Abstract:

All successful BCI systems rely on an efficient real-time feedback. For this
reason, the data processing methods must be also suitable for online and real-
time processing. This requires causal algorithms which can only use sample
values from the past and present but not from the future. Adaptive methods
typically fulfill this requirement, while minimizing also the time delay. The
data processing in BCI consists typically of two main steps, (i) signal process-
ing and feature extraction, and (ii) classification or feature translation.
It is the aim of this work to introduce adaptive methods for both
steps which are also closely related to two types of non-stationarities.

Abstract:

Brain-Computer Interfaces (BCIs) allow a user to control a computer application
by brain activity as acquired, e.g., by EEG. One of the biggest challenges in BCI research
is to understand and solve the problem of “BCI Illiteracy”, which is that BCI control does
not work for a non-negligible portion of users (estimated 15% to 30%). Here, we investigate
the illiteracy problem in BCI systems which are based on the modulation of sensorimotor
rhythms. In this paper, a sophisticated adaptation scheme is presented which guides the user
from an initial subject-independent classifier that operates on simple features to a subject-optimized
state-of-the-art classifier within one session while the user interacts the whole
time with the same feedback application and does not need to care about what is going
on behind the scenes. While initial runs use supervised adaptation methods for robust coadaptive
learning of user and machine, final runs use unsupervised adaptation and therefore
provide an unbiased measure of BCI performance. Using this approach, which does not
involve any offline calibration measurement, good performance was obtained by good BCI
participants (also one novice) after 3-6 minutes of adaptation. More importantly, the use of
machine learning techniques allowed users who were unable to achieve successful feedback
before to gain significant control over the BCI system. In particular, one participant had no
peak of the sensory motor idle rhythm in the beginning of the experiment, but could develop
such peak during the course of the session (and use voluntary modulation of its amplitude
to control the feedback application).

Abstract:

Several feature types have been used with EEG-based Brain Computer Interfaces.
Among the most popular are logarithmic band-power estimates with more or less
subject-specific optimization of the frequency bands. In this paper we introduce
a feature called Time Domain Parameter, that is defined by the generalization of
the Hjorth parameters. Time Domain Parameters are studied under two different
conditions. The first setting is defined when no data from a subject is available.
In this condition our results show that Time Domain Parameters outperform all
band power features tested with all spatial filters applied. The second setting is
the transition from calibration (no feedback) to feedback, in which the frequency
content of the signals can change for some subjects. We compare Time Domain
Parameters with logarithmic band power in subject-specific bands and show that
these features are advantageous in this situation as well.

Abstract:

Designing user interfaces which can cope with unconventional control properties is challenging, and conventional interface design techniques are of little help. This paper examines how interactions can be designed to explicitly take into account the uncertainty and dynamics of control inputs. In particular, the asymmetry of feedback and control channels is highlighted as a key design constraint, which is especially obvious in current noninvasive brain-computer interfaces. Brain-computer interfaces (BCIs) are systems capable of decoding neural activity in real time, thereby allowing a computer application to be directly controlled by thought. BCIs, however have totally different signal properties than most conventional interaction devices. Bandwidth is very limited and there are comparatively long and unpredictable delays. Such interfaces cannot simply be treated as unwieldy
mice. In this respect they are an example of a growing field of sensor-based interfaces which have unorthodox control properties. As a concrete example, we present the text entry application ‘Hex-o-Spell’, controlled via motor-imagery based electroencephalography (EEG).
The system utilises the high visual display bandwidth to help compensate for the limited control signals, where the timing of the state changes encodes most of the information. We present results showing the comparatively high performance of this interface, with entry rates exceeding seven characters per minute.

Abstract:

One basic rationale for Brain-Computer Interfaces (BCIs) is to enable severely paretic persons to interact again with their environment. While advancements of BCI techniques are significant in healthy volunteers, there are only few studies that investigated the applicability of BCIs in patients afflicted by spinal cord injury (SCI), and the spatiotemporal characteristics of sensorimotor cortical event-related potentials in these subjects is largely unknown. In this study we evaluated the feasibility and performance rate of the Berlin Brain-Computer Interface in a first-session setting in high-level SCI with tetraplegia.
In a one-dimensional online feedback four out of seven subjects were were able to control the BCI via attempted movements with their plegic limbs during the first session with a mean accuracy of 75%. Interestingly, subjects achieved an even higher performance rate of about 83 % (range: 74-95%) in a ‘cursor off’ mode, in which the feedback signal was provided only at the end of each trial. In contrast to a previous SCI-BCI study, topographical and temporal patterns of event related desynchronizations (ERDs) in the µ- and beta-frequency bands were well distinguishable in these patients.

Abstract:

The analysis of EEG signals for the use in BCI systems and for mental state monitoring applications is often impeded by artifacts caused by muscular activity or external technical sources. A promising approach for the reduction or removal of artifacts is based on methods of Blind Source Separation (BSS), which transform the original EEG signal into independent source components. In order to avoid the time-consuming hand rating of sources into artifactual and non-artifactual components, an automated method for their classification is proposed. Applying state of the art machine learning algorithms and nonlinear classification with a Support Vector Machine (SVM), the automated method shows a high level of agreement (90.5%) on unseen data with ratings of human experts.

Abstract:

Most P300 BCI approaches use the visual modality for stimulation. For use with ALS patients this might not be the preferable choice because of sight deterioration. Moreover, using a modality different from the visual one minimizes interference with possible visual feedback. Therefore, a multi-class brain-computer interface paradigm is proposed that uses spatially distributed, auditive cues. Ten subjects participated in an offline oddball task with the spatial location of the stimuli being a discriminating cue. Different inter-stimulus intervals of 1000 ms, 300 ms and 175 ms were tested. With averaging over multiple classifier outputs, selection scores went over 90% for most conditions; two subjects reached a 100% correct score. Corresponding information transfer rates were high, up to an average optimal score of 20.99 bits/minute for the 175 ms condition (best subject 37.80 bits/minute). We conclude that the proposed paradigm is successful for healthy subjects and shows promising results that may lead to a fast BCI that solely relies on the auditory sense.

Abstract:

Although the P300 spelling grid has been modestly successful even for patients, the visual flashes that are needed are not accessible for those with vision deterioration. For these potential users, tactile or auditory stimulation are the modalities of choice, but they are mostly binary in nature. In [1,2] we presented the offline results for a new auditory multi-class paradigm to overcome some of these difficulties. This paradigm is now ready to be used online and was successfully operated by two subjects to write short words.

Abstract:

Electroencephalographic signals are known to be
non-stationary and easily affected by artifacts, therefore their
analysis requires methods that can deal with noise. In this
work we present two ways of robustifying common spatial
patterns under a maxmin approach. The worst-case objective
function is optimized within a prefixed set of the covariance
matrices that is defined either very simply as identity matrices
or in a data driven way using PCA. We test common spatial
filters derived with these two approaches with real world
brain-computer interface (BCI) data sets in which we expect
substantial fluctuations caused by day-to-day (session transfer
problem). We compare our results with the classical common
spatial filters and show that both significantly improve the
performance of the latter.

Abstract:

EEG single-trial analysis requires methods that are robust against noise and disturbance. In this contribution, based on the framework of robust statistics, we propose a simple modification of Common Spatial Patterns by robustifying covariance estimators against outlying trials caused, for example, by artifacts. We tested the proposed robust filters with EEG recordings from 80 subjects and obtained, not only a significant improvement in performance, but for some subjects also better neuro-physiologically interpretable filters.

Abstract:

Brain-Computer Interfaces (BCIs) allow a user to control a computer application by brain
activity as acquired, e.g., by EEG. One of the biggest challenges in BCI research is to understand and
solve the problem of ‘BCI Illiteracy’, which is that BCI control does not work for a non-negligible
portion of subjects (estimated 15% to 30%). Here, we investigate the illiteracy problem in BCI systems
which are based on the modulation of sensorimotor rhythms. In this paper, a sophisticated adaptation
scheme is presented which guides the user from an initial subject-independent classifier that operates
on simple features to a subject-optimized state-of-the-art classifier within one session while the user
interacts the whole time with the same feedback application and does not need to care about what is
going on behind the scenes. While initial runs use supervised adaptation methods for robust co-
adaptive learning of user and machine, final runs use unsupervised adaptation and therefore provide an
unbiased measure of BCI performance. Using this approach, which does not involve any offline
calibration measurement, good performance was obtained by good BCI subjects (also one novice) after
3-6 minutes of adaptation. More importantly, the use of machine learning techniques allowed subjects
who were unable to achieve successful feedback before to gain significant control over the BCI system.
In particular, one subject had no peak of the sensory motor idle rhythm in the beginning of the
experiment, but could develop such peak during the course of the session (and use voluntary
modulation of its amplitude to control the feedback application).

Abstract:

Compared to invasive Brain-Computer Interfaces (BCI), non-invasive BCI systems based on Electroencephalogram (EEG) signals have not been applied successfully for precisely timed control tasks.
In the present study, however, we demonstrate and report on the interaction of subjects with a real device: a pinball machine. Results of this study clearly show that fast and well-timed control well beyond chance level is possible, even though the environment is extremely rich and requires precisely timed and complex predictive behavior. Using machine learning methods for mental state decoding, BCI-based pinball control is possible within the first session without the necessity to employ lengthy subject training. The current study shows clearly that very compelling control with excellent timing and dynamics is possible for a non-invasive BCI.

Abstract:

This paper will describe the outcomes of a study into the state-of-the art of user needs and user requirements of assistive applications for people with disabilities. The study was carried out within the framework of the TOBI (Tools for Brain Computer Interaction) project, a project (2008-2011) funded under the 7th Framework Programme of the EC. The project has the aim to develop practical technology for brain-computer interaction that will improve the quality of life of disabled people and the effectiveness of rehabilitation. TOBI includes leading European groups in BCI, human-computer interaction, intelligent robotics, and applied assistive technologies. It also includes rehabilitation clinics and commercial providers of assistive technology.

Non-invasive BCI are based on electroencephalogram (EEG) signals. The EEG is recorded through
electrodes placed on the user’s head. This technology is not invasive and only records the electrical
activity of the brain without interfering with it. TOBI is expected to have an impact by broadening the appropriate use of BCI assistive technology, by incorporating adaptive capabilities that augment those other assistive technologies they are combined with.
After a pre clinical validation the BCI based assistive solutions will be tested and evaluated in real life situations by different populations of end-users.

As a first step the project has identified potential users of BCI as people with severe motor disabilities. General inclusion and exclusion criteria have been identified to access in the future to the test phase, as well as specific inclusion/exclusion criteria for the different application areas Communication and Control, Entertainment, Motor substitution and Motor recovery. To favour the involvement of people with disabilities and to have a sound basis for the definition of user requirements based on user needs and experiences, a questionnaire has been designed to assess different aspects of the current situation of the respondents, their level of satisfaction with current solutions for their independence, their needs and preferences regarding technology based assistive solutions. Areas of independence explored through the questionnaire have included, among other, mobility, communication, computer access, environmental control. The assessment of important aspects in the evaluation of assistive technology was based on aspects such as: functionality, easiness of use, fatigue, design, comfort, impact on environment, etc.

The questionnaire has been delivered to a sample of 50 people across four European countries. The paper will focus specifically on user requirements for the BCI applications areas Communication, Entertainment, Environmental control. It will present the results of the survey, cross analysed with some person related factors, and discuss the impact of these outcomes on the user requirements for BCI applications as formulated within the consortium. Also the review of these outcomes by of teams of AT experts in leading European AT centres will be covered by the paper. Finally some comments will be made on how these user requirements have impacted on the evaluation activities as planned once the prototypes are available.

This work is supported by the European ICT Programme Project FP7-224631. This paper only reflects the authors' views and funding agencies are not liable for any use that may be made of the information contained herein.

Abstract:

EEG single-trial analysis requires methods that are robust with respect to noise, artifacts and nonstationarity among other problems. This work contributes by developing a minimax approach to robustify the common spatial patterns (CSP) algorithm. By optimizing the worst-case objective function within a prefixed set of the covariance matrices , we can transform the respective complex mathematical program into a simple generalized eigenvalue problem and thus obtain robust spatial filters very efficiently. We test our minimax CSP method with real world brain-computer interface (BCI) data sets in which we expect substantial fluctuations caused by day-to-day or paradigm-to-paradigm variability or different forms of stimuli. The results clearly show that the proposed method significantly improves the classical CSP approach in multiple BCI scenarios.

Abstract:

Implementation of Error Detection into the Graz-Brain-Computer Interface, The Interaction Error Potential
Alex Kreilinger, Christa Neuper, Gert Pfurtscheller, Gernot R. Müller-Putz
Graz University of Technology, Institute for Knowledge Discovery

Brain-Computer Interfaces (BCIs) are possible remaining means of communication for people with severe paralyses or who are in a locked-in state. The field of BCI-applications reaches from communication [1] to neuroprostheses [5] and wheelchair control [4]. However, there are still some severe restrictions for BCIs when compared to other assistive devices. Main problems are a low accuracy and an insufficient performance speed. A possible way to improve BCIs is the detection of errors after incorrect events in the EEG. Error correction can inhibit wrong classifications and hence is very useful to increase the accuracy and therewith also the bit rate. After execution or observation of false events characteristic waveforms (error potentials (ErrPs)) of the EEG over the location of the anterior cingulate cortex (ACC) can be recorded and adopted to detect errors. One of the most promising kind of ErrPs is the interaction ErrP [2,3] which can be measured after users observe errors committed by an interface that should translate their commands correctly.
This paper describes the recording of interaction ErrPs as a verification of previous studies dealing with this subject [2,3]. Here, the interaction ErrP is described as a reliable source to evaluate errors and to increase the precision of BCIs. The goal of this study was to retrace these findings and to additionally reduce the complexity of the recording setup, meaning to reduce the number of electrodes and keeping data processing as simple as possible. Under these conditions the error detection was implemented into the Graz-BCI-system based on motor imagery (MI) [6].
In summary, the findings about interaction ErrPs could be confirmed (in terms of the recorded waveforms). Furthermore, the detection rates for single trial error detection exceeded 70%:24% (true positive:false positive) in offline analysis, averaged over 13 subjects. After offline analysis online experiments were conducted that could not show the anticipated error detection rates but could still improve the bit rate significantly. This indicates that even with a limitation of complexity the error correction can improve the accuracy of BCI-systems, especially for people who only achieve a poor performance without error correction.


"This work is supported by the European ICT Programme Project FP7-224631. This paper only reflects the authors' views and funding agencies are not liable for any use that may be made of the information contained herein."

[1] N. Birbaumer, N. Ghanayim, T. Hinterberger, I. Iversen, B. Kotchoubey, A. Kübler, J. Perelmouter,
E. Taub, and H. Flor. A spelling device for the paralysed. Nature, 398:297–298, 1999.
[2] PW Ferrez. Error-related EEG potentials in brain-computer interfaces. PhD thesis, Ecole Polytechnique
Federale de Lausanne, 2007.
[3] PW Ferrez and J del R Millán. Error-related EEG potentials generated during simulated braincomputer
interaction. IEEE Transactions on Bio-Medical Engineering, 55(3):923–929, 2008.
[4] F Galán,MNuttin, E Lew, PW Ferrez, G Vanacker, J Philips, and J del R Millán. A brain-actuated
wheelchair: asynchronous and non-invasive brain-computer interfaces for continuous control of
robots. Clinical Neurophysiology, 119:2159–2169, 2008.
[5] GR Müller-Putz, R Scherer, G Pfurtscheller, and R Rupp. Brain-computer interfaces for control
of neuroprostheses: from synchronous to asynchronous mode of operation. Biomedical Engineering,
51(2):57–63, 2006.
[6] C Neuper, GR Müller-Putz, R Scherer, and G Pfurtscheller. Motor imagery and EEG-based
control of spelling devices and neuroprostheses. Progress in Brain Research, 159:393–409, 2006.

Abstract:

P300 Brainpainting: Evaluation of a novel BCI application with ALS patients and healthy controls

J Münßinger1, S Halder1, S Kleih1, A Furdea1, V Raco1, A Hösle2, A Kübler1,3
1 Institute of Medical Psychology and Behavioural Neurobiology, University of Tübingen, Tübingen, Germany
2 Babenhausen, Germany
3 Department of Psychology I, Biological Psychology, Clinical Psychology, and Psychotherapy , University of Würzburg, Würzburg, Germany

To date, brain-computer interfaces (BCIs) are primarily used to enable (completely) paralyzed patients to communicate. However, these applications do not allow them to communicate in a creative manner; which many amyotrophic lateral sclerosis (ALS) patients would consider an increase of their quality of life. The current P300-Brainpainting application is intended to enable the patients to express themselves creatively by means of painting pictures only using their brain activity.

Abstract:

Validation of SMR BCI Performance Categorization Using fMRI

S Halder 1,4, D Agorastos 1, R Veit 1, B Blankertz 2,5, T Dickhaus 2, E Hammer 1, S Kleih 1, S Lee 1, C Sannelli 2, B Varkuti 1, A Kübler 1,3
1 Institute of Medical Psychology and Behavioural Neurobiology, University of Tübingen, 72074Tübingen, Germany
2 Machine Learning Laboratory, Berlin Institute of Technology, Franklinstr. 28/29 10587 Berlin, Germany
3 Department of Psychology I, University of Würzburg, Marcusstr. 9-11, 97072 Würzburg, Germany
4 Wilhelm-Schickard Institute for Computer Engineering, University of Tübingen, Sand 13, 72076 Tübingen, Germany
5 Fraunhofer FIRST (IDA), Kekuléstr. 7, 12489 Berlin, Germany.

Brain-Computer Interfaces (BCIs) enable paralyzed people to communicate with their environment. Differences in performance between users and sessions remain largely unexplained, as does the question as to why communication in the complete locked-in-state (CLIS) has not been possible. A reliable performance indicator would allow an analysis of subject-to-subject and session-to-session performance differences and serve as and indicator of the capacity to use a BCI during the progression of the disease. BCIs based on modulation of the sensory motor rhythms (SMR) require several training sessions or at least a 30 minute calibration period, depending on the design of the system. Additionally the preparation of a high-density electroencephalography (EEG) cap takes a comparable amount of time. Therefore, indicators allowing us a fast screening of healthy participants or the selection of suitable training programs are particularly important.

Abstract:

Introduction
The EU-project “Tools for Brain-Computer Interaction” (TOBI) aims at developing practical technology for brain-computer interaction, i.e., non-invasive brain-computer interfaces (BCI) combined with other assistive technologies (AT) that will improve the quality of life of disabled people. An important concern of TOBI is the integration and participation of people with disabilities from the very beginning of the project.
The aim of the study was to define firstly, the needs (a person’s wants and necessities with respect to different aspects of independence) and secondly, the requirements (instrumental needs demanding specific functions/characteristics from the product or solution) of potential users with regard to assistive applications based on BCI.

Material
The TOBI members developed a questionnaire to assess i) satisfaction with current AT solutions in the areas of manipulation, mobility, communication, computer access, and environmental adaptation/control, ii) the three most important areas where participants wanted to improve their independence, iii) whether participants had independent access and used AT for access to devices for communication and entertainment as well as their desire to gain access, iv) the overall satisfaction with the current AT solutions. Variables were rated on a ten-point Likert scale (1=“not at all satisfied”, 10=“absolutely satisfied”). The importance of various aspects of AT was assessed on a four-point Likert scale from “very important” (4) to “not at all important” (1).

Participants
Participants (N = 77, 23 female) were from Austria (12%), Italy (41%), and Germany (47%) and diagnosed with spinal cord injury (37%), neurological/neuromuscular diseases (47%), or cerebrovascular disorders (16%). The majority of the participants (69%) was almost or completely tetraplegic.

Results
Overall satisfaction (M=7.12) and satisfaction in the different areas of independence was high. However, 16% (communication) to 30% (manipulation) were dissatisfied with their current solutions. Lowest dissatisfaction ratings were found for mobility aids (8%).
The majority of the participants had independent access to different devices for communication and entertainment. However, depending on the device 10% to 22% would have liked to have access to e-media and even more would have liked to use AT for access to the different devices. Participants who were impaired most severely (tetraplegic and only one channel to control AT´s) were in a worse situation. Sixteen to 63% of these participants had no independent access and wished to use AT to get access to the different devices.
“Mobility” (52%) was the aspect of life in which the majority of the participants wanted to improve their independence followed by “activities of daily living” (46%) and “occupation/ employment” (33%). Participants who used communication aids had needs, which were partially different from those of the rest of the participants. They wanted to improve their independence in personal expression (32%) and social interaction (24%).
Considering the adoption of a new AT solution, participants rated “functionality” (M=3.74) as the most important aspect followed by “possibility of independent use” (M=3.67) and “easiness of use” (M=3.60).

Conclusions
There is the need for better or/and alternative AT solutions in the areas where BCI can contribute with applications for manipulation, communication, and environmental control/entertainment.
The main lesson for TOBI is: (1) To develop simple (easiness of use) and effective (functional/ robust) BCI applications. (2) If communication aids are needed, to provide devices which enable people to communicate their thoughts and wishes and support their interaction with significant others. (3) To provide AT solutions with which users are as independent as possible from external support.

Abstract:

1 Introduction
A non-invasive brain-computer interface (BCI) is a system that translates user's intent, coded by spatiotemporal neural activity (usually EEG), into a control signal without using activity of any muscles or peripheral nerves. The user's involvement in current BCI systems is highly demanding in terms of cognitive attention and effort, since he/she needs to continuously generate mental commands for the brain-actuated device. In contrast, we discuss a new brain interaction approach where the user only monitors the performance of a semi-autonomous system. In this approach, the system carries out its task automatically (e.g., a wheelchair navigating autonomously) and occasionally receives corrective signals derived from the user's EEG whenever the operator wishes to deliver key decisions or improve the system's performance. In particular, we are interested in exploring the use of event-related potentials related to error detection (ErrP) and slow cortical potentials related to the anticipation of future events (i.e., contingent negative variation, CNV). These signals can replace or complement asynchronous control commands issued by traditional BCI [Gallan et al., 2008, for example]. The representation of this approach is illustrated in Figure 1.

2 Cognitive Monitoring
Several studies have shown error related EEG patterns elicited by subject-generated errors , erroneous feedback or errors generated during Brain-machine interaction. We propose to use these signals to provide a more natural mean of interaction for BCIs. Under this framework ErrPs are used to confirm or reject decisions taken by an artificial autonomous system.


To test this hypothesis we recorded EEG signals while subject monitors 1-D movement of a cursor towards a target . Figure 2(a) shows the ERP difference of signals elicited by erroneous and correct cursor movements. We have shown that it is possible to recognize these potentials in single trial. Moreover, system performance can possibly improved based on the recognition of these signals at least in two ways; on the one hand erroneous decisions can be corrected [Ferrez and Millan, 2008], on the other hand, the autonomous system can be updated so as to decrease the likelihood of actions that elicit error potentials, e.g. using ErrPs as negative reinforcers [Chavarriaga et al., 2007].
We have also explored the use of other cognitive signals for interaction. In particular, we have studied EEG signals related to the anticipation of future events, namely the Contingent-negative variation.
Gangadhar et al, [Gangadhar et al., 2009] have shown recognition of relevant/irrelevant situation based on Go-NoGo CNV paradigms both on-line and off-line analysis (c.f. Fig 2(b)). This signal may thus be used by the user to trigger specific behaviors in an asynchronous manner (e.g. in the scenario of wheelchair control to perform docking in front of a desk as opposed to avoid it).

3 Extended multimodal feedback
Efficient Brain-machine interaction requires the system to provide the user with information about its internal states. For instance, a moving robot may inform what action it is about to perform asking the user to confirm or reject such action via ErrPs. To this end, we have studied the use of different feedback modalities (i.e. visual, auditory and tactile) in realistic navigation tasks using virtual reality environments. This study shows that it is feasible to use ErrPs elicited by different feedback modalities in Brain Machine interactive systems [Perrin et al., 2008]. Moreover, simulated experiments in ErrP-based Human-robot interaction have shown that current ErrP classification performance is enough to successfully perform navigational tasks (c.f. Figure 2(c)). Experiments using real robots and online recognition of error potentials are currently undergoing.

Abstract:

People with impaired motor abilities suffer from significant deterioration of their quality of life due to the numerous physical and psychological barriers imposed by their situation on the conventional human communication and control channels. Despite common thinking, physically impaired people strive to partially restore or replace their communication and control pathways, in order to escape isolation and marginalization, and reach a satisfactory level of independence, social integration and eventually, an improved quality of life. Brain-Computer Interface (BCI) technology provides a promising additional option for recovering basic communication capabilities and operating devices that allow users to gain and maintain control of their environment.
TOBI (Tools for Brain-Computer Interaction) is an European project that aims to develop practical BCI Assistive Technology (AT), namely, to build non-invasive BCI prototypes based on electroencephalographic (EEG) signals, thus providing disabled users with a novel means of interaction with the outside world. Communication and control enhancement is specifically addressed within the TOBI framework, targeting a measurable middle-term impact in terms of preclinical validation on users with different levels of motor impairments.
State of the art BCI systems demonstrate the potential of using BCI as an additional communication and control channel. An increasing number of BCI applications has been reported (e.g., virtual keyboards, web browsers, brain-actuated mobile robots, neuroprostheses), yet, current BCI prototypes suffer
certain limitations. Interaction aspects of relevant applications, such as “ease of use” and “user friendliness” have been neglected, as research has been mainly focused on the machine learning and signal processing challenges of BCI technology. As a result, interested parties (clinics or patients) cannot benefit without constant expert guidance. TOBI aims at alleviating such limitations, by creating standalone BCI applications along with novel user training protocols, thus bringing BCI technology out of the lab and into the real world. The main short-term goal of TOBI is the integration of BCI control into currently existing, popular among the disabled community AT systems, that have proven to be useful. Embedding the additional BCI communication option is believed to augment the range of applications that users can currently access with AT products.
The AT software selected for our BCI experimentation is QualiWORLD (QualiLife SA, Lugano, Switzerland). QualiWORLD is a comprehensive platform allowing disabled users to gain access to many applications (text editor, web browser, email, etc.), while replacing standard mouse and keyboard by a variety of computer access solutions (Auto-Scan Mode, mouse alternatives, gesture recognition). It can be personalized to the specific user disabilities and preferences. QualiWORLD helps disabled persons to achieve greater independence by allowing them to communicate, have more control over their living environment, and participate in state of the art rehabilitation and educational programs. Furthermore, a considerable amount of users is already familiar with it, as it is particularly popular in the AT community.
A Mental Imagery (MI) based BCI system has been interfaced with QualiWORLD as a first demo of BCI integration in commercial AT software. Preliminary results in coupling BCI with QualiWORLD prove the feasibility of using BCI as an additional communication and control pathway for disabled people. Users have
been able to edit a text and browse a photo album using only one BCI command, along with the QualiWORLD Virtual Keyboard and Auto-Scan mouse mode. More specifically, after a short period of MI training, the user is able to select the application’s contents (virtual keys, images) while the QualiWORLD Auto-Scan sequentially highlights them, by performing the predefined MI. Thereby, the user’s EEG brain activity is classified by the BCI and the resulting class probabilities are reduced into a single binary (yes/no) command (key-press). This binary command is translated by the QualiWORLD into a selection event. Problematic is the fact that the BCI subject is only receiving a discrete feedback via the QualiWORLD and not a continuous one, as usually in BCI research. So,
the user is unaware of the development of his/her brain patterns over time, and how close he/she is to the decision border at a given time. This issue will be addressed in the near future, so that the QualiWORLD can provide visual feedback to the user concerning the confidence of the decision making process (probability distribution visualization). The goal is, on one hand, to demonstrate the suitability of BCI for controling state of the art AT products and, on the other hand, to provide cues on how these products can adapt to the BCI interaction channel, leading to more efficient user interfaces.
Future work will address the coupling of more complex applications with the BCI, the introduction of multiple BCI commands, and the evaluation of the potential of this novel interaction techniques by disabled users. Special attention will be paid to increase the robustness of the system in order to achieve satisfactory user experience and to establish BCI as a key player in the AT market. Towards more robust BCI, TOBI will take advantage of novel Human-Computer Interaction principles, take into account the inherent properties of EEG/BCI signals (low Signal-to-Noise-Ratio, non-stationarity, low output bitrate, lagged
dynamics) and address mental state and performance monitoring, as well as on-line adaptation issues. Finally, novel interaction objects for various tasks will be benchmarked, aiming a multimodal approach.

This work is supported by the European ICT Programme Project FP7-224631. This paper only reflects the authors’ views and funding agencies are not liable for any use that may be made of the information contained herein.

Abstract:

THE P300- BRAIN-COMPUTER INTERFACE BROWSER: A MUSCLE-INDEPENDENT SURFING TOOL FOR PARALYZED PEOPLE

Carolin Ruf1, Emily Mugler1, Sebastian Halder1, Michael Bensch2 & Andrea Kübler3

1 Institute of Medical Psychology and Behavioral Neurobiology, University of Tübingen,
2 Wilhelm-Schickard-Institute for Computer Science, University of Tübingen
3 Department of Psychology I, Biological Psychology, Clinical Psychology and Psychotherapy, University of Würzburg


EEG-based-Brain-Computer Interfaces (BCIs) can be used by paralyzed people for communication. To increase the general usefulness of BCI systems applications for particular activities are needed. The presented study evaluated the efficacy of a BCI application for surfing the web. A matrix paradigm based on the event-related potential P300 was used to control the BCI web browser. Ten healthy subjects and three paralyzed patients diagnosed with amyotrophic lateral sclerosis (ALS) performed web surfing tasks in several sessions. All participants were asked to evaluate the BCI browser after use. The healthy subjects achieved an average accuracy of 90 % and an information transfer rate (ITR) of 16.5 bits/minute when controlling the web browser. The ALS patients used the browser with an average accuracy of 72% and an ITR of 7 bits/minute. The patients indicated that they would use the BCI browser in everyday life and would participate in more BCI web browser sessions. The results confirmed a decreased ITR in people with neurological disease as compared to healthy controls. The lower P300 amplitude and a longer latency in the ALS patients may account for this difference. This aspect has to be taken into account when designing BCI protocols for patients. Nevertheless, accuracy in patients was still high enough to control the browser reliably.

Abstract:

Brain-Computer Interfaces (BCI) provide muscle-independent communication for paralyzed patients. Individuals differ in their ability to use a BCI. This study examines the relation between motivation and the P300 amplitude within a BCI controlled by event-related potentials (ERP). In two experimental groups participants received 25 or 50 Cent for each correct letter selection; the control group was not rewarded. Motivation was assessed with a BCI adapted questionnaire and a visual analogue scale. BCI performance was defined as the percentage of correctly selected characters (group mean = 99%). At Cz the P300 amplitude was positively correlated to self-rated motivation (r=.50). Offline analysis revealed that highly motivated participants would have needed fewer trials for a discriminable ERP and thus, would have been able to communicate faster with the ERP-BCI. These results indicate that motivation may contribute to variance in BCI performance and has to be monitored in BCI settings.

Abstract:

People with amyotrophic lateral sclerosis (ALS) lose their motor activity and so their ability to talk in the course of their disease. Brain-Computer Interfaces (BCIs) provide an alternative communication channel because they rely on brain signals and are thus muscle independent. However, individuals differ in their ability to use a BCI. To investigate the relevance of psychological influencing variables such as motivation in patients with ALS this study examined the relation between motivation and the ability to learn using a BCI and the P300 amplitude measured within a BCI controlled by event-related potentials (ERP). Motivation was manipulated with a 20 Euro gift certificate for an internet store. In the first run twelve ALS patients spelled a 14 character sentence without receiving a reward. In the second run they were promised a gift certificate for trying particularly hard to spell the sentence correctly. Motivation was assessed with a BCI-adapted questionnaire and a visual analogue scale. BCI performance was defined as the overall percentage of correctly selected characters (correct response rate=CRR). Three patients were not able to finish the session and were excluded from analysis. Average CRR across all runs and patients was 93%; four patients had a CRR of 100%. The gift certificate did not affect motivation but BCI performance. We found a trend for CRR being higher after motivation (96%) than before motivation (89%, Z=-1.84, p=.07). The results indicate that motivation may explain some of the variance in BCI performance and should be monitored in BCI settings.

Abstract:

P300 BCI Performance Prediction Using an Auditory Standard Oddball

Halder, S. a
Hammer, E.-M. a
Kleih, S. C. a
Kübler, A. b,a

a Institute for Medical Psychology and Behavioural Neurobiology, University of Tübingen, Tübingen, Germany
b Department of Psychology I, Biological Psychology, Clinical Psychology and Psychotherapy, University of Würzburg, Würzburg, Germany

Brain-Computer Interfaces (BCIs) enable paralyzed people to communicate with their environment. Differences in performance between users and sessions remain largely unexplained, as does the question as to why communication in the complete locked-in-state (CLIS) has not been possible. A reliable performance indicator would allow an analysis of subject-to-subject and session-to-session performance differences and serve as and indicator of the capacity to use a BCI during the progression of a disease.
A study with 40 healthy participants was conducted to determine the viability of performance indicators. All participants performed a single 20 symbol visual (VP300) and auditory P300 (AP300) BCI session. Additionally, an auditory oddball was recorded from each subject. Online feedback performance (<100% VP300, <70% AP300) was used to separate groups of good and bad performers (BP).
Average performance of 94.5% (12 subjects in BP group) using the VP300 and 63% (16 subjects in BP group) using AP300 were achieved. Mean P300 amplitude in the auditory oddball was significantly higher in good as compared to bad performers (Wilcoxon rank test, p < 0.05). Using the amplitudes of two samples at 395 ms on CPz and CP1 correlations (Pearson) with performance of r=0.57 were found. This result shows the viability of the auditory standard oddball in to predict individual BCI performance and suggests that the long term tracking of the P300 elicited by the auditory oddball will lead to a better understanding of BCI performance degradation in the CLIS.


Acknowledgements

Funded by DFG KU 1453/3-1. This work is supported by the European ICT Program Project FP7-224631. This paper only reflects the authors’ views and funding agencies are not liable for any use that may be made of the information contained herein.

Abstract:

P300 BCI's robustness as a communication device needs to be tested against the concurrent psychological states of users. We investigated the influence of emotional state and subjective depressed mood of 40 healthy participants on their letter spelling performance. Emotional state was induced by presenting images from the International Affective Picture System (IAPS). Depressed mood was assessed using subjects’ self-rating of the Centre for Epidemiologic Study Depression scale (CES-D). The study employed a within-subject design and each subject undertook three letter-spelling sessions associated to pleasant, unpleasant, and neutral valence. P300 event-related potential (ERP) amplitude is compared between conditions to provide a psychophysiological index to performance. No difference was observed between the affective conditions in terms of performance. Furthermore, there was no difference in P300 ERP amplitudes across subjects between the emotional conditions. Given the past findings of P300 amplitude suppression due to emotional experience, it is likely that our emotional manipulation was not sufficiently strong for this study although it successfully induced emotional states in the normative direction. More importantly however, this study demonstrated a significant correlation between the self-rated depressed mood and BCI performance. More depressed mood is related to poorer overall performance. We can conclude that depressed mood hinders the performance of P300 BCI. More sophisticated emotional induction is needed to investigate the influence of emotional states on BCI performance.

Abstract:

Introduction
Brain-computer interfacing (BCI) aims at providing paralyzed patients with a communication device that obviates the need of using the usual motor pathways. A large number of BCI systems is based on motor imagery for encoding the user's intention. Motor imagery typically leads to event-related desynchronization (ERD) of the 10Hz mu-rhythm in the motor cortex associated to the respective limb. This EEG phenomenon can be used for feedback control for most subjects by a classifier that was individually trained on the subject's EEG [1]. We introduce the goalkeeper paradigm that aims at improving online BCI performance by subject training under time pressure conditions.

Methods
Multi-channel EEG of 8 BCI-experienced subjects was acquired while they were playing 3 runs (100 trials each) of a BCI-controlled computer game that imitated the task of a goalkeeper during a penalty kick. During a trial, a ball was moving from the top of the screen towards one of its bottom corners. Using two different types of motor imagery (chosen from left hand, right hand and foot) the subjects had to control the horizontal movements of a bar at the bottom of the screen in order to catch the ball. The speed of the ball increased linearly from trial to trial and over the 3 runs. Subjects had to catch the ball within 2500ms (at the beginning of run 1) to 1250ms (at the end of run 3). Late arrival in a correct corner or arrival in a wrong corner were interpreted as misses.
In order to achieve a constant goalkeeping performance, the subjects were thus required to generate faster and/or stronger ERD responses in the later runs to steer the bar quickly into the correct corner. In an offline analysis, the goalkeeping performance, the reaction times (defined as the time needed to reach the correct corner) and EEG features were analyzed in relation to the block design of the experiment.

Results
The goalkeeper paradigm effectively increased time pressure over the 3 runs. Performance was measured in terms of balls caught within the first 1250ms. 7 out of 8 subjects managed to respond with increased performance from run 1 to 3 (avg. of 33.8 balls caught in run 1 to 41.6 in run 3, see Fig. 1).
A close analysis of time-frequency EEG features between successful trials of run 1 and 3 revealed different strategies of the subjects, e.g. earlier ERD or stronger ERD in the alpha band under time pressure. As a side effect, the training introduced for some subjects an additional ERD in the beta band (which had not been used for feedback). Earlier re-synchronization (ERS) could be observed for some subjects in run 3, where trials were shorter.

Acknowledgements
This work is supported by a BMBF grant No. 01GQ0850, a DFG grant MU 987/3-1, and by the European ICT Programme Project FP7-224631. This abstract only reflects the authors' views. Funding agencies are not liable for any use that may be made of the information contained herein.

References
[1] Müller KR, Tangermann M, Dornhege G, Krauledat M, Curio G, Blankertz B. Machine learning for real-time single-trial eeg-analysis: from brain-computer interfacing to mental state monitoring. J Neurosci Methods, 167(1):82–90, 2008.

Abstract:

The Effect of Emotions on P300 Brain-Computer Interface (BCI) Performance

S. Lukito1, S. Halder2, P. Bretherton3, C. Vögele3, A. Kübler1,2
1University of Würzburg, Germany
2University of Tübingen, Germany
3Roehampton University, London, UK


The rapid development of BCI technology and research has not been matched by research into its clinical applicability. It is important to know for instance whether BCI performance is affected by its users’ cognitive and emotional state. This question is all the more important because potential users of BCI in the clinical setting are patients in the locked-in-state, among whom an increased incidence of emotional problem and depression is reported.

This study investigated the effect of subjective feeling of depression, as assessed by a self-rated questionnaire, and emotional priming on the oddball BCI letter-spelling performance of forty healthy participants. In conjunction to observing the letter selection performance, the study also investigates whether there is a difference between P300 event-related potential (ERP) parameters across the affective valence conditions. Participants completed three blocks of letter-spelling task corresponding to three emotional valence categories (positive, neutral, and negative). During each block, participants were instructed to spell the word BRAINPOWER whilst being exposed to five images, taken from the International Affective Picture System (IAPS), prior to each letter selection. Participants were also requested to imagine recent or past emotional experience relevant to the valence condition of each block throughout the task. We found significant negative correlation between the subjective feeling of depression and BCI performance. However, evidence of the effect of emotional priming on performance is lacking, even though participants’ rating of valence and arousal of the pictures were at normative values. From these results, we cautiously conclude that negative mood indeed hampers BCI performance. To investigate further the influence of emotion on BCI performance, more sophisticated approaches to manipulation of emotion will be used in further studies.

Abstract:

Implementation of SMR Based Brain Painting

S. Halder1, A. Furdea1, R. Leeb2, G. Müller-Putz2, A. Hösle3, A. Kübler1,4
1 Universitätsklinikum Tübingen, Institute of Medical Psychology and Behavioral Neu-
robiology, Gartenstr. 29, 72074 Tübingen, Germany
2 Institute for Knowledge Discovery, Graz University of Technology, Krenngasse 37 8010 Graz, Austria
3 87727 Babenhausen, Germany
4 Universität Würzburg, Lehrstuhl für Psychologie I, Marcusstr. 9-11, 97070 Würzburg, Germany

Current brain-computer interface (BCI) systems are mostly used for communication with late-stage motoneuron disease patients. These systems offer only restricted possibilities to their users to express themselves creatively. Nonetheless, many patients consider artistic activity to be a valuable aspect of their lives.
We previously extended our P300 BCI that is being used by amyotrophic lateral sclerosis (ALS) patients for communication, to enable the use of a painting application. This was achieved by mapping the individual fields of the control matrix to painting functions. These can be used for e.g., cursor control and placing various figures on the virtual canvas used for painting. When using a P300 BCI though, the user is restricted to a predefined step intervals when moving a cursor or changing the size of objects on the canvas. This limitation was overcome by designing a new painting application that is controllable with a sensorimotor rhythm (SMR) BCI based on the detection of event-related desynchronization and synchronization (ERD/S) of those rhythms.
In this design command icons are placed in six hexagons that are arranged in a circle (modeled after the hex-o-spell interface [1]). The BCI is used to control an arrow which extends from the center of these hexagons to select the intended command or to rotate the arrow further to the next hexagon. Several commands transfer control from the menu to the canvas itself so that the BCI can be used to e.g., freely move the cursor to the desired position. Other commands which allow user adjustment are changing object size, object transparency and zooming. Returning control to the menu is achieved by using a non-control class that is recognized by the system when the user imagines neither of the two control classes.


[1] Klaus-Robert Müller, Michael Tangermann, Guido Dornhege, Matthias Krauledat, Gabriel Curio, and Benjamin Blankertz. Machine learning for real-time single-trial eeg-analysis: from brain-computer interfacing to mental state monitoring. J Neurosci Methods, 167(1):82–90, 2008 Jan 15.