EEG HEADSET FOR NEUROFEEDBACK THERAPY

Enabling Easy Use in the Home Environment

Joran van Aart, Eelco R. G. Klaver, Christoph Bartneck, Loe M. G. Feijs and Peter J. F. Peters

Department of Industrial Design, Eindhoven University of Technology, Den Dolech 2, Eindhoven, The Netherlands

Keywords: Headset, neurofeedback therapy, electroencephalography, EEG, brain waves, home environment, wellbeing,

impedance measurement.

Abstract: In this article we present the design of an EEG headset together with the context and vision that motivated

us to undertake the described design work. There is a variety of potential advantages and potential

difficulties associated with neurofeedback therapy. Both are analyzed informally and we argue in favour of

a change in perspective, moving away from treatment of illness towards prevention and giving the user an

active and responsible role. To structure the discussion we will deploy the closed loop diagram. We identify

elements from the world of gaming that will have added value over a pure training approach, notably

elements that improve enjoyment and motivation. We describe several of the design steps of the headset that

has been designed to achieve enjoyable neurofeedback therapy in the home environment and conclude with

an evaluation of this headset.

1 INTRODUCTION

With neurofeedback therapy we can make

brainwave patterns explicit by for example a

computer screen, patterns we normally cannot

influence since we are unable to see or feel them.

This feedback provides us the ability to influence

and change them. With neurofeedback therapy we

are literally reconditioning and retraining the brain

(Hammond, 2007).

In order to provide a structure and discuss the

wide variety of potential advantages and potential

difficulties associated with neurofeedback therapy,

we start by introducing the closed loop diagram of

Figure 1. In this diagram, G(s) is the user, patient,

sportsman, etcetera and H(s) is the external training

equipment together with services of the therapist.

To give an easy biofeedback example, let u be a

desirable value for the BMI (Body Mass Index) and

let y be the patient's or user's weight. Then H

includes the calculation y/(length

), but also a

display function. By informing the user of the

difference between desired and actual BMI, the user

is supposed to eat and move wisely and adjust his

weight. This example actually has been realized as

the Smart Mirror by Philips Electronics (Van

Splunter, 2002).

Figure 1: Closed loop diagram with user G(s) and training

equipment and services H(s).

Feedback theory has been the backbone of

mechanical and electrical engineering since almost a

century - and it still is. It is outside the scope of this

article to try and develop precise models of the

controlled system and the feedback function, we just

use it to structure the debate. One remark is in order:

it is of the utmost importance to be aware of the

complexity of the human mind-body system G(s).

Its subsystems include the skeletal, digestive,

muscular, lymphatic, endocrine, nervous, and

cardiovascular, urinary and reproductive systems.

The nervous system includes the brain and its higher

functions such as perception, cognition, and control.

The subsystems are not only coupled, they also have

internal feedback loops. This being said, we can

begin.

van Aart J., R. G. Klaver E., Bartneck C., M. G. Feijs L. and J. F. Peters P. (2008).

EEG HEADSET FOR NEUROFEEDBACK THERAPY - Enabling Easy Use in the Home Environment.

In Proceedings of the First International Conference on Bio-inspired Systems and Signal Processing, pages 23-30

DOI: 10.5220/0001065200230030

 SciTePress

2 OPEN QUESTIONS

In this section we begin with some preliminary

observations and open questions regarding

neurofeedback (NF). Some of the questions are

speculative and possibly controversial. We take

them as a source of inspiration; several of the issues

embedded in these questions will be addressed more

seriously in the subsequent sections.

Neurofeedback therapy (Evans, 1999) is about

training the mind in a non-intrusive way using EEG

signals, stimulating or discouraging certain brain

functions. Experiences with neurofeedback therapy

might seem promising, however for some reason it

hasn’t been adopted by the general public yet. One

of the causes might be that in present day society,

there seems to be a tendency to focus on the physical

aspect of health and on curing the symptoms rather

than fighting the cause. For example, people with

ADHD are drugged with Ritalin, which reduces

symptoms of hyperactivity, but doesn’t cure it. In

contrast, the main focus of neurofeedback is on

mental issues.

Another problem might be that neurofeedback

therapy is based on a ‘mind over matter’ perspective,

implying that physical and mental wellbeing are

interconnected (Fox, 1999; Kendell, 2001).

However, for some reason it seems difficult to

accept this perspective. Prejudices have been

formed, perhaps as a result of the natural fear of the

unknown. Still, although neurofeedback therapy has

shown to have a positive influence on numerous

disorders (Lubar, 1995), proof of absence of possible

side-effects has not been supplied yet. Whether this

is a potential issue for the therapy not to be accepted,

remains open for debate. However, it has been said

that if some or another medication could be as

broadly and effectively applicable as neurofeedback

therapy, it would already be available at every

pharmacy in the world. (Roskamp, 2007). Maybe we

should focus on a vision in which neurofeedback

therapy could eventually become as easy as taking

an aspirin.

Professional athletes use neurofeedback to

improve their already exceptional performances by

entering a state of ‘relaxed focus’, a moment of

optimal performance reached by this so-called peak-

performance training. This raises the question

whether neurofeedback could actually be applied to

elevate our general wellbeing in the way athletes do.

Heinrich et al (2007) formulate the following

promising statement: ‘In addition, ‘normal’ subjects

may improve cognitive functions (e.g., attention and

working memory) and performance in real-life

situations by means of NF’.

3 ANALYSIS

EEG products are mainly distinct in the number of

electrodes that are used for measurements. Products

range from the ElectroCap with 19 sensors (Figure

2) (ElectroCap International Inc., 2004), to two-

channel products only using 2 sensors and clips on

each earlobe. However, in either case scalp contact

is optimized by removing dead skin cells and parting

hair out of the way. Conducting gel is applied in

most cases, although products with dry sensors are

being developed as well. Recent products include

the g.EEGcap (Guger Technologies OEG, 2007) and

other techniques for measuring, like Project Epoc

(Emotiv Systems Inc., 2007) and SMART

BrainGames PlayStation System (SMART

BrainGames, 2007).

Figure 2: ElectroCap (image courtesy of ElectroCap

International Inc., 2004)

Referring to Figure 1 we decompose both G(s)

and H(s) to identify various difficulties and

opportunities. This also allows us to position our

own contribution, which includes the design of the

headset reported in section 5.

First we consider H(s), consisting of (1) the

services provided by the therapist and (2) the

training equipment. There are several good reasons

why the therapist is involved. These include the

professional judgment on the needs and on the

progress of the patient, checking for counter-

indications together with the personal attention and

motivation given to the patient by the therapist. But

at present the therapist is also needed to (a) position

the EEG cap or headset onto the patient's head, (b)

BIOSIGNALS 2008 - International Conference on Bio-inspired Systems and Signal Processing

add conducting gel and (c) set-up the equipment.

The latter reasons a-c, although indispensible at

present, form at the same time a bottleneck. They

are a bottleneck in the following sense:

 the therapist gets involved only after the user

has been diagnosed to be ill or has been

classified according to a certain disorder. The

user has become a patient. In other words,

feedback training is not used as a preventive

tool, which could help a user to maintain a

healthy equilibrium;

 the therapist is needed to take care of

positioning the cap, applying conducting gel

and setting-up equipment, in other words:

without the therapist the neurofeedback

training cannot happen. The user cannot train

at home. This limitation in turn, limits the

number of training sessions, both in frequency

and duration;

There is one conclusion which we draw

immediately, namely that if the practical difficulties

related to the cap, the gel and the equipment can be

solved or relieved, home usage and preventive usage

are much more likely to happen.

But the contribution of the therapist is larger.

There is no substitute for his or her professional

judgment and personal attention. So the therapist

keeps a role, although perhaps not during each

session. This leaves a question about the

unsupervised sessions. The question is how the user

will keep him or herself motivated to train regularly.

Our answer is simple: training should be fun.

Phrased differently, there have to be motivational

elements built-in to the training equipment (gaming,

flow). We shall address this in the next section.

Now let's have a closer look at the user. G(s) is

an intertwined mind-body complex. The output of

the training equipment enters the user through

suitable interface elements first at his or her bodily

level: ears, eyes, perhaps touch. This is not very

problematic because excellent interface elements

such as loudspeakers, earphones, video screens etc.

are readily available. There is also no lack of

pleasant and aesthetic audio-video content. The

user's output, the EEG signal, is also wiretapped at a

physical level. This is a source of difficulties, as

already mentioned. That is why we consider the

design of an easy-to-use headset to be an essential

step.

At a higher level inside the user there are

processes of perception, cognition, volition and

consciousness going on. It is at this level that issues

of attention and attention deficit become noticeable,

as in ADHD. But it is also at this level where fun,

beauty and motivation occur.

4 PRACTICAL AND

MOTIVATIONAL ELEMENTS

To enable neurofeedback training in the home

context, the product should be designed according

specific requirements, aiming at improving the

training methods and enabling users to operate

independently. These requirements should include

the ability to (a) easily locate sensor positions, (b)

apply sensors and gel without help, (c) measure

electrode impedance and to act accordingly and (d)

easily clean and reuse the system.

Bringing neurofeedback training to the home

environment would also benefit the financial aspect,

reducing the cost of training as a result of decreasing

expert involvement. Furthermore, only an EEG

measurement device and software would have to be

acquired, since we can assume that most households

already have a computer.

In addition to the requirements, the training

software should be intrinsically motivating (playful)

(Rauterberg, 2004) as mentioned before. For

designing training software, we formulated the

following assumptions based on observations and

experiences:

 A

: that it is helpful to give the user rewards

based on performance;

 A

: that it is helpful to simulate elements from

an assumed context of use;

 A

: that it is helpful to provide the user with

quantitative performance data.

Taking a closer look at these assumptions, we

find several opportunities to apply concepts of

gaming to the proposed neurofeedback training

software, for example:

 reward schemes with levels, credits, bonuses,

etc. (A

);

 sound generation and rendering of high-

resolution real-time environments (A

);

 statistics and graphs etc. (A

);

Acknowledging this overlap and the fact that

intrinsic motivation is found in gaming (Malone,

1980), implies stimulating concentration and

motivation. We suggest to incorporate gaming

aspects in neurofeedback training software (e.g.

‘Brainball’ of Hjelm (Hjelm, 2003) is considered to

be an interesting example).

EEG HEADSET FOR NEUROFEEDBACK THERAPY - Enabling Easy Use in the Home Environment

We aim for neurofeedback training applied in

such a practical and motivational way that it can be

considered a game. In the next section we describe

the creation of a product in which we applied the

aforementioned shift from treatment to play, in an

attempt to show it is not only desirable but also

achievable. However, we have to consider the fact

that the rich media and gaming experience people

are used to, has its influence on the expectancy for

neurofeedback software; we could say that

neurofeedback training has to keep pace with

gaming.

5 DEVELOPMENT OF EEG

HEADSET

In order to reach the aims as discussed earlier, we

developed a headset which gives the user the

opportunity to retrieve EEG signals from his scalp in

a convenient way. Combined with neurofeedback in

the form of a 3D environment, this headset is the

first step to the vision that combines gaming, EEG

signals and therapy within the home environment. In

contrast to the traditional ElectroCap, the headset

(Figure 3) can be put on the head with ease and

without professional help. Operations as locating

contact points, attaching sensors and applying

conducting gel are all integrated in the headset,

enabling neurofeedback applications for non-expert

users in the home environment. The product

includes a headset, electronics, sensors, hard- and

software.

Figure 3: EEG headset.

The headset has a flexible construction to adapt

to a variety of head dimensions. The material

flexibility enables a one-size fits all design,

reckoning with a head breadth variety of 1.3 inch.

This figure is based on difference in dimensions of

children age 6 (5

percentile girls, 5.1 inch) to adults

(95

percentile male, 6.4 inch) (Woodson, Tillman,

& Tillman, 1992). The mechanical properties of the

material make sure that the headset is stable on the

head, which is necessary for the sensors to work

properly. Soft cushion pads on the sides provide the

headset with more stability and comfort.

Embedded electronics take care of the 2-channel

EEG signal amplification, enhancing signal strength.

Two sensors (positioned on points C3 and C4 of the

international 10-20 system) measure EEG signals

while the sensor positioned on the Cz location acts

as a ground (a common reference electrode

placement). The fourth sensor is placed near the Fz

position and acts as an active electrode, according to

the ‘driven right leg’ (DRL) principle (Webster,

1998) to minimise distortion. This position has been

chosen for its symmetric location relative to points

C3 and C4, preventing topographic distortion.

Moreover, it increases the stability of the headset by

restricting the headset to tilt when nodding or

moving the head otherwise.

The sensors (Figure 4) can be clicked in and out

of slots in the headset. Those slots enable two

degrees of freedom (translation and rotation),

enabling the sensors to adapt to different head

shapes and dimensions autonomously. A flexible

material between the sensor and the headset allows

for vertical sensor translation and applies a force,

pushing the sensor on the scalp. Additionally, the

shape of the sensor slots enables a 20˚ rotation to

both sides. The combination of translation and

rotation assures that the sensors can adapt to

different head shapes and helps to maintain a stable

signal.

Figure 4: EEG headset sensor.

The detachable sensors measure EEG signals on the

scalp. Therefore the scalp has to be scratched in

BIOSIGNALS 2008 - International Conference on Bio-inspired Systems and Signal Processing

order to gently remove dead skin cells that disturb

the signals, this is achieved by simply twisting a part

of the sensor called ‘scratcher’. Furthermore, a

container for conducting gel is implemented in the

scratcher. The gel can simply be applied by pushing

a plunger. By doing so, the gel connects the tin (Sn)

electrode with the scalp, bridging the gap caused by

the hair. Currently, the sensors need to be filled with

gel using a syringe, but in the future gel capsules or

disposable sensors will be developed to nullify this

inconvenience.

The 2 channel EEG signal retrieved from the

sensors is magnified by the analogue pre-amplifier

in the headset. This signal is converted using an A/D

converter, which is located in an external box. The

box also contains a power source and an opto-

coupler for safety reasons. The signal is send to a

computer, which feeds the retrieved bio-signals back

through a 3D environment. This 3D environment

includes different sessions for neurofeedback

training, where the user learns to train specific brain

capacities using goals and instant brain feedback

(e.g. the intensity of a fire is related to the level of

concentration, see Figure 5).

Figure 5: Screenshot of 3D environment ‘AdventurePark’.

6 EVALUATION

To test the quality of the headset, we developed an

initial functional prototype since the actual headset

has not been manufactured yet. This prototype was

used in an evaluation of the headset principle.

6.1 Research Question

An impedance limit of 5 kΩ is often mentioned for

clinical use of the EEG, in order to prevent signal

distortions (Heinrich, 2007). In this project however,

we allow ourselves a 10 kΩ impedance limit

because of two reasons. Firstly, we assume that the

distortion level depends on the length of the wires,

which is reduced by embedding the preamplifier in

the headset. Secondly, the embedded DRL system

filters a lot of distortion, ensuring a relatively high

quality EEG signal. This leads to the following

formulation of the research question:

All sensors of the headset have impedance lower

than 10 kΩ when placed on the head of the

participants.

6.2 Method

We conducted an experiment in which we compared

the impedance values of the sensors against the

requirements of 10 kΩ.

6.2.1 Apparatus and Measurements

To prevent for extreme cases influencing the

reliability of the experiment, head dimensions of the

participants have been measured by the

experimenters (Figure 6). A tape measurer has been

used to measure the ear to ear distance along the

skull (from the points of attachment of the ears along

the centreline of the international 10-20 electrode

placement system). For the head breadth, measured

straight above the ears, a self-made sliding calliper

has been used (Figure 7).

Figure 6: Head dimension measurement method.

Figure 7: Self-made sliding calliper.

The impedance of the sensors C3, C4 and Fz were

measured relatively to sensor Cz. For this, the 2-

EEG HEADSET FOR NEUROFEEDBACK THERAPY - Enabling Easy Use in the Home Environment

Channel EEG MindSurfer Standard hardware has

been used. The value was determined by reading an

analogue gauge implemented in the Jukebox mental

training software installed on a notebook (NEC

Versa P520). The gauge was calibrated on forehand

with resistors with fixed values.

The functional prototype (Figure 8) was made of

4mm thick Plexiglass and shaped according the

headset design.

6.2.2 Participants and Procedure

Twenty students (17 male, 3 female, age 20-24)

participated in this test. The tests took place in an

cutoff room. After welcoming the participants, they

were asked to sit in a comfortable chair.

Figure 8: Picture user test participant.

After a short introduction the participant was

asked to put on the headset. The experimenters

assisted in positioning the headset and parting the

hair out of the way. Participants gently scratched

their own head with the sensor scratchers and

applied the gel. The participants were able to see the

impedance value on the notebook screen and were

instructed to optimise the conductance by trying to

bring this value down to less than 10 kΩ. They were

told to achieve this by using the sensor scratchers

and by shaking the headset slightly, allowing the gel

to settle. After optimising the conductance for about

one minute, the researcher recorded the impedance

value of each sensor. This value indicates the

impedance between sensor Cz and sensors C3, C4

and Fz. 2-Channel EEG hardware was used, whereas

three values had to be recorded. Therefore the C3

sensor wire was connected to the Fz sensor after

recording its value, the C4 wire was disconnected at

that time.

After each test, all gel was removed from the

sensors and new gel was inserted, to ensure the same

conditions for all participants.

6.3 Results

We calculated the mean and standard deviations

of all measurements and report on them in Table 1

and illustrate them in

Figure 9.

A t-test showed that the mean impedances for C3

(t(19)=-4.616, p<.001), C4 (t(19)=-4.082, p=.001)

and Fz (t(19)=-7,452, p<.001) were significantly

lower than 10 kΩ.

The head dimensions of the participants are

within the 5

and 95

percentile as defined in the

Human Factors Design Handbook (Woodson et al.,

1992).

Table 1: Head dimensions and sensor impedances.

mean std dev

Head breadth (cm) 15.01 0.52

Ear to ear distance (cm) 31.75 1.40

Sensor C3 (kΩ) 6.70 3.20

Sensor C4 (kΩ) 7.10 3.18

Sensor Fz (kΩ) 5.40 2.76

Figure 9: Median, quartile ranges and outliers of result set.

6.4 Conclusions

Based on the results of this evaluation we can state

that the functional prototype reaches impedances

significantly lower than the desired 10 kΩ limit. In

addition, based on our experience there seem to be

some aspects that would probably decrease the

impedance even more, although it should be

mentioned that no extensive testing of these aspects

has been done. Firstly, based on our own experience,

it seems that having some experience with the

headset plays an important role in creating good

conductivity. Given a certain amount of training,

most users are likely to achieve impedance levels

BIOSIGNALS 2008 - International Conference on Bio-inspired Systems and Signal Processing

below 5k. Secondly, the impedance measurements

were recorded after approximately one minute of

having the gel applied on the head, yet our

experience gives the impression that the impedance

tends to decrease a bit over time.

The results are considered promising and in

manufacturing the headset, there should be aimed

for the same characteristics as the functional

prototype. However, this doesn’t imply that the

manufactured headset will actually behave in the

exact same way and therefore we propose an

evaluation of the final product as well.

7 DISCUSSION

We identified several possible issues holding back

the development and implementation of

neurofeedback therapy. New aims of neurofeedback

are suggested, including: 1) a focus

shift in healthcare from cure to prevention,

2) increasing the focus on mental wellbeing in

healthcare, 3) elevating the standard of living by

enabling users to consciously train brain signals 4)

implementing gaming approaches in neurofeedback

to increase intrinsic motivation.

In an attempt to make neurofeedback training

more accessible by combining the therapy with

gaming in the home environment, we designed a

headset that can be used all by oneself in

combination with a 3D gaming environment for

desktop pc. We evaluated a prototype of the headset

and proved that the impedances of the sensors were

significantly below 10 kΩ. Of course, future work

will have to imply the actual realization and testing

of this headset, more iteration steps will have to be

made before starting large volume production.

To our knowledge, this headset is one of the first

attempts to apply enjoyable neurofeedback in the

home environment. It can be used without

supervision of a medical expert and can be operated

by a single user, lowering the practical barriers of

neurofeedback therapy, combined with motivational

elements in the form of an entertaining 3D game.

Hopefully, this will increase the societal acceptance

of neurofeedback

We argue to continue developing headsets

implementing EEG sensors, in order to stimulate the

ease of use. Future work could include gel capsules

to prevent hassle with syringes and cleaning, or even

dry EEG sensors. Furthermore, a focus should

remain on using gaming as motivational tool in

neurofeedback therapy and to support society

adopting neurofeedback training to increase overall

mental wellbeing. Neurofeedback therapy already

exists for over a decade; still the general public is

unaware of the broad spectrum of opportunities

neurofeedback could provide. Curing mental

illnesses is of course a big opportunity, but

neurofeedback has the potential to go much further.

Instead of curing, it could prevent mental illness to

happen or even exceed current human capabilities.

For this to be realized, the neurofeedback research

community should focus on practical and

motivational issues that hold back the

implementation of neurofeedback therapy, and

create a shift towards a society that greatly benefits

from its possibilities.

ACKNOWLEDGEMENTS

We would like to thank Joep Frens, Chet Bangaru,

Jacques Terken and Geert van den Boomen from the

Department of Industrial Design, Eindhoven

University of Technology, who helped shaping our

thoughts and enabled the design of the headset.

Furthermore we would like to thank Geert-Jan

Driessen, Pierre Cluitmans and Frans Tomeij for

sharing their knowledge and insights on the matter.

The Mind Connection in Maastricht has been a

contributor of the project.

REFERENCES

ElectroCap International Inc. (2004). ElectroCap

International Inc. Retrieved April 3, 2007, from

http://www.electro-cap.com

Emotiv Systems Inc. (2007). Project Epoc. Retrieved July

13, 2007, from http://www.emotiv.com

Evans. (1999). Introduction to quantitative EEG and

neurofeedback.

Fox, K. R. (1999). The influence of physical activity on

mental well-being. Public Health Nutr, 2(3A), 411.

Guger Technologies OEG (2007). g.EEGcap. Retrieved

July 13, 2007, from http://www.gtec.at/products/

Accessories/gEEGcap.htm

Hammond, D.C. (2007). What is neurofeedback. Journal

of Neurotherapy, 10(4), 25.

Heinrich, H., Gevensleben, H., Strehl, U. (2007).

Annotation: neurofeedback - train your brain to train

behaviour. J Child Psychol Psychiatry, 48(1), 3-16.

Hjelm, S. I. (2003). Research + design: the making of

Brainball. interactions, 10(1), 26-34.

Kendell, R. E. (2001). The distinction between mental and

physical illness. Br J Psychiatry, 178, 490.

Lubar, J. F. (1995). Evaluation of the effectiveness of

EEG neurofeedback training for ADHD in a clinical

setting as measured by changes in T.O.V.A. scores,

EEG HEADSET FOR NEUROFEEDBACK THERAPY - Enabling Easy Use in the Home Environment

behavioral ratings, and WISC-R performance.

Biofeedback Self Regul, 20(1), 83.

Malone. (1980). What Makes Things Fun to Learn?: A

Study of Intrinsically Motivating Computer Games.

Ph.D. dissertation, Department of Psychology,

Stanford University.

Rauterberg, M. (2004). Positive effects of entertainment

technology on human behaviour. In Rene Jacquart

(Ed.), Building the Information Society (pp. pp. 51-

58): IFIP, Kluwer Academic Press.

Roskamp, H. (2007, feb/mar 2007). Wordt baas in eigen

brein. Bright, 14, 57-61.

SMART BrainGames (2007). SMART BrainGames

PlayStation System. Retrieved July 13, 2007, from

http://www.smartbraingames.com/product.asp?id=1289

Splunter, J. van (2002). Building Technology for People.

Retrieved October 29, 2007, from

http://ww5.newscenter.philips.com/assets/Downloadab

lefile/CEBIT-Asia-Keynote-speech-Van-Splunter(1)-

3734-1440.pdf

Webster, J. G. (1998). Medical Instrumentation (3rd ed.).

New York: John Wiley & Sons.

Woodson, W. E., Tillman, B., & Tillman, P. (1992).

Human Factors Design Handbook (2nd ed.): McGraw

Hill.

BIOSIGNALS 2008 - International Conference on Bio-inspired Systems and Signal Processing