Showing papers by "Bertrand Thirion published in 2014"
TL;DR: It is illustrated how scikit-learn, a Python machine learning library, can be used to perform some key analysis steps and its application to neuroimaging data provides a versatile tool to study the brain.
Abstract: Statistical machine learning methods are increasingly used for neuroimaging data analysis. Their main virtue is their ability to model high-dimensional datasets, e.g. multivariate analysis of activation images or resting-state time series. Supervised learning is typically used in decoding or encoding settings to relate brain images to behavioral or clinical observations, while unsupervised learning can uncover hidden structures in sets of images (e.g. resting state functional MRI) or find sub-populations in large cohorts. By considering different functional neuroimaging applications, we illustrate how scikit-learn, a Python machine learning library, can be used to perform some key analysis steps. Scikit-learn contains a very large set of statistical learning algorithms, both supervised and unsupervised, and its application to neuroimaging data provides a versatile tool to study the brain.
1,418 citations
TL;DR: It is shown that in general Ward’s clustering performs better than alternative methods with regard to reproducibility and accuracy and that the two criteria diverge regarding the preferred models (reproducibility leading to more conservative solutions), thus deferring the practical decision to a higher level alternative, namely the choice of a trade-off between accuracy and stability.
Abstract: Analysis and interpretation of neuroimaging data often require one to divide the brain into a number of regions, or parcels, with homogeneous characteristics, be these regions defined in the brain volume or on on the cortical surface. While predefined brain atlases do not adapt to the signal in the individual subjects images, parcellation approaches use brain activity (e.g. found in some functional contrasts of interest) and clustering techniques to define regions with some degree of signal homogeneity. In this work, we address the question of which clustering technique is appropriate and how to optimize the corresponding model. We use two principled criteria: goodness of fit (accuracy), and reproducibility of the parcellation across bootstrap samples. We study these criteria on both simulated and two task-based functional Magnetic Resonance Imaging datasets for the Ward, spectral and K-means clustering algorithms. We show that in general Ward’s clustering performs better than alternative methods with regards to reproducibility and accuracy and that the two criteria diverge regarding the preferred models (reproducibility leading to more conservative solutions), thus deferring the practical decision to a higher level alternative, namely the choice of a trade-off between accuracy and stability.
288 citations
Posted Content•
TL;DR: Scikit-learn as mentioned in this paper contains a very large set of statistical learning algorithms, both supervised and unsupervised, and its application to neuroimaging data provides a versatile tool to study the brain.
Abstract: Statistical machine learning methods are increasingly used for neuroimaging data analysis. Their main virtue is their ability to model high-dimensional datasets, e.g. multivariate analysis of activation images or resting-state time series. Supervised learning is typically used in decoding or encoding settings to relate brain images to behavioral or clinical observations, while unsupervised learning can uncover hidden structures in sets of images (e.g. resting state functional MRI) or find sub-populations in large cohorts. By considering different functional neuroimaging applications, we illustrate how scikit-learn, a Python machine learning library, can be used to perform some key analysis steps. Scikit-learn contains a very large set of statistical learning algorithms, both supervised and unsupervised, and its application to neuroimaging data provides a versatile tool to study the brain.
173 citations
TL;DR: P predictive models have been used on neuroimaging data to ask new questions, i.e., to uncover new aspects of cognitive organization, and a statistical learning perspective on these progresses and on the remaining gaping holes is given.
Abstract: Functional brain images are rich and noisy data that can capture indirect signatures of neural activity underlying cognition in a given experimental setting. Can data mining leverage them to build models of cognition? Only if it is applied to well-posed questions, crafted to reveal cognitive mechanisms. Here we review how predictive models have been used on neuroimaging data to ask new questions, i.e., to uncover new aspects of cognitive organization. We also give a statistical learning perspective on these progresses and on the remaining gaping holes.
108 citations
University of Düsseldorf1, Allen Institute for Brain Science2, Washington University in St. Louis3, University of Southern California4, University of Minnesota5, University of California, Berkeley6, Maastricht University7, University of Oslo8, French Institute of Health and Medical Research9, Cuban Neuroscience Center10, French Institute for Research in Computer Science and Automation11, RWTH Aachen University12, University of Cologne13, Katholieke Universiteit Leuven14, Montreal Neurological Institute and Hospital15
TL;DR: To arrive at fully interoperable atlases of the human brain will still require much work at the frontiers of data acquisition, analysis, and representation, among them the latter may provide the most challenging task.
82 citations
TL;DR: In this paper, the authors address the question of which clustering technique is appropriate and how to optimize the corresponding model, and use two principled criteria: goodness of fit (accuracy), and reproducibility of the parcellation across bootstrap samples.
Abstract: Analysis and interpretation of neuroimaging data often require one to divide the brain into a number of regions, or parcels, with homogeneous characteristics, be these regions defined in the brain volume or on on the cortical surface. While predefined brain atlases do not adapt to the signal in the individual subjects images, parcellation approaches use brain activity (e.g. found in some functional contrasts of interest) and clustering techniques to define regions with some degree of signal homogeneity. In this work, we address the question of which clustering technique is appropriate and how to optimize the corresponding model. We use two principled criteria: goodness of fit (accuracy), and reproducibility of the parcellation across bootstrap samples. We study these criteria on both simulated and two task-based functional Magnetic Resonance Imaging datasets for the Ward, spectral and K-means clustering algorithms. We show that in general Ward’s clustering performs better than alternative methods with regards to reproducibility and accuracy and that the two criteria diverge regarding the preferred models (reproducibility leading to more conservative solutions), thus deferring the practical decision to a higher level alternative, namely the choice of a trade-off between accuracy and stability.
64 citations
01 Jan 2014
TL;DR: This work develops a method for the joint estimation of activation and HRF by means of a rank constraint, forcing the estimated HRF to be equal across events or experimental conditions, yet permitting it to differ across voxels.
Abstract: Despite the common usage of a canonical, data-independent, hemodynamic response function (HRF), it is known that the shape of the HRF varies across brain regions and subjects. This suggests that a data-driven estimation of this function could lead to more statistical power when modeling BOLD fMRI data. However, unconstrained estimation of the HRF can yield highly unstable results when the number of free parameters is large. We develop a method for the joint estimation of activation and HRF by means of a rank constraint, forcing the estimated HRF to be equal across events or experimental conditions, yet permitting it to differ across voxels. Model estimation leads to an optimization problem that we propose to solve with an efficient quasi-Newton method, exploiting fast gradient computations. This model, called GLM with Rank-1 constraint (R1-GLM), can be extended to the setting of GLM with separate designs which has been shown to improve decoding accuracy in brain activity decoding experiments. We compare 10 different HRF modeling methods in terms of encoding and decoding scores on two different datasets. Our results show that the R1-GLM model outperforms competing methods in both encoding and decoding settings, positioning it as an attractive method both from the points of view of accuracy and computational efficiency.
55 citations
04 Jun 2014
TL;DR: A wide variety of solvers are explored and their convergence properties on fMRI data are exhibited and a variant of smooth solvers is introduced and it is shown that it is a promising approach in these settings.
Abstract: Learning predictive models from brain imaging data, as in decoding cognitive states from fMRI (functional Magnetic Resonance Imaging), is typically an ill-posed problem as it entails estimating many more parameters than available sample points. This estimation problem thus requires regularization. Total variation regularization, combined with sparse models, has been shown to yield good predictive performance, as well as stable and interpretable maps. However, the corresponding optimization problem is very challenging: it is non-smooth, non-separable and heavily ill-conditioned. For the penalty to fully exercise its structuring effect on the maps, this optimization problem must be solved to a good tolerance resulting in a computational challenge. Here we explore a wide variety of solvers and exhibit their convergence properties on fMRI data. We introduce a variant of smooth solvers and show that it is a promising approach in these settings. Our findings show that care must be taken in solving TV-l1 estimation in brain imaging and highlight the successful strategies.
47 citations
14 Sep 2014
TL;DR: It is shown on real data that this approach provides significantly higher classification accuracy than directly using Pearson's correlation, and a non-parametric scheme for identifying significantly discriminative connections from classifier weights is proposed.
Abstract: We present a Riemannian approach for classifying fMRI connectivity patterns before and after intervention in longitudinal studies. A fundamental difficulty with using connectivity as features is that covariance matrices live on the positive semi-definite cone, which renders their elements inter-related. The implicit independent feature assumption in most classifier learning algorithms is thus violated. In this paper, we propose a matrix whitening transport for projecting the covariance estimates onto a common tangent space to reduce the statistical dependencies between their elements. We show on real data that our approach provides significantly higher classification accuracy than directly using Pearson’s correlation. We further propose a non-parametric scheme for identifying significantly discriminative connections from classifier weights. Using this scheme, a number of neuroanatomically meaningful connections are found, whereas no significant connections are detected with pure permutation testing.
42 citations
TL;DR: A novel classification framework based on group-invariant graphical representations is introduced, allowing to overcome the inter-subject variability present in functional magnetic resonance imaging (fMRI) data and to perform multivariate pattern analysis across subjects.
Abstract: In brain imaging, solving learning problems in multi-subjects settings is difficult because of the differences that exist across individuals. Here we introduce a novel classification framework based on group-invariant graphical representations, allowing to overcome the inter-subject variability present in functional magnetic resonance imaging (fMRI) data and to perform multivariate pattern analysis across subjects. Our contribution is twofold: first, we propose an unsupervised representation learning scheme that encodes all relevant characteristics of distributed fMRI patterns into attributed graphs; second, we introduce a custom-designed graph kernel that exploits all these characteristics and makes it possible to perform supervised learning (here, classification) directly in graph space. The well-foundedness of our technique and the robustness of the performance to the parameter setting are demonstrated through inter-subject classification experiments conducted on both artificial data and a real fMRI experiment aimed at characterizing local cortical representations. Our results show that our framework produces accurate inter-subject predictions and that it outperforms a wide range of state-of-the-art vector- and parcel-based classification methods. Moreover, the genericity of our method makes it is easily adaptable to a wide range of potential applications. The dataset used in this study and an implementation of our framework are available at http://dx.doi.org/10.6084/m9.figshare.1086317.
39 citations
French Institute for Research in Computer Science and Automation1, IBM2, Heidelberg University3, King's College London4, Trinity College, Dublin5, Université de Montréal6, Charité7, University of Vermont8, French Institute of Health and Medical Research9, Paris Descartes University10, University of Toronto11, University of Nottingham12, Montreal Neurological Institute and Hospital13, Dresden University of Technology14, University of California, Berkeley15
TL;DR: A new approach is introduced to overcome the limitations of standard methods, in which active voxels are detected according to a consensus on several random parcellations of the brain images, while a permutation test controls the false positive risk.
TL;DR: A scalable analysis tool that can deal with non-parametric statistics on high-dimensional data by combining a MapReduce framework (TomusBLOB) with machine learning algorithms (Scikit-learn library) and demonstrating the scalability and the reliability of the framework in the cloud with a 2 weeks deployment on hundreds of virtual machines.
Abstract: Brain imaging is a natural intermediate phenotype to understand the link between genetic information and behavior or brain pathologies risk factors. Massive efforts have been made in the last few years to acquire high-dimensional neuroimaging and genetic data on large cohorts of subjects. The statistical analysis of such data is carried out with increasingly sophisticated techniques and represents a great computational challenge. Fortunately, increasing computational power in distributed architectures can be harnessed, if new neuroinformatics infrastructures are designed and training to use these new tools is provided. Combining a MapReduce framework (TomusBLOB) with machine learning algorithms (Scikit-learn library), we design a scalable analysis tool that can deal with non-parametric statistics on high-dimensional data. End-users describe the statistical procedure to perform and can then test the model on their own computers before running the very same code in the cloud at a larger scale. We illustrate the potential of our approach on real data with an experiment showing how the functional signal in subcortical brain regions can be significantly fit with genome-wide genotypes. This experiment demonstrates the scalability and the reliability of our framework in the cloud with a two weeks deployment on hundreds of virtual machines.
14 Sep 2014
TL;DR: In this paper, the authors present post-processing techniques that automatically sparsify brain maps and separate regions properly using geometric operations, and compare these techniques according to faithfulness to data and stability metrics.
Abstract: Functional Magnetic Resonance Images acquired during resting-state provide information about the functional organization of the brain through measuring correlations between brain areas. Independent components analysis is the reference approach to estimate spatial components from weakly structured data such as brain signal time courses; each of these components may be referred to as a brain network and the whole set of components can be conceptualized as a brain functional atlas. Recently, new methods using a sparsity prior have emerged to deal with low signal-to-noise ratio data. However, even when using sophisticated priors, the results may not be very sparse and most often do not separate the spatial components into brain regions. This work presents post-processing techniques that automatically sparsify brain maps and separate regions properly using geometric operations, and compares these techniques according to faithfulness to data and stability metrics. In particular, among threshold-based approaches, hysteresis thresholding and random walker segmentation, the latter improves significantly the stability of both dense and sparse models.
TL;DR: This paper provides a detailed analysis of a new generative statistical model based on dense deformable templates that represents several tissue types observed in medical images and uses a stochastic algorithm for the estimation of the probabilistic atlas given a dataset.
Abstract: Computerized anatomical atlases play an important role in medical image analysis. While an atlas usually refers to a standard or mean image also called template, which presumably represents well a given population, it is not enough to characterize the observed population in detail. A template image should be learned jointly with the geometric variability of the shapes represented in the observations. These two quantities will in the sequel form the atlas of the corresponding population. The geometric variability is modeled as deformations of the template image so that it fits the observations. In this paper, we provide a detailed analysis of a new generative statistical model based on dense deformable templates that represents several tissue types observed in medical images. Our atlas contains both an estimation of probability maps of each tissue (called class) and the deformation metric. We use a stochastic algorithm for the estimation of the probabilistic atlas given a dataset. This atlas is then used for atlas-based segmentation method to segment the new images. Experiments are shown on brain T1 MRI datasets.
14 Sep 2014
TL;DR: In this article, the authors estimate the amount of variance that is fit by a random effects subspace learned on other images, and show that a principal component regression estimator outperforms other regression models and that it fits a significant proportion (10% to 25%) of the between-subject variability.
Abstract: Inter-subject variability is a major hurdle for neuroimaging group-level inference, as it creates complex image patterns that are not captured by standard analysis models and jeopardizes the sensitivity of statistical procedures. A solution to this problem is to model random subjects effects by using the redundant information conveyed by multiple imaging contrasts. In this paper, we introduce a novel analysis framework, where we estimate the amount of variance that is fit by a random effects subspace learned on other images; we show that a principal component regression estimator outperforms other regression models and that it fits a significant proportion (10% to 25%) of the between-subject variability. This proves for the first time that the accumulation of contrasts in each individual can provide the basis for more sensitive neuroimaging group analyzes.
Posted Content•
TL;DR: This work presents post-processing techniques that automatically sparsify brain maps and separate regions properly using geometric operations, and compares these techniques according to faithfulness to data and stability metrics.
Abstract: Functional Magnetic Resonance Images acquired during resting-state provide information about the functional organization of the brain through measuring correlations between brain areas. Independent components analysis is the reference approach to estimate spatial components from weakly structured data such as brain signal time courses; each of these components may be referred to as a brain network and the whole set of components can be conceptualized as a brain functional atlas. Recently, new methods using a sparsity prior have emerged to deal with low signal-to-noise ratio data. However, even when using sophisticated priors, the results may not be very sparse and most often do not separate the spatial components into brain regions. This work presents post-processing techniques that automatically sparsify brain maps and separate regions properly using geometric operations, and compares these techniques according to faithfulness to data and stability metrics. In particular, among threshold-based approaches, hysteresis thresholding and random walker segmentation, the latter improves significantly the stability of both dense and sparse models.
14 Sep 2014
TL;DR: This work proposes a novel learning scheme for functional connectivity based on sparse Gaussian graphical models that aims at minimizing the bias induced by the regularization used in the estimation, by carefully separating the estimation of the model support from the coefficients.
Abstract: The estimation of functional connectivity structure from functional neuroimaging data is an important step toward understanding the mechanisms of various brain diseases and building relevant biomarkers. Yet, such inferences have to deal with the low signal-to-noise ratio and the paucity of the data. With at our disposal a steadily growing volume of publicly available neuroimaging data, it is however possible to improve the estimation procedures involved in connectome mapping. In this work, we propose a novel learning scheme for functional connectivity based on sparse Gaussian graphical models that aims at minimizing the bias induced by the regularization used in the estimation, by carefully separating the estimation of the model support from the coefficients. Moreover, our strategy makes it possible to include new data with a limited computational cost. We illustrate the physiological relevance of the learned prior, that can be identified as a functional connectivity atlas, based on an experiment on 46 subjects of the Human Connectome Dataset.
01 Jan 2014
TL;DR: The authors reviewed how predictive models have been used on neuroimaging data to uncover new aspects of cognitive organization, and gave a statistical learning perspective on these progresses and on the remaining gaping holes.
Abstract: Functional brain images are rich and noisy data that can capture indirect signatures of neural activity underlying cognition in a given experimental setting. Can data mining leverage them to build models of cognition? Only if it is applied to well-posed questions, crafted to reveal cognitive mechanisms. Here we review how predictive models have been used on neuroimaging data to ask new questions, i.e., to uncover new aspects of cognitive organization. We also give a statistical learning perspective on these progresses and on the remaining gaping holes.
TL;DR: A simple method of rapid diagnostic classification for the clinic using Support Vector Machines (SVM) and easy to obtain geometrical measurements that, together with a cortical and sub-cortical brain parcellation, create a robust framework capable of automatic diagnosis with high accuracy.
Abstract: Magnetic Resonance Imaging (MRI) has been gaining popularity in the clinic in recent years as a safe in-vivoimaging technique. As a result, large troves of data are being gathered and stored daily that may be used asclinical training sets in hospitals. While numerous machine learning (ML) algorithms have been implemented forAlzheimer's disease classi cation, their outputs are usually dicult to interpret in the clinical setting. Here, wepropose a simple method of rapid diagnostic classi cation for the clinic using Support Vector Machines (SVM) 1 and easy to obtain geometrical measurements that, together with a cortical and sub-cortical brain parcellation,create a robust framework capable of automatic diagnosis with high accuracy. On a signi cantly large imagingdataset consisting of over 800 subjects taken from the Alzheimer's Disease Neuroimaging Initiative (ADNI)database, classi cation-success indexes of up to 99.2% are reached with a single measurement.Keywords: Alzheimer's disease, machine learning, mild cognitive impairment, support vector machines, fastclinical diagnosis.
TL;DR: This paper provides a detailed analysis of a generative statistical model based on dense deformable templates that represents several tissue types observed in medical images that is used to guide the tissue segmentation of new images.
Abstract: Automatic anatomical brain image segmentation is still a challenge. In particular, algorithms have to address the partial volume effect (PVE) as well as the variability of the gray level of internal brain structures which may appear closer to gray matter (GM) than white matter (WM). Atlas based segmentation is one solution as it brings prior information. For such tasks, probabilistic atlases are very useful as they take account of the PVE information. In this paper, we provide a detailed analysis of a generative statistical model based on dense deformable templates that represents several tissue types observed in medical images. The inputs are gray level data whereas our atlas is composed of both an estimation of the deformation metric and probability maps of each tissue (called class). This atlas is used to guide the tissue segmentation of new images. Experiments are shown on brain T1 MRI datasets. This method only requires approximate pre-registration, as the latter is done jointly with the segmentation. Note however that an approximate registration is a reasonable pre-requisite given the application.
01 Jan 2014
TL;DR: A novel analysis framework is introduced, where the amount of variance that is fit by a random effects subspace learned on other images is estimated; it is shown that a principal component regression estimator outperforms other regression models and that it fits a significant proportion (10% to 25%) of the between-subject variability.
Abstract: Inter-subject variability is a major hurdle for neuroimaging group-level inference, as it creates complex image patterns that are not captured by standard analysis models and jeopardizes the sensitivity of statistical procedures. A solution to this problem is to model random subjects effects by using the redundant information conveyed by multiple imaging contrasts. In this paper, we introduce a novel analysis framework, where we estimate the amount of variance that is fit by a random effects subspace learned on other images; we show that a principal component regression estimator outperforms other regression models and that it fits a significant proportion (10% to 25%) of the between-subject variability. This proves for the first time that the accumulation of contrasts in each individual can provide the basis for more sensitive neuroimaging group analyzes.