Dynamics of cortical activation during task-switching paradigm through MEG analysis: An extensive study.

The overall aim of this thesis is a better understanding of the main brain mechanisms through different magnetoencephalography (MEG) analysis methods, during a task-switching experiment. MEG is a non-invasive functional neuroimaging technique that relies on the measurement – outside of the head – of the magnetic field produced by neuronal activity. Because of its high temporal and spatial resolution yet complex and expensive system, MEG is still considered today a cutting-edge technology whose use is still mainly for research purposes and not a widely established diagnosis method like fMRI or EEG. This work has the objective to localize and investigate in space, time and frequency, the origin and propagation of the cortical signal, implementing specific functions provided mostly by the MNE-Python package. MNE-Python is the most known and up-to-date tool containing a set of algorithms that address and solve the mathematical challenges that MEG imaging is characterized by. The first part of this work focuses on introducing the main concepts of MEG and the linked challenges that need to be tackled to accurately estimate patients’ cortical activation. Later in the dissertation, the most important concepts about cognitive learning are broadly described and a precise illustration of the held task-switching experiment is provided. The last two chapters of this thesis gaze on different approaches to investigate the dynamics of cortical activations over time by looking directly at the spatiotemporal source estimates. Particularly, analyses have been carried on exploring different aspects of the signals: from a straightforward cortical source estimate to a more advanced machine learning regression. This project was done in collaboration with the Athinoula A. Martinos Center for Biomedical Imaging (Boston) and the Brain & Vision Research Laboratory (Biomedical Engineering Department - Boston University).