sensors

Article
An Unsupervised Method for Artefact Removal in
EEG Signals
Angel Mur *, Raquel Dormido and Natividad Duro
Department of Computer Sciences and Automatic Control, Universidad Nacional de Educación a
Distancia (UNED), Juan del Rosal 16, 28040 Madrid, Spain; [email protected] (R.D.); [email protected] (N.D.)
* Correspondence: [email protected]; Tel.: +34-91-398-7192




Received: 9 April 2019; Accepted: 16 May 2019; Published: 18 May 2019


Abstract: Objective: The activity of the brain can be recorded by means of an electroencephalogram
(EEG). An EEG is a multichannel signal related to brain activity. However, EEG presents a wide variety
of undesired artefacts. Removal of these artefacts is often done using blind source separation methods
(BSS) and mainly those based on Independent Component Analysis (ICA). ICA-based methods are
well-accepted in the literature for filtering artefacts and have proved to be satisfactory in most scenarios
of interest. Our goal is to develop a generic and unsupervised ICA-based algorithm for EEG artefacts
removal. Approach: The proposed algorithm makes use of a new unsupervised artefact detection,
ICA and a statistical criterion to automatically select the artefact related independent components
(ICs) requiring no human intervention. The algorithm is evaluated using both simulated and real
EEG data with artefacts (SEEG and AEEG). A comparison between the proposed unsupervised
selection of ICs related to the artefact and other supervised selection is also presented. Main results:
A new unsupervised ICA-based algorithm to filter artefacts, where ICs related to each artefact
are automatically selected. It can be used in online applications, it preserves most of the original
information among the artefacts and removes different types of artefacts. Significance: ICA-based
methods for filtering artefacts prevail in the literature. The work in this article is important insofar
as it addresses the problem of automatic selection of ICs in ICA-based methods. The selection is
unsupervised, avoiding the manual ICs selection or a learning process involved in other methods.
Our method is a generic algorithm that allows removing EEG artefacts of various types and, unlike
some ICA-based algorithms, it retains most of the original information among the artefacts. Within
the algorithm, the artefact detection method implemented does not require human intervention either.

Keywords: artefacts; EEG; artefact detection; artefact removal; ICA

1. Introduction
The activity of the brain can be recorded by means of electroencephalogram (EEG). EEG is
a noninvasive method that uses a set of sensors (electrodes) distributed along the scalp. The artefacts
are signals recorded by the EEG with no connection to a particular brain activity. Artefacts can be of
physiological or nonphysiological origin. Physiological artefacts are generated by the patient (e.g., ocular
movements, eye blinks and muscular activity) and nonphysiological artefacts can arise from outside the
body (e.g., equipment, environment).
Artefacts can be marked or detected. Once the artefacts are known, they can be directly removed
by rejecting the segments of the signal containing the artefacts. In this case, there could be a considerable
loss of useful information from the EEG signal. Rejecting contaminated EEG segments can be applied
when segments contain excessive interference [1,2], but, in general, it is desirable to filter only the
artefacts while retaining as much information of the EEG signal as possible [3]. This filtering without
rejecting the cognitive part is important in the brain-computer interfaces and short recordings [4,5].

Sensors 2019, 19, 2302; doi:10.3390/s19102302 www.mdpi.com/journal/sensors

Sensors 2019, 19, 2302 2 of 22

Urigüen et al. [1] provides a detailed review about EEG artefact removal methods. There are different
approaches such as regression [6], ocular artefact correction [7], filtering [8] or BSS techniques [9,10].
In general, the BSS methods are commonly used and mainly those based on ICA [9,11]. These methods
take advantage of the independence between the artefacts and the brain activity.
The BSS methods can also be classified as semiautomated or automated. Semiautomated methods
require visual inspection by experts either to detect artefacts or to classify the resulting components
as artefacts components or cognitive components. This strategy applies for offline applications and,
in general, it is a tedious procedure. Furthermore, the criterion used by the expert could not be
uniform during the analysis. To avoid these problems, an automatic procedure is preferred. However,
an automatic artefact elimination process is not a trivial problem since artefacts are mixed with each
other and with the EEG signal in very different ways.
The first step to remove an artefact is to detect it. There are different approaches for detecting
artefacts. For example, statistical methods and/or threshold values can be used [2,12–14]. Classification
methods are an alternative to detect artefacts, for example, DETECT [15] and the unsupervised method
for event detection (UMED) [16]. These methods detect artefacts by means of an identification or
characterization procedure.
Once an artefact has been detected, an ICA can be applied to the EEG signal to obtain a set of ICs.
Some of these components are related to the brain activity and the rest to the artefact. A method must
divide the components into those two groups. Then, the components of the brain activity are used to
rebuild a corrected EEG signal by eliminating the contributions of the artefactual sources.
When an EEG portion presents an artefact, the cognitive part whose origin is in the brain is
considered independent of the artefact whose origin is outside the brain. On the other hand, in the
ICA algorithms, the number of desired ICs is an input argument. When a number of 2 is selected,
the result, in most cases, is a poor approximation of the cognitive part and the artefact. Therefore,
to obtain an acceptable result, it is better to work with a higher number of ICs.
The EEG signal is captured using a set of N electrodes. The ICA methods offer a maximum
number of ICs equal to the number of electrodes. Since, in general, the optimum number of ICs is
greater than N, it is necessary to work with the highest possible number of components (that is N).
However, when the number of components increases, the problem of how to automatically group the
cognitive components and the components related to the artefact arises.
MARA [17] is a technique built on ICA that allows users to automatically determine those
components of ICA that do not come from brain activity. For this purpose, MARA uses a classifier
(a Linear Programming Machine) that uses some features that allow discrimination of an ICA component
whose origin is not in brain activity. It can handle any type of artefact.
The SASICA software [18] provides a practical guide to select artefactual ICs of EEG. Normally,
this selection needs to be improved analysing each IC. Consequently, this filtering process is semiautomatic.
wICA [19] is an ICA-based method for filtering artefacts that includes a wavelet threshold of ICs
as an intermediate step. This filtering is automatic, but the result depends on some parameters to
determine the wavelet threshold.
There are other recent research papers that present artefact denoising of EEG signals. Melia et al. [20]
presents an algorithm for removing peak and spike noise from EEG. It is based on filtering and thresholding
the analytic signal envelope. Mahajan et al. [21] presents an unsupervised algorithm that uses modified
multiscale sample entropy and Kurtosis to automatically identify the independent eye blink artefactual
components and, subsequently, denoise these components using biorthogonal wavelet decomposition.
This method does not require manual identification for artefactual components. The method FORCe [22] is
an artefact removal method developed for use in brain-computer interfacing. It is based on a combination
of wavelet decomposition, ICA and thresholding. The EEMD-ICA approach [23] removes artefacts,
preserving more information than other classical ICA methods. Chen et al. [24] proposes to filter both
ocular and muscle artefacts by exploiting diverse statistics. Somers et al. [25] presents a generic algorithm
based on the multichannel Wiener filter for removal of EEG artefacts of various types.
Sensors 2019, 19, 2302 3 of 22

Mannan et al. [26] and KafiulIslam et al. [27] provide a comparative study of various EEG
artefact removal techniques. Mannan et al. [26] also review the schemes developed for validating the
performances of algorithms with simulated and real EEG data.
The main goal of this paper is to present a generic and unsupervised ICA-based method to
automatically remove artefacts (UAR). The method combines a new unsupervised artefact detection
algorithm, ICA and a new unsupervised algorithm to select ICs related (or not) to brain activity.
The UAR offers a solution to the problem of automatic selection of ICs in ICA-based method using
a statistical criterion to select the ICs for artefact removal while preserving most of the original
information between artefacts. It is not limited to a specific number or type of artefact and it can also
be applied in online applications.
The performance evaluation of artefact removal methods is problematic due to the lack of
standard quantitative metrics in the literature to measure both the amount of artefact removal and
the distortion [1,26,27]. Most of the methods are often tested on real EEG, and their performances
are evaluated in terms of some qualitative plots analysed by visual inspection. UAR results will be
evaluated by visual inspection on AEEG and SEEG. A metric will also be used to evaluate the SEEG
filtering. Comparisons are also presented between UAR, MARA, SASICA and wICA.
There is a large number of algorithms developed for artefact removal from EEG signals. However,
there is no universal solution available yet and, in general, it is not fair to say which method
performs best. The filtering quality of a particular method varies according to the conditions of
use (type of artefacts, number of electrodes, etc.). For example, an ICA-based method can filter
electromyographic (EMG) artefacts worse than other methods when the number of electrodes is small,
but can obtain a better performance with a high number of electrodes. Regarding the number of
electrodes, Chen et al. [28] proposes to use single channel techniques for the removal of muscle artefacts
instead of using multichannel approaches. This study used a maximum of 23 channels.
For these reasons, in this paper it is shown that the quality of the proposed UAR filtering is at
least as good as that of other methods (in our case comparing UAR with MARA, SASICA and wICA).
The paper is organized as follows. Section 2 presents some background necessary to implement
the UAR. Section 3 describes the new artefact detection method used in UAR. The proposed method to
filter artefacts in EEG recordings is presented in Section 4. In Section 5, the method is tested using
some EEG recordings (both AEEG and SEEG) with different types of artefacts. Finally, in Sections 6
and 7, a discussion and conclusions of the paper are presented, respectively.

2. Background
This section briefly explains a method to detect events (UMED), the ICA algorithm and the
two-sample variance test.

2.1. Unsupervised Method for Event Detection

The Unsupervised Method for Event Detection (UMED) method [16] is used to detect events in
a multichannel signal in an unsupervised way. In this method, the multichannel signal is divided into
a set of intervals created from a sliding window (with a sliding distance of d samples). Each interval is
represented by a vector of autoregressive coefficients AR (2) (two for each channel) [29].
Using several windows, the hierarchical algorithm [30] and the S_Dbw index [31], the optimal
window Lw that classifies the intervals in an unsupervised way is found. Those consecutive intervals
belonging to the same group form a state. The transition between two states represents an event.
UMED performs an unsupervised classification to detect events. In this paper, a new method for
artefact detection that avoids this unsupervised classification is presented in Section 3. The implemented
method makes use of the optimal window Lw and the autoregressive coefficients AR defined in UMED.
AR coefficients are reliable feature sets to detect artefacts in EEG recordings [32].
The UMED method is also part of the unsupervised method for event detection and classification
of multichannel signals (UMEDC) [33].
Sensors 2019, 19, 2302 4 of 22

2.2. ICA
Given a set of n mixed signals formed by combining k independent signals with p samples, its ICA
model is defined as M = A × S, where M is a n × p data matrix, A is a n × k mixing matrix and S is
a k × p matrix of independent signals.
The objective of ICA is to calculate A and S knowing only M. ICA does not need any knowledge
concerning the nature of the source signals or their proportions. To estimate A, ICA requires the pure
signals in S to be truly independent and non-Gaussian. Both conditions are usually met when the
sources are real signals.
The independence in ICA can be reached by maximizing the non-Gaussianity of the components or
by minimizing the mutual information [34]. Around this concept, different ICA algorithms have been
developed: FastICA [35], JADE [36], InfoMax [37], Stochastic Non-Negative Independent Components
Analysis [38], RADICAL [39], etc. For EEG artefact removal, SOBI, InfoMax and FastICA are mainly
used [40,41].
The FastICA algorithm outperforms most of the commonly used ICA algorithms in convergence
speed. Such property becomes useful in this article since ICA is applied individually for each artefact.
FastICA decompositions are somewhat unstable (for the same signal, the ICs are not always reproduced
in the same way). However, this fact is not a problem for the proposed UAR algorithm.
The FastICA algorithm applied to an artefact requires a minimal window length so that some ICs
can capture characteristics of the artefact. In general, a window length above 10 s could be enough.
In this paper we will use a window length of 26 s (see explanations in Section 5.3).
Given an EEG signal of N electrodes (E1 , . . . , EN ), the FastICA outputs a set of ICs (CO1 , . . . , CON )
and its corresponding mixing matrix A. For a specific independent component, CO, we denote its
projection over an electrode E as Projection (E,CO). If the signal of the electrode E is ME , then
X
ME = Pro jection(E, COi ) (1)
i

2.3. Two-Sample F-Test for Equal Variances

The two-sample F-Test [42] is used to test if the variances of two populations (X1 and X2) are equal.
The test statistic is a ratio of the two sample variances (v1 and v2). This ratio is called the F-statistic and
follows an F distribution: F = v1/v2. The null hypothesis H0 assumes that the variances are equal and
the ratio is one. The alternative hypothesis H1 assumes that v1 and v2 are different, and that the ratio
deviates from unity. H0 is rejected if F is either too large or too small. The test uses a significance level
(normally at the 5% or 1%) when rejecting H0. This test is based on two assumptions: (a) the samples
are normally distributed and (b) the samples are independent of each other.
In this paper, R1 denotes when the two-sample F-Test rejects the null hypothesis at the 1%
significance level and R0 otherwise.

3. Method for Artefact Detection Used in UAR

In this section, we present a new artefact detection method (AD_UAR). It has been developed
based on our previous UMED algorithm. An AEEG is used to show how it works.

3.1. EEG Data, UMED and the Artefact Detection Method

In this paper, the test signal is an 8-min EEG recording used in [15,16] to show the performance of
DETECT toolbox and the UMED method. The data are sampled at 256 Hz using a 64-channel Biosemi
Active Two System. All 64 channels are used. This signal presents different kinds of physiological
artefacts: Jaw Clench, Jaw Movement, Eye Blink, Eye Left Movement and Eye Up Movement. Figure 1
represents the distribution of the electrodes on the scalp. The signals of the AF7 and AF8 electrodes
will be used in the graphical representations in this paper.
of DETECT toolbox and the UMED method. The data are sampled at 256 Hz using a 64-channel
Biosemi Active Two System. All 64 channels are used. This signal presents different kinds of
physiological artefacts: Jaw Clench, Jaw Movement, Eye Blink, Eye Left Movement and Eye Up
Movement. Figure 1 represents the distribution of the electrodes on the scalp. The signals of the
AF7Sensors
and AF82019, 19, 2302
electrodes will be used in the graphical representations in this paper. 5 of 22

Figure 1. Distribution of the electrodes on the scalp. AF7 and AF8 are the electrodes of the signals
used in the graphical representations.

In Section 2.1 UMED’s basic principles have been presented (full details can be found in [16]).
Using UMED on the AEEG described above (Lw = 155 samples and d = 32 samples) some clusters of
intervals are found. As Figure 2 shows, these groups can be represented by means of a Principal
Component Analysis (PCA) [43]. Each cluster can be distinguished by a specific shape and colour
(full details about
Figure Figure 2 can
1. Distribution of thebe found in
electrodes on [16]).
the scalp. AF7 and AF8 are the electrodes of the signals used
Figure 1. Distribution of the electrodes on the scalp. AF7 and AF8 are the electrodes of the signals
The most numerous group G1 consists of intervals that represent the cognitive activity of the
in the graphical representations.
used in the graphical representations.
brain without artefacts. The other groups correspond to different types of artefacts: winks, face
In Section 2.1 UMED’s basic principles have been presented (full details can be found in [16]).
movements,
In etc.2.1
Section AsUMED’s
we will analyse
basic later, you
principles cannot rule out finding some intervals in G1 in
with a
Using UMED on the AEEG described abovehave
(Lw =been presented
155 samples and(full details
d = 32 can some
samples) be found
clusters[16]).
small
Usingofartefact
UMEDactivity.
intervals on
arethe AEEG
found. As described above
Figure 2 shows, (Lw groups
these = 155 samples and d = 32by
can be represented samples)
means ofsome clusters of
a Principal
intervals are found. As Figure 2 shows, these groups can be represented by means of a distance)
The new
Component AD_UAR
Analysis is based
(PCA) [43].on
Eachan analysis
cluster can of
be the distances
distinguished (for
by a example,
specific shape Euclidean
and colour Principal
(full
between all the
details about
Component intervals and
Figure(PCA)
Analysis their
2 can be[43]. centroid.
found
Eachin [16]). A histogram is used to carry out this analysis.
cluster can be distinguished by a specific shape and colour
(full details about Figure 2 can be found in [16]).
The most numerous group G1 consists of intervals that represent the cognitive activity of the
brain without artefacts. The other groups correspond to different types of artefacts: winks, face
movements, etc. As we will analyse later, you cannot rule out finding some intervals in G1 with a
small artefact activity.
The new AD_UAR is based on an analysis of the distances (for example, Euclidean distance)
between all the intervals and their centroid. A histogram is used to carry out this analysis.

Figure 2. Groups of intervals found using unsupervised method for event detection (UMED) and
Figure 2. Groups of intervals found using unsupervised method for event detection (UMED) and
displayed by means of a PCA. The Lw has 155 samples. The first two principal components contain
displayed by means of a PCA. The Lw has 155 samples. The first two principal components contain
55% of the full information.
55% of the full information.
The most numerous group G1 consists of intervals that represent the cognitive activity of the brain
Figure artefacts.
without 3 showsThethe otherhistogram that can
groups correspond be modelled
to different by a Generalized
types of artefacts: Extreme etc.
winks, face movements, Value
Distribution.
As we willWe consider
analyse later, two thresholds:
you cannot T1finding
rule out and T2.some
T1 points at the
intervals in Gbeginning of artefact
1 with a small the extreme tail of
activity.
The new AD_UAR is based on an analysis of the distances (for example,
the histogram. T2 value is chosen so that beyond T2 the number of intervals is the same as within Euclidean distance)
between all the intervals and their centroid. A histogram is used to carry out this analysis.
SensorsFigure
2019, 19,2.x;Groups
Figure doi: FORof
3 shows PEER
the REVIEW
intervals found
histogram thatusing
can beunsupervised
modelled by amethod for event
Generalized www.mdpi.com/journal/sensors
detection
Extreme Value(UMED) and
Distribution.
displayed
We considerby means
two of a PCA.
thresholds: TheTL2w. has
T1 and 155 samples.
T1 points The first of
at the beginning twotheprincipal components
extreme tail contain
of the histogram.
55% of the full information.
T2 value is chosen so that beyond T2 the number of intervals is the same as within [T1 , T2 ]. All the
intervals containing artefacts (included from G2 to G7 ) are beyond T2 . Some intervals with a low
Figureactivity
artefact 3 shows the
can be histogram
found next to T2that can test
. For our be signal
modelled byofaT1Generalized
the value is 1.33 and of TExtreme
2 is 1.66. Value
Distribution.
FigureWe consider
4 shows twochannel,
an EEG thresholds: T1 and T
the distances of2.the
T1 intervals
points attothe beginning
their of the
centroid and extreme tail of
the thresholds
T and T
the histogram.
1 2 . All the channels have been used to calculate the distances. It is
T2 value is chosen so that beyond T2 the number of intervals is the samesimple to infer thatasthe
within
distance between the intervals and their centroid is a good quantitative measure to detect artefacts.
Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors
[T1, T2]. All the intervals containing artefacts (included from G2 to G7) are beyond T2. Some interv
with a low artefact activity can be found next to T2. For our test signal the value of T1 is 1.33 and
T2 is 1.66. Sensors 2019, 19, 2302 6 of 22

Figure 3. Histogram of interval distances and its generalized extreme value distribution.

Figure 4 shows an EEG channel, the distances of the intervals to their centroid and the
thresholds T1 and T2. All the channels have been used to calculate the distances. It is simple to infer
that the distance between the intervals and their centroid is a good quantitative measure to
detect artefacts.
Given a specific threshold T (for example T = T1), any piece of the signal formed by consecutive
intervals with a distance greater or equal to T is a potential portion of the signal
containing artefacts.
The artefact detection method (AD_UAR) can be summarized as follows.
Given an EEG signal of N channels, a sliding window Lw, a sliding distance d, a number of
channels (or electrodes) n (1 <= n<= N) and a threshold T:
1) Calculate the intervals (and their autoregressive vectors) of the AEEG using Lw, d and n.
2) Calculate the distances between the intervals and their centroid.
3) Determine potential portions of signal containing artefacts by selecting consecutive intervals
3. Histogram
Figure greater
with a distance of interval
or equal to T. distances and its generalized extreme value distribution.
Figure 3. Histogram of interval distances and its generalized extreme value distribution.

Figure 4 shows an EEG channel, the distances of the intervals to their centroid and
thresholds T1 and T2. All the channels have been used to calculate the distances. It is simple to in
that the distance between the intervals and their centroid is a good quantitative measure
detect artefacts.
Given a specific threshold T (for example T = T1), any piece of the signal formed by consecut
intervals with a distance greater or equal to T is a potential portion of the sig
containing artefacts.
The artefact detection method (AD_UAR) can be summarized as follows.
Given an EEG signal of N channels, a sliding window Lw, a sliding distance d, a number
channels (or electrodes) n (1 <= n<= N) and a threshold T:
Figure 4. An EEG channel, the distances of the intervals to their centroid and the thresholds T1 (in black)
Figure 4. An EEG channel, the distances of the intervals to their centroid and the thresholds T (in
1) Calculateand theT intervals
(in red). (and their autoregressive vectors) of the AEEG using Lw, d and n. 1

black)2 and T (in red).

2

2) CalculateGiven
the distances between the intervals and their centroid.
a specific threshold T (for example T = T1 ), any piece of the signal formed by consecutive
3) Determine
intervalspotential
with a distanceportions
greater orof signal
equal to T iscontaining artefacts
a potential portion by selecting
of the signal consecutive interv
containing artefacts.
with a distance The artefact
greater detection
ordoi:
equal method
to T. (AD_UAR) can be summarized as follows.
Sensors 2019, 19, x; FOR PEER REVIEW www.mdpi.com/journal/sensors
Given an EEG signal of N channels, a sliding window Lw , a sliding distance d, a number of
channels (or electrodes) n (1 <= n<= N) and a threshold T:
(1) Calculate the intervals (and their autoregressive vectors) of the AEEG using Lw , d and n.
(2) Calculate the distances between the intervals and their centroid.
(3) Determine potential portions of signal containing artefacts by selecting consecutive intervals
with a distance greater or equal to T.

3.2. Selection of d, T and Lw

The sliding distance d has to be small, but it is not necessary to choose d = 1 to obtain a good
detection. In this paper we have chosen d = 32 samples.
To set the value of T let us think about some possible situations that could happen in the
detection process.
 3.2. Selection of d, T and Lw
The sliding distance d has to be small, but it is not necessary to choose d = 1 to obtain a good
detection. In this paper we have chosen d = 32 samples.
Sensors 2019,To
19,set
2302the value of T let us think about some possible situations that could happen in the7 of 22
detection process.
The artefacts can be mixed in a different way for each channel. This means the artefact
The artefacts
recorded at eachcanelectrode
be mixed in a (it
varies different way for or
can be stronger each channel.
weaker) This means
depending on thethe artefact
electrode recorded at
position.
each When
electrode varies (it
the number of can be stronger
electrodes orand
n is high weaker) depending
the portion on thecontaining
of the signal electrodetheposition. When
artefact is not the
significantly different from the signal without artefact (on most electrodes), the AD_UAR
number of electrodes n is high and the portion of the signal containing the artefact is not significantly could not
always
different frombe precise enough
the signal to detect
without artefacts.
artefact (on most electrodes), the AD_UAR could not always be
Figure 5 shows two
precise enough to detect artefacts. types of detection at the beginning of an artefact. When the artefact is well
represented in most of the channels, the AD_UAR offers a “good detection”. There is a “late
Figure 5 shows two types of detection at the beginning of an artefact. When the artefact is well
detection” when the artefact weakly changes the AEEG. For example, if the beginning of the
represented in most of the channels, the AD_UAR offers a “good detection”. There is a “late detection”
artefact is weak. A similar phenomenon can also appear at the end of the artefact with an “early
whendetection”.
the artefact weakly changes the AEEG. For example, if the beginning of the artefact is weak.
A similar phenomenon can also appear at the end of the artefact with an “early detection”.

Figure 5. 5.An
Figure Anartefact
artefact and twotypes
and two typesofofdetection.
detection.

To avoid these
To avoid possible
these detection
possible detectionproblems
problemsthe
the following solutions
following solutions are
are proposed.
proposed.
• Option A (OA): Select in AD_UAR a threshold of T= T1. That is, T is set to the threshold where
• Option A (OA): Select in AD_UAR a threshold of T= T1 . That is, T is set to the threshold where
the tail of the Generalized Extreme Value Distribution begins.
the tail of the Generalized Extreme Value Distribution begins.
• Option B (OB): Increase the limits of the detected artefact portion.
• Option
OABtries
(OB):toIncrease
find all the
the limits of the
potential detected
artefacts. artefact
Any portion.
threshold between T1 and T2 could be a
candidate for T. However, as it is not possible to know a priori which is the best one for a signal,
OA tries to find all the potential artefacts. Any threshold between T1 and T2 could be a candidate
threshold T = T1 is chosen. In this way, portions of signal with distances between T1 and T2 will be
for T. However, as it is not possible to know a priori which is the best one for a signal, threshold T = T1
analysed using UAR. Although it is expected that most of these portions of signals will not need to
is chosen. In this
be filtered, it isway, portions
guaranteed of some
that signal with distances
intervals between
with a potential T1 and
artefact willTnot
2 will
be be
lost.analysed
However,using
UAR.thisAlthough it is expected
will imply more processing. that most of these portions of signals will not need to be filtered, it is
guaranteed that some
Threshold T1 = intervals
1.33 shouldwith a potential
not vary artefact
significantly withwill not signal.
another be lost.Thus,
However,
this valuethis willbeimply
could
morechosen
processing.
for another AEEG with 64 electrodes without the need to analyse a new Generalized
Extreme Value
Threshold T1 = Distribution.
1.33 should not Thisvary
is explained by the
significantly following
with another paragraph
signal. Thus, of Ref. [16]. “The
this value could be
artefacts are very well-differentiated from the cognitive EEG part without
chosen for another AEEG with 64 electrodes without the need to analyse a new Generalized Extreme artefacts. This cognitive
Valuepart is only related to the thinking and its high variability is concentrated in a unique cluster that is
Distribution. This is explained by the following paragraph of Ref. [16]. “The artefacts are very
the biggest. This cluster is not affected using different subjects since the AR coefficients are scale-
well-differentiated from the cognitive EEG part without artefacts. This cognitive part is only related to
and location-invariant.”
the thinking and its high variability is concentrated in a unique cluster that is the biggest. This cluster
If the number of channels n is high, OA could not be enough to correctly detect the beginning
is notand
affected
the endusing different
of some subjects
artefacts. Thissince
is sothe ARthe
since coefficients
Euclideanare scale- is
distance and location-invariant.”
less sensitive to small
If the number
changes. of channels
This problem n is using
is solved OA could not be enough to correctly detect the beginning and
high, OB.
the end ofWith some OB,artefacts. This is artefacts
all the potential so sincedetected
the Euclidean distance
are widened. For is less sensitive
example, in Figure to 5small changes.
the “late
detection”isissolved
This problem expanded using to OB.
the left by means of the AT1 portion. If the AT1 size is appropriate, the
SensorsOB,
With 2019,all
19,the potential
x; doi: FOR PEERartefacts
REVIEW detected are widened. For example, www.mdpi.com/journal/sensors
in Figure 5 the “late detection”
is expanded to the left by means of the AT1 portion. If the AT1 size is appropriate, the possible “late
detections” are corrected. In general, for any type of detection, a small part of the signal without artefact
will be added to the detected artefact portion. In Figure 5, the “corrected late detection” adds the EA1
portion to the left of the artefact. Analogously, it is understood that the detected end of the artefact
expands to the right in order to correct “early detections”.
It is not possible to automatically know how good the detection is, and consequently all the
potential artefacts detected are widened in the same way. The UAR algorithm will filter each artefact
using the new limits established by OB. The UAR filtering will hardly modify the «EA» portions.
 detected end of the artefact expands to the right in order to correct “early detections”.
It is not possible to automatically know how good the detection is, and consequently all the
potential artefacts detected are widened in the same way. The UAR algorithm will filter each
artefact using the new limits established by OB. The UAR filtering will hardly modify the
«EA» portions.
Sensors 2019,
When 19, 2302
the number of channels is high and the artefact appears in one electrode or in a very 8 of 22
small number of electrodes, the AD_UAR algorithm (using all the channels) may not detect it, even
using a T = T1 for the OA explained before. In this case, it is better to use AD_UAR for each channel
When the number
individually of channels
(or in small groups of is high and the
channels) andartefact
then fuse appears in oneofelectrode
the results or in a very
all the channels (or small
number
groups). In this paper, we have preferred to test the UAR algorithm using all the channels for the using
of electrodes, the AD_UAR algorithm (using all the channels) may not detect it, even
= T1 for the
a T AD_UAR OA explained before. In this case, it is better to use AD_UAR for each channel individually
algorithm.
Finally,
(or in small the choice
groups of Lw = 155
of channels) andis then
motivated by the
fuse the UMED
results result.
of all the With another
channels (or AEEG,
groups).Lw In
canthis
be paper,
we different. However,
have preferred if UMED
to test the UARis not used to determine
algorithm using allLthe
w, the windowfor
channels Lw the
= 155 could alsoalgorithm.
AD_UAR give good
filtering
Finally, results since OB
the choice of corrects
Lw = 155 theislack of detection
motivated accuracy.
by the UMEDThe accuracy
result. Withof another
the detection
AEEG, canL can
w
decrease if the window is not optimal, for example, to detect consecutive artefacts. In our article,
be different. However, if UMED is not used to determine Lw , the window Lw = 155 could also give
accuracy is not decisive because consecutive intervals are considered as a single artefact and also
good filtering results since OB corrects the lack of detection accuracy. The accuracy of the detection
both the beginning and the end of an artefact are extended by OB.
can decrease if the window
Consequently, is not optimal,
in this paper, we will workfor example, to detect
with d = 32, n = 64, consecutive
Lw = 155 and artefacts. In our
T = T1 = 1.33. The article,
accuracy is not decisive because consecutive intervals are considered
selection of the increased limits for OB will be specified in the next section. as a single artefact and also both
the beginning and the end of an artefact are extended by OB.
4. Consequently,
Unsupervised Method
in thisfor Artefact
paper, will work with d = 32, n = 64, Lw = 155 and T = T1 = 1.33.
we Removal
The selection of the increased limits for OB will be
This section explains the new Unsupervised specified
Method in the next
for Artefact section.
Removal (UAR) proposed in
this paper.
4. Unsupervised Method for Artefact Removal
4.1. Previous Notation
This section explains the new Unsupervised Method for Artefact Removal (UAR) proposed in
Figure 6 shows some pieces around a detected artefact (AT) used in the UAR algorithm. When
this paper.
an artefact is detected using the AD_UAR algorithm, OB adds two portions of the signal around AT
4.1.(AT
Previous Notation
1 and AT 2). The UAR algorithm will filter the portion ART = AT1 + AT + AT2. ICA_P is the portion

of the signal where the ICA (FastICA) performs. The reference (REF) is a reference portion used by
Figure 6 shows some pieces around a detected artefact (AT) used in the UAR algorithm. When
UAR to carry out the two-sample F-Test (see Section 2.3).
an artefact is detected
In this paper, theusing the of
duration AD_UAR algorithm,
ICA_P is 26 OB adds in
s (see explanations two portions
Section 5.3), of
ATthe signal around AT
1 = AT2 = 1 s and
(ATthe
1 and AT 2 The UAR algorithm will filter the portion ART = AT 1 + AT + AT
length of REF is equal to the length of ART. If a new artefact starts inside AT2, then isAT
). 2 . ICA_P the
2 isportion

of the signal where the ICA (FastICA) performs. The reference (REF) is a reference
selected to a point less than or equal to the beginning of this new artefact. The AT1 of the new portion used by
UAR artefact willout
to carry startthe
at that F-Test (see Section 2.3).
selected point.
two-sample

Figure 6. Notation of some parts around a detected artefact (for an EEG channel) used in the
Figure 6. Notation
automatically removeofartefacts
some parts around
(UAR) a detected artefact (for an EEG channel) used in the
algorithm.
automatically remove artefacts (UAR) algorithm.
In this paper, the duration of ICA_P is 26 s (see explanations in Section 5.3), AT1 = AT2 = 1 s and
4.2. The UAR Method
the length of REF is equal to the length of ART. If a new artefact starts inside AT2 , then AT2 is selected
to a pointOnce
lessanthan
artefact has been
or equal to thedetected
beginningusing
of the
thisAD_UAR algorithm
new artefact. the1 of
The AT basic
theidea
newofartefact
the UARwill start
method is to calculate
at that selected point. an ICs set by selecting a portion of signal according to the schema of Figure 6.
Then, the ICs are arranged in descending order from the one that most influences the artefact to the
4.2.Sensors
The UAR Method
2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors

Once an artefact has been detected using the AD_UAR algorithm the basic idea of the UAR
method is to calculate an ICs set by selecting a portion of signal according to the schema of Figure 6.
Then, the ICs are arranged in descending order from the one that most influences the artefact to the less.
Finally, following the ordered ICs found, the filtering process subtracts iteratively, for each electrode
and only in the ART part, the projection of each IC. The two-sample F-Test (see Section 2.3) allows
comparing each subtraction with REF and stopping the iterative process.
The UAR method can be summarized as follows.
Given an AEEG of N electrodes:
Step 1: Artefact Detection. The potential artefacts are detected by means of the AD_UAR method.
Step 2: ICs. For each artefact and according to the notation in Section 4.1, the FastICA algorithm
is applied to obtain the N ICs. The ICs and ICA_P have the same length.
Sensors 2019, 19, 2302 9 of 22

Step 3: ICs in order. Each IC is projected on all the electrodes. The projection of the electrode
with, for example, the maximum absolute value (or the maximum variance) in the artefact portion,
is selected. Then, the ICs are sorted in a descending way comparing the selected maxima. The ordered
ICs are denoted with: C1 , . . . , CN .
Step 4: Filtering process. For each electrode in the set (E1 , . . . , EN ), the variance of REF is compared
with the variance in ART each time the projection of a Ci is filtered starting in C1 .
For example, with the electrode E (see Algorithm 1).

Algorithm 1. The filtering process with the electrode E

N_ART ← ART; C ← C1 ; STOP ← 0; FIRST ← 0; BEFORE ← 0; i ← 1;
while STOP ← 0
N_ART ← N_ART - Projection (E, C, ART) //Filtering process
if variance(REF) > variance(N_ART) and R1
STOP ← 1 and ART ← N_ART
if BEFORE ← 1
ART ← OLD_ART
end
end
if R0
if FIRST← 0
OLD_ART ← N_ART
end
FIRST ← 1 and BEFORE ← 1
else
FIRST ← 0 and BEFORE ← 0
end
i ← i + 1; C ← Ci
end

Explanation: The filtering process for an artefact and an electrode E consists in subtracting the
projection of each independent component following the ordered set of ICs (C1 , . . . , CN ). The Projection
(E, C, ART) denotes the selection of the “ART portion” from the projection of C over E. In each iteration
this subtraction is saved in N_ART. The algorithm stops (STOP ← 0) when the variance of reference
REF is bigger than the variance of N_ART and the two-sample F-Test (between R and N_ART) rejects
the null hypothesis at the 1% significance level (meaning the result of the test is R1). In this case,
the ART portion is replaced by the result of the final subtraction (ART ← N_ART). If previously there
is a consecutive sequence of R0 (BEFORE ← 1), then the ART portion is replaced by the first portion
(FIRST ← 0) of that sequence (ART ← OLD_ART).
The UAR algorithm works each artefact individually respecting the AD_UAR detection order.
Each detected portion may represent one or some consecutive artefacts.

5. Testing the UAR Algorithm

In this section, we filter the AEEG described in Section 3.1 using the UAR algorithm. Then, using
that AEEG, we compare UAR with MARA, SASICA and wICA. We also use a SEEG that has been
created by mixing a real EEG signal without artifacts (REEG) and some simulated artifacts. Finally,
using that SEEG signal we compare UAR with MARA, SASICA and wICA.

5.1. Filtering an AEEG Using UAR

The parameters used for the AD_UAR method are d = 32; n = 64; Lw = 155; and T = T1 = 1.33 for
the AD_UAR method, and for the rest of the UAR algorithm: length{ICA_P} = 26 s; length{AT1 } =
length{AT2 } = 1 s; length{REF} = length{ART}; and 1% significance level for the two-sample F-Test.
 Figures 8 and 9 show the entire channels AF8 and AF7 without and with filtering.
The parameters used for the AD_UAR method are d = 32; n = 64; Lw = 155; and T = T1 = 1.33 for
the AD_UAR method, and for the rest of the UAR algorithm: length{ICA_P} = 26 s;
length{AT1} = length{AT2} = 1 s; length{REF} = length{ART}; and 1% significance level for the
two-sample F-Test.
Sensors 2019, 19, 2302 10 of 22
Figure 7 shows three artefacts sequences on the AF8 channel. They have been filtered
with UAR.
Figures
Figure 78shows
and 9 show the entiresequences
three artefacts channels on
AF8 and
the AF7
AF8 without
channel. andhave
They withbeen
filtering.
filtered with UAR.

Figure 7. Three sequences of artefacts in the AF8 channel. In red, the filtered signal. The X axes
show the samples.

As displayed in Figure 8, after filtering in the AF8 channel, a small artefact appears near the
sample 40,000. This artefact has not been filtered because it has not been detected. The reasons for
this are threefold: the artefact is small, it appears in a very small number of channels and the
number of electrodes is high. The AF7 electrode is not far from AF8; this artefact has lost its
7. Three sequences
Figure 7.
Figure ofofartefacts inin
thethe
AF8 channel. In red, the the
filtered signal. The X axes
X show
importance in Three sequences
relation to the rest artefacts
of the signal. AF8
This channel.
artefact In
hasred, filtered
a maximum signal.
amplitudeThe in axesarea
the
the samples.
show the samples.
of AF8 and is losing strength when we move away from that area (see Figure 10).
The figures
Figures show
8 and the quality
9 show of the UAR performance.
the8,entire
As displayed in Figure after channels
filtering in AF8
theand
AF8AF7 without
channel, a and
smallwith filtering.
artefact appears near the
sample 40,000. This artefact has not been filtered because it has not been detected. The reasons for
this are threefold: the artefact is small, it appears in a very small number of channels and the
number of electrodes is high. The AF7 electrode is not far from AF8; this artefact has lost its
importance in relation to the rest of the signal. This artefact has a maximum amplitude in the area
of AF8 and is losing strength when we move away from that area (see Figure 10).
The figures show the quality of the UAR performance.

Sensors 2019, 19, x 11 of 22

TheAF8
Figure8.8.The
Figure AF8channel
channelofofAEEG
AEEGwith
withand
andwithout
withoutartefacts. TheXXaxes
artefacts.The axesshow
showthe
thesamples.
samples.

Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors

Figure 8. The AF8 channel of AEEG with and without artefacts. The X axes show the samples.

Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors

Figure 9. The AF7 channel of AEEG with and without artefacts. The X axes shows the samples.
Figure 9. The AF7 channel of AEEG with and without artefacts. The X axes shows the samples.

As displayed in Figure 8, after filtering in the AF8 channel, a small artefact appears near the
sample 40,000. This artefact has not been filtered because it has not been detected. The reasons for
this are threefold: the artefact is small, it appears in a very small number of channels and the number
 Sensors 2019, 19, 2302 11 of 22

ofFigure
electrodes is AF7
9. The high.channel
The AF7 electrode
of AEEG withisand
not without
far fromartefacts.
AF8; thisThe artefact
X axeshas lost its
shows theimportance
samples. in
relation to the rest of the signal. This artefact has a maximum amplitude in the area of AF8 and is
Figure 9.when
losing strength The AF7
we channel of AEEG
move away fromwith
thatand without
area artefacts.
(see Figure 10).The X axes shows the samples.

FigureFigure
Figure 10.
10. The The
The undetected
10.undetected
undetected artefact
artefact in
inthe
in the
artefact EEG
the EEG portion
portion
EEG portion [37,500
[37,500 39,800]
39,800]of
[37,50039,800] ofthe
of thechannels
the channels
channelsAF8 and
AF8
AF8 AF7.
and
and AF7.
AF7.

The figures show the quality of the UAR performance.

5.2. Filtering
5.2. Filtering an AEEGan AEEG
usingusing
MARA MARA
5.2. Filtering an AEEG Using MARA
The EEGLAB
The EEGLAB is anisinteractive
an interactive toolbox
toolbox forfor processingelectrophysiological
processing electrophysiological data dataincorporating
incorporating
ICA and artefact
The EEGLAB
ICA and artefact rejection.
is anThis
rejection. This
interactivetoolbox
toolboxtoolbox presents
presents some
for some
processingextensions for artefact
electrophysiological
extensions removal, for example,
data incorporating
for artefact removal, for example,
the MARA
ICA and extension
artefact [17]. This
rejection. MARA is a linear
toolbox presentsclassifier
some that learns for
extensions from expertremoval,
artefact ratings by
for extracting
example,
the MARA extension [17]. MARA is a linear classifier that learns from expert ratings by extracting
six MARA
the featuresextension
from the [17].
spatial,
MARAthe spectral
is a linearand the temporal
classifier domains.
that learns This
from This
expertclassifier
ratings and
by UAR pursue
extracting six
six features from the spatial, the spectral and the temporal domains. classifier and UAR pursue
the same objectives. However, MARA is a supervised method.
features from the spatial, the spectral and the temporal domains. This classifier and UAR pursue the
the same objectives. However, MARA is awith supervised method.
same In this section,
objectives. we compare
However, MARAUAR MARAmethod.
is a supervised using the AEEG described in Section 3.1. Instead
Inofthis section,
creating a we
new compare
metric, bothUAR
methods with MARA
will be using
visually the AEEG described in Section 3.1. Instead
compared.
In this section, we compare UAR with MARA using the AEEG described in Section 3.1. Instead of
of creating
creating
a new
Figure
a new
metric,
11 shows
metric,
both
the methods
both EEG
methods signalwill bevisually
willinbethe
visually compared.
AF8 channel
compared. after using the MARA method. When
Figure
comparing11 shows
Figure Figures
11 the
shows EEG
8 and
the 11,signal
EEG it signal ininthe
is observed theAF8
that
AF8UARchannel after better
performs
channel after usingthe
using the
for someMARA
MARA method.
important
method. When
artefacts
When
comparing
(see Figures
the red 8 and
arrows). 11, it
However,is observed
MARA is that
able UAR
to filter performs
the small better
artefact for
nearsome
comparing Figures 8 and 11, it is observed that UAR performs better for some important artefacts (see sampleimportant
40,000. artefacts
(see the
thered
redarrows). However,MARA
arrows). However, MARA is is able
able to filter
to filter the the
small small artefact
artefact near sample
near sample 40,000. 40,000.

Figure 11. The AF8 channel of AEEG after using the MARA method. The red arrows point to some
artefacts that have not been well filtered.
Figure Figure
11. The The AF8
11. AF8 channel
channel of AEEG
of AEEG after
after usingthe
using theMARA
MARA method.
method. The
Thered
redarrows
arrowspoint to some
point to some
Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors
artefacts that have not been well
artefacts that have not been well filtered. filtered.

The quality of the MARA result depends on its training process but it is also
Sensors 2019, 19, x; doi: FOR PEER REVIEW
important the way in
www.mdpi.com/journal/sensors
which the ICs are obtained. Figure 12 shows the MARA result after filtering only the EEG portion in
the interval [33,000 49,000]. In this case, the filtering seems to be better although an artefact remains in
the interval [45,000 49,000].
This example shows that the UAR strategy to obtain the ICs around each artefact (with an ICA_P
portion) seems more effective. It also preserves most of the original information between artefacts. On the
other hand, UAR needs more processing and only filters the artefacts that have been previously detected.
portion in the interval [33,000 49,000]. In this case, the filtering seems to be better although an
ICA_P portion)
artefact remains in seems more effective.
the interval It also preserves most of the original information between
[45,000 49,000].
artefacts.
This On the other
example hand,
shows thatUAR needsstrategy
the UAR more processing
to obtain and onlyaround
the ICs filters the
eachartefacts
artefact that have
(with an
been previously detected.
ICA_P portion) seems more effective. It also preserves most of the original information between
artefacts. On the other hand, UAR needs more processing and only filters the artefacts that have
Sensors 2019, 19, 2302 12 of 22
been previously detected.

Figure 12. The EEG portion [33,000 49,000] of the AF8 channel before and after using the MARA method.

Figure 12. The EEG portion [33,000 49,000] of the AF8 channel before and after using the MARA method.
5.3.Figure
Filtering an AEEG using SASICA
12. The EEG portion [33,000 49,000] of the AF8 channel before and after using the MARA method.
5.3. Filtering
The EEGLABan AEEGalso
Using
hasSASICA
the SASICA extension [18]. The following parameters were used:
5.3. Filtering
Autocorrelation,
The EEGLAB an AEEG
Focal using
also has theSASICA
Components, Correlation
SASICA extension [18].with other channels
The following parametersandwere
ADJUST
used:selection. These
Autocorrelation,
parameters
Focal The allow Correlation
Components,
EEGLAB selecting
also has the ICsother
with
the related
SASICA toextension
channels theand
artefact.
ADJUST Figure
[18]. 13
selection.
The shows
Thesethe
following AF8 channel
parameters
parameters allow
wereof AEEG
selecting
used:
after
the ICsusing
relatedthetoSASICA
the method
artefact. Figure (in
13 red).
shows It is
the noted
AF8 that
channel SASICA
of AEEG
Autocorrelation, Focal Components, Correlation with other channels and ADJUST selection. These selects
after more
using ICs
the than
SASICA necessary
method
and,
(in consequently,
red).
parameters It isallow thereSASICA
notedselecting
that is thea loss of information
selects
ICs relatedmore
to the after
ICsartefact. removing
than necessary
Figure 13 the
and, components.
consequently,
shows To achieve
there
the AF8 channel is of better
a loss
AEEG of
filtering,
information it is necessary
after removing tothe analyse the
components. selected
To ICs.
achieve Then,
better each component
filtering, it
after using the SASICA method (in red). It is noted that SASICA selects more ICs than necessary is necessaryis accepted
to analyse or
the rejected;
selected
but this
ICs.
and, Then,is each
no longer
consequently, a generic
component
there and of
isis aaccepted
loss automatic
or
information process.
rejected; but this removing
after is no longerthea generic and automatic
components. process.
To achieve better
filtering, it is necessary to analyse the selected ICs. Then, each component is accepted or rejected;
but this is no longer a generic and automatic process.

Figure 13.The
Figure 13. TheAF8
AF8channel of AEEG
channel withwith
of AEEG (in black) and without
(in black) artefacts
and without (in red) (in
artefacts afterred)
using SASICA.
after using
The X axis shows the samples.
SASICA. The X axis shows the samples.

5.4. Figure 13.

Filtering an The
AEEG AF8 channel
Using wICAof AEEG with (in black) and without artefacts (in red) after using
Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors
SASICA. The X axis shows the samples.
The AEEG has also been filtered using the wICA method [19]. This method uses stationary
wavelets to19,enhance
Sensors 2019, independent
x; doi: FOR PEER REVIEWcomponents analysis artefact removal. The following parameters
www.mdpi.com/journal/sensors
have been used: a level set for stationary wavelet transform equal to 5, a threshold multiplier equal
to 1 and “coif5” as wavelet family. Figure 14 shows the AF8 channel of AEEG after using the wICA
method (in red). In this example wICA filters more than necessary and, consequently, there is a lost of
information. Another choice of the parameters could achieve better filtering. However if this happens
this method is no longer a generic and automatic process.
method
have been(inused:
red). In this example
a level wICA filters
set for stationary moretransform
wavelet than necessary and,
equal to 5, aconsequently, there is aequal
threshold multiplier lost
of
to 1information.
and “coif5” Another
as waveletchoice of Figure
family. the parameters
14 shows could achieve
the AF8 better
channel filtering.
of AEEG afterHowever
using the if this
wICA
happens
method (inthisred).
method is no
In this longerwICA
example a generic andmore
filters automatic process. and, consequently, there is a lost
than necessary
of information. Another choice of the parameters could achieve better filtering. However if this
Sensors 2019, 19, 2302 13 of 22
happens this method is no longer a generic and automatic process.

Figure 14. The AF8 channel of AEEG with (in black) and without artefacts (in red) after using wICA.
The X axis shows the samples.
Figure 14. The AF8 channel of AEEG with (in black) and without artefacts (in red) after using wICA.
Figure 14. The AF8 channel of AEEG with (in black) and without artefacts (in red) after using wICA.
5.5. Filtering The
a SEEG
X axis using UAR
shows the samples.
The X axis shows the samples.
5.5. Filtering
In this section,a SEEG
we Using
filter UARthe artefacts of a SEEG. The SEEG has 64 channels, a sampling
5.5. Filtering a
In SEEG
this using
frequency of 256 Hz and duration
section, weUAR
filter of 1 min.
the artefacts of aThe
SEEG.simulated
The SEEG artefacts represent
has 64 channels, somefrequency
a sampling eyes and EMG
artefacts.of 256 Hz and
Thesection, duration
eyes artefacts of
have1 min.
been The simulated
obtained artefacts represent some eyes and EMG artefacts.
In this we filter the artefacts of using
a SEEG. a moving
The SEEG average
hasfilter over somea real
64 channels,
The eyes artefacts have been obtained using a moving average filter over some real noisy eyes artefacts.
noisy
sampling
eyes artefacts.
frequency of 256Muscle
Hz and artefacts areof obtained by band-pass filtering random noise between 20EMG
and
Muscle artefacts are duration 1 min. The
obtained by band-pass simulated
filtering randomartefacts represent
noise between 20 andsome
60 Hz.eyes
Theseand
60 Hz. These
artefacts.artefactsartefacts
The eyes beenhave
haveartefacts been
have
projected projected
been
over theobtained
channelsover thetoachannels
using
trying moving
imitate the trying
averagetofilter
behaviour ofimitate
over
some the
real behaviour
some of
real noisy
artefacts.
some real artefacts.
Figure
eyes artefacts. Figure
15Muscle
shows the 15 shows
simulated
artefacts the simulated
artefacts.
are obtained artefacts. filtering random noise between 20 and
by band-pass
60 Hz. These artefacts have been projected over the channels trying to imitate the behaviour of
some real artefacts. Figure 15 shows the simulated artefacts.

Figure 15. The simulated artefacts.

Figure 15. The simulated artefacts.
Figure 16 shows a channel of the SEEG and the distances of the SEEG intervals to their centroid
for AD_UAR. The centroid has been calculated using all the intervals. The sliding window size is
Figure 16 shows a channel of the SEEG and the distances of the SEEG intervals to their centroid
Lw = 155 and the sliding distanceFigure
is d = 32.
15.The
Thepotential
simulated artefacts.
portions of signal containing artefacts appear
for AD_UAR. The centroid
when selecting has intervals
consecutive been calculated using
with a distance all the
greater intervals.
or equal to T1 =The
1.33.sliding window
The artefacts are size is
Lw = Figure
155clearly
and the sliding
16selected.
shows distance
a channel is SEEG
of the d = 32.andThethepotential
distancesportions
of the SEEGof signal containing
intervals to their artefacts
centroid
appear when selecting
Before applyingconsecutive
UAR to SEEG, intervals with
an EEG portion a distance
without greater
artefacts of 30 s wasor equal
added
for AD_UAR. The centroid has been calculated using all the intervals. The sliding window size is at theto T 1 = 1.33. The
beginning
artefactsofare
SEEG. In this
clearly way, UAR can filter the artefacts near the beginning of SEEG.
selected.
Lw = 155 and the sliding distance is d = 32. The potential portions of signal containing artefacts
Sensors
appear2019,
when19, x; selecting
doi: FOR PEER REVIEW
intervals with a distance greater orwww.mdpi.com/journal/sensors
consecutive equal to T1 = 1.33. The
artefacts are clearly selected.
Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors
Sensors 2019, 19, x 14 of 22

Before applying UAR to SEEG, an EEG portion without artefacts of 30 s was added at the
beginning of 2019,
Sensors SEEG. In this way, UAR can filter the artefacts near the beginning of SEEG.
19, 2302 14 of 22

Figure 16. A SEEG channel, the distances of the intervals to their centroid and the thresholds T1 (in
black) and T2 (in red). The black arrows point at some intervals not related to the
simulated artefacts.

Figure 16. A SEEG channel, the distances of the intervals to their centroid and the thresholds T1
Figure 17 A
Figure 16. shows
SEEG the entire theAF8 channel ofintervals
SEEG without and with andfiltering. Following
T1 (in the
(in black) and Tchannel, distances of the to their centroid the thresholds
2 (in red). The black arrows point at some intervals not related to the simulated artefacts.
ordered ICs and
black) for each
T2 (in artefact,
red). UARThe obtains resultspoint
black arrows that approximate
at some intervals REEG. not The related
statistical
to criterion
the
of UAR stops
Figure the filtering
17 shows
simulated artefacts. algorithm.
the entire AF8 channel Figure
of SEEG16 shows
without and some examples
with filtering. in
Following which
the UAR
ordered has
ICs for each artefact, UAR obtains results that approximate REEG. The
successfully removed different artefacts. Artefacts used in SEEG are in black, the corresponding statistical criterion of UAR
stops the filtering algorithm. Figure 16 shows some examples in which UAR has successfully removed
portions of REEG
Figure 17 shows in green and the
the entire result
AF8 of theoffiltering
channel using UAR
SEEG without andinwithred. filtering.
The FastICA algorithm
Following the
different artefacts. Artefacts used in SEEG are in black, the corresponding portions of REEG in green
has used
orderedandan ICA_P
ICstheforresult
each of 26 s.
of artefact,
the filtering UAR
usingobtains results
UAR in red. that approximate
The FastICA algorithm hasREEG. used anThe
ICA_Pstatistical
of 26 s. criterion
of UAR stops the filtering algorithm. Figure 16 shows some examples in which UAR has
successfully removed different artefacts. Artefacts used in SEEG are in black, the corresponding
portions of REEG in green and the result of the filtering using UAR in red. The FastICA algorithm
has used an ICA_P of 26 s.

Figure 17. The AF8 channel of SEEG with (in black) and without artefacts (in red) after using UAR.
Figure 17. The AF8 channel of SEEG with (in black) and without artefacts (in red) after using UAR.
The X axis shows the samples.
The X axis shows the samples.
Differences between the filtered portions and their corresponding REEG portions are evaluated
Differences ofbetween
by means the Mean Square Error (MSE):
the filtered portions and their corresponding REEG portions are
evaluated
Figureby
17.means of the
The AF8 Mean
channel of Square Error
SEEG with (inM(MSE):
P
black)
[FP and ]2 artefacts (in red) after using UAR.
without
− RP n=1 EEG EEG
MSE = (2)
The X axis shows the samples. ∑ 𝐹𝑃 M 𝑅𝑃
𝑀𝑆𝐸 = (2)
where FPEEG is a filtered portion of M samples and RP𝑀 EEG is its corresponding REEG portion.
Differences between the filtered portions andthetheir corresponding
between the filteredREEG
portionsportions are
where FPEEG In is Figure 18, following
a filtered portion the order from
of M samplesleftandto right,
RPEEG isMSEs
its corresponding REEG and
portion.
evaluated bycorresponding
their means of thepartsMean Square
of REEG are Error
7.6, 5.6(MSE):
and 10.1 for the first row and 7.4, 4.7 and 5.9 for the
In Figure 18,
second row. following the order from left to right, the MSEs between the filtered portions and
their corresponding parts of REEG are 7.6, 5.6 ∑ and 𝐹𝑃10.1 for 𝑅𝑃the first row and 7.4, 4.7 and 5.9 for the
𝑀𝑆𝐸 = (2)
second row. 𝑀
where FPEEG is a filtered portion of M samples and RPEEG is its corresponding REEG portion.
In Figure 18, following the order from left to right, the MSEs between the filtered portions and
Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors
 Sensors 2019, 19, 2302 15 of 22
Sensors 2019, 19, x 15 of 22

Figure 18. Some portions of SEEG with artefacts of the AF8 channel (in black), the corresponding
REEG portions (in green) and the result of the filtering after using UAR (in red).

The quality of the UAR result depends on the quality of the ICs which in turn depends on the
size of ICA_P whereFigure Some
Some portions
Figure 18.FastICA is ofapplied.
portions ofSEEG
SEEGwith artefacts
The
with of the
SEEG
artefacts AF8 channel
allows
of the AF8 (in black),
the
channel testingthe corresponding
(in black), of
the different REEG
corresponding sizes for ICA_P.
portions
REEG (in green)
portions and theand
(in green) result
the of the filtering
result after using
of the filtering afterUAR
using(in UARred).
(in red).
For each size, the MSE between the result of UAR over the entire SEEG and REEG (using the AF8
channel) has been The calculated.
The quality
quality of
of the From
theUAR
UAR resulta size
result of on
depends
depends ICA_P
onthe
the of 20
quality
quality sthethe
of of the quality
ICsICs
which
which of
in turn the filtering
depends
in turn depends onfor
on the the most of the
size
of ICA_P where
size of ICA_P where FastICA is
FastICAapplied. The
is applied. SEEG allows
The SEEG the testing
allows reached of different
the testingwithin sizes
of differentfor ICA_P. For each
artefacts is acceptable. However, the best result has been thesizes for ICA_P.
interval [25,30] seconds
size,each
For the MSE
size, between
the MSEthe result of
between theUAR over
result of the
UAR entire
overSEEG and REEG
the entire SEEG(using the AF8
and REEG channel)
(using has
the AF8
where the MSEbeen between
calculated. the
From afiltering
size of ICA_Presults
of 20 and
s the REEG
quality of are
the low
filteringandfor
channel) has been calculated. From a size of ICA_P of 20 s the quality of the filtering for most of the
steady.
most of Figure
the artefacts 19
is shows the
MSE (for the entire
acceptable.
artefacts AF8 channel)
However,
is acceptable. the
However, between
best result has been
the best the
result hasfiltering
reachedbeenwithin
reached results
the and
intervalthe
within REEG
[25,30] seconds
interval from
[25,30]where different
secondsthe sizes
MSE between the filtering results and REEG are low and steady. Figure 19 shows the MSE (for the
of ICA_P. where the MSE between the filtering results and REEG are low and steady. Figure 19 shows the
entire(for
MSE AF8the
channel)
entirebetween the filtering
AF8 channel) results
between theand REEG from
filtering different
results sizes of
and REEG ICA_P.
from different sizes
of ICA_P.

Figure 19. The mean squared errors (MSEs) (for the entire AF8 channel) between the filtering results
and REEG from different sizes of ICA_P.
Figure 19. Figure
The 19. The
mean mean squared
squared errorserrors (MSEs)
(MSEs) (forthe
(for the entire
entireAF8 channel)
AF8 between
channel) the filtering
between theresults
filtering results
and REEG
This studyfrom different
justifies thesizes of ICA_P.
selection of ICA_P = 26 s that has been previously used in UAR. With
and REEG from different sizes of ICA_P.
this selection, the filtering is very similar to REEG. In Figure 14, the potential artefacts are portions
This study justifies the selection of ICA_P = 26 s that has been previously used in UAR. With
of consecutive intervals with a distance greater or equal to T1 = 1.33. Some intervals between T1 and
this selection, thethe
This study filtering is very similar to REEG. In Figure 14,has
the potential artefacts are portions
inof UAR. With
T2, notjustifies selection
related to the simulated of ICA_P
artefacts, are also= selected;
26 s that
for example,been the previously
intervals 53, 142used
and 331
consecutive intervals with a distance greater or equal to T1 = 1.33. Some intervals between T1 and T2 ,
(seethe
this selection, the black arrows).
filtering is The
veryUAR methodtotries
similar to filterIn
REEG. them but the
Figure statistical
14, criterion ofartefacts
the potential UAR stopsare portions
not related to the simulated artefacts, are also selected; for example, the intervals 53, 142 and 331
the algorithm at the beginning. Consequently, UAR only filters the simulated artefacts of SEEG.
of consecutive intervals
(see the with The
black arrows). a distance
UAR method greater
tries toor equal
filter themto butTthe
1 = statistical
1.33. Some intervals
criterion between T1 and
of UAR stops
the
T2, not related algorithm at the
to the simulated
5.6. Filtering beginning. Consequently,
a SEEG Signalartefacts, UAR only filters the simulated artefacts
using MARAare also selected; for example, the intervals 53, 142 and 331 of SEEG.

(see the black arrows).

The SEEGThe UAR
signal method
has been filteredtries
usingtoMARA.
filter them
Figure but the statistical
20 shows criterion
the AF8 channel of UAR stops
of SEEG
the algorithm
afteratusing
the beginning. Consequently,
MARA (in red). In this example,UAR MARA only
has filters thesimilar
behaviour simulated artefacts
to the one shown of
in SEEG.
Sensors 2019, 19, x; doi: FOR PEER REVIEW www.mdpi.com/journal/sensors
5.6. Filtering a SEEG Signal using MARA
 Sensors 2019, 19, 2302 16 of 22

Sensors 2019, 19, x 16 of 22

5.6. Filtering a SEEG Signal Using MARA
Sensors 2019, 19, x 16 of 22
Section
The5.2. MARA
SEEG does
signal hasnot
beenselect all the
filtered necessary
using MARA.ICs to filter
Figure the artefacts
20 shows the AF8and, consequently,
channel the
of SEEG after
filtering is not(incomplete; especially the EMG has
filtering. The similar
MSE between theshown
filtering result and
Sectionusing
5.2. MARA
REEG
MARA
is 13.42,
red). In this
does
which
not
is bigger
example,
select theMARA
all UAR
than
necessary
where MSE
behaviour
ICs
=
to filter thetoartefacts
4.1.
the one
and, in Section 5.2.
consequently, the
MARA does not select all the necessary ICs to filter the artefacts and, consequently, the filtering is not
filtering is not complete; especially the EMG filtering. The MSE between the filtering result and
complete; especially the EMG filtering. The MSE between the filtering result and REEG is 13.42, which
REEG is is bigger
13.42,than
which
UARis where
biggerMSE
than= UAR
4.1. where MSE = 4.1.

Figure 20. The AF8 channel of SEEG with (in black) and without artefacts (in red) after using
MARA. The X axis shows the samples.

5.7. Filtering a SEEG Signal Using SASICA

20. The AF8 channel of SEEG with (in black) and without artefacts (in red) after using MARA.
FigureFigure
20. The AF8 channel of SEEG with (in black) and without artefacts (in red) after using
TheXSEEG
The signalthehas
axis shows been filtered using SASICA. Figure 21 shows the AF8 channel of SEEG
samples.
MARA. The X axis shows the samples.
after using SASICA (in red). In this example, SASICA has used the same parameters than
5.7. Filtering a SEEG Signal Using SASICA
Section 5.3.
5.7. Filtering a SEEG
The SEEG Signal
filtering
signal aUsing
hashas been SASICA
better behaviour
filtered usingthan the one
SASICA. shown
Figure in Section
21 shows 5.3.channel
the AF8 However, the MSE
of SEEG after
between the filtering
using
The SASICA
SEEG signal hasresult
(in red). been andexample,
In this REEG isSASICA
filtered 16.1. Bigger
using SASICA.than MARA
has used the sameand
Figure UAR.
21parameters
shows than
the Section
AF8 5.3. of SEEG
channel
after using SASICA (in red). In this example, SASICA has used the same parameters than
Section 5.3.
The filtering has a better behaviour than the one shown in Section 5.3. However, the MSE
between the filtering result and REEG is 16.1. Bigger than MARA and UAR.

Figure 21. The

Figure 21. TheAF8
AF8channel of SEEG
channel withwith
of SEEG (in black) and without
(in black) artefacts
and without (in red)(in
artefacts after using
red) SASICA.
after using
The X axis shows the samples.
SASICA. The X axis shows the samples.

The filtering
5.8. Filtering hasSignal
a SEEG a better wICA than the one shown in Section 5.3. However, the MSE between
behaviour
Using
the filtering result and REEG is 16.1. Bigger than MARA and UAR.
The SEEG signal has been filtered using wICA. Figure 22 shows the AF8 channel of SEEG after
5.8.
usingFiltering
wICA a(in SEEG
red).Signal
In thisUsing wICAwICA has used the same parameters than Section 5.4.
example,
The SEEG
The filtering hashas
signal a different behaviour
been filtered than the
using wICA. one 22
Figure shown
showsinthe
Section 5.4. In of
AF8 channel this case,after
SEEG the
Figure 21. isThe
filtering not AF8 channel
complete; of SEEG
especially with
the EMG (infiltering.
black) and
The without
MSE artefacts
between the (in red) result
filtering after using
and
using wICA (in red). In this example, wICA has used the same parameters than Section 5.4.
REEG
SASICA. isThe
The 19.67; bigger
X axis
filtering athan
shows
has SASICA,
the MARA
samples.
different behaviour andthe
than UAR.
one shown in Section 5.4. In this case, the filtering
is not complete; especially the EMG filtering. The MSE between the filtering result and REEG is 19.67;
5.8. Filtering
bigger a SEEG
than Signal
SASICA, Usingand
MARA wICA
UAR.
The SEEG signal has been filtered using wICA. Figure 22 shows the AF8 channel of SEEG after
using wICA (in red). In this example, wICA has used the same parameters than Section 5.4.
The filtering has a different behaviour than the one shown in Section 5.4. In this case, the
filtering is not
Sensors 2019, complete;
19, x; doi: FOR especially
PEER REVIEW the EMG filtering. The MSE between the filtering result and
www.mdpi.com/journal/sensors
 Sensors 2019, 19, 2302 17 of 22
Sensors 2019, 19, x 17 of 22

Figure
Figure 22. The
22. The AF8AF8 channel
channel ofofSEEG
SEEGwith
with(in
(in black)
black) and
andwithout
withoutartefacts
artefacts(in(in
red) after
red) using
after wICA.
using wICA.
The X axis shows the samples.
The X axis shows the samples.
6. Discussion
6. Discussion
The UAR method allows automated filtering of artefacts. This method combines unsupervised
artefact
The UAR detection,
method FastICA
allowsand unsupervised
automated ICs selection.
filtering of artefacts.EachThis
artefact detected
method with AD_UAR
combines is
unsupervised
filtered individually when the previous artefacts have been removed.
artefact detection, FastICA and unsupervised ICs selection. Each artefact detected with AD_UAR is Most of the original information
between
filtered artefacts is when
individually preserved. the previous artefacts have been removed. Most of the original
The focus of the
information between artefacts is present study is on physiological artefacts. Although UAR could also be used to
preserved.
filter nonphysiological artefacts, they usually are avoided through an appropriate filtering as a previous
The focus of the present study is on physiological artefacts. Although UAR could also be used
step to UAR. For example, an interference of the electrical network can be corrected by means of
to filter nonphysiological artefacts, they usually are avoided through an appropriate filtering as a
a band-stop filter with a narrow stopband (notch filter). The AEEG in Section 5 has been previously
previous
filteredstep
usingto aUAR.
notch For filterexample,
to remove an interference
potential of the electrical network can be corrected by
60 Hz interferences.
means ofInaan band-stop filter with
online application, a narrow
UAR_AD stopband
obtains the AR (notch filter). The
of the intervals as theAEEG
AEEG in Section 5When
is acquired. has been
previously
it obtainsfiltered using
a set of a notch intervals
consecutive filter to remove potential
with a distance to60theHz interferences.
centroid greater than T1 , it detects
a potential artefact. Then, a ICA_P of 26 s is selected. It is convenient to add a as
In an online application, UAR_AD obtains the AR of the intervals the AEEG
portion of EEG is acquired.
without
When it obtains
artefacts beforeathe setstarting
of consecutive
point of theintervals
acquisition. with a distance
In this to the centroid
way, it is possible the selectiongreater
of ICA_PthanforT1, it
an artefact near the starting point.
detects a potential artefact. Then, a ICA_P of 26 s is selected. It is convenient to add a portion of
EEG without In an online application,
artefacts before the it is starting
sufficientpoint
to use ofa processor where theInfiltering
the acquisition. this way,processing time of the
it is possible
an artefact with UAR is less than the acquisition
selection of ICA_P for an artefact near the starting point. of 2 × d samples; this results from the fact that after
an artefact there is always at least one interval without artefact. In the most stringent case, we find
In an online application, it is sufficient to use a processor where the filtering processing time of
after the artefact two intervals where the second is a potential artefact. While filtering occurs, these
an artefact with UAR is less than the acquisition of 2 × d samples; this results from the fact that after
2 × d samples must be saved. This is possible using ping-pong storage.
an artefact Thethere
AD_UARis always at least
is the result of aone interval without
simplification of UMED. artefact.
AD_UAR In has
thebeen
mostdesigned
stringent case, we
to operate in find
afteran the artefact twoway
unsupervised intervals
with UAR. where the secondanisoptimal
In AD_UAR, a potential artefact.
threshold fromWhile
whichfiltering occurs,
the intervals withthese
2 × dartefacts
samplesappearmust is benotsaved. ThisInisour
defined. possible
work, we using
selectping-pong
a minimum storage.
threshold T1 = 1.33 (the start of the
The tail)
right AD_UAR
for our is the resultof of
distribution a simplification
64 electrodes. of UMED.
This implies that theAD_UAR has been
selected portions designed
of the signal to
operate
usinginT1anmay unsupervised
or may not have wayanwith UAR.If In
artefact. UAR AD_UAR, an optimal
does not filter a portion,threshold from
that portion doeswhich
not the
intervals with artefacts appear is not defined. In our work, we select a minimum threshold T1or= 1.33
present artefacts. AD_UAR detects potential signal portions to be filtered and it is UAR which filters
(the not (following
start of the arightstatistical
tail) criterion)
for our such portions. of 64 electrodes. This implies that the selected
distribution
portions of the signal using T1 may or maytonot
The AD_UAR is a detection method work havewithan UAR in an If
artefact. unsupervised
UAR does way and itadoes
not filter portion,
not make sense to compare it with other detection methods since it does not work with an optimal
that portion does not present artefacts. AD_UAR detects potential signal portions to be filtered and
threshold. This method proposes possible portions with artefacts to UAR and the statistical criterion of
it is UAR which filters or not (following a statistical criterion) such portions.
UAR decides whether it is necessary to filter or not. That is, “true detection” occurs once UAR has
The AD_UAR
worked on a portionis aselected
detection method toThis
by AD_UAR. work withmore
implies UAR in an unsupervised
processing, way andthat
but has the advantage it does
not make sense to compare it with other detection methods
AD_UAR and UAR can work in an unsupervised way and in online applications. since it does not work with an optimal
threshold. This method proposes possible portions with artefacts to UAR and the statistical
criterion of UAR decides whether it is necessary to filter or not. That is, “true detection” occurs once
UAR has worked on a portion selected by AD_UAR. This implies more processing, but has the
advantage that AD_UAR and UAR can work in an unsupervised way and in online applications.
AD_UAR takes advantage of the AR coefficients, since, on the one hand, they are useful to
Sensors 2019, 19, 2302 18 of 22

AD_UAR takes advantage of the AR coefficients, since, on the one hand, they are useful to detect
artefacts in a simple way and, on the other, they allow us to generalize the results. These characteristics
also justify why this method has been used in UAR.
In Figure 8, a small artefact not detected to be filtered is shown. This artefact has not interfered
with the filtering of others artefacts. When the number of electrodes is high, that kind of artefact
should not affect the global information collected from all the channels. In this case, it is not necessary
to filter it. However, the AD_UAR method can improve its performance by working with a small
number of electrodes or with each electrode individually. Consequently, for better detection, instead of
using AD_UAR on all the channels, it can be applied to small groups of electrodes and then merge the
potential artefacts detected. When the AD_UAR works with fewer channels, the Euclidean distance is
more sensitive to small changes.
In theory, any ICA-based method can filter any type of artefact as long as it works with the right
number of electrodes. Consequently, UAR can also correctly filter any type of artefact.
The filtering quality of a particular method varies according to the conditions of use. The quality
of a method can change if, for example, the number of electrodes, the type of artefact, the selected ICA
method, the internal parameters of each method, etc. also change.
In general, when working with a number of ICs less than the optimal (e.g., ICs from a portion of
EEG signal that has an artefact) there is a possibility that some ICs related to the artefact have some
cognitive activity. Therefore, it is important to work with a high number of electrodes and, in this way,
the possible cognitive influence in the ICs related to the artefacts will be very small or nonexistent.
From there, it only remains to select those ICs related to the artefacts to perform the filtering. We use
a statistical criterion to obtain a good result selecting the ICs related to the artefact. We cannot quantify
if some ICs related to the artefact have any cognitive influence. However, we can affirm that with
a high number of electrodes, that cognitive influence will be nonexistent or very small.
UAR always works in the same way for each type of artefact. ICA is used to obtain the ICs for
each artefact, (in our case using FastICA to obtain 64 ICs). Regardless of the content of the ICs, the only
difference between artefacts is the number of ICs that represent each artefact. For example, a wink
can be related to a small number of ICs. However, for an EMG artefact or a mixture or succession of
artefacts with EMG, the number of related ICs increases significantly.
UAR sorts the ICs and uses their projections, one by one, to filter the selected artefact. Using
a statistical criterion, the filtering process is stopped. This criterion (a two-sample F-Test working at 1%
significance) is also a way of quantitatively assessing the quality of the result establishing the same
filtering objective for any type of artefact. In UAR, this process is the same for any artefact and is
independent of the type of artefact. UAR does not care about the type of artefact, it only considers a set
of ICs to work with. All these reasons clearly justify why the evaluation of our filtering method does
not require testing all the existing artefacts. We believe that UAR is well explained with the artefacts
found in the real and simulated recordings used in this paper. More types of artefacts would not add
additional information about how our methodology works.
Neither UAR nor AD_UAR depend on the origin of the test data. Consequently, the filtering
process is a subject-independent algorithm. In this way, the AEEG used in this paper to test the
algorithm is sufficient to analyse and evaluate it.
It is understood that UAR is only for artefacts that are mixed to the EEG signal. This method is
not valid for artefacts where there is nothing to filter; for example, a disconnected electrode.
The AEEG has allowed us to show how UAR filters some real artefacts. In other EEG recordings,
more types of artefacts exist. As explained above, this fact is not a problem for UAR because its
filtering process does not depend on the type of artefact. When, in an AEEG, there is a large portion
contaminated by artefacts, UAR filtering may not be so effective. The UAR needs a REF to filter
an artefact. If the entire signal is contaminated, it would not be possible to select such REF and,
therefore, UAR could not be used. However, if it is possible to select a REF before the large artefact,
then it would be preferable to divide the artefact into segments and apply UAR sequentially.
Sensors 2019, 19, 2302 19 of 22

Our method has the following characteristics.

• It uses an unsupervised detection method to detect potential signal portions of artefacts.

• The filtering process is unsupervised (it does not depend on a learning process and automatically
selects ICs using a statistical criterion).
• Its strategy to obtain the ICs around each artefact (with an ICA_P portion) improves the quality
of the ICs and facilitates its ordering and selection. Most of the original information between
artefacts is preserved.
• It is easy to implement and can be used in online applications.
• It can be used with different types of artefacts including a portion of consecutive artefacts and any
mixture of artefacts.
• To obtain good results it is necessary to work with a significant number of electrodes (e.g., 64).

The main characteristic is the automatic unsupervised selection of the ICs. It is not necessary
to make a manual selection or a learning process (for an automatic selection using a machine
learning algorithm).
UAR results can be evaluated by visual inspections of Figures 7–9 of an AEEG and Figures 15
and 16 of a SEEG. MSE has also been used to evaluate the filtering of SEEG.
The quality of the proposed UAR filtering is at least as good as that of other methods. We have
compared UAR with MARA. Both are ICA-based methods and pursue the same goal: to automatically
select ICs. MARA needs less processing but it is a supervised method and the quality of the ICs
selection depends on the quality of a learning process. On the other hand, UAR is an unsupervised
method and the ICs selection quality depends on a statistical criterion. Both methods have been tested
using an AEEG and the results can be visually analysed. In general, UAR provides a better quality
result than MARA. Graphics and MSE for the SEEG filtering show this fact.
We have also compared SASICA and UAR. In [18], different methods to select artefactual ICs are
evaluated. They show that “their selection was not perfectly consistent, showing that there are inherent
limitations to the precision of artefact selection using ICA” and, consequently, “no automated method
can accurately isolate artefacts without supervision”. Some of those methods work only with some
specific type of artefact. Unlike UAR, in general, SASICA does not automatically select the correct ICs
to filter artefacts and needs supervision.
With the wICA method, the worst results have been obtained. In theory, the quality of the wICA
filtering depends on the choice of certain parameters. Therefore, if we use the same parameters for any
signal, the quality of the result may vary. If the parameters need to be changed to improve the quality,
then, unlike UAR, wICA cannot be considered as a generic and automatic method.
An ICA-based method used to filter artefacts begins with the detection of the artefacts (this stage
is not always done), the ICs are obtained and selected for filtering and finally the quality of the result
can be analysed through a visual inspection or a quantitative evaluation. These three processes are
independent. However, in our work, these stages are related to each other by the statistical criterion.
The statistical criterion determines if there is filtering or not (that is, whether there is an artefact or not),
it controls the filtering of the ICs projections since it decides when the process stops and, finally, it is
also a way to evaluate the result, establishing the same objective for any type of artefact. Instead of
using an index after filtering, the statistical criterion is itself an indirect form of quantitative evaluation
of the result.
The performance of the statistical criterion to select ICs has been evaluated quantitatively with the
help of a SEEG and the MSE metric. The MSE shows the difference between the filtering result using
UAR for each artefact and its corresponding part of REEG. The good filtering quality of the different
analysed artefacts shows that UAR correctly selects the ICs in an unsupervised way. However, the
quality of the result also depends on how the ICs have been obtained. The size of ICA_P is important
for ICs to capture the characteristics of the artefacts.
Sensors 2019, 19, 2302 20 of 22

All UAR steps have been selected to use UAR in online applications. However, for off-line
applications, UAR could benefit from data reduction techniques, such as PCA [43]. These would be
applied to the vectors of autoregressive coefficients that represent the intervals obtaining new vectors.
In theory, with these new vectors, AD_UAR would improve the quality of artefact detection.
ICA-based methods are well-accepted in the literature for filtering artefacts and it is not the
objective of our article to evaluate their performance in relation to other methods. ICA-based methods
prevail in the literature [1,44], and have proved to be satisfactory in most scenarios of interest.
For example, Ref. [44] presents a comparison of artefact removal algorithms including ICA-based
methods. Consequently, UAR is important insofar as it addresses the problem of automatic selection
of ICs in ICA-based methods.
The UAR strategy to obtain the ICs around each artefact (with an ICA_P portion) improves the
quality of the ICs and facilitates its selection. It also allows UAR to preserve most of the original
information between artefacts. Figure 6 shows the ICA_P portion used to determine the ICs. However,
the filtering of the ICs projections is only carried out on the area of the artefact (AT1 + AT + AT2 ).
This means that the portion between the origin of ICA_P and the origin of AT1 is not altered at all.
Depending on the detection of the artefact and the selection of AT1 (or AT2 ), the filtering may slightly
alter a small portion called EA1 (or EA2 ), (see Figure 5) unrelated to the artefact. Unlike other methods
that do not perform detection and filter using ICs throughout the signal, our algorithm retains most of
the original information among the filtered artefacts.
A two-sample F-Test for equal variances is used to select the ICs related to the artefact.
Other statistical tests or a combination of them can also be used. For example, the two-sample
Kolmogorow-Smirnov test [45]. This test analyses if two sets of data X1 and X2 are from the same
continuous distribution. It tests the null hypothesis that data in X1 and X2 come from populations
with the same distribution. However, the two-sample F-Test has given good results working in UAR.
It is especially effective when different consecutive artefacts appear.
The common issue among the ICA-based methods is the algorithm to get the ICs. It is also the part
of the algorithm that needs more time. The processing time difference between UAR and other possible
methods can be estimated. The processing time of an ICA algorithm depends on the window size used.
Suppose TD denotes the processing time of an ICA over a window of D seconds. For an AEEG with
NP artefacts, UAR will work with MP portions of potential artefacts with MP >=NP. If we compare
UAR with another algorithm that applies ICA in the whole signal of DT seconds, then the processing
time difference is approximately (MP × T26 ) − TDT .
In an online application, artefact detection depends on the number of electrodes. To detect a small
artefact that is present in a few electrodes, we suggest to work with groups of electrodes smaller than 64.
This increases the sensitivity of the detection. The ideal would be to work with a number of electrodes
such that the detection of all the artefacts is coincident with the opinion of an expert. As a future task,
a clinical study would allow us to study different groups of electrodes and choose the optimal group
(or the optimal groups) so that the detection of the artefacts is more accurate.

7. Conclusions
The presented UAR method is an automatic, reliable tool developed to remove artefacts in
an unsupervised way. It combines an unsupervised artefact detection method with an automatic
selection of ICs after using the FastICA algorithm for each artefact individually. This method can be
used in online applications.
The two-sample F-test used to select the ICs also offers a statistical criterion to justify the filtering
process. Furthermore, the results of the algorithm can be visually analysed.
The UAR filtering on EEG test signals with different types of artefacts has been correct. This
means that the unsupervised strategy proposed for selecting the ICs works. As UAR is unsupervised,
it should be able to filter any artefact, regardless on their number or type.
Sensors 2019, 19, 2302 21 of 22

Author Contributions: A.M. researched the literatures, designed the algorithm, analysed the results and wrote
the manuscript; R.D. and N.D. wrote and revised the manuscript. All authors have approved the final manuscript.
Funding: This work was supported in part by the Spanish Ministry of Economy and Competitiveness under
Project CICY DPI2017-84259-C2-2-R, Project DPI2014-55932-C2-2-R and the Project CICYT RTI2018-094665-B-I00.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Urigüen, J.A.; Garcia-Zapirain, B. EEG artefact removal—State-of-the-art and guidelines. J. Neural Eng. 2015,
12, 031001. [CrossRef]
2. Nolan, H.; Whelan, R.; Reilly, R.B. FASTER: Fully Automated Statistical Thresholding for EEG artefact
Rejection. J. Neurosci. Methods 2010, 192, 152–162. [CrossRef] [PubMed]
3. Sörnmo, L.; Laguna, P. Bioelectrical Signal Processing in Cardiac and Neurological Applications; Academic Press:
Cambridge, MA, USA, 2005.
4. Lance, B.J.; Kerick, S.E.; Ries, A.J.; Oie, K.S.; McDowell, K. Brain-computer interface technologies in the
coming decades. Proc. IEEE 2012, 100, 1585–1599. [CrossRef]
5. James, C.J.; Gibson, O.J. Temporally constrained ICA: An application to artefact rejection in electromagnetic
brain signal analysis. IEEE Trans. Biomed. Eng. 2003, 50, 1108–1116. [CrossRef] [PubMed]
6. Pham, T.T.H.; Croft, R.J.; Cadusch, P.J.; Barry, R.J. A test of four EOG correction methods using an improved
validation technique. Int. J. Psychophysiol. 2011, 79, 203–210. [CrossRef] [PubMed]
7. Croft, R.J.; Barry, R.J. Removal of ocular artefact from the EEG: A review. Neurophysiol. Clin. 2000, 30, 5–19.
[CrossRef]
8. Sweeney, K.T.; Ward, T.E.; McLoone, S.F. Artefact removal in physiological signals-practices and possibilities.
IEEE Trans. Inf. Technol. Biomed. 2012, 16, 488–500. [CrossRef] [PubMed]
9. James, C.J.; Hesse, C.W. Independent component analysis for biomedical signals. Physiol. Meas. 2005, 26,
R15. [CrossRef] [PubMed]
10. Vigário, R.; Oja, E. BSS and ICA in Neuroinformatics: From Current Practices to Open Challenges. IEEE Rev.
Biomed. Eng. 2008, 1, 50–61. [CrossRef]
11. Jung, T.P.; Makeig, S.; Humphries, C.; Lee, T.W.; McKeown, M.J.; Iragui, V.; Sejnowski, T.J. Removing
electroencephalographic artefacts by blind source separation. Psychophysiology 2000, 37, 163–178. [CrossRef]
[PubMed]
12. Junghöfer, M.; Elbert, T.; Tucker, D.M.; Rockstroh, B. Statistical control of artefacts in dense array EEG/MEG
studies. Psychophysiology 2000, 37, 523–532. [CrossRef] [PubMed]
13. Oostenveld, R.; Fries, P.; Maris, E.; Schoffelen, J.-M. FieldTrip: Open source software for advanced analysis of
MEG, EEG, and invasive electrophysiological data. Comput. Intell. Neurosci. 2011, 2011, 156869. [CrossRef]
14. Delorme, A.; Makeig, S. EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including
independent component analysis. J. Neurosci. Methods 2004, 134, 9–21. [CrossRef] [PubMed]
15. Lawhern, V.; Hairston, W.D.; Robbins, K. DETECT: A MATLAB Toolbox for Event Detection and Identification
in Time Series, with Applications to artefact Detection in EEG Signals. PLoS One 2013, 8, e62944. [CrossRef]
[PubMed]
16. Mur, A.; Dormido, R.; Vega, J.; Duro, N.; Dormido-Canto, S. Unsupervised event characterization and
detection in multichannel signals: An EEG application. Sensors 2016, 16, 590. [CrossRef] [PubMed]
17. Winkler, I.; Haufe, S.; Tangermann, M. Automatic Classification of artefactual ICA-Components for artefact
Removal in EEG Signals. Behav. Brain Funct. 2011, 7, 30. [CrossRef] [PubMed]
18. Chaumon, M.; Bishop, D.V.; Busch, N.A. A practical guide to the selection of independent components of the
electroencephalogram for artifact correction. J. Neurosci. Methods 2015, 250, 47–63. [CrossRef] [PubMed]
19. Castellanos, N.P.; Makarov, V.A. Recovering EEG brain signals: Artifact suppression with wavelet enhanced
independent component analysis. J. Neurosci. Methods 2006, 158, 300–312. [CrossRef] [PubMed]
20. Melia, U.; Clariá, F.; Vallverdú, M.; Caminal, P. Filtering and thresholding the analytic signal envelope
in order to improve peak and spike noise reduction in EEG signals. Med. Eng. Phys. 2014, 36, 547–553.
[CrossRef] [PubMed]
21. Mahajan, R.; Morshed, B. Unsupervised Eye Blink artefact Denoising of EEG Data with Modified Multiscale
Sample Entropy, Kurtosis and Wavelet-ICA. IEEE J. Biomed. Heal. Informatics. 2014, 2194, 1–8.
Sensors 2019, 19, 2302 22 of 22

22. Daly, I.; Scherer, R.; Billinger, M.; Muller-Putz, G. FORCe: Fully online and automated artefact Removal for
brain-Computer interfacing. IEEE Trans. Neural Syst. Rehabil. Eng. 2014, 23, 725–736. [CrossRef] [PubMed]
23. Zeng, K.; Chen, D.; Ouyang, G.; Wang, L.; Liu, X.; Li, X. An EEMD-ICA Approach to Enhancing artefact
Rejection for Noisy Multivariate Neural Data. IEEE Trans. Neural Syst. Rehabil. Eng. 2015, 24, 630–638.
[CrossRef] [PubMed]
24. Chen, X.; Liu, A.; Chen, Q.; Liu, Y.; Zou, L.; McKeown, M.J. Simultaneous ocular and muscle artifact removal
from EEG data by exploiting diverse statistics. Comput. Biol. Med. 2017, 88, 1–10. [CrossRef]
25. Somers, B.; Francart, T.; Bertrand, A. A generic EEG artifact removal algorithm based on the multi-channel
Wiener filter. J. Neural Eng. 2018, 15, 036007. [CrossRef]
26. Mannan, M.M.N.; Kamran, M.A.; Jeong, M.Y. Identification and Removal of Physiological Artifacts from
Electroencephalogram Signals: A Review. IEEE Access 2018, 6, 30630–30652. [CrossRef]
27. Islam, M.K.; Rastegarnia, A.; Yang, Z. Methods for artifact detection and removal from scalp EEG: A review.
Neurophysiol. Clin./Clin. Neurophysiol. 2016, 46, 287–305. [CrossRef] [PubMed]
28. Chen, X.; Liu, A.; Chiang, J.; Wang, Z.J.; McKeown, M.J.; Ward, R.K. Removing Muscle Artifacts from EEG
Data: Multichannel or Single-Channel Techniques? IEEE Sens. J. 2016, 16, 1986–1997. [CrossRef]
29. Burg, J.P. Maximum Entropy Spectral Analysis. In Proceedings of the 37th Annual International SEG Meeting,
Oklahoma City, OK, USA, 31 October 1967.
30. Rokach, L.; Maimon, O. Chapter 15—Clustering Methods; Springer: Boston, MA, USA, 2005; pp. 321–352.
31. Halkidi, M.; Vazirgiannis, M. Clustering validity assessment: Finding the optimal partitioning of data set.
In Proceedings of the 2001 IEEE International Conference on Data Mining, San Jose, CA, USA, 29 November–2
December 2001; pp. 187–194.
32. Lawhern, V.; Hairston, W.D.; McDowell, K.; Westerfield, M.; Robbins, K. Detection and classification of
subject-generated artifacts in EEG signals using autoregressive models. J. Neurosci. Methods 2012, 208,
181–189. [CrossRef] [PubMed]
33. Mur, A.; Dormido, R.; Vega, J.; Dormido-Canto, S.; Duro, N. Unsupervised event detection and classification
of multichannel signals. Expert Syst. Appl. 2016, 54, 294–303. [CrossRef]
34. Wang, G.; Ding, Q.; Hou, Z. Independent component analysis and its applications in signal processing for
analytical chemistry. Trends Anal. Chem. 2008, 27, 368–376. [CrossRef]
35. Hyvärinen, A.; Oja, E. A Fast Fixed-Point Algorithm for Independent Component Analysis. Neural Comput.
1997, 9, 1483–1492. [CrossRef]
36. Cardoso, J.F.; Souloumiac, A. Blind beamforming for non-gaussian signals. In IEE Proceedings F (Radar and
Signal Processing); IET Digital Library: Zürich, Switzerland, 1993; Volume 140, pp. 362–370.
37. Bell, A.J.; Sejnowski, T.J. An information-maximization approach to blind separation and blind deconvolution.
Neural Comput. 1995, 7, 1129–1159. [CrossRef] [PubMed]
38. Astakhov, S.A.; Stögbauer, H.; Kraskov, A.; Grassberger, P. Monte Carlo algorithm for least dependent
non-negative mixture decomposition. Anal. Chem. 2006, 78, 1620–1627. [CrossRef]
39. Learned-Miller, E.G.; Fisher III, J.W. ICA Using Spacings Estimates of Entropy. J. Mach. Learn. Res. 2003, 4,
1271–1295.
40. Delorme, A.; Palmer, J.; Onton, J.; Oostenveld, R.; Makeig, S. Independent EEG sources are dipolar. PLoS ONE
2012, 7, e30135. [CrossRef] [PubMed]
41. Albera, L.; Kachenoura, A.; Comon, P.; Karfoul, A.; Wendling, F.; Senhadji, L.; Merlet, I. ICA-based EEG
denoising: A comparative analysis of fifteen methods. Bull. Polish Acad. Sci. Tech. Sci. 2012, 60, 407–418.
[CrossRef]
42. Snedecor, G.W.; Cochran, W.G. Statistical Methods, 8th ed.; Iowa State University Press: Iowa City, IA,
USA, 1989.
43. Jolliffe, I.T. Principal Component Analysis, Second Edition. Encycl. Stat. Behav. Sci. 2002, 30, 487.
44. Urigüen, J.A.; Garcia-Zapirain, B. Validation EEG artefact removal. J. Med. Imag. Health Inform. 2017, 7,
174–180. [CrossRef]
45. Massey, M.J. The Kolmogorov-Smirnov Test for Goodness of Fit. J. Am. Stat. Assoc. 1951, 46, 68–78. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).