Home | Agenda | Meeting Information | Registration and Housing | Students | Abstracts | Workshops | CME Information | Satellite Meetings | Participants | Sponsors | Job Board | Contact Us |

BCI Meeting Logo

June 3-7, 2013
Asilomar Conference Center, Pacific Grove, California

Research Sessions

Tuesday Morning:

Research Session 1: Scientific and Technical Advances

8:45 Welcome, Introduction of Major Sponsors
Future Technical Challenges
Paper 180Bringing Big Data to Neural Interfaces
Iyad Obeid*, Temple University; Joseph Picone, Temple University
The purpose of this paper is to present a new community-wide research entity to be launched called the Neural Engineering Data Consortium (NEDC). The purpose of the NEDC is to accelerate Brain Computer Interface research by creating, curating, and archiving massive neural datasets for the neural engineering community. The need for massive datasets is justified by the innate variability in neuronal activity. By pooling the resources of the neural engineering community and making such massive datasets available, we will enable investigators to improve BCI performance metrics and increase robustness. A proof of concept data corpus comprising 12,000 clinical EEGs is presently under development.
Paper 29EEG Correlates of Performance during Long-term use of a P300 BCI by Individuals with Amyotrophic Lateral Sclerosis
Yalda Shahriari*, Old Dominion University ; Theresa Vaughan, Wadsworth Center; Daniel Corda, Wadsworth Center; Debra Zeitlin, Helen Hayes Hospital; Jon Wolpaw, Wadsworth Center, Albany, New York ; Dean Krusienski, Old Dominion University
People with amyotrophic lateral sclerosis (ALS) are using BCI24-7, a P300-based brain-computer interface (BCI) system, independently, in their homes, for work and play. At the same time speed and reliability remain important issues for these independent users. This study seeks to correlate the EEG in six frequency bands (0-30 Hz), collected from eight electrode locations, as features in a linear model to predict if a P300-based BCI session will be successful. Data were collected from six home users during a copy-spelling calibration task. These data were divided into sessions with accuracy greater than or less than 70%. The prediction accuracy for session performance using information from the frequency bands was 82.72%. Better understanding of which EEG features are correlated with success could lead to better performance and greater system reliability.
Paper 115Using Multiple Reward Related Signals in the Adaptation of Neuoprosthetic Decoders
Scott Roset*, University of Miami; Noeline Prins, University of Miami; Shijia Geng, University of Miami; Hernan Gonzalez, University of Miami; Babak Mahmoudi, University of Miami; Eric Pohlmeyer, University of Miami; Justin Sanchez, University of Miami, Neuroprosthetics Research Group
Using neuroprosthetics during daily living presents new challenges that affect design choices in neural decoders. New classes of architectures for neural decoding that are based on reinforcement learning (RL) are being investigated. RL decoders use experience to help shape and adapt the decoder such that the benefits of tasks are maximized for the user. Since RL is based on reward feedback, reward signals are a critical part of the functionality. In this work, we investigate and compare 4 sources of reward related signals that can provide feedback for RL based decoders: the external environment, error-related potentials (ErrP) in EEG and LFPs, and single neuron activity in the Nucleus Accumbens (NAcc).

Panel Discussion: Future Scientific or Technical Challenges
Jonathan Wolpaw

Coffee Break

Future Equipment Innovations

Paper 103Modeling the Electroencephalogram using Intracranial Signals
Komalpreet Kaur*, Old Dominion University; Jerry Shih, Mayo Clinic; Dean Krusienski, Old Dominion University
This study presents models of electroencephalographic (EEG) event-related potentials (ERPs) using intracranially recorded ERPs from electrocorticography (ECoG) and stereotactic depth electrodes in the hippocampus. The patients had medically-intractable epilepsy and underwent temporary placement of an intracranial electrode arrays to localize seizure foci. Six patients performed one experimental session using the P300 Speller paradigm controlled by scalp-recorded EEG prior to the ECoG grid implantation, and one identical session controlled by ECoG after the grid implantation. All patients were able to achieve excellent spelling accuracy using EEG, and four of the patients achieved roughly equivalent performance in the intracranial sessions. Each EEG-ERP was modeled from the intracranial ERPs. The results indicate that EEG-ERPs can be accurately estimated from the intracranial ERPs for the patients that exhibited stable ERPs over the respective sessions. The resulting models provide a better understanding of the EEG-ERPs and can potentially be used to improve noninvasive BCI methods.
Paper 101Field Study of an fNIR-Based Brain Computer Interface for Communication
Melody Jackson*, Georgia Tech, Atlanta, GA; Ian McClendon, Georgia Institute of Technology; Kuniaki Ozawa, Tokyo Women's Christian University
Functional Near Infrared (fNIR) Brain-Computer Interfaces use light in the infrared spectrum to determine activity levels in the brain by detecting hemodynamic changes. Hitachi Japan developed the Kokoro-Gatari device, a simple fNIR-based communication system for people with locked-in syndrome. Following a successful study in Japan, we performed a year-long U.S.-based study in the home environments of 35 people with ALS, to determine the long-term effectiveness of the Kokoro-Gatari as a communication device. Accuracy ranged from 32% to 100% with an average of 70% overall. We also examined average usage (compliance with the study), regional differences, and learning effects.
Paper 105Transcranial Doppler Ultrasonography-Driven Online Augmentative and Alternative Communication Aid
Jie Lu*, University of Toronto; Tom Chau, University of Toronto
Although there are currently many signal detection modalities available for BCIs, they all have limitations that preclude their usage in practical assistive devices. To overcome these problems, we consider transcranial Doppler ultrasonography (TCD) as a BCI modality. TCD is a medical imaging technique that is used to measure cerebral blood flow velocity. It is robust, portable and inexpensive, making it a strong candidate for practical BCI applications. In this study, we focus on analyzing the ability of 12 able-bodied participants in using the online TCD-driven BCI for the purpose of controlling a scanning keyboard through two mental tasks. The activation mental task is a repetitive mental spelling of the intended word. The rest mental task is a visual tracking of the TCD signal feedback. These two tasks are simple to perform and have shown an overall offline validation accuracy of 89.03 ± 6.74% from preliminary results. Classification of these two mental tasks is achieved using Naïve Bayes and a set of time-domain user-dependent features. Using session-dependent classifiers, we were able to achieve an online classification accuracy of 80.17 ± 11.43%. The online TCD-driven BCI will be an asset, as a communication aid, for individuals with severe physical disabilities. Since TCD is portable and affordable, it has the potential to be successfully integrated into a home-use communication device.
Paper 106Real-Time Estimation and 3D Visualization of Source Dynamics and Connectivity Using Wearable EEG
Tim Mullen*, SCCN/INC/UCSD; Christian Kothe, SCCN/INC; Mike Chi, SCCN/INC; Alejandro Ojeda, SCCN/INC; Trevor Kerth, Cognionics, Inc; Scott Makeig, UC San Diego; Tzyy-Pin Jung,
This report summarizes our recent efforts to deliver real-time data extraction, preprocessing, artifact rejection, source reconstruction, multivariate dynamical system analysis (including spectral Granger causality) and 3D visualization within the SIFT and BCILAB toolboxes. We report the application of such a pipeline to EEG data obtained from wearable high-density (32-64 channel) dry EEG systems.

Panel Discussion: Future Equipment Innovations
Nick Ramsey


11:45 End

Tuesday Afternoon:

Research Session 2: Innovative BCI Applications and Protocols

13:15 Intro
Defining the Next Great BCI Applications
Paper 60Detecting cognitive states for enhancing driving experience
Ricardo Chavarriaga*, EPFL-CNBI; Lucian Gheorghe, EPFL-CNBI; Huaijian Zhang, EPFL-CNBI; zahra Khaliliardal, EPFL-CNBI; Jose del R. Millan, EPFL
Intelligent cars exploit environmental information to support drivers by providing extra information and assisting complex maneouvers. They can also take into account the internal state of the driver by means of decoding cognition-related brain activity. Here we show the feasibility of successfully classify EEG correlates of anticipation, movement preparation and error processing while subjects drive in a realistic car simulator.
Paper 82EEG-predictors of covert vigilant attention
Adrien Martel*, TU Berlin; Benjamin Blankertz, Berlin Institute Technology, Berlin, Germany; Sven Dähne, TU Berlin
A brain computer interface (BCI)-based cognitive state monitoring system able to determine the current, and predict the near-future development of the brain's attentional processes would bear great theoretical and practical implications. The present study investigated the evolution of neurophysiological signals preceding an omission error during a covert sustained attention task. The findings confirm the presence of characteristic EEG signals anteceding inadequate levels of attention.
Paper 62Iterative EEG-based Natural Image Search under RSVP
Marija Uscumlic*, CNBI EPFL; Ricardo Chavarriaga, EPFL-CNBI; Jose del R. Millan, EPFL
This work extends previous studies on using EEG decoding for automatic image retrieval. We propose an iterative way to integrate the information obtained from the EEG decoding and image processing methods. In the light of real-world BCI applications, we demonstrated that a limited number of EEG channels provide sufficient information about the subject's preference to be exploited in image retrieval by the proposed synergistic scenario. Furthermore, to meet a more realistic scenario we used natural images (i.e., images of objects in their natural environment).
Paper 39EEG-based Communication with Patients in Minimally Conscious State
Gernot Müller-Putz*, Graz University of Technology, Graz, Austria; Christoph Pokorny, ; Daniela Klobassa, Medical University of Graz; Gerald Pichler, Albert Schweitzer Klinik Graz; Petar Horki, Graz University of Technology
We investigated an EEG-based approach for communication with minimally conscious state (MCS) patients. In a mental imagery paradigm, the patients were instructed to perform imagined sports, navigation and feet movements. Classification accuracies above chance level were reached by three of the four patients performing mental imagery tasks, indicating the feasibility of this paradigm for communication with MCS patients.

Panel Discussion: Defining the Next Great BCI Applications
Lee Miller

Engineering the BCI User
Paper 76Facilitating effects of transcranial direct current stimulation on EEG-based motor imagery BCI for stroke rehabilitation
Kai Keng Ang*, I2R, A*STAR; Cuntai Guan, I2R, A*STAR; Kok Soon Phua, I2R, A*STAR; Chuanchu Wang, I2R, A*STAR; Ling Zhao, NUHS; Wei Peng Teo, NUHS; Effie Chew, NUHS
This clinical trial investigates the facilitating effects of combining tDCS with EEG-based motor imagery Brain-Computer Interface (MI-BCI) robotic feedback compared to sham-tDCS for upper limb stroke rehabilitation. 32 hemiparetic stroke patients were recruited and screened for their ability to use EEG-based MI-BCI. Subsequently, 17 of these patients who passed screening and gave further consent were randomized to receive 20 minutes of tDCS or sham-tDCS prior to 10 sessions of 1-hour MI-BCI with robotic feedback for 2 weeks. The offline and online accuracies of detecting motor imagery from idle condition for the calibration session and the evaluation part of the 10 rehabiltiation sessions were respectively assessed. The results showed that there were no significant difference in the accuracies of the calibration session from both groups, but the online accuracies of the evaluation part of 10 rehabilitation sessions of the tDCS group were significantly higher than the sham-tDCS group. Hence the results suggest towards tDCS effect in modulating motor imagery in stroke.
Paper 81Motivation and SMR-BCI: The Fear to Fail affects BCI Performance
Sonja Kleih*, ; Tobias Kaufmann, University of Würzburg; Eva Hammer, University of Würzburg; Iolanda Pisotta, Fondazione Santa Lucia, IRCCS; Floriana Picchiori, Fondazione Santa Lucia; Donatella Mattia, IRCCS Fondazione Santa Lucia; Andrea Kübler, Julius-Maximilians-University Würzburg, Würzburg, Germany
Components of motivation have been shown to affect performance when using a Brain-Computer Interface (BCI) based on sensorimotor rhythms (SMR). However, usually reported results are based on relatively small sample sizes of healthy adults. Therefore, neither conclusions about motivation effects on BCI performance in larger samples nor in clinical samples can be drawn. In this study we correlated the subjective ratings of motivation of N=51 healthy participants and N=11 stroke patients with their SMR BCI performance and found that incompetence fear or the fear to fail was significantly related to lower performance.
Paper 79Investigation of the utility of mind-body awareness training in the early learning of a 1D sensorimotor rhythm based brain-computer interface
Alexander Doud*, University of Minnesota ; Kaitlin Cassady, ; Karl Lafleur, ; Bin He, University of Minnesota, Minneapolis, MN
In the present study we present the initial BCI training period of a cohort of 5 subjects with regular exposure to yoga, meditation, or a combination of both practices. The investigation of these mind-body awareness training practices may provide insight into valuable strategies for reducing barriers to BCI fluency that limit the use of these systems by some individuals. The investigated subjects showed rapid training times, and were able to achieve competence in the use two differentiable control signals, left hand vs right hand and both hands vs rest. Subjects were able to achieve = 80% accuracies in these traditional BCI cursor tasks using standard electrode configurations and with as little as 33 minutes of training time.
Paper 75Dynamic Stopping in a Calibration-less P300 Speller
Pieter-Jan Kindermans, Ghent University; Benjamin Schrauwen, Ghent University
Even though the P300 based speller has proved to be usable by real patients, it is not a user-friendly system. The necesarry calibration session and slow spelling make the system tedious. We present a machine learning approach to P300 spelling that enables us to remove the calibration session. We achieve this by a combination of unsupervised training, transfer learning across subjects and language models. On top of that, we can increase the spelling speed by incorporating a dynamic stopping approach. This yields a P300 speller that works instantly and with high accuracy and spelling speed, even for unseen subjects.

Panel Discussion: Engineering the User
Eric Sellers

Coffee Break

Recognizing Success: BCI Performance Metrics
Paper 187Autonomy and Social Inclusion for the Severely Disabled: The BrainAble Prototype
Josef Faller*, Graz University of Technology; Sergi Torrellas, Barcelona Digital; Ursula Costa, ; Eloy Opisso, Guttmann Institute; Juan Manuel Fernández, ; Christoph Kapeller, g.tech; Clemens Holzner, G.Tec medical engineering GmbH; Josep Medina, ; Clare Carmichael, ; Günther Bauernfeind, Graz University of Technology; Felip Miralles, Barcelona Digital; Christoph Guger, Guger Technologies, Graz, Austria; Reinhold Scherer, ; Gernot Müller-Putz, Graz University of Technology, Graz, Austria
Severely disabled individuals often suffer from high care-giver dependence and are at risk of social exclusion. The prototype developed in the EU Project BrainAble (http://www.brainable.org/) offers access to common smart-home devices and popular Internet services, using electroencephalography (EEG) and non-EEG inputs. Here we describe our user-centered design approach and the software architecture of the system.
Paper 186The Brindisys project: Brain Computer Interfaces as assistive technology for people with ALS
Francesca Schettini*, Fondazione Santa Lucia, IRCCS; Angela Riccio, IRCCS Fondazione Santa Lucia, Rome; Luca Simione, Department of Psychology, "Sapienza", University of Rome; Giulia Liberati, Eberhard-Karls-University, Tübingen ; Mario Caruso, DIAG, Sapienza University of Rome , Italy; Barbara Calabrese, "Magna Graecia" University of Catanzaro Italy ; Nicola Ielpo, "Magna Graecia" University of Catanzaro Italy; Arrigo Palumbo, "Magna Graecia" University of Catanzaro Italy; Vittorio Frasca, Department of Neurology and Psychiatry, "Sapienza" University of Rome, Italy; Massimo Mecella, DIAG, Sapienza University of Rome , Italy; Francesco Amato, "Magna Graecia" University of Catanzaro Italy; Alessia Pizzimenti, Crossing Dialogues Association, Rome, Italy; Maurizio Inghilleri, Department of Neurology and Psychiatry, "Sapienza" University of Rome, Italy; Febo Cincotti, Fondazione Santa Lucia, IRCCS, Rome, Italy
The Brindisys project aims at designing and developing a general assistive tecnology to support communication and autonomy in people with Amyotrophic Lateral Sclerosis (ALS) from the onset of the disease to the locked-in phase. The prototype consists of a specific interface and applications allowing for communication and environmental control that can be managed both with conventional/assistive input devices and with a P300-based Brain Computer Interface (BCI). This work describes the functionalities of the current prototype and reports a preliminary assessment with end users.
Paper 183Correctly Applying Performance Metrics to Neuroprosthetic Control Interfaces
Charlie Matlack*, University of Washington; Howard Chizeck, University of Washington; Chet Moritz, University of Washington
There are nearly as many, if not more, performance metrics used with neuroprosthetic control interfaces as there are published studies. Many metric implementations contain logical inconsistencies or unfounded assumptions. These must be addressed for clinically-relevant comparative analyses of innovations to be feasible. We examine performance metrics that are used in neuroprosthetic control studies, and identify problems in common implementations of these metrics. For the popular Fitts' Law metric, we show that in a brain-controlled task performed by a macaque, target size impacts movement times in a way not captured by the index of difficulty measure. We also show that the information transfer rate (ITR) is often incorrectly assessed, and we examine the dependence of the information transfer rate on task design to correctly measure the ITR.

Panel Discussion: Recognizing Success: BCI Performance Metrics
Andrea Kübler


17:45 End

Wednesday Morning:

Research Session 3: Clinical Translation and Dissemination

8:45 Intro
BCIs for Users with Impairments
Paper 1An Improved Auditory Streaming BCI with Voice Stimuli
Erin Ricci, Wadsworth Center, Albany, NY, USA; Sameah Haider, Wadsworth Center, Albany, NY, USA; Theresa Vaughan, Wadsworth Center; Jeremy Hill*, Wadsworth Center
We have previously shown that it is possible to build EEG brain-computer systems based on voluntary shifts of covert attention between simultaneous streams of auditory stimuli. The system exploits not only the later event-related potential components (P3) which are strongest in response to rare "oddball" stimuli, but also the early (N1 or Nd) components that are attention-modulated on every stimulus occurrence. So far, however, such systems have been based on abrupt artificial stimuli (short discrete beeps or pulses). This creates two problems. First, many subjects find the stimuli annoying, intrusive or otherwise unpleasant. Second, the abstract nature of the stimuli makes the system unintuitive to many users. We would like to build a system in which the stimuli are natural and also semantically indicative of the purpose of the corresponding interface selection. This greatly simplifies the instructions for the users: when they want to say "yes" using the BCI, they must listen to a voice repeating the word "yes", and when they want to say "no", they listen to a voice repeating the word "no". We set out to assess, in a within-subject design, whether these new voice stimuli or the beep stimuli of previous studies allow better performance in an 8-channel EEG-based BCI. We find no significant penalty (in fact, a non-significant advantage) for the voice stimuli (mean ± std online performance = 76% ± 11) in comparison with the beep stimuli (73% ± 11).
Paper 5"Eyes-closed" SSVEP-based BCI for Binary Communication of Individuals with Impaired Oculomotor Function
Jeong-Hwan Lim; Han-Jeong Hwang, ; Chang-Hee Han, ; Chang-Hwan Im*, , Hanyang University, Korea
In this study, we propose a new paradigm for steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI), which can be potentially suitable for disabled individuals with impaired oculomotor function. The proposed electroencephalography (EEG)-based BCI system allows users to express their binary intentions without a need to open their eyes. A pair of glasses with two light emitting diodes (LEDs) flickering at different frequencies was used to present visual stimuli to 12 participants with their eyes closed. Through offline experiments performed with 11 healthy participants, we confirmed that human SSVEP could be modulated by visual selective attention to a specific light stimulus penetrating through the eyelids. After customizing the parameters of the proposed SSVEP-based BCI paradigm based on the offline analysis results, binary intentions of five healthy participants and one ALS patient were classified in real time. The average ITR of five healthy participants reached 10.83 bits/min, and the ITR of the ALS patient was 2.78 bits/min, demonstrating the feasibility of our BCI paradigm.
Paper 26Enhancing Brain-Computer Interface Performance in an ALS population: Checkerboard and Color paradigms
David Ryan, ETSU; Kenneth Colwell, Duke University; Sandy Throckmorton, Duke University; Leslie Collins, Duke University; Eric Sellers, East Tennessee State University
A brain-computer interface (BCI) speller provides non-muscular communication via detection of EEG features. In a non-disabled population, a Checkerboard (CB) stimulus presentation has been shown to improve BCI performance over the standard Row/Column (RC) paradigm. Another improvement is a gray-to-color (CL) paradigm that presents perceptually-salient targets defined by nine unique colors. The current study examines the RC, CB, and CL paradigms in an amyotrophic lateral sclerosis (ALS) population (N=7). Pilot data suggest improved performance of CB and CL over RC. The results suggest matrices including CB and CL provide more efficient communication and higher user satisfaction in an ALS population.
Paper 22BCI-controlled Videogame for Cerebral Palsy Children
Jeng-Ren Duann*, China Medical University; Jung-Chin Chou, China Medical University; Sheng-Chuen Liang, Chine Medical University; Ching-Hung Lin, National Chiao Tung University; Jin-Chern Chiou, National Chiao Tung University
We describe here a brain-computer interface (BCI) based videogame by integrating an 8-channel wireless EEG system, which samples and wirelessly transfers the EEG and EOG signals to a personal computer (PC). The PC runs a shooting videogame and an EEG and EOG translator to extract the features of eye blinkings as well as player's attention level so as to control the shooting and bombing of the videogame. Sixteen cerebral palsy children were trained to play this game for one month (one hour per week) and participated in the Central Taiwan Assistive Device Videogame Contest hold in Taichung on Dec. 29, 2012.

Panel Discussion: BCIs for Users with Impairments
Theresa Vaughan
Catherine Wolf

Coffee Break

Defining BCI Translational Issues
Paper 23BCI for stroke rehabilitation: a randomized controlled trial of efficacy.
Floriana Pichiorri*, Fondazione Santa Lucia, IRCCS; Giovanni Morone, Fondazione Santa Lucia, IRCCS; Iolanda Pisotta, Fondazione Santa Lucia, IRCCS; Manuela Petti, IRCCS Fondazione Santa Lucia; Marco Molinari, Fondazione Santa Lucia, IRCCS; Laura Astolfi, IRCCS Fondazione Santa Lucia; Febo Cincotti, Fondazione Santa Lucia, IRCCS, Rome, Italy; Donatella Mattia, IRCCS Fondazione Santa Lucia
A novel sensorimotor BCI prototype was developed to boost motor recovery of the upper limb in stroke patients. After preliminary testing, the prototype was installed in a rehabilitation ward and validated as an add-on to standard therapy in a randomized controlled trial, involving 26 unilateral subacute stroke patients. Clinical benefits and resting state brain network reorganization closer to normal were observed in the target BCI group.
Paper 7Just a switch: Timing characteristics of ECoG based Assistive Technology control
Erik Aarnoutse*, UMC Utrecht; Mariska Vansteensel, ; Martin Bleichner, UMC Utrecht; Zac Freudenburg, ; Nick Ramsey,
In the development of an implantable ECoG-based BCI system, we have investigated the possibilities of generating quick shifts between high and low levels of brain activity ("brain clicks") using voluntarily modulated brain activity. We conclude that it is possible to produce ECoG-based 'clicks' using both motor execution and covert serial subtraction. These 'clicks' can serve as an input for commercially available assistive technology devices, potentially making communication, as well as environmental control, available for locked in patients.
Paper 15Stability of Local Field Potentials Over 11 Months of Brain-Machine Interface Use
Marc Slutzky*, Northwestern University; Robert Flint, Northwestern University; Michael Scheid, Northwestern University; Zachary Wright, Northwestern University
Local field potentials (LFPs) are derived from many thousands of neurons. As such, they may enable long-lasting and stable control signals for brain-machine interfaces (BMIs). Here we assess the stability of LFPs in primary motor cortex of 2 monkeys during 2-D cursor control using an LFP-based BMI. Using a biomimetic BMI decoder without retraining or adaptation, monkeys exhibited high performance that remained stable for over 11 months. We examined the stability of the LFP features by computing decoders of the brain-controlled cursor velocity from individual features in each session and using them to decode the velocity in the last session. Many of the LFP features showed high correlation with the cursor velocity which grew increasingly stable for over 11 months. This suggests that the monkeys learned a stable mapping between motor cortical field potentials and outputs, and that LFPs will provide a highly stable signal source for BMIs.

Panel Discussion: Discussion of Implantable Brain Machine Interface Systems for the Restoration of Communication and Mobility
C. Chestek, J. Henderson, W. Wang, A. Sachs, B. Ajiboye, L. Hochberg


11:45 End

Home | Agenda | Meeting Information | Registration and Housing | Students | Abstracts | Workshops | CME Information | Satellite Meetings | Participants | Sponsors | Job Board | Contact Us |