Generic Dry-Contact Ear-EEG

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Generic Dry-Contact Ear-EEG

A. R. Bertelsen1, H. Bladt1, C. B. Christensen1, S. L. Kappel1,2, H. O. Toft3, M. L. Rank3, K. B. Mikkelsen1 and P.


Kidmose1,4

recording, reference and ground electrode placed on the same


Abstract— Generic dry-contact ear-EEG allows for discreet,


user-friendly, unobtrusive, cost-effective and convenient earpiece. Despite the promising results, the silicone electrodes
recordings of EEG in real-life settings. In this study we introduce required conductive gel for good electrode-skin contact and
a new generic earpiece design with larger internal ear electrode were susceptible to degradation, causing increased electrode
distances, resulting in an increased spatial coverage compared impedance. To address these issues, Goverdovsky et al. [7]
to previous generic earpiece designs. The signal quality of ear- developed a memory foam based device with conductive
Fpz, within-ear (the measuring and reference electrode located silver-coated nylon cloth electrodes allowing for more
in the same ear) and cross-ear (the measuring electrodes located comfortable and long-term monitoring. Using only a small
in one ear and the reference electrode in the opposite ear) amount of saline solution, to obtain good electrode-skin
electrode configurations of the developed generic earpiece was contact, they recorded ASSR with similar SNR as that of
evaluated with auditory steady-state responses (ASSR) and
conventional scalp EEG. The recordings were performed with
compared to dry-contact cEEGrid. Ten subjects with different
an electrode on the earpiece referred to an electrode on the
ear sizes were included. The recordings were performed in a
sleep setup, where the subjects were lying on a bed and the effect same-side mastoid, and with the ground electrode on the same-
of sleeping position (back vs. sides) was investigated. We found side ear-lobe. Both Looney et al. [7] and Nquyen et al. [8]
that the generic earpiece attained statistically significant ASSRs investigated the use of memory foam based earpieces with
with ear-Fpz, within-ear and cross-ear electrode configurations. electrodes made of conductive fabric for sleep assessment.
However, the dry-contact cEEGrid achieved significantly higher These designs included placement of the recording and
average ASSR signal-to-noise ratio (SNR) compared to the reference electrodes in the opposite ear, and a connecting cable
generic earpiece. Additionally, this study showed no significant between the earpieces. They succeeded in monitoring different
difference between back and side positions for the ear-EEG. stages of sleep using a single-channel ear-EEG.
The aim of the current study was to develop a generic
I. INTRODUCTION earpiece with dry electrodes for recording of ear-EEG with
Ear-EEG is one among several methods of wearable EEG recording and reference electrodes placed in the same ear [9].
monitoring that has emerged over the past decade[1, 2]. In Due to the short inter-electrode distances, it is a
contrast to conventional EEG which is traditionally performed challenging task to record EEG signals with high SNR within-
in the laboratory using EEG-caps with wet electrodes, ear- ear. Previously, it was demonstrated that increasing the spatial
EEG enables discreet and minimally intrusive long-term distance by combining a recording electrode in the ear canal
recordings outside the laboratory, using dry-electrodes placed with a reference electrode in the concha leads to improved
in the outer ear [3]. Discrete and mobile EEG has many ASSR SNR when using customized earpieces [6, 10, 11].
applications; one important example is sleep monitoring [4], Based on this finding, it was decided that the generic earpiece
which requires a robust and comfortable device to ensure design used in the current study should include electrodes in
unbiased recordings. both canal and concha.
Traditionally, ear-EEG is recorded from electrodes The study consisted of three steps: a design process
imbedded in a customized earpiece which is modeled from a developing generic earpieces, a comfort study evaluating the
3D scanned ear impression of the subject’s ear and produced comfort of the earpieces during sleep, and finally an
by 3D printing. This production process is cumbersome due to assessment of signal quality based on ASSR recordings. In
the many steps needed, resulting in a long manufacturing order to evaluate the performance of the earpieces, ASSR was
process with high production costs. To solve these issues, concurrently measured using a dry-contact version of the
several ear canal based generic earpiece designs have been cEEGrid electrode grid.
proposed to achieve a more cost-effective ear-EEG device. II. METHODS
Lee et al. [5] investigated an earphone shaped device with one
carbon nanotube polydimethylsiloxane (CNT/PDMS)-based
electrode in each ear canal. During the evaluation of the A. Design process
device, they recorded various brain phenomena including the To achieve unobtrusive EEG recordings, a generic ear-
auditory steady-state response (ASSR) and found a significant EEG earpiece must be both comfortable to wear and ensure
response for 40 Hz modulated pure tone stimuli. Kidmose et reliable electrical contact between the electrodes and the
al. [6] developed a silicone-based generic ear-plug-like surface of the body. This must be achieved across the large
earpiece with three conductive silicone electrodes in the ear anatomical diversity of normal ears, such that the generic
canal. They succeeded in recording an ASSR with both

1
Department of Engineering, Electrical and Computer Engineering, Katubedda, Sri Lanka. 3 UNEEG medical, Lynge, Denmark. 4 Corresponding
Aarhus University, Finlandsgade 22, 8200 Aarhus N, Denmark. 2 Department author: pki@eng.au.dk.
of Electronic & Telecommunication Engineering, University of Moratuwa,

978-1-5386-1311-5/19/$31.00 ©2019 IEEE 5613

Authorized licensed use limited to: Durban University of Technology. Downloaded on March 07,2022 at 06:31:28 UTC from IEEE Xplore. Restrictions apply.
ERA cR3 cR2
ERB
cR4 cR1
cR4a

ERE
ERK cR4b
cR8
cR5
ERT ERC cR7
cR6
(a) (b) (c) (a) (b) (c)
Fig. 1. (a) Photo of a right ear generic earpiece with electrodes – the black Fig. 2. Photos of the test setup. (a) The generic earpiece and cEEGrid
dots are electrodes. (b) 3D model of generic earpiece for the right ear with mounted on the test subject. (b) A recording performed on the back with
electrode labels and positions. (c) 3D model of the cEEGrid inspired device stimuli in left ear. (c) A recording performed on the left side with stimuli in
for the right ear with electrode labels and positions. the right ear.

earpiece can be used in the majority of subjects. In a design in the earpieces, was used as a reference device to evaluate the
process, managed using The Double Diamond model, several signal quality of the generic earpiece. The dry-contact
generic earpiece designs were developed. The first step of the cEEGrid was composed of a silicone rubber sheet containing
design process was a literature review of the anatomy, ten Ø 3.5 mm IrO2 dry-contact electrodes [9], as shown in Fig.
morphology and average size of human ears. This review 1c. The cEEGrid was attached around the ear with silicone
revealed that even though the human ear has some general adhesive, Adapt Medical Adhesive 7730 (Hollister).
recognizable anatomical features, the ear morphology is as
D. Subjects and Experimental setup
individual as fingerprints [12]. The size range of the average
human ear was used as a starting point in the design process, Ten subjects (four female and six males), with normal
and subsequent design optimization included a hands-on hearing, participated in the study. The recordings were
iterative process assessing the fit of the earpieces inside human performed in a laboratory setting with the subjects lying on a
ears. Due to the large size range, the earpieces were developed bed. To evaluate the electrode-skin contact of the earpieces,
in three sizes – small, medium and large. two ASSR recordings were performed with the subject lying
Initially, three different prototype earpiece designs were on the back; one with sound stimuli being presented to the left
developed. From these, the final generic earpiece design was ear, and one with sound stimuli being presented to the right
selected based on an evaluation of comfort using a semi- ear. Thereafter, two ASSR recordings were made with the
quantitative questionnaire completed by six test subjects after subject lying first on their right side and then left side (Fig. 2),
two nights of sleep with each earpiece. to investigate potential effects of lying on the earpiece. To
comfort subjects lying on their side, only one of the hearing
B. Generic earpiece design cups was used from the headphone. During the recordings, the
The earpiece design which scored highest in the comfort subjects were carefully instructed to close their eyes and relax.
study had a concha part with an outward curve following the Sound stimuli were presented to the subjects by one
cavity of the concha. This part acted like a spring when the hearing cup from a DT 770 Pro 32 OHM headphone
earpiece was inserted into the ear (Fig. 1a-b). This earpiece (Beyerdynamic GmbH & Co. KG) with an ESI U46 XL
design was chosen for the study of the EEG signal quality. The soundcard (ESI Audiotechnik GmbH) at a sampling frequency
generic earpiece was made in biocompatible silicone Detax of 48 kHz. The ASSR recordings were performed using 6
softwear 2.0 (Detax). Based on earlier observations that minutes long broadband CE-chirp stimuli [14] presented at a
electrodes located in the concha are more prone to loose repetition rate of 80 Hz at 45 dB relative to individual
contact compared to electrodes located in the ear canal [9], the sensation levels, which were determined behaviorally prior to
electrode-skin contact was optimized by adding bulges at the the ASSR recordings. A g.tec trigger box (g.TRIGbox, g.tec
electrode positions. Moreover, a recess was added to the edge medical engineering GmbH) was used to trigger the EEG data
of the earpiece at ‘crus of helix’ to ensure good comfort. collection with a trigger frequency of 1 Hz. EEG was recorded
Finally, a small plate was added to the concha part to enhance concurrently from the generic earpiece and cEEGrid using a
the insertion procedure of the earpiece. The ear canal part of 136 channel Refa EEG amplifier (TMSi) sampling at a rate of
the earpiece was connected to the concha part by a flexible 1024 Hz. The amplifier was grounded with a wet conductive
narrow link to accommodate the morphological variation in wristband. In addition to the 32 electrodes inserted in the
the first bend of the ear canal. A cylindrical conduit was added generic earpieces and the cEEGrids, a single Ø 3.5 mm IrO2
to the ear canal part to prevent acoustic occlusion of the ear dry-contact electrode was placed at Fpz, thereby enabling re-
and allow passage of electrode cables. Six Ø 2.6 mm IrO2 dry- referencing to an electrode far away from the ears. A custom
contact electrodes were positioned at ExA, ExB, ExC and ExT routine written in Matlab (Matlab R2018a, Mathworks) was
in concha, and ExE and ExK in the ear canal [2, 9], where x used to present the sound stimuli and control the EEG
denotes the left (L) or the right (R) ear (Fig. 1b). recording.
E. Data analysis
C. Dry-contact cEEGrid
The EEG data were analyzed after performing the recordings.
A C-shaped electrode grid with a geometry similar to the
First, the data were high-pass filtered with a 4th order
cEEGrid [13], using the same type of dry-contact electrodes as

5614

Authorized licensed use limited to: Durban University of Technology. Downloaded on March 07,2022 at 06:31:28 UTC from IEEE Xplore. Restrictions apply.
Table I. Number of ear-Fpz electrode configurations in left (L) and right (R) earpieces with a significant ASSR SNR in recordings made with the subjects
lying on their back along with mean and standard deviation across subjects. The first column for each subject represents the recording, where the subject was
stimulated in the left ear, whereas the second column for represents the recording where the subject was stimulated in the right ear.
Subject 1 2 3 4 5 6 7 8 9 10 Mean ± SD
L 5 4 3 4 3 4 3 3 4 4 1 1 3 3 2 3 4 3 3 2
Ear-Fpz config. 3.4 ± 1.1
R 4 6 4 4 2 5 2 5 4 4 4 4 3 3 2 4 3 3 4 5
Table II Number of within-ear (Left and Right earpiece, respectively) and cross-ear electrode configurations with a significant ASSR SNR from recordings
made with the subjects lying on their back along with mean and standard deviation across subjects. The first column for each subject represents the recording,
where the subject was stimulated in the left ear, whereas the second column for represents the recording where the subject was stimulated in the right ear.

Subject 1 2 3 4 5 6 7 8 9 10 Mean ± SD
L 4 3 2 5 3 7 6 6 5 3 0 2 1 3 1 6 0 9 2 5
Within-ear 3.6 ± 2.3
R 6 8 3 4 1 8 1 3 0 4 5 5 2 2 2 2 3 3 3 4
Cross-ear 20 32 11 16 6 20 7 16 16 16 4 5 5 9 8 16 12 15 12 11 12.9 ± 6.7

Butterworth filter with a cut-off frequency of 1 Hz to reduce good within-ear electrode configurations were found in each
offset drift, secondly 50 Hz and 100 Hz notch filter were subject.
applied to reduce power line interference. After filtering, a Table III summarizes the ASSR SNR found using
data set was constructed containing all possible bipolar electrodes within or around a single ear (within-ear and within-
electrode derivations and split into epochs of four second’s cEEGrid) or from both ears (cross-ear and cross-cEEGrid).
duration. Epochs with a maximum absolute value above 100 The SNR was calculated as the average of the three highest
V were rejected, and datasets with less than 15 remaining SNR values found for each of the devices. One-sided paired t-
epochs were left out of the analysis. The epochs were time tests were used to test for significant differences in average
averaged, and Fourier transformed. Hereafter, the SNR of all ASSR SNR between the generic earpiece and the dry-contact
datasets were calculated as the ratio between the amplitude at cEEGrid.
the repetition rate of 80 Hz and the average noise amplitude  For the earpieces, the average ASSR SNR was found to be
10 Hz around the repetition rate. An F-test was applied, to significantly higher for cross-ear electrode configurations
determine the statistical significance of the ASSR. compared to within-ear electrode configurations (p=0.001). In
contrast, no significant difference was found between the
III. RESULTS within-cEEGrid and cross-cEEGrid configurations.
Table I summarizes the recordings in which the subjects When comparing the two devices, the average SNR of the
were lying on their backs, showing the number of good (i.e. within-cEEGrid electrode configurations was found to be
significant ASSR) ear-Fpz electrode configurations in left and significantly higher than the SNR of within-ear electrode
right earpieces for each subject, along with the mean and configurations (p<0.001). Similarly, the SNR of the cross-
standard deviation (SD). All subjects had at least four ear-Fpz cEEGrid electrode configurations was found to be
electrode configurations with a significant ASSR, one in each significantly higher than the SNR of the cross-ear
ear. On average a significant ASSR was recorded from 3.4 configurations (p<0.001).
electrodes per earpiece. In addition to the assessment of the signal quality from the
The number of good within-ear and cross-ear electrode generic earpieces, it was also tested whether the subject
configurations for each subject is shown in Table II along with position had any effect on the number of good electrode
mean and standard deviation across subjects. Every subject configurations or on the measured ASSR SNR. A one-sided
had at least three good within-ear electrode configurations and paired t-test was used to compare the results obtained when the
four good cross-ear electrode configurations. On average, 3.6 subjects was lying on the back vs when they were lying on

Table III. Average ASSR SNR in dB of the three highest SNR for within-ear, cross-ear, within-cEEGrid and cross-cEEGrid electrode configurations for each
subject. Along with the mean SNR for within-ear, cross-ear, within-cEEGrid and cross-cEEGrid electrode configurations with standard derivation (SD) across
all subjects. The first row for each electrode configuration represents the back recording, in which the subject was stimulated in the left ear, whereas the second
row represents the back recording in which the subject was stimulated in the right ear. Significant difference between within-ear, cross-ear, within-cEEGrid
and cross-cEEGrid are marked with .

Subject 1 2 3 4 5 6 7 8 9 10 Mean ±SD Significant


L 28.6 10.3 19.6 19.0 13.8 21.9 9.7 7.8 12.0 8.8
Within-ear 16.3 ± 6.5
R 13.6 14.3 25.4 28.5 11.1 14.1 18.8 23.8 14.9 10.0
L 28.3 17.3 22.2 17.3 16.8 13.5 14.0 11.0 18.6 16.2
Cross-ear 19.3 ± 5.4
R 14.7 20.3 28.6 28.6 18.1 13.3 23.5 26.3 20.4 16.3
L 29.9 17.8 31.4 34.1 18.6 27.6 19.8 15.8 30.5 14.1
Within-cEEGrid 24.1 ± 7.0
R 17.8 22.0 31.3 29.4 18.8 24.9 32.6 24.0 30.1 10.8
L 29.3 20.3 31.1 33.6 17.4 27.5 21.1 14.4 29.4 19.4
Cross-cEEGrid 24.2 ± 5.9
R 18.1 23.9 26.6 28.5 18.2 23.4 32.2 23.0 29.9 15.9
5615

Authorized licensed use limited to: Durban University of Technology. Downloaded on March 07,2022 at 06:31:28 UTC from IEEE Xplore. Restrictions apply.
their sides. Only the earpieces that were not covered by the REFERENCES
headphone cup was included in the analysis. No significant [1] D. Looney, P. Kidmose, C. Park, M. Ungstrup, M. L. Rank, K.
effect of subject position on either the number of significant Rosenkranz, and D. P. Mandic, “The In-the-Ear Recording
ear-Fpz electrode configurations per earpiece (p=0.5) or the Concept: User-Centered and Wearable Brain Monitoring,” IEEE
Pulse, vol. 3, no. 6, pp. 32-42, 2012.
maximum ASSR SNR (p=0.08) was found. Subject 8 had no [2] P. Kidmose, D. Looney, M. Ungstrup, M. L. Rank, and D. P.
good ear-Fpz electrode configurations in the recording on the Mandic, “A Study of Evoked Potentials From Ear-EEG,” IEEE
right side and was therefore not included in the comparison of Transactions on Biomedical Engineering, vol. 60, no. 10, pp.
the maximum SNR. 2824-2830, 2013.
[3] S. L. Kappel, and P. Kidmose, “Real-Life Dry-Contact Ear-EEG,”
IV. DISCUSSION AND CONCLUSION IEEE Engineering in Medicine and Biology Society. IEEE
Engineering in Medicine and Biology Society, vol. 2018, pp.
The study showed that the developed generic earpiece 5470, 2018.
obtained good skin-electrode contact as evaluated through [4] K. B. Mikkelsen, D. B. Villadsen, M. Otto, and P. Kidmose,
measurement of ASSR relative to a forehead reference. “Automatic sleep staging using ear-EEG,” Biomedical
engineering online, vol. 16, no. 1, pp. 111-15, 2017.
Furthermore, the earpieces enabled recording of statistically [5] J. Hoon Lee, S. Min Lee, H. Jin Byeon, J. Sook Hong, K. Suk
significant ASSRs when both recording- and reference Park, and S.-H. Lee, “CNT/PDMS-based canal-typed ear
electrodes were placed in the same earpiece. However, electrodes for inconspicuous EEG recording,” Journal of neural
compared to previous results for dry-contact ear-EEG, based engineering, vol. 11, no. 4, pp. 046014, 2014.
on customized earpieces [9], we found that the developed [6] P. Kidmose, D. Looney, L. Jochumsen, and D. P. Mandic, “Ear-
EEG from generic earpieces: a feasibility study,” Annual
generic earpieces had a lower number of good electrodes and International Conference of the IEEE Engineering in Medicine
a lower ASSR SNR. and Biology Society. IEEE Engineering in Medicine and Biology
The number and position of electrodes with a good skin Society. Annual Conference, vol. 2013, pp. 543, 2013.
contact were found to vary considerably between subjects. [7] V. Goverdovsky, D. Looney, P. Kidmose, and D. P. Mandic, “In-
These findings are in accordance with the fact that there is Ear EEG From Viscoelastic Generic Earpieces: Robust and
Unobtrusive 24/7 Monitoring,” IEEE Sensors Journal, vol. 16,
large inter-subject variation in ear morphology as well as no. 1, pp. 271, 2016.
earlier observations of inter subject variation in ASSR ear- [8] A. Nguyen, R. Alqurashi, Z. Raghebi, F. Banaei-kashani, A. C.
topography [10, 11]. This inter-subject variation in electrode Halbower, and T. Vu, “A Lightweight and Inexpensive In-ear
performance underlines the importance of using multiple Sensing System For Automatic Whole-night Sleep Stage
electrode positions in generic earpieces, to ensure a sufficient Monitoring,” in Proceedings of the 14th ACM Conference on
number of good electrodes in all subjects. Embedded Network Sensor Systems CD-ROM, Stanford, CA,
USA, 2016, pp. 230-244.
The average ASSR SNR for within-ear electrode [9] S. L. Kappel, M. L. Rank, H. O. Toft, M. Andersen, and P.
configurations was found to be significantly lower than the Kidmose, “Dry-Contact Electrode Ear-EEG,” IEEE Transactions
SNR for cross-ear configurations. This was expected, since the on Biomedical Engineering, pp. 1-1, 2018.
electrode distances of cross-ear configurations were much [10] S. L. Kappel, C. B. Christensen, K. B. Mikkelsen, and P.
larger than the electrode distances of within-ear Kidmose, "Reference configurations for ear-EEG steady-state
responses." pp. 5689-5692.
configurations. Along the same lines, within-cEEGrid also [11] C. B. Christensen, R. K. Hietkamp, J. M. Harte, T. Lunner, and P.
outperformed within-ear, likely for the same reason. Kidmose, “Toward EEG-Assisted Hearing Aids: Objective
Surprisingly, the cross-cEEGrid performance was Threshold Estimation Based on Ear-EEG in Subjects With
significantly higher than the cross-ear performance. This is not Sensorineural Hearing Loss,” Trends in hearing, vol. 22, pp. 1-
caused by increased electrode distance but may be due to the 13, 2018.
considerably larger number of possible cross-cEEGrid [12] B. Moreno, A. Sanchez, and J. F. Velez, "On the use of outer ear
images for personal identification in security applications,"
electrode combinations compared to the generic earpiece (100
Proceedings IEEE 33rd Annual 1999 International Carnahan
vs 36). In future studies, a rigorous train-test paradigm to Conference on Security Technology (Cat. No.99CH36303), IEEE,
alleviate this problem should be included. 1999, pp. 469-476.
In conclusion, it was possible to record cortical responses [13] M. G. Bleichner, B. Mirkovic, and S. Debener, “Identifying
auditory attention with ear-EEG: cEEGrid versus high-density
with the developed generic dry-contact earpiece. While it was
cap-EEG comparison,” Journal of neural engineering, vol. 13, no.
possible to record responses both cross and within ears, we 6, pp. 066004, 2016.
found that the primary driver in signal quality from generic [14] C. Elberling, “A direct approach for the design of chirp stimuli
earpieces is the inter-electrode distance, similar to what has used for the recording of auditory brainstem responses,” The
been seen before [15]. This underlines the importance of Journal of the Acoustical Society of America, vol. 128, no. 5, pp.
2955-2964, 11, 2010.
extending the generic earpiece to also comprise electrodes in [15] K. B. Mikkelsen, S. L. Kappel, D. P. Mandic, and P. Kidmose,
the concha and cymba regions of the ear. This observation “EEG Recorded from the Ear: Characterizing the Ear-EEG
coupled with the fact that deformation of the outer ear (tested Method,” Frontiers in neuroscience, vol. 9, pp. 438, 2015.
by lying on the side) did not lead to a significant degradation
of electrode contact, leads us to believe that further iterations
on the earpiece design can bring generic earpiece performance
on par with customized devices, and that these earpieces would
be usable for sleep recordings. Subjects generally found the
earpieces to be comfortable to wear, with several subjects even
falling asleep during the recordings.

5616

Authorized licensed use limited to: Durban University of Technology. Downloaded on March 07,2022 at 06:31:28 UTC from IEEE Xplore. Restrictions apply.

You might also like