My research activities have been focused on the development of statistical estimation methods as well as probabilistic and biophysical models for the analysis of different Neuroimaging modalities and their integration. This entails bridging the gap between data produced by recording machines and the actual underlying neuronal activity by solving the forward (from neuronal activity to data) and inverse problems (from data to neuronal activity) that arise in each case. I have played out this goal to different extents in functional techniques like Electro and Magneto encephalography (EEG/MEG) and functional Magnetic Resonance Imaging (fMRI). Particularly, I have worked intensively in the development of the so called Brain Electromagnetic Tomography (BET). This allows reconstructing 3D images of electric activity inside the brain from EEG/MEG measurements with a high temporal resolution. BET can also benefit from the high spatial resolution of fMRI and the detailed anatomical connectivity information from Diffusion Weighted Magnetic Resonance Imaging (DWMRI) data. To achieve this, I have used Neural Mass Models and information from DWMRI to study the interactions between electric, metabolic and haemodynamic variables and their coupling with the common underlying excitatory and inhibitory neuronal activity.
Along years, I have also realised that it is not spatial (where) or temporal (when) localization of neural activity per se that causes brain function, but rather the way (how) the neural masses are dynamically (causally) interconnected as time evolves. Therefore, I am currently interested in the development of Bayesian Models to estimate fluctuations of the time varying (non-stationary) Dynamical Causal Networks (DCNs) underlying on-going EEG/MEG and resting state fMRI. These fluctuations reflect changes in the dynamical regimes of the brain, which can be interpreted as fluctuations in its cognitive/physiological state. The tracking and prediction of brain state changes by means of such dynamical models can be applied to address important problems in a wide range of biomedical applications, such as the detection of changes in the physiological state of a patient (diagnosis), or the follow-up of the response to a therapy (treatment). These possibilities can also be exploited to develop state-of-the-art therapeutic techniques like Neurofeedback and Brain-Computer Interfaces (BCI). Due to the high complexity of the above models (many parameters vs. scarce data) a key aspect of this enterprise is multimodal integration. I am particularly interested in the use of information about the space-time structure of interregional coupling as provided by DWMRI, to inform Bayesian as well as Biophysical generative models of EEG and fMRI. These models can then be inverted to obtain robust estimates of the time-varying DCNs underlying brain function.
Forward and inverse problems of EEG, MEG and fMRI, functional Neuroimaging techniques, Multimodal Integration, Computational Neuroscience, Bayesian Probability Theory, Machine Learning, Image Processing and analysis, Causality Theory, Deterministic and Stochastic Nonlinear Dynamics, Complex Systems, Time Series Analysis, Markov Chain Monte Carlo Methods, Differential Equations, Neuronal Network Modelling