Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Co-Planar Stereotaxic Atlas of the Human Brain—3-Dimensional Proportional System: An Approach to Cerebral Imaging, J. Talairach, P. Tournoux. Georg Thieme Verlag, New York (1988), 122 pp., 130 figs. DM 268

The power of feedback.

An overview of the self-organizing map algorithm, on which the papers in this issue are based, is presented in this article.

The Self-Organizing Map

In this paper we review classification algorithms used to design brain–computer interface (BCI) systems based on electroencephalography (EEG). We briefly present the commonly employed algorithms and describe their critical properties. Based on the literature, we compare them in terms of performance and provide guidelines to choose the suitable classification algorithm(s) for a specific BCI.

/pdf/a-review-of-classification-algorithms-for-eeg-based-brain-p1h22hb1wr.pdf

A review of classification algorithms for EEG-based brain–computer interfaces

This paper deals with pattern rejection strategies for self-paced brain-computer interfaces (BCI). First, it introduces two pattern rejection strategies not used yet for self-paced BCI design: 1) the rejection class (RC) strategy and 2) thresholds on reliability functions (TRF) based on the automatic multiple-threshold learning algorithm. Second, it compares several rejection strategies using several classifiers, on motor imagery data, in order to identify their most desirable properties. Results showed that nonlinear classifiers led to the most efficient self-paced BCI. Concerning the reject option, RC outperformed a specialized reject classifier which outperformed TRF. Overall, the best results were obtained using the RC reject option and non-linear classifiers such as a Gaussian support vector machine, a fuzzy inference system or a radial basis function network.

/pdf/pattern-rejection-strategies-for-the-design-of-self-paced-38n8agf4ik.pdf

Pattern rejection strategies for the design of self-paced EEG-based Brain-Computer Interfaces

While promising for many applications, Electroencephalography (EEG)-based Brain-Computer Interfaces (BCIs) are still scarcely used outside laboratories , due to a poor reliability. It is thus necessary to study and fix this reliability issue. Doing so requires to use appropriate reliability metrics to quantify both signal processing and user learning performances. So far, Classification Accuracy (CA) is the typical metric used for both aspects. However, we argue in this paper that CA is a poor metric to study how well users are learning to use the BCI. Indeed CA is notably unspecific, discrete, training data and classifier dependent, and as such may not always reflect successful EEG pattern self-modulation by the user. We thus propose new performance metrics to specifically measure how distinct and stable the EEG patterns produced by the user are. By re-analyzing EEG data with these metrics, we indeed confirm that CA may hide some learning effects or hide the user inability to self-modulate a given EEG pattern.

https://hal.archives-ouvertes.fr/hal-01519478/document

Online classification accuracy is a poor metric to study mental imagery-based bci user learning: an experimental demonstration and new metrics

Objective. Going adaptive is a major challenge for the field of Brain-Computer Interface (BCI). This entails a machine that optimally articulates inference about the user's intentions and its own actions. Adaptation can operate over several dimensions which calls for a generic and flexible framework. Approach. We appeal to one of the most comprehensive computational approach to brain (adaptive) functions: the Active Inference (AI) framework. It entails an explicit (probabilistic) model of the user that the machine interacts with, here involved in a P300-spelling task. This takes the form of a discrete input-output state-space model establishing the link between the machine's (i) observations – a P300 or Error Potential for instance, (ii) representations – of the user intentions to spell or pause, and (iii) actions – to flash, spell or switch-off the application. Main results. Using simulations with real EEG data from 18 subjects, results demonstrate the ability of AI to yield a significant increase in bit rate (17%) over state-of-the-art approaches, such as dynamic stopping. Significance. Thanks to its flexibility, this one model enables to implement optimal (dynamic) stopping but also optimal flashing (i.e. active sampling), automated error correction, and switching off when the user does not look at the screen anymore. Importantly, this approach enables the machine to flexibly arbitrate between all these possible actions. We demonstrate AI as a unifying and generic framework to implement a flexible interaction behaviour in a given BCI context.

/pdf/active-inference-as-a-unifying-generic-and-adaptive-1c8iqzt3lj.pdf

Active inference as a unifying, generic and adaptive framework for a P300-based BCI.

Although Mental Imagery based Brain-Computer Interfaces (MI-BCIs) seem to be very promising for many applications, they are still rarely used outside laboratories. This is partly due to suboptimal training protocols, which provide little help to users learning how to control the system. Indeed, they do not take into account recommendations from instructional design. However, it has been shown that MI-BCI performances are significantly correlated to certain aspects of the users' cognitive profile, such as their Spatial Abilities (SA). Thus, it remains to be elucidated whether training the SA of BCI users would also improve their BCI control performance. Therefore, we proposed and validated an SA training that aimed at being included in an MI-BCI training protocol. Our pre-studies indeed confirmed that such a training does increase people's SA abilities. We then conducted a pilot study with 3 participants, one with a standard MI-BCI training protocol, one with the proposed SA training integrated into a standard MI-BCI training, and another control integrating another training, here verbal comprehension tasks, into a standard MI-BCI training. While such a small population cannot lead to any strong result, our first results show that SA training can indeed be integrated into MI-BCI training and is thus worth being further investigated for BCI user training.

/pdf/towards-a-spatial-ability-training-to-improve-mental-imagery-13hj2md4t3.pdf

Towards a spatial ability training to improve Mental Imagery based Brain-Computer Interface (MI-BCI) performance: A Pilot study

Neurofeedback (NF) allows to exert self-regulation over specific aspects of one's own brain activity by returning information extracted in real-time from brain activity measures. These measures are usually acquired from a single modality, most commonly electroencephalography (EEG) or functional magnetic resonance imaging (fMRI). EEG-fMRI-neurofeedback (EEG-fMRI-NF) is a new approach that consists in providing a NF based simultaneously on EEG and fMRI signals. By exploiting the complementarity of these two modalities, EEG-fMRI-NF opens a new spectrum of possibilities for defining bimodal NF targets that could be more robust, flexible and effective than unimodal ones. Since EEG-fMRI-NF allows for a richer amount of information to be fed back, the question arises of how to represent the EEG and fMRI features simultaneously in order to allow the subject to achieve better self-regulation. In this work, we propose to represent EEG and fMRI features in a single bimodal feedback (integrated feedback). We introduce two integrated feedback strategies for EEG-fMRI-NF and compare their early effects on a motor imagery task with a between-group design. The BiDim group (n=10) was shown a two-dimensional (2D) feedback in which each dimension depicted the information from one modality. The UniDim group (n=10) was shown a one-dimensional (1D) feedback that integrated both types of information even further by merging them into one. Online fMRI activations were significantly higher in the UniDim group than in the BiDim group, which suggests that the 1D feedback is easier to control than the 2D feedback. However subjects from the BiDim group produced more specific BOLD activations with a notably stronger activation in the right superior parietal lobe (BiDim > UniDim, p < 0.001, uncorrected). These results suggest that the 2D feedback encourages subjects to explore their strategies to recruit more specific brain patterns. To summarize, our study shows that 1D and 2D integrated feedbacks are effective but also appear to be complementary and could therefore be used in a bimodal NF training program. Altogether, our study paves the way to novel integrated feedback strategies for the development of flexible and effective bimodal NF paradigms that fully exploits bimodal information and are adapted to clinical applications.

/pdf/learning-2-in-1-towards-integrated-eeg-fmri-neurofeedback-2o4re4oqa2.pdf

Fabien Lotte

Papers

Pattern rejection strategies for the design of self-paced EEG-based Brain-Computer Interfaces

Online classification accuracy is a poor metric to study mental imagery-based bci user learning: an experimental demonstration and new metrics

Active inference as a unifying, generic and adaptive framework for a P300-based BCI.

Towards a spatial ability training to improve Mental Imagery based Brain-Computer Interface (MI-BCI) performance: A Pilot study

Learning 2-in-1: Towards Integrated EEG-fMRI-Neurofeedback