Concordia University Physics Department | McGill University Biomedical Engineering Department
Multimodal Functional Imaging Laboratory
Welcome to the Multimodal Functional Imaging Laboratory, directed by Dr. Christophe Grova. Learn about our ongoing research themes.
RESEARCH THEME 1:
TAKING BENEFIT FROM DIRECT MEASUREMENTS OF NEURONAL ACTIVITY
ElectroEncephaloGraphy (EEG) and MagnetoEncephaloGraphy (MEG) explore directly bioelectric activity generated by the neurons (at a millisecond scale), but only through scalp recordings, therefore providing only limited spatial representations. We have demonstrated expertise in EEG/MEG source localization, proposing new methods to reconstruct accurately along the cortical surface the generators of EEG/MEG scalp recordings and their spatial extent, at every millisecond. We proposed and evaluated the Maximum Entropy on the Mean method (MEM) as one of the only source localization method able to retrieve the underlying spatial extent of the generators along the cortical surface, while other methods are only able to reconstruct accurately the location of the peak of underlying activity. We validated MEM performances using realistic simulations before applying it on EEG/MEG clinical epilepsy data. We demonstrated the clinical yield of EEG/MEG to localize epileptic discharges on large groups of patients, validating its accuracy with reference intracranial EEG recordings, now incorporating source imaging in clinical practice. Extending MEM in the time-frequency domain, we were able to localize accurately high frequency oscillations and rhythmic patterns at seizure onset, but also when assessing the underlying structure of brain networks at rest. Applying EEG/MEG source imaging on resting state oscillations, we showed that reduction of long-range connectivity with the focus could predict postsurgical outcome, even when no discharges could be detected. We have carefully validated the ability of MEG resting state source imaging to recover patterns of oscillations and functional connectivity patterns in comparison with Gold Standard intracranial EEG data. Our next challenges will consist in further investigating the ability EEG/MEG source imaging from resting state data to identify functional connectivity patterns and resting state networks, identifying specific biomarkers to characterize sleep quality in healthy controls and to predict postsurgical outcome in patients with epilepsy. Our next step is to acquire simultaneously MEG or high density EEG data with invasive intracranial EEG, as the ideal framework to validate our proposed methods and to improve the ability of MEM to localized deep seated brain activity.
The MEM software package has been made available to the community through the Brainstorm Software
This video shows the steps and the results of Electroencephalography (EEG) source localization for a visual half field paradigm. EEG is a technique that measures the neuronal activity in the brain using electrodes that are pasted on the scalp. It can measure brain signals at every millisecond, however, without advanced mathematical and physical tools it could not provide 3-dimensional (3D) information on the brain. Our research involves fusing the EEG signals -measured with a high-density EEG system (256 electrodes, EGI)- with the 3D anatomical information of the brain, obtained with Magnetic Resonance Imaging, to localize the neuronal activity on the cortex.
Complete multimodal investigation involving MEG source localization, fMRI response and intra-cranial EEG investigation.
KEY PUBLICATIONS
RESEARCH THEME 2:
ESTABLISHING NEW BIOMARKERS FROM MULTIMODAL ANALYSIS OF CONNECTOR HUBS
Functional connectivity consists in analysing the underlying network structure of functional brain activity during rest. Functional connectivity analyses have been proposed to address several important questions raised in neurosciences and medicine, showing usually highly sensitive but poorly specific reorganizations when characterizing pathological conditions. Using slow fluctuations of resting state fMRI data, we reported brain network reorganizations that are epilepsy specific. Brain regions interacting between several of those networks, the so-called connector hubs, are essential to maintain the integrity of such network architecture. They are also efficient in term of metabolism when compared to non-hub regions. We proposed a sparse decomposition method estimating maps of connector hubs together with the fMRI networks involved in each hub, reporting specific hub reorganizations in epilepsy . We also evaluated the reorganization of fMRI networks and hubs during a nap following total sleep deprivation and reported key findings correlating glucose/oxygen metabolism to hubness in the healthy brain. Our next challenges will consist in estimating connector hubs from resting state source imaging of EEG/MEG data, shedding light on the neuronal origins of these networks and hubs detected using resting state fMRI, while investigating the relationship between hubness estimated from functional data (fMRI, EEG/MEG), structural connectivity patterns (estimated using diffusion MRI) and glucose/oxygen metabolism estimated using FDG PET in healthy and pathological conditions (epilepsy).
The SPARK software package, which is used to estimate connector hubs, can be found on our Github
KEY PUBLICATIONS
NREM sleep brain networks modulate cognitive recovery from sleep deprivation
Lee, K., Wang, Y. et al.
Bioarxiv Preprint
2024
RESEARCH THEME 3:
DEVELOPING PERSONALIZED AND WEARABLE IMAGING TO STUDY HEALTH IN EVERYDAY LIFE
Combining EEG/MEG source imaging and simultaneous EEG-fMRI, we reported excellent spatial concordance between EEG/MEG sources and hemodynamic responses to epileptic discharges, localizing accurately the epileptic focus, validating our findings with intracranial EEG. To further investigate neurovascular coupling processes during prolonged whole night recordings, we decided to add wearable imaging to our multimodal framework: fNIRS which consists in sending and receiving infra-red light through optic fibres attached to the head to monitor cortical fluctuations of oxy-hemoglobin and deoxy-hemoglobin concentrations. We developed personalized simultaneous EEG/fNIRS, optimizing the position of fNIRS sensors on the scalp, maximizing light sensitivity to reach specific cortical targets. These personalized locally dense montages allow accurate 3D reconstruction of fNIRS hemodynamic signals along the cortex using MEM. We reported promising results in epilepsy, now proposing whole night monitoring to study sleep/epilepsy interactions. We developed a hierarchical Bayesian model to assess fNIRS changes associated with modulation of brain excitability. We then implemented a similar hierarchical Bayesian model, showing for the very first time how sleep stages are interacting with the hemodynamic responses elicited by epileptic discharges, suggesting the occurrence more transient hypoxic events during deep sleep.
It is worth mentioning that EEG/fMRI studies usually involve short duration scans (2-3h) and sleep protocols are notoriously complex: strict immobility imposed by MRI limits acquisition duration and the noisy MRI scanner is not an ideal environment to fall asleep. Therefore, our overarching objective is to evaluate whole night EEG/fNIRS as a unique wearable neuroimaging technique mapping fNIRS hemodynamic signals along the cortical surface during the whole night, to evaluate if epileptic discharges, but also spontaneous fNIRS oscillations, are associated with preserved or impaired neurovascular coupling and oxygen metabolism, and how vigilance states (awake / sleep states) interfere with these phenomena. In addition to our applications in epilepsy, we are developing the very first atlas of EEG/fNIRS sleep physiology in young healthy subjects. Our preliminary results are suggesting promising sleep specific fNIRS features and notably spectral signatures, suggesting that fNIRS could become a new wearable technique to image sleep activity in a complementary manner than when using conventional EEG.
Our fNIRS software package, NIRSTORM, is made available on our Github with our tutorial available on the Brainstorm site