Virtual Reality Frameworks for Non-invasive Intercranial Explorations

The objective of this project is to take the first step towards the definition of a model of visual-functional reconstruction of the brain's electrophysiologic activity without having access to patient-specific structural MRI. The proposed approach relies on the combination of electroencephalography (EEG) source localization and machine-learning algorithms for the definition of a standard based on the comparison of discriminating features between brain models. EEG source localization (ESI) aims at identifying the areas of the brain responsible for potential variations detected by electrodes placed on the scalp of the patient, and involves the resolution of two fundamental problems: the forward problem and the inverse problem. ESI has several clinical application ranging from the pre-surgical localization of epileptic seizure-onset zone, or the treatment of psychological disorders via neurofeedback. The localization of the sources occurs by first calculating the electrical potentials resulting from hypothetical neuronal source activity distributions (forward problem), that are then compared with the real data recorded by the EEG to obtain an estimate of the activity that best fit these data (inverse problem). The inverse problem finds its complexity in the ill-posedness of the problem itself. Only by placing reliable a-priori constraints is therefore possible to accurately define an electric source, thus solving the inverse problem. The forward problem, instead is well-posed: it is unequivocally determined by the morphologic and electric properties of the patient's head, that can be obtained e.g. by magnetic resonance imaging (MRI). By obtaining an accurate estimate of the electrical sources it is possible to contextualize them in a three-dimensional model derived from the image processing of a structural MRI. However, the realization of a specific model for a patient for which an EEG source localization is performed is expensive in terms of time, cost and availability of medical equipment. We will therefore build a system capable of sidestepping the need for patient specific MRI head models. We will leverage a structural MRI dataset and build a new algorithm capable of comparing the discriminating brain features derived from the mapping of the scalp of a patient with those related to the three-dimensional models generated by the MRI dataset, and then associate the model that best fits the patient's brain. This will lay the foundations for a new method of EEG source imaging more convenient in economic terms and more reliable in terms of real-time visualization of electrical sources.