Section: New Results

Multi-Imaging Modalities

Coupling functional and structural models

A nested cortex parcellation combining analysis of MEG forward problem and diffusion MRI tractography

Participants : Anne-Charlotte Philippe, Maureen Clerc, Théodore Papadopoulo, Rachid Deriche.

Understanding the relationship between structure and function is a major challenge in neuroscience. Diffusion MRI (dMRI) in the only non-invasive modality allowing to have access to the neural structure. Magnetoencephalography (MEG) is another non-invasive modality that allows a direct access to the temporal succession of cognitive processes. Functional cortex parcellation being one of the most important ways to understanding structure-function relationship, we propose an innovative method merging MEG and dMRI to parcellate the cortex. The combination of MEG forward problem and connectivity information reveals cortical areas generating a similar magnetic field at sensors while having a similar connectivity. Results show suitable clusters that forecast interesting studies for inter- and intra- subjects comparisons of the cortex parcellations. The automatic nested cortex parcellation we propose could be a first step to analyse sources that are seeds of long or short range connectivity and to differentiate these connectivities in the white matter

This work has been published in [31] .

dMRI tractography of WM fibers to recover the anatomical connectivity supporting a MEG epileptic network

Participants : Anne-Charlotte Philippe, Maureen Clerc, Théodore Papadopoulo, Rachid Deriche.

Cerebral organization is determined by segregated and integrated regions both functionally and anatomically. These cerebral networks are the foundations of the execution of major part of cognitive processes. Information about the structure of the white matter (WM) and the functionality of networks are both needed to understand these cerebral networks.

This work proposes an efficient method to inform a given functional network on its anatomical support: how many anatomical connections exist between functionally connected regions and what are their geometries . Diffusion MRI being the only non invasive method allowing to have access to the micro-structure of the WM, we used diffusion information to underline the degree of connectivity between functionally connected regions while taking advantage of WM fibers reconstruction to determine the way taken by the anatomical network supporting the functional network.

Due to the complex dynamical alteration of epilepsy, the study of large-scale functional connectivity is difficult. But diffusion imaging studies have shown alterations of the WM between epileptic zones and connected areas. This methodology allows to add qualitative (degree of connectivity) and geometrical (WM fibers reconstruction) information on the anatomical network supporting an epileptic network mostly determined by magneto-encephalography (MEG).

This work has been published in [35] .

Whole cortex parcellation combining analysis of MEG forward problem, structural connectivity and Brodmann's atlas

Participants : Anne-Charlotte Philippe, Maureen Clerc, Théodore Papadopoulo, Rachid Deriche.

Functional cortex parcellation is one of the most important ways to understand the link between structure and function in the brain. Brodmann's atlas remains a fundamental pillar to understand this relationship because its areas are defined by similar cytoarchitecture and functional imaging notably had revealed that they correspond, entirely or in part, to functional areas. So, its integration to diffusion MRI (dMRI) data is pertinent, dMRI being the only non invasive and in-vivo imaging modality able to have access to a detailed geometric description of the anatomical connectivity between brain areas. In this work, we propose to define a new connectivity profile of cortical sources based on the Brodmann's atlas. After its registration to T1 and diffusion weighted images of the same subject, we reconstructed the brain surfaces and considered the cortical sources to be the vertices of the white matter/ grey matter boundary mesh. We performed a probabilistic tractography taking each cortical sources as seeds and the L Brodmann's areas as L targets. Thus, we obtained the connectivity profile of a cortical source: a vector v of size L where v (l) is the degree of connectivity of the source to the l th Brodmann's area. Then, we developped a cortical parcellation method jointly analyzing the MEG forward problem and the connectivity profiles based on Brodmann's atlas of cortical sources. We computed the leadfield matrix that relates the sources to the MEG sensors. We applied a k-means algorithm to the leadfield matrix to cluster sources having a close magnetic field to the MEG sensors. Then, in each leadfield-based cluster, we clustered sources via their connectivity profile based on Brodmann's atlas. This automatic parcellation is an efficient preprocessing to compute a MEG inverse problem on functional data informed by its structural connectivity.

This work has been published in [32] .

Study of the brain connectivity in an Immersive Space

Participants : Anne-Charlotte Philippe, Jean-Christophe Lombardo [Dream Project-Team, Inria, Sophia Antipolis, Méditerranée, France] .

Virtual reality is a powerful tool for scientific visualization. When the amount and complexity of the visualized data grows, standard visualization applications on desktop computers become inefficient. In this work, we present the use of a CAVE like VR facility in a neuroscientific context. The aim is to have a better understanding of the brain connectivity. Both anatomical and functional data are attached to a mesh representing the brain surface.

Specific tools developed for this study and the way we used them are presented in [36] emphasizing drawbacks and advantages of virtual reality in a scientific visualization context.

This work has been published in [36] .

Cortex parcellation via diffusion data as prior knowledge for the MEG inverse problem

Participants : Anne-Charlotte Philippe, Maureen Clerc, Théodore Papadopoulo, Rachid Deriche.

In this work, we present a new approach to the recovery of dipole magnitudes in a distributed source mo-del for magnetoencephalographic (MEG) imaging. This method consists in introducing prior knowledge regarding the anatomical connectivity in the brain to this ill-posed inverse problem. Thus, we perform cortex parcellation via structural information coming from diffusion MRI (dMRI), the only non-invasive modality allowing to have access to the structure of the WM tissues. Then, we constrain, in the MEG inverse problem, sources in the same diffusion parcel to have close magnitude values. Results of our method on MEG simulations are presented and favorably compared with classical source reconstruction methods.

This work is currently under submission.

Fractality in the neuron axonal topography of the human brain based on 3-D diffusion MRI

Participants : Panayotis Katsaloulis [Institute of Physical Chemistry "Demokritos" (IPC),National Center for Scientific Research "Demokritos", Greece] , Aurobrata Ghosh, Anne-Charlotte Philippe, Astero Provata [Institute of Physical Chemistry "Demokritos" (IPC),National Center for Scientific Research "Demokritos", Greece] , Rachid Deriche.

In this work, we conduct a group study, with 18 subjects, to validate the computational robustness of the fractal dimension of the neuron axonal topography in the human brain that is derived from diffusion MRI (dMRI) acquisitions. We extend the work done in a previous paper by some of the current authors where the fractal dimension of the neuron axonal topography from dMRI data was computed from 2-D regions of interest. The fractal dimensions Df of the entire 3-D volume of the brain is here estimated via the Box Counting, the Correlation DImension and the Fractal Mass Dimension methods. 3-D neuron axon data are obtained using tractography algorithms on Diffusion Tensor Imaging of the brain. We find that all three calculations of Df give consistent results across subjects, namely, they demonstrate fractal characteristics in the short and medium length scales: different fractal exponents prevail at different length scales, an indication of multifractality. We surmise that this complexity stems as a collective property emerging when many local brain units performing different functional tasks and having different local topologies are recorded together.

This work has been published in [15] .