BBCI Winter School on Neurotechnology 2014

The aim of this lecture is to provide an illustrative tutorial on the methods for single-trial ERP analysis. Basic concepts of feature extraction and classification will be explained in a way that is accessible to participants from non-technical areas for BCI research in order to facilitate the interdisciplinary exchange. The tutorial will provide the foundation for subsequent more advanced data analysis lectures.
[ presentation | supplementary material ]

This lecture proposes an accessible introduction to the design of Brain-Computer Interfaces (BCI) based on oscillatory EEG activity (e.g., motor imagery), notably from a signal processing point of view. In particular, it first presents the basic feature extraction and classification tools to design such a BCI. The lecture then describes the use of spatial filters, both simple static ones (e.g., Laplacian) as well as advanced supervised ones (e.g., Common Spatial Patterns and variants) to enhance the performance and the robustness of the whole BCI. A few supervised temporal filters will be considered as well. Alternative EEG features representation are then exposed as promising additions to basic features. This notably includes features measuring the EEG signals complexity, and more importantly, features measuring how EEG signals from different brain areas are synchronized. This lecture will ends by briefly showing the audience that designing oscillatory activity-based BCI is not all about signal processing. Indeed, considering the user and how to train him/her to control the BCI is also a key point for successful BCI design.
[ presentation ]

The sounds around us are an amazingly rich source of information, and human brains are well-equipped to extract meaningful sound features within few milliseconds from the surrounding auditory scene. Sound stimuli leave a trace in the electroencephalogram (EEG) of a listener, and some of the characteristics of this trace change dependent on whether the listener has attended the sound or not. As brain-computer interfaces (BCIs) can be designed to exploit attention-dependent differences, they can also operate in the auditory modality. Such auditory BCIs are a relatively novel line of research. Either combined with other modalities or relying on sound stimuli alone, they may provide an interface even for users, who have lost control over their gaze direction or lid closure. Furthermore, as BCI systems deliver "for free" an objective metric of the attention level of a listener, auditory BCIs might prove useful in clinical routines in the future. After a brief walk through the human auditory system, the lecture will review existing auditory BCI paradigms, with a special emphasis on the potential of spatially distributed stimuli, workload and usability. In the context of spelling applications, the lecture will explain, how auditory BCIs can be used practically.
[ presentation ]

Brain Computer interface (BCI) technology exploits a variety of brain signals to create new artificial channels which can provide people with a new way to interact with the external world. As such, BCI systems can operate external application devices that are intended to restore, replace, enhance and even improve brain function. Replacement and restoration of lost motor functions are the goals of most of the current BCI research development and application, with the ultimate aim to improve the quality of life of severely disabled people. More recently, BCI technology has attract attention as a potential tool to support functional rehabilitation after brain injury such as stroke, trauma… by offering an on-line feedback about brain signals associated with mental practice, motor intention/attempt, and thus helping to guide neuroplasticity to improve recovery. The aim of this lecture is to provide a state of art of the current BCI applications in different clinical fields, with a special emphasis on the role of the BCI outputs to access assistive technology devices and their role as a novel class of therapeutic methods. An overview of the crucial issues to translate BCIs from the lab to the real scenario usage will close the lecture.
[ Presentation not (yet) available, since it includes unpublished material. ]

A variety of causal inference methods has been introduced to neuroimaging in recent years, including Causal Bayesian Networks, Dynamic Causal Modeling (DCM), Granger Causality, and Linear Non-Gaussian Acyclic Models (LINGAM). While all these methods aim to provide insights into how brain processes interact, they are based on rather different concepts of causality. In this talk, I will review the theoretical foundations of each of these methods, describe their inherent assumptions, and discuss the resulting consequences for the analysis and interpretation of neuroimaging data.
[ presentation ]

Modern theories of cognition emphasize the role of bodily interactions with the environment. Eye movements are a prime example of such an intimate relation of sensory processing and motor behavior. In their lifetime humans perform more eye movements than any other type of behavior. Hence, they provide a unique window for observation of cognitive processes. Fueled by recent technological and algorithmic advances the combination of electrophysiological methods (EEG, MEG) with the study of eye movements the investigation of computational properties and physiological mechanisms of the control of eye movements has moved into the focus of research interest. In fact, eye movements can be predicted to a substantial degree based on the concept of salience maps incorporating low level image properties. This is maintained across repeated presentation of identical stimuli. Importantly, manipulating image properties reveals that this predictive power is at least in part a true causal mechanism. Yet, fMRI and clinical studies show that the physiological substrate is not located in early visual cortex, but higher-level areas. This is compatible with the observation of the emotions' Impact on Viewing Behavior under Natural Conditions. Finally, we can demonstrate that overt visual attention is a causal factor of perceptual awareness. In sum, these studies contribute to the pragmatic turn in cognitive science and advocate an embodied view of cognition.
[ presentation ]

EEG and MEG measure brain electrical activity indirectly from outside the head, where each sensor measures a superposition of activity from the entire brain/cortex rather than only from its closest sourrounding. This limits the signal-to-noise ratio (SNR) of the measurements and prohibits the straightforward localization of the underlying brain activity. To perform localization, the physical mapping from brain electrical activity to EEG potentials/MEG magnetic fields has to be reversed, which is only possible using prior knowledge on the properties of the sources. A different approach to recovering EEG/MEG source activity is statistical source separation. Here, the data are factorized into source time series (components) and their corresponding static EEG potential/MEG field maps (patterns) based on assumptions such as mutual independence or class discriminability of the source time series. Although no physical model is employed in source separation methods, each component can be localized in a subsequent step. We will review established inverse source reconstruction and source separation algorithms employing various assumptions on the number of active sources, the spatial structure and the temporal dynamics of the source activity.
[ presentation ]

Like other new and promising developments in scientific research, neurotechnologies like Brain-Computer Interfacing (BCI) and Deep Brain Stimulation (DBS) provide cause for considering their potential philosophical, ethical and societal consequences. Especially over the last few years, there has been an enormous growth in publications that examine the exploration and application of neurotechnology with human value systems. I will review some of the issues related to BCI and DBS, focusing on personal identity, agency and mental competence, as these can be of great relevance to an individual’s moral and legal responsibility for decisions and actions.
[ presentation ]

The combination of multiple neuroimaging modalities has become an important field of research. While the technical challenges associated with multimodal neuroimaging have been mastered more than a decade ago, analysis techniques for multimodal neuroimaging data are still being developed. This tutorial will cover data driven analysis techniques for multimodal neuroimaging, including recent advances in multimodal brain-computer-interfaces and in integration of neural bandpower signals with hemodynamic signals. A special focus will be placed on simple and efficient subspace methods that are useful in all stages of multimodal neuroimaging analyses, starting from basic preprocessing and artifact removal to integration of multiple modalities with complex spatiotemporal coupling dynamics.
[ presentation ]

EEG is a highly complex signal. One of the main challenges of EEG analysis is to robustify against artifacts, non-stationarities and task unrelated variability. This holds in particular for EEG experiments outside a controlled lab environment. The tutorial will report on a broad line of algorithmic research to analyse and compensate non-stationarity in EEG thriving for more robust analysis.
[ presentation ]