CBM342 BRAIN COMPUTER INTERFACE AND APPLICATIONS

UNIT I INTRODUCTION TO BCI

Fundamentals of BCI – Structure of BCI system – Classification of BCI – Invasive, Non-
invasive and Partially invasive BCI – EEG signal acquisition - Signal Preprocessing – Artifacts
removal.

1.1 BRAIN COMPUTER INTERFACE

 Current Human-Computer Interaction (HCI): Human controls virtual or physical objects
using muscular activity.
Examples:
 Mouse (hand/finger movements)
 Keyboard (finger/hand movements)
 Joystick (hand/arm movements)
 Steering wheel, buttons, and pedals (hand/arm/feet/leg movements)

Brain-Computer Interface (BCI): A device that utilizes brain activity for direct control of
physical or virtual objects without using muscular activity or body movements.
Some Applications of BCI
✦ Improved communication and control for paralyzed and locked-in patients (e.g. stroke,
ALS, spinal injury patients)
✦ Applications in health and safety
E.g. Early detection, diagnosis, and treatment of symptoms
E.g. Alertness monitoring in critical occupations (e.g. night drivers, pilots, railway
“engineers”)
✦ Computer-aided education and learning
E.g. Brain-activity based presentation of material?
✦ Augmented cognition (brain-body actuated control)
E.g. Air Force research using hybrid brain-body interfaces for speeding up responses during
flight
✦ Entertainment and Security
 E.g. Video games, TV/web browsing for patients,
E.g. Better lie detection devices and “brain fingerprinting
✦BCIs in Sci-Fi Fictions
Ex : Donovan’s Brain, 1953, Johnny Mnemonic, 1995 , The Matrix, 1999

1.2 Fundamentals of Brain Computer Interface (BCI)

 A BCI is a system that measures central nervous system (CNS) activity and converts it into
artificial output that replaces, restores, enhances, supplements, or improves natural CNS
output and thereby changes the ongoing interactions between the CNS and its external or
internal environment.

 BCI is simply a hardware and software communications system that enables humans to
interact with their surroundings by directly acquiring and analyzing neural signals between
the brain and the computer
 A Brain Computer Interface (BCI – also called Brain Machine Interface) represents a non-
muscular channel for sending messages, “mental commands”, to an automated system such
as a robot, prosthesis or a cursor on a computer screen. (ie) BCI connects the brain with a
computer system.

 Diagram of holiday greeting that Herbert Jasper sent to Hans Berger in 1938.
 Unlike the conventional systems which are controlled by computer, the BCI is controlled by
human brain signal. Basically, BCIs are of active types which are controlled by means of
endogenous tasks such as motor imagery and mental arithmetic operations, and reactive types
that are controlled using external stimulation like auditory, visual and haptic.
  BCI introduces a direct communication channel between the brain and the external world,
providing a special communication and control channel for people with disabilities, but also a
new control channel for those without disabilities.
 The system does actually not use normal output pathways of the central nervous system, as
nerves or muscles do, but relies only on the identification and interpretation of the
physiological activity patterns in different areas of the brain.
 At present, BCI have been proposed as a tool for diagnosing, treating and following up many
other neurophysiological and neuropsychological disorders
 The various applications developed require different areas for signal recording and different
signal quality is needed; hence several recording methods are suitable for use:
1. EEG (ElectroEncephaloGraphy) :
2. fMRI (functional Magnetic Resonance Imaging)
3. MEG (MagnetoEncephaloGraphy)
4. PET (Positron Emission Tomography)
5. Optical Imaging
6. ECoG (ElectroCorticoGraphy)

1.3 Structure of BCI system

 BCI systems offer a new way to communicate between the human brain and a computer. A
BCI system analyses the brain physiological activity recorded by electrodes in order to
understand the user’s intention.
 Brain Computer Interfaces (BCI) is a system that allow communication between the brain
and various machines.
They work in 3 main steps:
1. Collecting brain signals
2. Interpreting them
3. Outputting commands to a connected machine, according to the brain signal received.
Figure 1.1 illustrates the components of a generic BCI. The aim is to translate brain activity into control
commands for devices and/or stimulate the brain to provide sensory feedback or restore neurological
function. One or more of the following processing stages are typically involved:
1. Brain recording: Signals from the brain are recorded using either invasive or non-invasive
recording techniques.
 1. Invasive Approaches:
2. Non-Invasive Brain Imaging:
3. Partially Invasive
Invasive
They are implanted directly into the brain during neurosurgery. The micro electrodes are placed
directly into the cortex, measuring.
Recording Activities of Neurons inside the Brain using Electrodes and Electrode Arrays
* Typically only in animals (rats and monkeys)
 Recording Electrical Activity from the Brain Surface (Electrocorticography or ECoG)
* In humans (patients scheduled for brain surgery)
 Implants and Neural Stimulation
* In animals and humans (e.g., Parkinson’s patients)
Non Invasive
The sensors are placed on the scalp to measure the electrical potentials produced by the brain
(EEG) or the magnetic field (MEG).
1. fMRI (Functional Magnetic Resonance Imaging): Measures changes in blood flow due to
increased brain activity * Good spatial resolution but too slow for real-time BCI
2. MEG (MagnetoEncephaloGraphy): Measures changes in magnetic fields due to neural activity
* Good spatiotemporal resolution but expensive and cumbersome
3. EEG (ElectroEncephaloGraphy): Measures voltage changes at the scalp due to neural activity
* Good temporal resolution but poor spatial resolution * Inexpensive and therefore most
common in current BCIs

Semi Invasive
 The electrodes are placed on the exposed surface of the brain electrocorticography (ECoG)
It is called semi invasive but it still requires a craniotomy to implant the electrodes. For this
reason it is used only when surgery is necessary for medical reasons (epilepsy for example).
 First, the block performs a pre-processing in order to reject artefacts and increase the signal
to noise ratio (SNR). It is also used to ensure that the extracted features are not contaminated
by
1. EMG (electromyography - electrical activity produced by skeletal muscles),
 2. EOG (electrooculography - measurement of eye movements) or some other non-CNS
(central nervous system) artefacts.

2. Signal processing: Raw signals are preprocessed after acquisition (e.g., by band pass filtering) and
techniques for artifact reduction and feature extraction are used.
3. Pattern recognition and machine learning: This stage generates a control signal based on patterns in
the input, typically using machine- learning techniques.
 The features extraction block identifies the parameters from the pre-processed signals thus
discriminating between different classes of commands. Some typical features used in
developed applications are the root mean square amplitude and power density in certain area
of the spectrum. The system can associate the recorded signal with a feature by using a
special classifier.
 After the class feature is identified, the system can associate it with a command for the
application. Typical classifiers for BCI applications are:
1. Linear Support Vector Machine (LSVM),
2. Gaussian Support Vector Machine (GSVM),
3. Neural Network (NN),
4. Fisher Linear Discriminant (FLD) and
5. Kernel Fisher Discriminant (KFD).
 The signal processing block processes all the recorded data and transforms the signals in
commands for the application.
 The system must process all the data very fast in order to provide a real-time operation.
In order to produce a command, the user must execute a specific activity. Thus the
system can associate the produced signals with the generated command.

4. Sensory feedback: The control signal from the BCI causes a change in the environment. (e.g.,
movement of a prosthetic arm or a wheelchair, change in the grip of a prosthetic hand). Some of these
changes can be seen, heard, or felt by the user but in general, one can use sensors in the Environment such
as tactile sensors, force sensors, cameras, and microphones, and use the information from these sensors to
provide direct feedback to the brain via stimulation.
 Figure 1.1. Basic components of a brain- computer interface (BCI)
5. Signal processing for stimulation: Before stimulating a particular brain region, it is important to
synthesize an activity pattern for stimulation that mimics the type of activity normally seen in the brain
region and that will have the desired effect. This requires a good understanding of the brain area being
stimulated and the use of signal processing (and potentially machine learning) to home in on the right
stimulation patterns.
6. Brain stimulation: The stimulation pattern received from the signal processing component is used in
conjunction with invasive or noninvasive stimulation techniques to stimulate the brain.

1.4 Classification of BCI

The three types of BCIs have been developed namely
1. Invasive BCIs,
2. Non-invasive BCIs
3. Partially-invasive BCIs,
Figure 1.3 Three different ways to detect the brain’s electrical activity: EEG, ECoG, and
intracortical
Brain neuro Signal Recording
 The various applications developed require different areas for signal recording and different
signal quality is needed; hence several recording methods are suitable for use:
1. EEG (electroencephalography) : Recording brain activity are based on detecting changes
in electrical potentials in neurons,
2. FMRI (functional magnetic resonance imaging) : Detecting neural activity indirectly by
measuring changes in blood flow that result from increased neural activity in a region
3. MEG (magnetoencephalography): by measuring the minute changes in the magnetic field
around the skull caused by neural activity
4. PET (positron emission tomography), It uses a radioactive substance called a tracer to look
for disease or injury in the brain
5. Optical imaging via multi-photon microscopy can measure activity-dependent changes in
fluorescent signals directly from individual neurons, astrocytes and blood vessels in the
living brain.
6. ECoG (electrocorticography) is the process of recording electrical activity in the brain by
placing electrodes in direct contact with the cerebral cortex or surface of the brain
 Figure 1.4 Recording sites for electrophysiological
signals
used by BCI systems.
(a) Electroencephalographic activity (EEG) is
recorded by electrodes on the scalp.
(b) Electrocorticographic activity (ECoG) is
recorded by electrodes on the cortical surface.
(c) Neuronal action potentials (spikes) or local
field potentials (LFPs) are recorded by electrode
arrays inserted into the cortex
1.5 Invasive BCI
 Invasive BCIs are implanted directly into the
grey matter of the brain during neurosurgery.
 Using chips implanted against the brain that have
hundreds of pins less than the width of a human
hair protruding from them and penetrating the
cerebral cortex, scientists are able to read the firings of hundreds of neurons in the brain.
 The recording itself is not painful because the brain has no internal pain receptors.
 In the case of humans, invasive recordings are taken only in clinical settings such as
during brain surgery or when patients are being monitored for abnormal brain activity.
 A major advantage of invasive recordings is that they allow recording of action potentials
at the millisecond timescale.
 Invasive recording in both humans and animals is based on the technology of electrodes.

1.5.1 Microelectrodes
 Electrodes with tip sufficiently small to penetrate a single cell in order to obtain potential
from within the cell are called as microelectrode. The size of the microelectrode must be
small with respect to the cell dimensions to avoid cellular injury and cell damage thereby
changing the cell’s behavior.
 A microelectrode is simply a very fine wire or other electrical conductor used to make
contact with brain tissue.
  A typical electrode is made of tungsten or platinum-iridium alloy and is insulated except
at the tip, which measures around 1μm in diameter (a neuron’s cell body diameter is in
the range of tens of μm).
 In addition to small in size, the electrode used for measuring intracellular potential must
also be strong enough so that it can penetrate the cell membrane and remain mechanically
stable.
 Microelectrodes have tip diameters ranging from approximately 0.05 to 10μm.
 Microelectrodes are of two types-
1. Metal Microelectrode and
2. Micropipet Microelectrode
1. Metal Microelectrode:
• Metal microelectrodes are formed by electrolytic ally etching the tip of fine tungsten or
stainless steel wire or platinum iridium alloy to the desired size. Then the wire is coated
with an insulating material almost to the tip.
• Some electrolytic processing like chloriding the tip is also carried out to reduce the tip
impedance. The electrode-electrolyte interface takes place where the metal tip contacts the
electrolyte either inside or outside the cell.
• The etched metal needle is then supported in a larger metallic shaft that can be insulated.
This shaft serves as a mechanical support for the microelectrode and lead wire is
connected to it.
• The microelectrode and the supporting shaft are insulated by a thin film of polymeric
material or varnish. Only the extreme tip of the electrode remains uninsulated.

Metal Microelectrode
 Figure: Position of Electrode
2. Micropipet Microelectrode
 Non-metallic microelectrode is called micropipette
 Consists of a glass micropipet with diameter of tip about 1micrometre.
 The micropipet is filled with an electrolyte usually 3 M KCI which is compatible with the
cellular fluid.
 In intracellular recordings neuroscientists, use a glass micropipette electrode filled with a
weak electrolyte solution similar in composition to intracellular fluid.

Figure A glass micropipette electrode

 1.5.2 Intracellular Recording
• The most direct way of measuring the activity of a neuron is through intracellular
recording, which measures the voltage or current across the membrane of the neuron.
• The most common technique, known as patch clamp recording uses a glass micropipette.
• To record from the cell, the glass microelectrode is placed next to the cell and, using
gentle suction, a piece of the cell membrane (a “patch”) is drawn into the electrode tip,
forming a high-resistance seal with the cell membrane.
1.5.3 Extracellular Recording
• One of the most common types of invasive recordings, performed especially in the intact
brains of animals, is extracellular recording of a single neuron
• The depth of the microelectrode is adjusted until it comes close enough to a cell body to
pick up the electrical fluctuations caused by action potentials generated by the cell.
Figure (a) Intracellular recording using the patch clamp technique
(b) Intracellular versus extracellular recording of spikes
1.5.4 Tetrodes
• It is possible to record from multiple neurons simultaneously by using more than one
electrode
• One common configuration is called a tetrode, where four wires are tightly wound together in
a bundle.
• It may be possible to record from up to 20 neurons with a single tetrode.
1.5.5 Multielectrode Arrays
• To record from larger n umbers of neurons,
microelectrodes can be arranged in a grid-like
structure to form a multi-electrode array of m × n
electrodes.
• The most common types of implantable arrays are
microwire, silicon-based, and flexible
microelectrode arrays.
• Microwire arrays use tungsten, platinum alloy, or steel electrodes.
• Silicon-based arrays include the so-called Michigan and Utah arrays.
• Flexible arrays rely on polyimide, parylene, or benzocyclobutene rather than silicon for
recording.
• The major advantage of multielectrode arrays over more conventional single-electrode
systems is increased spatial resolution;
• Implantable arrays also have their disadvantages, especially if the implanted device remains
in the brain tissue for a long time.
 • non-neuronal cells known as glial cells surround the implanted device, leading to the
formation of first scar tissue and then an insulating sheath around the array, resulting in
significant reduction in recorded signal quality.

1.6 Non-Invasive BCIs: EEG-based Systems

 The sensors are placed on the scalp to measure the electrical potentials produced by the brain
Electroencephalography (EEG) or the magnetic field (MEG).

1.6.1 Electroencephalography (EEG)

 Electroencephalography (EEG) is a popular noninvasive technique for recording signals from
the brain using electrodes placed on the scalp. The electrodes can read brain signals.
 EEG signals reflect the summation of postsynaptic potentials from many thousands of
neurons that are oriented radially to the scalp.
 EEG signals reflect the summation of postsynaptic potentials from many thousands of
neurons that are oriented radially to the scalp.
 The measured signals are in the range of a few tens of microvolts, necessitating the use of
powerful amplifiers and signal processing to amplify the signal and filter out noise. The weak
amplitude of the underlying brain signal also means that EEG signals can be easily corrupted
by muscle activity and contaminated by nearby electrical devices
 The electrical recordings from the surface of the brain or from the outer surface of the head
demonstrate continuous electrical activity in the brain. The intensities of the brain waves on
the surface of the scalp range from 0-300μV and their frequencies range from 0.5 Hz to
100Hz.
 Brain waves are irregular and no general pattern can be discerned in the EEG. However, at
other times distinct patterns will appear. In normal persons it is classified as alpha, beta,
theta, gamma and delta waves.
EEG Recorder:

Chart drive EEG

Electrode Pre- and writer representation
Montage amplifier unit on chart
selector and Filters

EEG
Computer representation
Subject head Analog to on computer
with 10-20 Digital
electrode Converters
system (ADC) CRO
Oscilloscope

 In a typical setup, each EEG electrode is connected to one input of a differential

amplifier, and the other input is connected to a reference electrode – for instance, nasion
or linked mastoids (average of the two mastoids).
 The amplification of voltage between the active electrode and the reference is typically
1,000–100,000 times.
 The amplified signal is passed through an anti-aliasing filter and then digitized via an
A/D (analog to digital) converter at sampling rates of up to 20 kHz depending on the
application (typical sampling rates for BCI applications are in the range of 256 Hz–
1kHz).
 After digitization, the EEG signal may be additionally filtered by a 1–50 Hz band pass
filter. This excludes noise and movement artifacts in the very low and very high
frequency ranges.
 An additional notch filter is typically used to remove “line noise” caused by the electrical
power supply
Advantages
1. Temporal resolution is good.
2. Easy measurement
Disadvantages
1. Tangential Currents of the scalp and currents originating deep in the brain are also not
detected by EEG.
2. Poor spatial resolution
1.6.2 Magnetoencephalography (MEG)
 Magnetoencephalography (MEG) measures the magnetic fields produced by electrical
activity in the brain using SQUIDs (superconducting quantum interference devices).
 Both MEG and EEG signals originate from the net effect of ionic currents flowing in the
dendrites of neurons due to synaptic inputs from other neurons.
 These currents produce an orthogonally oriented magnetic field.
Advantages of MEG
1. MEG offers high temporal resolution
2. The magnetic fields produced by neural activity are not distorted by the intervening
organic matter
3. MEG offers better spatial resolution than EEG

Limitations of MEG
1. More expensive than EEG systems
2. Bulky
3. Not portable
4. Require a magnetically shielded room to prevent interference from external magnetic
signals,

1.6.3 Functional Magnetic Resonance Imaging (fMRI)

 fMRI indirectly measures neural activity in the brain by detecting changes in blood flow
due to increased activation of neurons.
 Neural activity, consume more oxygen, triggers a dilation of local capillaries, result ing in
an increased inflow of highly oxygenated blood.
 The fact that de- oxygenated hemoglobin is more magnetic than oxygenated hemoglobin
is exploited in fMRI to generate images of different cross sections of the brain.
 The signal recorded by fMRI is called the blood oxygenation level dependent (BOLD)
response.
 This hemodynamic response is comparatively slow – it appears several hundred
milliseconds after neural activity and peaks at 3–6 seconds, before falling back to
baseline in another 20 seconds.
Advantages
 Higher spatial resolution than other noninvasive techniques
Limitation
 Poor temporal resolution
1.6.4 Functional Near Infrared (fNIR) Imaging
 Functional near infrared (fNIR) imaging is an optical technique for measuring changes in
blood oxygenation level caused by increased neural activity in the brain.
 This type of imaging is based on detecting near- infrared light absorbance of hemoglobin
in the blood with and without oxygen.
 infrared light can penetrate the skull and enter a few centimeters into the cortex
 Infrared emitters placed on the scalp send infrared light through the skull; this light is
partly absorbed and partly reflected back through the skull, where it is detected by
infrared detectors.
 Similar to EEG, using a number of evenly spaced “optodes” across the head allows one to
construct a two- dimensional map of neural activity across the brain surface.
Advantages
1. Can image deep regions of the brain
2. Subjects are not restricted in their movement
3. Not as susceptible to muscle artifacts compared to EEG
4. Less expensive than fMRI
5. Portable
Limitations
1. More prone to noise
2. Less spatial resolution.
1.6.5 Positron Emission Tomography (PET)
 PET measures emissions from radioactively labeled, metabolically active chemicals that
have been injected into the bloodstream for transportation to the brain.
 The labeled compound is called a radiotracer
 Sensors in the PET scanner detect the radioactive compound as they make their way to
various areas of the brain as a result of metabolic activity caused by brain activity
 The most commonly used radiotracer is a labeled form of glucose.
 Advantage
 The spatial resolution of PET can be comparable to fMRI, but
Limitation
 The temporal resolution is typically quite low
 Need to inject radioactive chemicals into the body
 Rapid decay of radioactivity, which limits the amount of time available for experiments

1.7 PARTIALLY-INVASIVE BCI or SEMI-INVASIVE

 Partially invasive BCI devices are implanted inside the skull but rest outside the brain rather than
within the grey matter.
 They produce better resolution signals than non-invasive BCIs where the bone tissue of the cranium
deflects and deforms signals and have a lower risk of forming scar-tissue in the brain than fully-
invasive BCIs.

1.7.1 Electrocorticography (ECoG)

 Electrocorticography (ECoG) measures the electrical activity of the brain taken from
beneath the skull in a similar way to noninvasive electroencephalography, but the
electrodes are embedded in a thin plastic pad that is placed above the cortex, beneath the
dura.

 Electrocorticography (ECoG) is a technique for recording brain signals that involves

placing electrodes on the surface of the brain.
  Typically, a grid or strip of m × n electrodes is implanted, where the values of m and n
vary between 1 and 8.
 ECoG electrodes can be tipped with carbon, platinum, or gold alloy, and are typically 2–5
mm in diameter.
 The spacing between grid electrodes is usually 10 mm to 1 cm.
 ECoG electrodes can record the electrical fluctuations caused by the coherent activity of
large populations of neurons.
 ECoG is a partially invasive compromise between invasive multi-electrode arrays and
noninvasive EEG
Advantages
• Safer than arrays implanted inside the brain
• Less likely to wear out
• Offers greater spatial resolution than non-invasive techniques
• Higher Amplitude
• Broader spectral bandwidth
• Less vulnerability to artifacts
Limitations
• Can be currently used only in surgical settings.
• Only surgically relevant portions of the brain can be recorded
• there may be interference due to drugs or patient-related conditions such as seizures.

1.7.2 MicroECoG
• The relatively large size of ECoG electrodes is being addressed by researchers using
microECoG electrodes.
• These microelectrodes are only a fraction of a millimeter in diameter and spaced only 2–3
mm apart in a grid.
• This opens up the possibility of decoding fine movements, such as the movements of
individual fingers, or even speech, without actually penetrating the brain.

1.7.3 Optical Recording

The most prominent of these are imaging techniques based on
• voltage-sensitive dyes and
 • two-photon fluorescence microscopy.
1.7.3.1 voltage-sensitive dyes
• operate on the principle that once neurons are stained with a voltage-sensitive dye, their
electrical activity can be imaged because the dye responds to changes in membrane
potential by changing its absorption and/or fluorescence.
• Styryl or oxonol dyes have been used to stain the upper layers of sensory cortex.
• A photodiode array which acts as detector in receives light from many neurons and thus
the recorded optical signals correspond to summed responses from several
simultaneously active neurons.
• Voltage- sensitive, dye-based optical imaging is particularly useful for imaging
macroscopic features of the brain.
1.7.3.2 Two-photon calcium imaging
• for more targeted imaging of neurons, the technique of two-photon microscopy has been
used
• The technique is based on the fact that electrical activity in neurons is typically associated
with changes in calcium concentration.
• The technique of two photon calcium imaging involves:
(1) using pressure ejection to load neurons with fluorescent calcium-indicator dyes
(2) monitoring changes in calcium fluorescence during neural activity using two-photon
microscopy.

1.8 EEG SIGNAL ACQUISITION

ELECTROENCEPHALOGRAPH (EEG)
Electroencephalography: This is the technique by which the electrical activities of the brain are
studied.
Electroencephalograph: This is the instrument by which the electrical activities of the brain are
recorded.
Electroencephalogram: This is the record or graphical registration of electrical activities of the
brain.
EEG helps in diagnosing disorders or abnormalities related to brain such as
 Tumors: Electrical activity absent in part of hemisphere.
  Epilepsy: Symptom of brain damage due to defects in the birth delivery or head injury
during accident
 Sleep disorders: Discharge of large groups of neurons also due to brain tumor
 Brain activity: To observe and analysis of brain responses to the sensory stimulus.
EEG Electrodes
EEG electrodes transform ionic currents from cerebral tissue into electrical current used in EEG
preamplifier. 5 types of electrodes are used.
1. Scalp: silver pads, discs or cups, stainless steel rods, chlorided silver wires.
2. Sphenoidal: Alternating insulated silver and bare wire and chlorided tip inserted through muscle
tissue by a needle.
3. Nasopharyngeal: Silver rod with silver ball at the tip inserted through the nostril.
4. Electrocorticographic: Cotton wicks soaked in saline solution that rests on brain surface.
5. Intracerebral: sheaves of Teflon coated gold or platinum wires used to stimulate the brain.
1.8.1 EEG Waves:
Electroencephalography (EEG) is a popular noninvasive technique for recording signals from the
brain using electrodes placed on the scalp.
EEG signals reflect the summation of postsynaptic potentials from many thousands of neurons that
are oriented radially to the scalp.
The electrical recordings from the surface of the brain or from the outer surface of the head
demonstrate continuous electrical activity in the brain. The intensities of the brain waves on the
surface of the scalp range from 0-300μV and their frequencies range from 0.5 Hz to 100Hz.
Brain waves are irregular and no general pattern can be discerned in the EEG. However, at other
times distinct patterns will appear. In normal persons it is classified as alpha, beta, theta, gamma and
delta waves.
Table 1. Brain Rhythms and Respective Frequency Band
S.No Name Rhythm Frequency Band (Hz) /
Amplitude ( 100µV - 200 µV ) Waveform
1. Delta (δ)
Associated with deep
sleep Sleeping adults,
Dreaming, infants
 2. Theta (θ)
Associated with
movements or intention to
move Drowsiness
3. Alpha (α) or Mu
Associated with
Unfocusing attention
(relaxation) Reflective
and restful
4. Beta (β) Associated with
alertness and heightened
mental activity (Busy and
Active mind
5. Gamma (γ)
Problem solving and
Concentration > 30 Hz

1.8.2 Placement of Electrodes:

In EEG, electrodes are placed in standard positions on the skull in an arrangement 10-20 system,
a placement scheme devised by the International Federation of societies of EEG. The electrodes
in this arrangement are placed as follows:
For the 10-20 system, the electrodes are labeled by their location on the different brain lobes,
under the scalp:
F for frontal
C for central
T for temporal
P for parietal
O for occipital
and numbered so that the Z (for zero) electrodes run down the center, odd electrodes are on the
left hemisphere, and even electrodes are on the right.
i. Draw a line on the skull from the nasion, the root of the nose, to the inion, ossification
center (bump) on the occipital lobe.
 Figure Placement of electrodes on the scalp for EEG recording
ii. Draw a similar line from the left preauricular (ear) point to the right preauricular point.
iii. Mark the intersection of these two lines as Cz which is the mid-point of the distance
between the nasion and inion (or) the distance between the auricular points.
iv. Mark points at 10, 20, 20, 20, 20 and 10% of the total nasion - inion distance. These
points are Fpz, Fz, Cz, Pz and Oz·
v. Mark points at 10, 20, 20, 20, 20 and 10% of the total distance between the preauricular
points.These points are T3, C3, Cz, C4 and T4. In these odd numbered points T3 and C3
are on the left and even numbered points C4 and T4 are on the right.
vi. Measure the distance between Fpz and Oz along the great circle passing through T3 and
mark points at 10, 20, 20, 20, 20 and 10% of this distance. These are the positions of Fp1,
F7, T3, T5 and O1.
vii. Repeat this procedure on the right side and mark the positions of Fp2, F8, T4, T6 and O2.
viii. Measure the distance between Fp1 and O1 along the circle passing through C3 and mark
points at 25%intervals. These points give the positions of F3, C3 and P3.
ix. Repeat this procedure on the right side and mark the positions of F 4, C4 and P4.
x. Check that F7, F3, Fz, F4 and F8 are equidistant along the transverse circle passing
through F7, Fz and F8 and check that T5, P3, Pz, P4 and T6 are equidistant along the
transverse circle passing through T5, Pz and T6.
 In the above Figure 1.3 the positions of the scalp electrodes are indicated. Further there are
nasopharyngeal electrodes Pg1 and Pg2 and ear electrodes A1and A2.
Before placing the electrodes, the scalp is cleaned, lightly abraded and electrode paste is
applied between the electrode and the skin. By means of this application of electrode paste, the
contact impedance is less than 10KΩ. Generally, disc like surface electrodes are used. In some
cases, needle electrodes are inserted in the scalp to pick up EEG.
1.8.3 EEG Recorder:
1) Montage selector:
 Montages are patterns of connections between the electrodes and the recording channels.
 The patient cable consists of 21 electrodes and is connected to the eight channel selector. The
electrodes are attached to the channel selector in groups of eight called a montage of
electrodes.
 The right ear electrode acts as reference electrode for the right brain electrodes and the left
ear electrode acts as reference electrode for the left brain electrodes.
 The montage selection switch is used for selecting a particular channel. Different channels
convey different information.
 Montages are always symmetrical and hence in the 10-20 electrode placement system the
electrodes are also placed symmetrically.
 The EEG signals are transmitted from the electrodes to the montage selector panel.
 The montage selector of an EEG machine is a large frame which consists of different
switches so as to allow the user to select the desired electrode pair.
2) Pre-amplifier
 As the EEG signals are having amplitude levels in microvolt range it is necessary to be amplified for
further processing.
 It is to ensure that the information from the EEG electrodes is not affected by any external noise. High
gain, high CMRR operational amplifiers are used as a preamplifier.
3) Filters and amplifiers
 The muscle artifacts (noise) are a major problem regarding the EEG waveform.
 These noises can make the representation spurious. So these noise to be filter out from the EEG signal.
 This function is done by a bank of filters in the EEG machine systems, which are selected according to the
need. Amplifiers are used here also to improve the amplitude levels of EEG waveform.
EEG RECORDER
 4)Signal processing unit
 Other facilities are also available to record evoked potentials from sensory parts of the brain such that
there are external stimuli like visual stimulus, audio stimulus and tactile (touch) stimulus. The time delay
between the stimulus and response can also be measured in the signal processing unit.
5) Analog to Digital Converters (ADC)
 Computers and oscilloscopes are used to analyze the EEG waveform.
 As the computers only accept digital data we have to convert the analog EEG information in to digital
form.
 The function of ADC is to convert the analog EEG signal to digital form. Thus the computer can store the
EEG waveform for future reference.
6) Writing recorder and paper drive
 The writing part of an EEG machine is usually consisting of an ink type direct writing recorder.
 In the eight channel pen recorder there are 8 pens such that a pen for each channel.
 The recorder will be a chart paper which is driven by a synchronous motor.
 For the clear representation of the EEG waveform an accurate and stable paper drive mechanism is
provided by the synchronous motor. Also there are provisions to control the paper speed.
EEG Uses
EEGs are used to diagnose conditions like:
 Brain tumors
 Brain damage from a head injury
 Brain dysfunction from various causes (encephalopathy)
 Inflammation of the brain (encephalitis)
 Seizure disorders including epilepsy
 Sleep disorders
 Stroke
An EEG may also be used to determine if someone in a coma has died or to find the right level
of anesthesia for someone in a coma.
1.9 Signal Preprocessing
 The signals that are picked up from the scalp are not necessarily an accurate
representation of the signals originating from the brain (e.g., missing spatial information).
 EEG data contains a lot of noise which can obscure weaker EEG signals (cf. true signal).
  Preprocessing aims at simplifying subsequent processing operations without losing
relevant information.
 An important goal of preprocessing is to improve signal quality by improving the so-
called signal-to-noise ratio (SNR).
 A bad or small SNR means that the brain patterns are buried in the rest of the signal (e.g.
background EEG), which makes relevant patterns hard to detect.
 The acquired EEG signals are highly contaminated by noise and artifacts. Thus, different
preprocessing algorithms are applied on the EEG signals prior to the extraction of
features to reduce the noise.
Below some preprocessing steps commonly used in MI-based BMIs are introduced.
1.9.1. Re-referencing
 Unipolar: When a reference electrode is set on one side of the earlobe (A1 or A2),
spatially symmetric responses appear as asymmetric spatial distribution. A method for
avoiding this is to record both A1 and A2 and set the average of A1 and A2 as the
reference.
 Bipolar: Another method is recalculating the voltage of each electrode at a potential
average over all electrodes is commonly used. Changing the reference electrode is called
re-referencing electrodes. In this case, all electrodes should be symmetrically located
over left and right hemispheres over left and right hemispheres.
 For Unipolar derivation: Now, let N be the number of electrodes and EEGn the
measured EEG at the nth electrode (n = 1,…, N). If the potential of each electrode is Vn
and the potential at the reference electrode is R, then
EEGn = Vn − R
Therefore, the EEG REEGn re-referenced by the average over all electrodes is given as

The point is that without using unmeasured Vn and R, it is possible to reset the reference
potential.
  For bipolar derivation: Let A and B be the potentials at electrodes A and B, respectively.
By definition, the bipolar derivation EEG A−B = A − B amounts to EEGA − EEGB, because
the EEGs at A and B in the monopolar derivation are given as
EEGA = A – R and EEGB = B − R, respectively.
1.9.2 Channel selection
 Multichannel EEG is generally used in BMIs where by performing EEG channel
selection
(1) Improves BMI performance by removing irrelevant channels and
(2) Enhances user convenience from the use of lesser channels
1.9.3 Spectral filtering
 Appropriate spectral filters, such as band-pass or low-pass filters, can enhance the
accuracy and robustness of BMI by reducing the influence of activities that are lying
outside of the frequency regions of interest
 For example, band-pass filters can reduce the negative effects of the power line noise,
EMG, EOG, and other brain activities lying outside the desired frequency band.
 Recent electroencephalograph is configured to acquire data at a sampling frequency
exceeding 1,000 Hz. Indeed, since the required bandwidth is from 1 Hz up to 50 Hz or
less, a high sampling rate such as 1,000 Hz is not necessary.
 A filter widely used in preprocessing of EEG is a Butterworth filter, which has an infinite
impulse response (IIR). Since an IIR filtering can bring phase distortion, zero phase
filtering should be implemented
1.9.4 Spatial filtering
 Various simple spatial filters are used in order to isolate the relevant spatial information
embedded in the signals. This is achieved by selecting or weighting the contributions
from the different electrodes.
 Two simple and popular spatial filters are the common average reference (CAR) and the
surface Laplacian filters.
 These two filters make it possible to reduce the background activity.
 The CAR filter subtracts the average value of all the electrodes from the electrode of
interest, while the Laplacian filter subtracts the average of the surrounding channels from
the channel of interest.
  The most popular spatial filtering algorithm, which is increasingly used for preprocessing
in BMI and has proved to be very efficient, is the common spatial pattern (CSP)
algorithm This algorithm is based on the decomposition of the EEG signals into spatial
patterns.
 These patterns are selected in order to maximize the differences between the classes
involved once the data have been projected onto these patterns.
 Determining these patterns is performed using a joint diagonalization of the covariance
matrices of the EEG signals from each class.
1.9.4 Other preprocessing algorithms
 Preprocessing algorithms - common spatial pattern (CSP) algorithm, stationary
subspace analysis
 Spectro-spatial filters - Moving average filtering, subsampling or baseline correction.
1.10 Artifacts removal or Artifact Reduction Techniques
 Eye movements, eye blinks, eyebrow movements, talking, chewing, and head movements
can all cause large artifacts in the EEG signal.
 Subjects are therefore typically instructed to avoid all movement, and powerful artifact
removal algorithms are used to exclude or filter out portions of the EEG signal corrupted
by muscle artifacts.
 Additional noise sources include changing electrode impedance and varying
psychological states of the user due to boredom, distraction, stress, or frustration
(e.g., caused by BCI mistranslation).
 Artifacts in BCIs are any undesirable signals that originate from outside the brain.
 In EEG BCIs, one often encounters 50/60Hz power- line noise and artifacts originating
from within the subject’s body.
 Artifacts originating from within the subject’s body may include:
(1) Rhythmic artifacts due to respiration and heartbeat
(2) Signal distortion or attenuation due to skin conductance changes
(3) Eye movement and eye blink artifacts which appear as high- amplitude deflections in
brain signals such as EEG with frequencies in the range 3 – 4Hz
 (4) muscle artifacts caused by movements of the head, face, jaw, tongue, neck, and other
parts of the body; EMG artifacts tend to occur maximally in the 30Hz or higher
frequency range.
A true brain- computer interface should possess the ability to eliminate or reduce such
artifacts so that the signals being used to control a device originate solely from the brain.
• Artifacts that originate from outside the body such as 50/60Hz power- line noise can
often be reduced by using a Faraday cage, a physical enclosure made of conducting
material, to block external electrical interference.
• When this is not possible, one can remove such noise in software using filtering
techniques
1.10.1 Thresholding
 The simplest method for automatic artifact rejection is thresholding
 If the magnitude or some other characteristic of a recorded EOG or EMG signal exceeds
a pre- determined threshold, the brain signals recorded during that epoch are deemed to
be contaminated and rejected.
 A similar thresholding technique can be applied directly to brain signals
 The subject is made to make various kinds of eye or body movements to calibrate the
threshold.
Limitation
 A major drawback of the thresholding method is that not all artifact- contaminated data
may be rejected by this method
 The goal of such artifact removal methods is to identify and excise artifacts from data
while preserving neurological phenomena useful for BCI
1.10.2 Band- Stop and Notch Filtering
 Band- stop filtering is a useful artifact reduction technique that attenuates the components
of a signal in a specific frequency band and passes the rest of the components of the
signal.
 Can be performed by first transforming the signal to the frequency domain (e.g., using
FFT), filtering out the desired frequency band, and using the inverse FFT to transform
back to the time domain.
  A commonly used band- stop filter is a notch filter set to the 59–61 Hz band (in the
United States) for filtering out the 60 Hz power- line noise artifact.
 Another band- stop filter set to a low frequency band (e.g., 1–4 Hz) is sometimes used in
EEG recordings to reduce EOG artifacts.
 Low- pass filtering is sometimes used to exclude EMG artifacts.
Limitation
 However, filtering approaches work only when the brain signal of interest does not fall
within the frequency range of artifacts.
 For example, low- pass filtering may remove EMG artifacts, but if the BCI utilizes high-
frequency components of the brain signal, such filtering may eliminate these useful
components as well.

1.10.3 Linear Modeling

 A simple way of modeling the effect of artifacts on a recorded brain signal is to assume
that the effect is additive.
 For example, if EEGi(t) is the EEG signal recorded from electrode i at time t, then a
model could be

Where is the uncontaminated (“true”) EEG signal from electrode i at time

t, EOG(t) is the recorded EOG signal at time t and K is a constant .Therefore,

Limitation

• Applying linear modeling for removing EMG artifacts is more difficult because EMG
artifacts arise from multiple muscle groups, and an additive model with a single EMG(t)
signal as for EOG may not be appropriate

1.10.4 Principal Component Analysis (PCA).

  One can use PCA to find the directions of maximum variance in the recorded brain
data.
 By projecting new data onto the eigenvectors, one can find a set of orthogonal
“components” of the brain signals recorded from a set of electrodes.
 PCA has been shown to be useful for removing EOG artifacts from EEG signals

Limitation

 However, the assumption that artifacts are uncorrelated with the brain signal may not
be appropriate in certain cases, and PCA may be unable to separate these artifacts
from the true brain signals.
1.10.5 Independent Component Analysis (ICA)
 ICA overcomes some of the shortcomings of PCA by seeking statistical independence
rather than decorrelation.
 ICA decomposes brain signals (e.g., EEG) from a set of electrodes into a set of
“components” that are as statistically independent
 By visually inspecting the components or automatic detection using a learned model
for artifacts, one can often identify components due to EOG, EMG, or other artifacts
1.10.6 Independent Component Analysis (ICA)
 ICA overcomes some of the shortcomings of PCA by seeking statistical independence
rather than decorrelation.
 ICA decomposes brain signals (e.g., EEG) from a set of electrodes into a set of
“components” that are as statistically independent
 By visually inspecting the components or automatic detection using a learned model
for artifacts, one can often identify components due to EOG, EMG, or other artifacts
1.10.5 Artifact reduction using ICA.
(A) Five seconds of EEG data
(B) Output of ICA when applied to
the data in (A).