Sparse coupled logistic regression to estimate co-activation and modulatory influences of brain regions

Accurate mapping of the functional interactions between remote brain areas with resting-state functional magnetic resonance imaging requires the quantification of their underlying dynamics. In conventional methodological pipelines, a spatial scale of interest is first selected and dynamic analysis then proceeds at this hypothesised level of complexity. If large-scale functional networks or states are studied, more local regional rearrangements are then not described, potentially missing important neurobiological information. Here, we propose a novel mathematical framework that jointly estimates resting-state functional networks and spatially more localised cross-regional modulations. To do so, the changes in activity of each brain region are modelled by a logistic regression including co-activation coefficients (reflective of network assignment, as they highlight simultaneous activations across areas) and causal interplays (denoting finer regional cross-talks, when one region active at time t modulates the t to t + 1 transition likelihood of another area). A two-parameter l1

Multiscale body maps in the human brain

Michel Akselrod

A large number of brain regions are dedicated to processing information from the body in order to enable interactions with the environment. During my thesis, I studied the functional organization of brain networks involved in processing bodily information. From the processing of unimodal low-level features to the unique experience of being a unified entity residing in a physical body, the brain processes and integrates bodily information at many different stages. Using ultra high-field functional Magnetic Resonance Imaging (fMRI), I conducted four studies to map and characterize multiscale body representations in the human brain. The goals of my thesis were first to extend the actual knowledge about primary sensorimotor representations, and second to develop novel approaches to investigate more complex and integrated forms of body representations. In studies I and II, I first investigated how natural touch was represented in the three first cortical areas processing tactile information. I applied a mapping procedure to identify in each of these three areas the somatosensory representations of 24 different body parts on hands, feet and legs at the level of single subjects. Using fMRI and resting-state data, I combined classical statistical analyses with modern methods of network analysis to describe the functional properties of the formed network. In study III, I applied these methods to investigate primary somatosensory and motor representations in a rare population of patients. Following limb loss, the targeted muscle and sensory reinnervation (TMSR) procedure enables the intuitive control of a myoelectric prosthesis and creates an artificial map of referred touch on the reinnervated skin. I mapped the primary somatosensory and motor representations of phantom sensations and phantom movements in TMSR patients. I investigated whether sensorimotor training enabled via TMSR was associated with preserved somatosensory and motor representations compared to healthy controls and amputee patients without TMSR. Finally in study IV, I studied brain regions involved in the subjective body experience. Following specific manipulations of sensorimotor information, it is possible to let participants experience a fake or virtual hand as their own and to give them the sensation of being in control of this hand. Using MR-compatible robotics and virtual reality, I investigated the brain regions associated with the alteration of the sense of hand ownership and the sense of hand agency. The present work provides important findings and promising tools regarding the understanding of brain networks processing bodily information. In particular, understanding the functional interactions between primary unimodal cortices and networks contributing to subjective body experience is a necessity to promote modern approaches in the fields of neuroprosthetic and human-machine interactions.

EPFL2017

Fermi resonance in CO2: Mode assignment and quantum nuclear effects from first principles molecular dynamics

Sara Bonella, Guillaume André Jean Fraux

Vibrational spectroscopy is a fundamental tool to investigate local atomic arrangements and the effect of the environment, provided that the spectral features can be correctly assigned. This can be challenging in experiments and simulations when double peaks are present because they can have different origins. Fermi dyads are a common class of such doublets, stemming from the resonance of the fundamental excitation of a mode with the overtone of another. We present a new, efficient approach to unambiguously characterize Fermi resonances in density functional theory (DFT) based simulations of condensed phase systems. With it, the spectral features can be assigned and the two resonating modes identified. We also show how data from DFT simulations employing classical nuclear dynamics can be post-processed and combined with a perturbative quantum treatment at a finite temperature to include analytically thermal quantum nuclear effects. The inclusion of these effects is crucial to correct some of the qualitative failures of the Newtonian dynamics simulations at a low temperature such as, in particular, the behavior of the frequency splitting of the Fermi dyad. We show, by comparing with experimental data for the paradigmatic case of supercritical CO2, that these thermal quantum effects can be substantial even at ambient conditions and that our scheme provides an accurate and computationally convenient approach to account for them. Published by AIP Publishing.

Amer Inst Physics2017

Multiscale body maps in the human brain

Michel Akselrod

EPFL2017