On an Computationally Empowered Virtual Reality System for Real Time Intracranial Neuronavigation

Physiological and electrical information about the brain can be obtained via a wide array of techniques such as the Electroencephalography (EEG), the Positron Emission Tomography (PET), Magnetic Resonance imaging (MRI) and Functional MRI (fMRI) which all have their trade-offs. One of the most popular acquisition techniques used both for research and for clinical application is the EEG which is an affordable and noninvasive technology that measures the electric potential on the scalp of the subject by means of a set of electrodes. The EEG is also used for EEG source imaging (ESI) which aims at computing the intracranial electrical activity from the electric potential measured on the surface of the head. The source imaging problem can be split into two key sub-problems, the forward problem (FP) and the inverse problem (IP). The forward problem is a computationally complex problem, whose goal is to map the electrical currents inside the brain to the readings acquired on the head surface with an EEG. The IP tries performs the opposite mapping and yields the position and intensity of the electrical activity inside the human brain based on the EEG measurements on the scalp surface. Numerical methods can solve the FP, typically boundary or finite element methods, and its accuracy only depends on the accuracy of the model and anatomical information of the subject. The IP, however, is an ill-posed problem whose solution is not unique and generally unstable. In this work we have designed and implemented an immersive virtual reality system, capable of offering a real-time navigation inside the electrical activity of the human brain, by leveraging neuroimaging and electromagnetic source imaging techniques. One of the challenges tackled in this thesis was the advancement of state of the art source imaging techniques for allowing the integration in a new virtual reality environment. In addition we had to address the high requirements imposed by a fluid virtual reality experience by developing computationally empowered algorithms designed to solve the inverse problem. The brain activity displayed to the end-user of the virtual reality system has been computed from a real-time ESI-inverted EEG reading or from pre-recorded data which can be of interest in a research or medical contest. The brain anatomical structure is commonly derived from MRI and fMRI images, with a procedure called brain segmentation. This procedure can produce 2D or 3D reconstruction of the human brain, which is essential in medical diagnoses or surgical planning. The 3D meshes used in the project come from an elaboration of real human fMRIs, produced with some of the most recent brain segmentation approaches. Part of the design focus was also dedicated to the creation of software pipelines and tools essential for satisfying the requirements of an immersive real-time neuroimaging experience. One of the key problems we have tackled has been to maintain stable performances even with high-density research EEG, to remain compatible with real-time visualization. The final system has been used in several demonstrations, using both live-recorded data and simulations. The virtual reality environment we have obtained can be adopted for teaching, research, or medical purposes. One of these applications is neurofeedback which is a therapeutic procedure that allows patients to regulate their brain activity by allowing them to visualize and navigate it in real-time.