Brain Computer Interface

ByMohammed AbdelAal

Definition

A Brain-Computer Interface (BCI) is a communication system that does not require any peripheral muscular activity. BCI systems enable a subject to send commands to an electronic device only by means of brain activity. Such interfaces can be considered as being the only way of communication for people affected by a number of motor disabilities.

Applications

● Communication

– Spelling Programs● Motor Restoration

– Controlling Robot Arm● Environment Control

– Controlling TV

– Controlling OS

● Locomotion

– Wheel Chair● Entertainment

– Play Games● Neuromarketing

– Emotion Recognition

Wheel Chair

Controlling Robot Arm

Spelling Programs

Entertainment

Neuromarketing

BCI Stages

● Signal acquisition: capture the brain signals and may also perform noise reduction and artifact processing.

● Preprocessing: prepare the signals in a suitable form for further processing.

● Feature extraction: map the brain signals onto a vector containing effective and discriminant features.

● Classification: classify the signals taking the feature vectors into account, and decipher the user’s intentions.

● Application interface: translate the classified signals into meaningful commands for any connected device.

BCI Stages

Brain Anatomy and Functions

● Every part of the brain do a specific function

Signal Acquisition Stage

● Electrophysiological (direct): it is generated by electro-chemical transmitters exchanging information between the neurons. The neurons generate ionic currents which flow within and across neuronal assemblies.

● Hemodynamic (Indirect): the blood releases glucose to active neurons at a greater rate than in the area of inactive neurons.

Electrophysiological

● Invasive: intracranial microarrays are implanted in the gray mater, that involves significant health risks.

● Partially-invasive: micro-electrods 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 and have a lower risks than fully invasive BCIs.

● Non-invasive: signals are captured by external devices (eg. Placing electrodes on surface of scalp), so there are no risks in this methods but it produce poor resolution signals.

1- Electroencephalography (EEG)

EEG is a non-invasive technique that records electrical activity along the scalp. EEG measures voltage fluctuations resulting from ionic current flows within the neurons of the brain.

● Advantages:

– Non-invasive

– Lower costs

– Portable

– High temporal resolution● Disadvantages:

– Low spatial resolution

– High noise ratio

2- Magnetoencephalography (MEG)

MEG is a non-invasive imaging technique that registers the brain’s magnetic activity by means of magnetic induction.

● Advantages:

– Non-invasive

– Better spatial resolution (vs EEG)

– Lower noise ratio (vs EEG)● Disadvantages:

– Too expensive

– Non-portable (too bulky)

3- Electrocorticography (ECoG)

ECoG is a technique that measures electrical activity in the cerebral cortex by means of electrodes placed directly on the surface of the brain.

● Advantages:

– Higher temporal and spatial resolution

– Higher amplitudes

– Lower noise ratio● Disadvantages:

– Invasive

4- Intracortical Neuron Recording

Intracortical neuron recording is a neuroimaging technique that measures electrical activity inside the gray matter of the brain.● Advantages:

– Higher temporal and spatial resolution● Disadvantages:

– Invasive

– Signal quality may be affected by the reaction of cerebral tissue to the implanted recording micro-electrode and by changes in the sensitivity of the micro-electrode.

– Periodic re calibrations of electrode sensitivity may be necessary.

5- Functional Magnetic Resonance Imaging (fMRI)

fMRI is a non-invasive neuroimaging technique which detects changes in local cerebral blood volume, cerebral blood flow and oxygenation levels during neural activation by means of electromagnetic fields.

● Advantages:

– Non-invasive

– High spatial resolution● Disadvantages:

– Too expensive

– Non-portable (too bulky)

– Low time resolution

6- Near Infrared Spectroscopy (NIRS)

NIRS is an optical spectroscopy method that employs infrared light to characterize non-invasively acquired fluctuations in cerebral metabolism during neural activity.

● Advantages:

– Non-invasive

– portable

– Low cost● Disadvantages:

– Low time resolution

– Low spatial resolution

Summary of neuroimaging methods

Neuroimaging method

Activity measured

Direct/ Indirect

Measurement

Temporal resolution

Spatial resolution Risk Portability

EEG Electrical Direct ~0.05 s ~10 mm Non-invasive Portable

MEG Magnetic Direct ~0.05 s ~5 mm Non-invasive Non-portable

ECoG Electrical Direct ~0.003 s ~1 mm Invasive Portable

Intracortical neuron

recordingElectrical Direct ~0.003 s

~0.5 mm (LFP) ~0.1 mm (MUA) ~0.05 mm (SUA)

Invasive Portable

fMRI Metabolic Indirect ~1 s ~1 mm Non-invasive Non-portable

NIRS Metabolic Indirect ~1 s ~5 mm Non-invasive Portable

EEG: 10-20 System

EEG signals are easily recorded in a non-invasive manner through electrodes placed on the scalp, for which that reason it is by far the most widespread recording modality. The electrodes placed over the scalp are commonly based on the International 10–20 system, which has been standardized by the American EEG Society.

EEG signals

EEG raw data

EEG rhythmic activity frequency bands

Band Frequency Normally

Delta (δ) up to 4 Hz babies, and adults in deep sleep state

Theta (θ) 4 – 7 Hz children, and adults in drowsy, meditative or sleep states

Alpha (α) 8 – 12 Hz closing the eyes and the body is relaxed

Beta (β) 12 – 30 Hz thinking and concentration with no motor activity

Gamma (γ) 30 – 100 Hz - maximal muscle contraction- perception of both visual and auditory stimuli- affected by artifacts such as EMG or EOG

Mu (μ) 7 – 13 Hz the body is physically at rest

Artifacts in BCIs

● Physiological artifacts are usually due to muscular, ocular and heart activity, known as electromyography (EMG), electrooculography (EOG), and electrocardiography (ECG) artifacts respectively.

● Technical artifacts are mainly attributed to power-line noises or changes in electrode impedances, which can usually be avoided by proper filtering or shielding.

Preprocessing Stage

Data from brain signals can be quite high-dimensional, and potentially full of artifacts. So, the aim of this stage is to enhance the quality of the recorded brain signal and to prepare it for further processing stages.

Features Extraction Stage

The aim of this stage is to identify and generate a set of representative features which target specific aspects of brain activity.

Features Extraction Stage (Challenges)

● Noise and outliers: BCI features are noisy or contain outliers because EEG signals have a poor signal-to-noise ratio.

● High dimensionality: In BCI systems, feature vectors are often of high dimensionality.

● Time information: BCI features should contain time information as brain activity patterns are generally related to specific time variations of EEG.

● Non-stationarity: BCI features are non-stationary since EEG signals may rapidly vary over time and more especially over sessions.

● Small training sets: The training sets are relatively small, since the training process is time consuming and demanding for the subjects.

Features Extraction Stage (Methods)

● Principal Component Analysis (PCA)

● Independent Component Analysis (ICA)

● Recursive Feature Elimination (RFE)

● Auto-Regressive Components (AR)

● Matched Filtering (MF)

● Wavelet Transform (WT)

● Common Spatial Pattern (CSP)

Classification Stage

The aim of the classification step in a BCI system is recognition of a user’s intentions on the basis of a feature vector that characterizes the brain activity provided by the feature step.

Classification Stage (Methods)

● Support Vector Machine (SVM)

● Linear Discriminant Analysis (LDA)

● Artificial Neural Network (ANN)

● K-Nearest Neighbor (k-NN)

● Hidden Markov Models (HMM)

● Bayesian Statistical

● Combinations of classifiers

– Boosting

– Voting

– Stacking

Types of BCIs

According to the nature of the signals used as input:

● Exogenous BCI: uses the neuron activity elicited in the brain by an external stimulus such as visually or auditory evoked potentials.

● Endogenous BCI: user can operate the BCI at free will (like moving a cursor to any point in a two-dimensional space)

Types of BCIs

According to the input data processing modality:

● Synchronous BCI: analyze brain signals during predefined time windows, and any brain signal outside the predefined window is ignored.

● Asynchronous BCI: continuously analyze brain signals no matter when the user acts.

Types of BCIs

According to the analysis time:

● Offline analysis: brain signals are acquired then analyzed in later time.

● Online analysis: EEG device is connected to BCI system directly, and brain signals are acquired and analyzed in the same time.

Datasets on the Internet

● BCI Competitions: II (2003), III (2004), and IV (2008)

● Multimedia Signal Processing Group (MMSPG)

● Neurotycho.org: a project that aims to share neural data

● Statistical Parametric Mapping (SPM): software for the analysis of brain imaging data sequences

● LINI: share experimental data recorded and used by the neuroimaging laboratory of UAM

● DEAP dataset: emotion recognition

● MAHNOB HCI Tagging Database: emotion & media tagging

● PhysioNet.org: EEG motor movement/imagery dataset

● Project BCI - EEG motor activity data set

BCI tools and frameworks

● EEGLAB: MATLAB toolbox

● BCILAB: MATLAB toolbox and EEGLAB plugin

● BioSig: open source software library for biomedical signal processing (C++, Octave and MATLAB)

● BCI2000: general-purpose system for BCI research

● OpenViBE: software platform dedicated to designing, testing and using BCI

● xBCI: platform for building an online BCI system

Commercial BCI systems

● Imec EEG Handset

● Neurosky Mindwave headset

● Emotive EPOC headset

Challenges to employing BCI control for real-world tasks

● The information transfer rate provided by BCIs is too low for natural interactive conversation, even for experienced subjects and well-tuned BCI systems

● The high error rate further complicates the interaction

● BCI systems cannot be used autonomously by disabled people, because BCI systems require assistants to apply electrodes or signal-receiving devices before the disabled person can communicate

● A BCI user may be able to turn the BCI system off by means of brain activity as input, but usually cannot turn it back on again, which is termed the “Midas touch” problem

● Handling BCI applications demands a high cognitive load that can usually be achieved by users in quiet laboratory environment, but not in the real world

References

● Luis Fernando Nicolas-Alonso and Jaime Gomez-Gil: "Brain Computer Interfaces, a Review," Sensors, vol. 12, no. 2, pp. 1211-1279, 2012.

● Fabien Lotte; M Congedo; A Lécuyer; F Lamarche and B Arnaldi: "A Review of Classification Algorithms for EEG-based Brain-Computer Interfaces," Journal of Neural Engineering, vol. 4, no. 2, 2007.