Guillaume Flandin

Functional connectivity: studying nonlinear, delayed interactions between BOLD signals.

Correlation analysis has been widely used in the study of functional connectivity based on fMRI data. It assumes that the relevant information about the interactions of brain regions is reflected by a linear relationship between the values of two signals at the same time. However, this hypothesis has not been thoroughly investigated yet. In this work, we study in depth the information shared by BOLD signals of pairs of brain regions. In particular, we assess the amount of nonlinear and/or nonsynchronous interactions present in data. This is achieved by testing models reflecting linear, synchronous interactions against more general models, encompassing nonlinear, nonsynchronous interactions. Many factors influencing measured BOLD signals are critical for the study of connectivity, such as paradigm-induced BOLD responses, preprocessing, motion artifacts, and geometrical distortions. Interactions are also influenced by the proximity of brain regions. The influence of all these factors is taken into account and the nature of the interactions is studied using various experimental conditions such that the conclusions reached are robust with respect to variation of these factors. After defining nonlinear and/or nonsynchronous interaction models in the framework of general linear models, statistical tests are performed on different fMRI data sets to infer the nature of the interactions. Finally, a new connectivity metric is proposed which takes these inferences into account. We find that BOLD signal interactions are statistically more significant when taking into account the history of the distant signal, i.e., the signal from the interacting region, than when using a model of linear instantaneous interaction. Moreover, about 75% of the interactions are symmetric, as assessed with the proposed connectivity metric. The history-dependent part of the coupling between brain regions can explain a high percentage of the variance in the data sets studied. As these results are robust with respect to various confounding factors, this work suggests that models used to study the functional connectivity between brain areas should in general take the BOLD signal history into account.

Multivariate Model Specification for fMRI Data

We present a general method-denoted MoDef-to help specify (or define) the model used to analyze brain imaging data. This method is based on the use of the multivariate linear model on a training data set. We show that when the a priori knowledge about the expected brain response is not too precise, the method allows for the specification of a model that yields a better sensitivity in the statistical results. This obviously relies on the validity of the a priori information, in our case the representativity of the training set, an issue addressed using a cross-validation technique. We propose a fast implementation that allows the use of the method on large data sets as found with functional Magnetic Resonance Images. An example of application is given on an experimental fMRI data set that includes nine subjects who performed a mental computation task. Results show that the method increases the statistical sensitivity of fMRI analyses.

Parcellation of brain images with anatomical and functional constraints for fMRI data analysis

We propose a methodology for brain parcellation with anatomical and functional constraints dedicated to fMRI data analysis. The aim is to provide a representation of fMRI data at any intermediate dimensionality between voxel and region of interest. In order to fill in the gap between these two approaches we developed an automatic parcellation of the 3D cortex with an adjustable resolution. The algorithm relies on an adaptation of the K-means clustering in a non convex domain with geodesic distances. Fine anatomical or functional constraints can be embedded through the use of weighted geodesic distances. The applications of such a method are principally connectivity studies, multivariate analyses and fusion with other modalities.

Model Based Spatial and Temporal Similarity Measures between Series of Functional Magnetic Resonance Images

We present a method that provides relevant distances or similarity measures between temporal series of brain functional images. The method allows to perform a multivariate comparison between data sets of several subjects in the time or in the space domain. These analyses are important to assess globally the inter subject variability before averaging subjects to draw some conclusions at the population level. We adapt the RV-coefficient to measure meaningful spatial or temporal similarities and use multidimensional scaling for visualisation.