● Introduction
● BCI for communication in paralysis
● BCI software
● Self-regulation of SCPs and training
● Spelling through brain-cmputer communication

Brain-Computer-Interface (BCI) ?
Subject/
Patient
EEG
Brain Computer
Interface control signal
Device
“A system for controlling a device e.g. computer, wheelchair or a neuroprothesis by human intention which does not depend on the brain’s normal output pathways of peripheral nerves and muscles”
[Wolpaw et al., 2002].
HCI – Human Computer Interface
DBI – Direct Brain Interface (University of Michigan)
TTD – Thought Translation Device (University of Tübingen)

Some examples of BCI applications
BCI_
BCI
Leeb et al., Computational Intelligence and
Neuroscience, 2007 (doi:10.1155/2007/79642)

Brain Computer Interfaces
● Allow patients to control a computer by concious changes of brain activity
● Provide a means of communication to completeley paralysed patients: amyotrophic lateral sclerosis (ALS), cerebral palsy, locked in syndrome
● Can be used to control a cursor, select symbols, control external devices like orthesis / prothesis (depending on type of BCI)
● Have a very low data rate, typical a few bit per second or less
● First results in the 1970ies (Vidal, visual evoked potentials, VEP-BCI)

Brain Computer Interfaces
Principles of operation:

Brain Computer Interfaces – Electrophysilogical
Activities used
● SCP Slow Cortical Potentials
● Mu Movement Imagination
● P300, SSVEP ERP-Analysis
● cortical neurons, direct brain interfaces
The control information is extracted from the real time EEG-recording http://www.wired.com/news/images/full/thoughtlock1_f.jpg

Brain Computer Interfaces – SSVEP
● Steady State Visual Evoked Potentials derived from the visual (occipital) cortex
● focussing attention to visual stimuli of different frequency shows up in the EEG freqeuncy bands
● relibable and high transfer rate, but some prerequisites (eyes) http://www.iua.upf.es/activitats/ semirec/semi-Reilly/

Brain Computer Interfaces – SCP BCIs
● detection of slow cortical potentials (SCPs)
● needs DC EEG Amplifiers
● first successful device end 1990‘s:
Niels Birbaumers Thought translation device intensive training was necessary to gain control over the SCP waves
SCPs:
DC-shifts, slow negativation of cortical areas
Preparation of movement and cognitive tasks,
Several hundert milliseconds before the task Patinet using TTD to write a letter http://www.heise.de/ct/06/18/088/bild1.jpg

Brain Computer Interfaces μ-rhythm BCIs
● μ–rhythm is the idle-rhythm of the motor cortex
● frequencies around 10 and 18 Hz.
● ERD / ERS – event related desynchronisation / synchronisation movements or imagination of movements inhibit the μ–rhythm
Berlin-BCI, http://www.fraunhofer.de/

Brain Computer Interfaces - P300 BCIs
● P300 wave – posivite component in the event related potential, 300ms after a stimulus
● natural response to events considered as important
● selection of a symbol: count the flashes, algorithm averages trails and finds a P300
P300 runtime user interface

Brain Computer Interfaces μ / P300 comparison
μ - BCIs P300 BCIs
Require training
2d-control possible movement imagination affected by movement do not require training
1D control only concentration / decision affected by distraction

• An overview of different approaches to BCIs developed at Institute of Medical Psychology and
• Thought-Translation Device (TTD).
• Brain-Controlled Web Browser.
• Visual and Auditory feedback modes.
• Oscillatory Features based Classification.

– Down regulation of brain potentials towards a positive amplitude.
– Communication through self-regulation of SCPs
– Also known as TTD

– EEG Amplifier
• EEG8 system
• G.tec amplifiers
• BrainAmp system
– Two Monitors
• One for operator (supervise the training)
• One for patient (feedback)
• Sampling frequency: 256Hz
• Digitized with 16 bits/sample
• Amplitude range: +(-) 1mV
• Low frequency cutoff: 0.01 Hz
• High frequency cutoff:: 40-70 Hz

TTD Feedback and Communication System
• The current version of TTD software is derived from BCI2000 standard

TTD Feedback and Communication System

Data acquisition and storage
Online signal processing
Classification
Feedback and application interface

● Research Platform for BCI Systems
● Written by Gerwin Schalk, Wadsworth Center, Albany (NY)
● Modular structure: Signal Aquisition, Signal Processing and User Application communicatie via TCP/IP
● Operator module used for configuration of the other modules
● various user tasks availbale: 1D/2D cursor, Speller, P300, SCP
● free for academic use
● driver for OpenEEG available http://www.bci2000.org/

• Filters
– Spatial, temporal, and spectral
• Online artifact detection and correction
• Classification
– Linear Discriminant Analysis (LDA)
– Simple Threshold Classification
– Support Vector Machine (SVM)
• MATLAB interfaces.

●
Slow event-related direct-current shifts of the EEG.
● They last from 0.3 seconds up to several seconds.
● O ccur as a result of external or internal events.
● Negative shift is related to excitability of neurons.
● Positive shifts are measured during the execution of cognitive tasks
● Healthy subjects as well as patients can learn to produce positive or negative SCPs
● Training requires feedback
• Visual
• Auditory

●
Recording site for feedback signal is usually Cz.
● EEG is usually recorded from 3-7 Ag/AgCI-electrodes placed at Cz, C3, C4, Fz, and Pz.
● vEOG is recorded using a bipolar channel for online and offline artifact correction.
● A fixed percentage of vEOG signal is subtracted from the SCP signal at Cz for EOG correction.
●
SCPs are calculated by applying a 500ms moving average to EEG signal.

●
With the visual feedback modality:
• Subjects viewed the course of their SCPs as the vertical movement of feedback cursor.
• Vertical cursor movement corresponded to the SCP amplitude.
• Task was to move the cursor towards the modality indicated by a red rectangle .

Self-Regulation of SCPs: Training Process
Target
Presentation interval
Selection interval
Response interval

Self-Regulation of SCPs: Training Process
Target
Presentation interval
Selection interval
●
First 2-4 sec of the trial.
●
Target is illuminated in red.
●
Allows the subject to prepare for the corresponding
SCP regulation.
Response interval

Self-Regulation of SCPs: Training Process
Target
Presentation interval
Selection interval
Response interval
●
Feedback is provided by the vertical position of the steady horizontally moving cursor.
●
Cortical negativity moved the cursor up.
●
Positivity moved the cursor down.
●
Center of the screen corresponded to the baseline.
●
Task is to move the cursor to the red area.

Self-Regulation of SCPs: Training Process
Target
Presentation interval
Selection interval
Response interval
●
A response is classified as correct if:
• Average potential carried the correct polarity.
• Or is inside the target boundaries of the required goal.
●
Classification methods, such as LDA or SVM can be used for improvement of the correct response rate.

Self-Regulation of SCPs: Performance
1 st Trial: Target presentation ,
Selection, Response
2nd Trial: Target presentation ,
Selection, Response
NthTrial: Target presentation ,
Selection, Response
●
Performance
• Percentage of correct responses in valid trials.
• After a rate of 75% correct responses, subjects were trained to select letters and write messages
● Subjects typically reach this level after 1 to 5 months of training, with 1 to 2 training days per week .
●
A training day comprises 7 to 12 runs , and a run comprises 70 to 100 trials.

Applications
Spelling by Brain-Computer Communication:
• A program driven by “yes” or “no” responses, which serve as
“select” or “reject” commands.
• Requires three intervals in one trial.
• Allows user to select letters from a language alphabet and to combine letters into words and sentences.
Presentation Selection Response

Applications
Spelling by Brain-Computer Communication:
• A program driven by “yes” or “no” responses, which serve as
“select” or “reject” commands.
• Requires three intervals in one trial.
• Allows user to select letters from a language alphabet and to combine letters into words and sentences.
Presentation Selection Response
• Presentation of the letter set.
• Displayed in target rectangles on the screen.

Applications
Spelling by Brain-Computer Communication:
• A program driven by “yes” or “no” responses, which serve as
“select” or “reject” commands.
• Requires three intervals in one trial.
• Allows user to select letters from a language alphabet and to combine letters into words and sentences.
Presentation Selection Response
• Feedback is provided.
• Self regulation of SCP amplitudes is used to select or reject the letter set.

Applications
Spelling by Brain-Computer Communication:
• A program driven by “yes” or “no” responses, which serve as
“select” or “reject” commands.
• Requires three intervals in one trial.
• Allows user to select letters from a language alphabet and to combine letters into words and sentences.
Presentation Selection Response
• Response interval indicating to user the result of the selection.
• Error correction is done using a “go-back” option.

Performance:
• Writing the most conveniently situated letter , “E,” takes 5 trials.
• Writing the most remote sign requires 9 trials, i.e. 36 – 45 sec.
Improvement
• A simple personal dictionary to make free spelling less time consuming.
• Contains words that are frequently used by patients.
• A complete word is suggested after at least two letters have been written.
• This word can then be chosen with a single selection response.