Functional Magnetic Resonance Imaging (fMRI), EEG and MEG are today’s gold standard approaches to study the human brain, in normal subjects or patients, because they are non invasive and relatively easy to use. However, the origin of their signals is nonetheless not fully understood, resulting in limitations which are not always taken into account [1].

The relationship between neuronal activity, metabolism, neurotransmission and associated vascular events, on which are based those imaging modalities is still poorly characterized. For instance the link between variations in blood flow and oxygen consumption during neuronal activation is yet a great subject of controversy. Similarly, the decoding of vascular effects into precise neuronal events remains elusive. Models allowing integrating information gathered at different scales and from different modalities are almost inexistent. Partial models have been proposed to interpret the BOLD (Blood oxygen level dependant) fMRI signal (so-called ‘balloon’ model [2]), especially for its initial dip or its late undershoot, but they have not been validated. Simultaneous recordings of electrical and BOLD signals in primates have given insights on the relationship between local potentials, multi-unit activity and the BOLD signal, potentially leading to a finer interpretation of fMRI signals.

The IFR 49 houses a unique range of imaging modalities (MEG, EEG, fMRI, diffusion MRI, ultra-high field MRI, EEG-fMRI, MR spectroscopy, etc) giving access to many different signals and scales. Our aim is to better understand the biological mechanisms underlying those brain biosignals, at a mesoscale and microscopic scale, so as to better exploit them.

1. Physiological models of brain activity and metabolism

In order to contribute to the physiological and biochemical interpretation of brain functional imaging signals (fMRI, NMR spectroscopy, EEG, optical imaging), several group’s of IFR49 developed models for brain physiological processes. The quantitative relationships between the physiological parameters of the model which have has been developed are still not properly understood. We will be pursuing our work in order to make the physiological models more quantitative.

Over the last few years increasingly accurate quantitative data has been published relating to the metabolic responses to cerebral tissue activity, data obtained using very different approaches, ranging from NADH fluorescence studies on sections of hippocampus to neuroimaging. This has raised several questions :

• Is it possible to incorporate this range of results to assist with modelling ?
• Is it possible to suggest new hypotheses for interpreting functional imaging data ?

With this in mind, we hope to construct a model which takes into account interactions between neurons and astrocytes. This model will aim first of all to report on in vivo data in animals, and on neuroimaging (fMRI) data in humans. This approach will allow an interpretation of the various aspects of the kinetics of tissue oxygen and extracellular lactate in animals. Finally, by applying the same type of model to brain functional imaging in humans, we will be testing the hypothesis that the early constituents of the BOLD signal in fMRI depend exclusively on the oxidative response of the neurons, whereas the later aspects of the BOLD signal are probably a combination of the responses of neurons and astrocytes [3].

Using multimodal imaging (fMRI, DOI, NMR spectroscopy) and modelling, we will be studying the relative contributions of neuronal and glial responses to the BOLD signal in fMRI. This innovative work could open the way to a closer analysis of the metabolic component of cerebral functional imaging signals.

2. Optical imaging and physiological models

Cerebral Diffusion Optical Imaging (DOI) has a limited spatial resolution in the order of a centimetre and would benefit from additional information taken from anatomical MRI. In this way, by merging the two methods, the spatial localisation of the sources of neuronal activity will be more accurate. Experimentally the integration of optical fibres in a system involving a high magnetic field and the simultaneous measurement of infrared optics and fMRI imaging can be complex. The fibres must be taken outside the Faraday cage and the detectors positioned on the head must not be sensitive to the magnetic field.

A second problem associated with optical imaging is the modelling of the propagation of infrared light and the tomographic reconstruction of images : the various cortical tissues, white matter, grey matter and cerebrospinal fluid, all have different properties of light absorption and diffusion. The modelling of light propagation (direct problem) is necessary to obtain functional images. This modelling is difficult, as the information about the spatial distribution of the tissues is not often available. The segmentation of the different tissues of the cerebral cortex, using anatomical MRI, associated with the fusion of aMRI and DOI will be used to realistically model the propagation of the optical signal. We will then develop a simulation model to deal with the direct problem (propagation of light on a segmented network) with the aid of a method using finite elements.

Finally, this direct problem will be used in solving the inverse problem, the tomographic reconstruction of optical images with prior anatomical information. MRI and DOI will allow us to generate images corresponding to different components of neuronal activity : blood flow and BOLD signal (fMRI), blood volume and level of oxygenation (DOI). However, we need to go one step further in order to complete our study and integrate all the physiological parameters measured using optics within the physiological models of the BOLD signal. We will use these measurements to construct a model which takes into account neuron-astrocyte interactions.

3. Multiscale Imaging : from synaptic functional imaging to VSD network and Bold imaging

One way to understand the integrative steps leading from neural activity to the Bold signal is to combine at the same time several technologies of observation, ranging from microscopic (intracellular recordings) to more macroscopic activity integration (MEA (multiple electrode array) & LFP (local field potential) recordings, extrinsic voltage sensitive dye (VSD) and Bold imaging).

One issue that remains largely unsolved (in cortical networks) is the level of correlation that can be detected as a function of the scaling of the signal. Another key issue, important in the field of coding, is to identify a possible dependency of these correlations on the sensory drive statistics.

These multiscale experiments will be realized at UNIC using two setups : one is designed to combine VSD (with a high speed camera) and intrinsic imaging with mutli-electrode Array (silicon probe) recordings (32 single unit and LFP channels) and the other one will combine intrinsic and VSD imaging with intracellular recordings. Activity will be recorded in the visual cortex of anaesthetized mammals in responses to “movies” of increasing complexity. Preliminary results obtained for natural scene viewing statistics show that fine spatial and temporal scale measurements induce highly decorrelated activities, whereas more macroscopic and low temporal frequency voltage or VSD measurements show strong long-distance coherence. PCA analysis will be used in order to extract the VSD and the intrinsic components from the same optical imaging fluorescence signal. The comparison of the dynamic of these different signals, at the same point in time and for the same visual context, will useful to understand how the cortical network integrates and computes visual information.

4. Diffusion MRI imaging

On the other hand, a fundamentally new paradigm has emerged to look at brain activity through the observation with MRI of the diffusion behavior of the water molecules [4]. Water molecules are used as a probe that can reveal microscopic details about the subtle brain tissue architecture and spatial organization [5]. It has been shown that the diffusion of water slightly slows down in the activated brain cortical areas. This slowdown, which occurs several seconds before the hemodynamic response detected by BOLD fMRI, has been described in terms of a phase transition of the water molecules from a somewhat fast to a slower diffusion pool in the cortex undergoing activation and tentatively ascribed to the membrane expansion of cortical cells which undergo swelling during brain activation. Such biophysical events have been monitored at the microscopic level by using invasive techniques in neuronal cell cultures or slices, but do not necessarily reflect physiological conditions. Observing changes in cortical cell configuration in intact animals or humans with diffusion MRI would, thus, have a tremendous impact, because it would be directly linked to neuronal events and bridge the gap with current approaches to obtain images of human brain activation, and potentially offer improved spatial and temporal resolution. This hypothetical mechanism remains, however, to be confirmed.

Our aim is to obtain a more quantitative understanding of the biophysical mechanisms at the origin of the diffusion MRI signals, theoretically, using a numerical simulator of the diffusion process, and experimentally by measuring the diffusion MRI signal at high spatial/temporal resolution in non-human primate brains, in combination with other state-of-the art electrophysiological and optical imaging techniques using experimental paradigms from a common library. Optical imaging allows to observe together vascular signals (similar to BOLD) and intrinsic signals linked to cortical column architecture (“Light Scattering”).

5. Ultra-high field MRI

An important asset for the teams of IFR 49 is the NeuroSpin plateform which is being equipped with unique MRI systems operating at very high magnetic field strengths not yet available elsewhere in the world, as well as related tools and an advanced computer platform. Future equipment includes a 17.6T 250mm horizontal magnet in 2009 and a 900mm 11.7T whole-body system to be delivered in 2012, both of which will be world premiers for such systems. With ultra high magnetic field (UHF) Magnetic Resonance Imaging and Spectroscopy the IFR 49 is thus in good position to explore the brain at spatial and temporal scales which may give access to a neural code, by pushing the current limits of brain imaging, from mouse to man, as far as possible with.

A first aim for UHF MRI is to obtain higher spatial and temporal resolution, and to gain at least one order of magnitude compared to conventional approaches, at least in some parts of the brain, such as the hippocampus (NeuroSpin, CENIR). This region is particularly important for early diagnosis of Alzheimer’s disease. Functionally, in humans at 11.7T the target is the ‘mesoscale’, clusters of thousands of neurons, while at 17.6T in animals MRI microscopy could reach the neuron level. To do so, new acquisition schemes must be devised to take into account the change in electromagnetic regimen imposed by ultra-high frequencies (heavily parallel reception and transmission, reconstruction, dedicated radiofrequency pulses and coil arrays, etc.). Hence, ultra-high resolution MRI of specific parts of the brain may be sometimes preferred to whole-brain MRI. Another issue is to investigate contrast mechanisms (such as the relationship between molecular diffusion and tissue microstructure and physiology) and to explore new avenues to generate novel contrasts with or without using dedicated tracers. Additionally, at UHF spectroscopic techniques allow detection and mapping of non-hydrogen nuclei (e.g. 31P, 23Na, 13C, 17O) and biologically relevant molecules (physiology, metabolism, neurotransmission).