Online Connectivity Estimation and Neurofeedback with TMS

Human capabilities are characterized by fast and parallel processing of information across local and long-range brain networks. Understanding the dynamic neural processing underlying our capabilities requires an electrophysiological technique capable of identifying the networks and their functions with a millisecond resolution. Today, electroencephalography (EEG) and magnetoencephalography (MEG) are the only noninvasive techniques that can estimate the neuronal activity with such resolution. On the contrary, techniques such as (functional) Magnetic Resonance Imaging (MRI/fMRI) measure comparatively slow, but with a much higher spatial resolution. Regarding the proposed project with the title Online Neuronal Connectivity Estimation and Neurofeedback with Transcranial Magnetic Stimulation we will establish new real-time methods to analyze and process MEG/EEG data for subsequent cortical stimulation scenarios. The overarching objective is to develop real-time computational tools for advancing our understanding of electrophysiological functions of neuronal networks and cortical stimulation in the human brain in health and disease. During the project methods for estimating functional connectivity in real-time MEG/EEG scenarios will be established. Furthermore, ways to integrate Transcranial Magnetic Stimulation (TMS) into a real-time scenario will be investigated. Such real-time scenarios can be found in neurofeedback and Brain Computer Interface (BCI) research. Both fields address data processing pipelines, which present specific stimuli to the subject, measure the corresponding brain activity, process the measured data in real-time and readjust the upcoming stimulus. This way the subject is part of a close-loop data processing pipeline with a direct feedback. This is for example useful when learning to regain motor function during stroke rehabilitation treatment. The proposed work will advance knowledge regarding the network activity in human with a millisecond resolution and thus contribute to one of the promising research field of our time, the decoding of the human brain.

Electroencephalography (EEG) and magnetoencephalographie (MEG) offer the possibility to measure and localize brain activity with a unique temporal resolution. The online or real-time processing of EEG and MEG data allows to not only inspect the data during a measurement but to dynamically react to the measurement results, e.g., by adapting the experimental design, delivering targeted brain stimulation or giving neurofeedback to the subject. Thereby, online processing can be used in a variety of applications such as brain-computer interfaces, neurological rehabilitation or to design novel experimental paradigms. Within the research project "ONCE-TMS / Online Neuronal Connectivity Estimation and Neurofeedback with Transcranial Magnetic Stimulation" algorithms that allow for the online estimation of EEG/MEG functional connectivity in large all-to-all networks were derived and implemented. In combination with algorithms enabling the online estimation of the brain sources underlying EEG/MEG, this allows the online analysis of the complex interplay of different brain regions. Being able to perform complex processing tasks such as source analysis or connectivity analysis online has the potential to facilitate important progress in all applications of EEG/MEG online processing. Within this research project, we laid ground for the use of online EEG/MEG connectivity analysis in combination with transcranial magnetic brain stimulation (TMS) and neurofeedback applications. This allows establishing closed-loop applications, in which the stimulation or neurofeedback is determined based on the measured EEG/MEG signal and can then be altered based on the measured effect of the stimulus. Besides the efficient implementation of EEG/MEG connectivity analysis, this required the implementation of various signal processing algorithms such as frequency analysis or common spatial patterns (CSP). As main outcome of this research project, we were for the first time able to demonstrate the online analysis of large all-to-all functional connectivity networks, derived a novel algorithm for the computation of CSPs, and demonstrated a closed-loop pipeline for EEG/MEG-based TMS in MNE Scan. Besides, we were able to demonstrate that basic EEG/MEG online processing with MNE Scan can also be performed on mobile and highly affordable single-board computers, such as the Raspberry Pi. All implementations within this research project were performed in the open-source C++ package MNE Scan and are available at https://mne-cpp.github.io/. This allows researchers to directly apply the results from this research project to perform closed-loop TMS and neurofeedback experiments.