Brain-Computer Interfaces 1: Methods and Perspectives

Brain–computer interfaces (BCI) are devices which measure brain activity and translate it into messages or commands, thereby opening up many investigation and application possibilities. This book provides keys for understanding and designing these multi-disciplinary interfaces, which require many fields of expertise such as neuroscience, statistics, informatics and psychology. This first volume, Methods and Perspectives, presents all the basic knowledge underlying the working principles of BCI. It opens with the anatomical and physiological organization of the brain, followed by the brain activity involved in BCI, and following with information extraction, which involves signal processing and machine learning methods. BCI usage is then described, from the angle of human learning and human-machine interfaces. The basic notions developed in this reference book are intended to be accessible to all readers interested in BCI, whatever their background. More advanced material is also offered, for readers who want to expand their knowledge in disciplinary fields underlying BCI. This first volume will be followed by a second volume, entitled Technology and Applications.

Table of Contents:
Foreword  xiii José DEL R. MILLAN Introduction  xv Maureen CLERC, Laurent BOUGRAIN and Fabien LOTTE Part 1. Anatomy and Physiology  1 Chapter 1. Anatomy of the Nervous System 3 Matthieu KANDEL and Maude TOLLET 1.1. General description of the nervous system 4 1.2. The central nervous system  5 1.2.1. The telencephalon 6 1.2.2. The diencephalon 10 1.2.3. The brain stem 12 1.3. The cerebellum 14 1.4. The spinal cord and its roots 15 1.5. The peripheral nervous system  18 1.5.1. Nerves 18 1.5.2. General organization of the PNS 19 1.5.3. The autonomic nervous system  20 1.6. Some syndromes and pathologies targeted by Brain–Computer Interfaces  21 1.6.1. Motor syndromes 21 1.6.2. Some pathologies that may be treated with BCIs  22 1.7. Conclusions  23 1.8. Bibliography  24 Chapter 2. Functional Neuroimaging 25 Christian BÉNAR 2.1. Functional MRI  26 2.1.1. Basic principles of MRI  26 2.1.2. Principles of fMRI 26 2.1.3. Statistical data analysis: the linear model  27 2.1.4. Independent component analysis 29 2.1.5. Connectivity measures  30 2.2. Electrophysiology: EEG and MEG 31 2.2.1. Basic principles of signal generation  31 2.2.2. Event-related potentials and fields  31 2.2.3. Source localization  32 2.2.4. Independent component analysis 34 2.2.5. Time–frequency analysis 34 2.2.6. Connectivity  35 2.2.7. Statistical analysis 36 2.3. Simultaneous EEG-fMRI 37 2.3.1. Basic principles  37 2.3.2. Applications and data analysis  37 2.3.3. Connections between EEG and fMRI  38 2.4. Discussion and outlook for the future  38 2.5. Bibliography  40 Chapter 3. Cerebral Electrogenesis  45 Franck VIDAL 3.1. Electrical neuronal activity detected in EEG  45 3.1.1. Action and postsynaptic potentials  46 3.1.2. Resting potential, electrochemical gradient and PSPs 47 3.1.3. From PSPs to EEG  48 3.2. Dipolar and quadrupole fields  51 3.2.1. Field created by an ion current due to the opening of ion channels  51 3.2.2. Factors determining the value of the potential created by an ion current 56 3.3. The importance of geometry 57 3.3.1. Spatial summation, closed fields and open fields  57 3.3.2. Effect of synapse position on the polarity of EEG 60 3.3.3. Effect of active areas’ position  61 3.4. The influence of conductive media  62 3.4.1. Influence of glial cells 62 3.4.2. Influence of skull bones  63 3.5. Conclusions  64 3.6. Bibliography  64 Chapter 4. Physiological Markers for Controlling Active and Reactive BCIs  67 François CABESTAING and Philippe DERAMBURE 4.1. Introduction  67 4.2. Markers that enable active interface control 72 4.2.1. Spatiotemporal variations in potential  72 4.2.2. Spatiotemporal wave variations 74 4.3. Markers that make it possible to control reactive interfaces  77 4.3.1. Sensory evoked potentials  77 4.3.2. Endogenous P300 potential  80 4.4. Conclusions  81 4.5. Bibliography  82 Chapter 5. Neurophysiological Markers for Passive Brain–Computer Interfaces  85 Raphaëlle N. ROY and Jérémy FREY 5.1. Passive BCI and mental states  85 5.1.1. Passive BCI: definition  85 5.1.2. The notion of mental states  86 5.1.3. General categories of neurophysiological markers 87 5.2. Cognitive load 87 5.2.1. Definition  87 5.2.2. Behavioral markers  87 5.2.3. EEG markers  87 5.2.4. Application example: air traffic control 88 5.3. Mental fatigue and vigilance 89 5.3.1. Definition  89 5.3.2. Behavioral markers  89 5.3.3. EEG markers  89 5.3.4. Application example: driving  90 5.4. Attention  90 5.4.1. Definition  90 5.4.2. Behavioral markers  91 5.4.3. EEG markers  91 5.4.4. Application example: teaching  92 5.5. Error detection 92 5.5.1. Definition  92 5.5.2. Behavioral markers  92 5.5.3. EEG markers  93 5.5.4. Application example: tactile and robotic interfaces  93 5.6. Emotions  94 5.6.1. Definition  94 5.6.2. Behavioral markers  94 5.6.3. EEG markers  94 5.6.4. Application example: communication and personal development  95 5.7. Conclusions  96 5.8. Bibliography  96 Part 2. Signal Processing and Machine Learning  101 Chapter 6. Electroencephalography Data Preprocessing  103 Maureen CLERC 6.1. Introduction  103 6.2. Principles of EEG acquisition  104 6.2.1. Montage  104 6.2.2. Sampling and quantification 105 6.3. Temporal representation and segmentation 105 6.3.1. Segmentation  106 6.3.2. Time domain preprocessing 106 6.4. Frequency representation 107 6.4.1. Fourier transform 107 6.4.2. Frequency filtering  108 6.5. Time–frequency representations 109 6.5.1. Time–frequency atom 109 6.5.2. Short-time Fourier transform 111 6.5.3. Wavelet transform 112 6.5.4. Time–frequency transforms of discrete signals 114 6.5.5. Toward other redundant representations 114 6.6. Spatial representations  115 6.6.1. Topographic representations 115 6.6.2. Spatial filtering 116 6.6.3. Source reconstruction 118 6.6.4. Using spatial representations in BCI 120 6.7. Statistical representations 121 6.7.1. Principal component analysis 121 6.7.2. Independent component analysis 122 6.7.3. Using statistical representations in BCI 122 6.8. Conclusions  123 6.9. Bibliography  124 Chapter 7. EEG Feature Extraction 127 Fabien LOTTE and Marco CONGEDO 7.1. Introduction  127 7.2. Feature extraction 127 7.3. Feature extraction for BCIs employing oscillatory activity  130 7.3.1. Basic design for BCI using oscillatory activity 130 7.3.2. Toward more advanced, multiple electrode BCIs  131 7.3.3. The CSP algorithm  133 7.3.4. Illustration on real data  135 7.4. Feature extraction for the BCIs employing EPs 137 7.4.1. Spatial filtering for BCIs employing EPs  138 7.5. Alternative methods and the Riemannian geometry approach 139 7.6. Conclusions  141 7.7. Bibliography  142 Chapter 8. Analysis of Extracellular Recordings  145 Christophe POUZAT 8.1. Introduction  145 8.1.1. Why is recording neuronal populations desirable? 146 8.1.2. How can neuronal populations be recorded?  146 8.1.3. The properties of extracellular data and the necessity of spike sorting  147 8.2. The origin of the signal and its consequences  148 8.2.1. Relationship between current and potential in a homogeneous medium  148 8.2.2. Relationship between the derivatives of the membrane potential and the transmembrane current  150 8.2.3. “From electrodes to tetrodes”  154 8.3. Spike sorting: a chronological presentation 155 8.3.1. Naked eye sorting 155 8.3.2. Window discriminator (1963)  155 8.3.3. Template matching (1964)  156 8.3.4. Dimension reduction and clustering (1965) 157 8.3.5. Principal component analysis (1968)  158 8.3.6. Resolving superposition (1972) 160 8.3.7. Dynamic amplitude profiles of action potentials (1973)  161 8.3.8. Optimal filters (1975) 162 8.3.9. Stereotrodes and amplitude ratios (1983)  165 8.3.10. Sampling jitter (1984)  168 8.3.11. Graphical tools  170 8.3.12. Automatic clustering 171 8.4. Recommendations 179 8.5. Bibliography  181 Chapter 9. Statistical Learning for BCIs 185 Rémi FLAMARY, Alain RAKOTOMAMONJY and Michèle SEBAG 9.1. Supervised statistical learning  185 9.1.1. Training data and the predictor function 186 9.1.2. Empirical risk and regularization 187 9.1.3. Classical methods of classification  190 9.2. Specific training methods 192 9.2.1. Selection of variables and sensors  192 9.2.2. Multisubject learning, information transfer 194 9.3. Performance metrics  194 9.3.1. Classification performance metrics 195 9.3.2. Regression performance metrics 196 9.4. Validation and model selection  197 9.4.1. Estimation of the performance metric  197 9.4.2. Optimization of hyperparameters  200 9.5. Conclusions  202 9.6. Bibliography 202 Part 3. Human Learning and Human–Machine Interaction 207 Chapter 10. Adaptive Methods in Machine Learning 209 Maureen CLERC, Emmanuel DAUCÉ and Jérémie MATTOUT 10.1. The primary sources of variability 209 10.1.1. Intrasubject variability  210 10.1.2. Intersubject variability  211 10.2. Adaptation framework for BCIs  213 10.3. Adaptive statistical decoding  214 10.3.1. Covariate shift  214 10.3.2. Classifier adaptation 216 10.3.3. Subject-adapted calibration 218 10.3.4. Optimal tasks 219 10.3.5. Correspondence between task and command 221 10.4. Generative model and adaptation  221 10.4.1. Bayesian approach  221 10.4.2. Sequential decision  224 10.4.3. Online optimization of stimulations  226 10.5. Conclusions  229 10.6. Bibliography 229 Chapter 11. Human Learning for Brain–Computer Interfaces  233 Camille JEUNET, Fabien LOTTE and Bernard N’KAOUA 11.1. Introduction  233 11.2. Illustration: two historical BCI protocols 235 11.3. Limitations of standard protocols used for BCIs 237 11.4. State-of-the-art in BCI learning protocols 238 11.4.1. Instructions  238 11.4.2. Training tasks 239 11.4.3. Feedback 239 11.4.4. Learning environment  242 11.4.5. In summary: guidelines for designing more effective training protocols 243 11.5. Perspectives: toward user-adapted and user-adaptable learning protocols  244 11.6. Conclusions  247 11.7. Bibliography 247 Chapter 12. Brain–Computer Interfaces for Human–Computer Interaction  251 Andéol EVAIN, Nicolas ROUSSEL, Géry CASIEZ, Fernando ARGELAGUET-SANZ and Anatole LÉCUYER 12.1. A brief introduction to human–computer interaction 251 12.1.1. Interactive systems, interface and interaction 252 12.1.2. Elementary tasks and interaction techniques 252 12.1.3. Theory of action feedback  253 12.1.4. Usability 254 12.2. Properties of BCIs from the perspective of HCI  255 12.3. Which pattern for which task? 257 12.4. Paradigms of interaction for BCIs 259 12.4.1. BCI interaction loop 259 12.4.2. Main paradigms of interaction for BCIs  260 12.5. Conclusions  265 12.6. Bibliography 266 Chapter 13. Brain Training with Neurofeedback  271 Lorraine PERRONNET, Anatole LÉCUYER, Fabien LOTTE, Maureen CLERC and Christian BARILLOT 13.1. Introduction  271 13.2. How does it work?  274 13.2.1. Design of an NF training program 274 13.2.2. Course of an NF session: where the eyes “look” at the brain 275 13.2.3. A learning procedure that we still do not fully understand 276 13.3. Fifty years of history 278 13.3.1. A premature infatuation 278 13.3.2. Diversification of approaches  279 13.4. Where NF meets BCI  281 13.5. Applications 283 13.6. Conclusions  287 13.7. Bibliography 288 List of Authors  293 Index 295 Contents of Volume 2 299

About the Author :
Maureen Clerc is Senior Researcher at Inria Sophia Antipolis, France. Laurent Bougrain is Assistant Professor at the University of Lorraine, France. Fabien Lotte is Junior Researcher at Inria Bordeaux, France.