the orientation distribution function (ODF) using a fast and robust spherical harmonic (SH) method. Then, we use a region-based statistical surface evolution on this image of ODFs to efficiently find coherent white matter fiber bundles. We show that our method is appropriate to propagate through regions of fiber crossings and we show that our results outperform state-of-the-art diffusion tensor (DT) imaging segmentation methods, inherently limited by the DT model. Results obtained on synthetic data, on a biological phantom, on real datasets and on all 13 subjects of a public QBI database show that our method is reproducible, automatic and brings a strong added value to diffusion MRI segmentation.

Recently, concerns have been raised that the correspondences computed by volumetric registration within homogeneous structures are primarily driven by regularization priors that differ among algorithms. This paper explores the correspondence based on geometric models for one of those structures, midsagittal section of the corpus callosum (MSCC), and compared the result with registration paradigms. We use geometric model called continuous medial representation (cm-rep) to normalize anatomical structures on the basis of medial geometry, and use features derived from diffusion tensor tractography for validation. We show that shape-based normalization aligns subregions of the MSCC, defined by connectivity, more accurately than normalization based on volumetric registration. Furthermore, shape-based normalization helps increase the statistical power of group analysis in an experiment where features derived from diffusion tensor tractography are compared between two cohorts. These results suggest that cm-rep is an appropriate tool for normalizing the MSCC in white matter studies.

The main goal of this work was to assess the accuracy of several well-known methods which provide global (BSI and SIENA) or local (Jacobian integration) estimates of longitudinal atrophy in brain structures using Magnetic Resonance images. For that purpose, we have generated realistic simulated images which mimic the patterns of change obtained from a cohort of 19 real controls and 27 probable Alzheimer’s disease patients. SIENA and BSI results correlate very well with gold standard data (BSI mean absolute error < 0.29%; SIENA < 0.44%). Jacobian integration was guided by both fluid and FFD-based registration techniques and resulting deformation fields and associated Jacobians were compared, region by region, with gold standard ones. The FFD registration technique provided more satisfactory results than the fluid one. Mean absolute error differences between volume changes given by the FFD-based technique and the gold standard were: sulcal CSF < 2.49%; lateral ventricles < 2.25%; brain < 0.36%; hippocampi < 1.42%.

An important issue in electroencephalographiy (EEG) experiments is to measure accurately the three dimensional (3D) positions of the electrodes. We propose a system where these positions are automatically estimated from several images using computer vision techniques. Yet, only a set of undifferentiated points are recovered this way and remains the problem of labeling them, i.e. of finding which electrode corresponds to each point. This paper proposes a fast and robust solution to this latter problem based on combinatorial optimization. We design a specific energy that we minimize with a modified version of the Loopy Belief Propagation algorithm. Experiments on real data show that, with our method, a manual labeling of two or three electrodes only is sufficient to get the complete labeling of a 64 electrodes cap in less than 10 seconds.

Many hearing aid users experience physical discomfort when wearing their device. The main contributor to this problem is believed to be deformation of the ear and ear canal caused by movement of the mandible. Physical discomfort results from added pressure on soft tissue areas in the ear. Identifying features that can predict potential deformation is therefore important for identifying problematic cases in advance. A study on the physical deformation of the human ear and canal due to movement of the mandible is presented. The study is based on laser scannings of 30 pairs of ear impressions from 9 female and 21 male subjects. Two impressions have been taken from each subject, one with open mouth, and one with the mouth closed. All impressions are registered using non-rigid surface registration and a shape model is built. From each pair of impressions a deformation field is generated and propagated to the shape model, enabling the building of a deformation model in the reference frame of the shape model. A relationship between the two models is established, showing that the shape variation can explain approximately 50% of the variation in the deformation model. An hypothesis test for significance of the deformations for each deformation field reveals that all subjects have significant deformation at Tragus and in the canal. Furthermore, a relation between the magnitude of the deformation and the gender of the subject is demonstrated. The results are successfully validated by comparing the outcome to the anatomy by using a single set of high resolution histological sectionings of the region of interest.

This paper proposes a novel approach that achieves shape registration by optimizing shape representation and transformation simultaneously, which are modeled by a constrained (GMM) and a regularized respectively. The problem is formulated within a Bayesian framework and solved by an expectation-maximum (EM) algorithm. Compared with the popular methods based on , its advantages include: (1) It can naturally deal with shapes of complex topologies and 3D dimension; (2) It is more robust against data noise; (3) The registration performance is better in terms of the generalization error of the resultant statistical shape model. These are demonstrated on both synthetic and biomedical shapes.

Volumetric layers are often encountered in medical images. Unlike solid structures, volumetric layers are characterized by double and nested bounding surfaces. It is expected that better statistical models can be built by utilizing the surface coupleness rather than simply applying the landmarking method on each of them separately. We propose an approach to optimizing the landmark correspondence on the coupled surfaces by minimizing the description length that incorporates local thickness gradient. The evaluations are performed on a set of 2-D synthetic close coupled contours and a set of real-world open surfaces, the skull vaults. Compared with performing landmarking separately on the coupled surfaces, the proposed method constructs models that have better generalization ability and specificity.

Tensor-based morphometry (TBM) studies anatomical differences between brain images statistically, to identify regions that differ between groups, over time, or correlate with cognitive or clinical measures. Using a nonlinear registration algorithm, all images are mapped to a common space, and statistics are most commonly performed on the Jacobian determinant (local expansion factor) of the deformation fields. In [14], it was shown that the detection sensitivity of the standard TBM approach could be increased by using the full deformation tensors in a multivariate statistical analysis. Here we set out to improve the common space itself, by choosing the shape that minimizes a natural metric on the deformation tensors from that space to the population of control subjects. This method avoids statistical bias and should ease nonlinear registration of new subjects data to a template that is ’closest’ to all subjects’ anatomies. As deformation tensors are symmetric positive-definite matrices and do not form a vector space, all computations are performed in the log-Euclidean framework [1]. The control brain that is already the closest to ’average’ is found. A gradient descent algorithm is then used to perform the minimization that iteratively deforms this template and obtains the mean shape.

We apply our method to map the profile of anatomical differences in a dataset of 26 HIV/AIDS patients and 14 controls, via a log-Euclidean Hotelling’s test on the deformation tensors. These results are compared to the ones found using the ’best’ control, . Statistics on both shapes are evaluated using cumulative distribution functions of the -values in maps of inter-group differences.