Postdoctoral position in graph signal processing for monitoring spatio-temporal dynamics
of brain activity

Telecom Bretagne, Brest, France

P.I. : Dr. Vincent Gripon

Context

The present proposal is funded by the Neural Communication project, which is part of the
Labex CominLabs. This project started in November 2015 for a duration of two years as a
continuation of the Neural Coding project, also founded by the Labex CominLabs from November
2012 to October 2015. The main goal of the project is to propose models for communications
between cortical areas in the brain and to develop novel methodologies for identifying
spatio-temporal patterns in brain activity.

The project takes advantage of the expertise of two internationnaly renowned teams. The first
one is led by Fabrice Wendling, director of research at LTSI in Rennes, and aims at proposing
innovative methods for extracting connectivity graphs from EEG signals. The second one
(host team) is led by Claude Berrou, professor at Télécom bretagne, and aims at proposing
novel models for information processing in neural networks. The host team is also active in
the new field of graph signal processing. The host team is composed of three permanent
researchers as well as two postdocs and six PhD students.

The proposed postdoctoral position is meant to take place in Télécom Bretagne, in Brest.
The candidate is expected to work in collaboration with people involved in the ERC project
Neucod led by Claude Berrou.


Proposal

It is now well-known that brain activity is structurally organised into functionally connected
networks. Using functional MRI on subjects at rest, resting state networks can be defined as
sets of spatially distinct brain areas with temporally correlated activity. While having been
extensively studied over long time periods (e.g. 6 minutes), such networks are now investigated
using dynamic analysis with a temporal resolution of a few dozens of seconds. Other works are
 interested in applying similar methodologies to monitoring the dynamics when performing cognitive
 tasks, as it has been shown that task-related functional networks are spatially similar to
resting-state networks.

Recently, a novel framework has been developed that aim at exploiting graphical structures
of the underlying topology the signals are projected onto, called graph signal processing.
As a result, it is possible to define ad hoc harmonic operators including convolutions and
wavelets. A major advantage of this approach is the ability to detect similar patterns in
distinct regions of the network, despite the inhomogeneity of the network topology. We are
interested in applying graph signal processing to brain imaging to study special temporal
dynamics, both in the resting state and task-related networks using EEG and fMRI.


Details

The head of the project is Vincent Gripon, permanent researcher with Télécom Bretagne.
Please send your curriculum vitae, contacts for two references, as well as a cover letter
to vincent.gripon@telecom-bretagne.eu.

Expected disposable salary is about 30k{\euro} per year, depending on experience. The
position starts in September 2016 (or sooner) and finishes in November 2017. The postdoc
position is accompanied with overheads for travel expenses covering the participation to
international conferences.

The candidate must have defended a Ph.D. related to neuroscience, neuroimaging and
familiar with signal processing. Programming skills are essential to this position.

The deadline for applying to the position is July 17th.