We worked towards our objectives and harvested numerous and important
results in the 4 main focus areas listed in our research program.
1] In Focus Area 1, we aimed to develop advanced dMRI methods for
structural connectivity mapping. We developed generative and
ground-breaking models for advanced acquisition and processing of dMRI
data [Refs 4,7,19,34,39,54]. We contributed to the challenge to recover
microstructure tissue parameters by developing the Diffusion
Microstructure Imaging in Python (Dmipy) toolbox [Refs 15, 25, 84] and
assessed some of its applicability [Refs.13,14,94]. To open the way to
new dMRI biomarkers, we investigated the use of high order diffusion
models and developed new high order invariants [Refs. 9,18,31,85]
applied to characterize some diseases [Ref. 31,95,76,80]. We have also
developed new concepts using spherical convolutional neural network
for dMRI fiber ODF estimation [Refs. 69,75], and new approaches
grounded on microstructure from dMRI applied them to Microstructure
based tractography [Refs. 3,50]. For validation of tractography and
dMRI results, we bridged the gap between Polarized Light Imaging and
dMRI, demonstrating a great promise to validate diffusion MRI
tractography thanks to multi-scale fiber tracking based on 3D-PLI
[Refs. 22,23,27,28,73,82,88].
2] In Focus Area 2, we aimed to develop advanced M/EEG source
reconstruction methods using spatial and temporal constraints and
explored in more depth how to use diffusion MRI data for spatially
regularizing the M/EEG forward and inverse problems
[Refs. 8,10,12,29,45,55,57,77,78,79,97]. We proposed groupwise
parcellation of the whole cortex based on structural connectivity
[Refs. 2,5,35,38,43], solved the cross-subject parcel matching problem
using Optimal Transport [Refs. 59,60], developed a novel atlas of
human cerebral cortex based on extrinsic connectivity [Refs 41, 43],
and assessed how robustly an atlas captures the network topology
across different subjects in a homogeneous cohort [Ref. 93].
] In Focus Area 3, we developed joint structure-function models from
which dynamical structural-functional brain connectivity networks can
be extracted. To unravel dynamical brain networks using both dMRI and
M/EEG data and build large brain effective network from EEG/MEG data
and dMRI information, we contributed to reconstruct the information
flow in the brain for visual and motor tasks [Refs. 6, 33] and to estimate axonal
conduction speed, conduction delay, and diameter using connectivity
informed maximum entropy on the mean [Refs, 48,49]. In addition, we
explored the structure-function relationship and developed a unified
framework for multimodal structure-function mapping based on
eigenmodes to predict the function given the structure
[Refs. 72,87]. Integrating the spatial information of dMRI and the
temporal one coming from fMRI has also been developed and validated
[Refs. 44,47,51,52,53,74,83]. Finally, we have succeeded to combine
functional MRI, diffusion MRI, and M/EEG to infer communication and
information flow between cortical regions, therefore opening the door
to the non-invasive exploration of information flow in the white
matter [Ref. 86].
4] In Focus Area 4, We assessed the applicability of microstructure
imaging [Refs 7, 15, 25, 84,.24,40] and the use of our new high order
invariants in some diseases such as mild Traumatic Brain Injury [56] and
Multiple Sclerosis [Refs. 95,96]. In addition, we also
investigated the impact of tractography filtering [Refs. 16.21,91] on
the structural networks of mTBI injury subjects.
Finally, we contributed to develop a set of publicly available software packages:
The Diffusion Microstructure Imaging in Python package (Refs 15, 25, 84],
The OpenMEEG package [Ref. 67], The Tractograms As Linear Operators
in Neuroimaging package (
https://pypi.org/project/cobcom-talon/(s’ouvre dans une nouvelle fenêtre)) and
the Python implementation of the graph alignment WL-align algorithm
[Ref. 93].
erall, and at the end of this project, we published over a hundred
papers in the most selective journal and conferences of the domain.
