Pone 0191111

RESEARCH ARTICLE
Localizing on-scalp MEG sensors using an

array of magnetic dipole coils
Christoph Pfeiffer1*, Lau M. Andersen2, Daniel Lundqvist2, Matti Hämäläinen2,3,4,5, Justin
F. Schneiderman6, Robert Oostenveld2,7
1 Department of Microtechnology and Nanoscience, Chalmers University of Technology, Gothenburg,
Sweden, 2 NatMEG, Department of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden,
3 Department of Neuroscience and Biomedical Engineering, Aalto University, Aalto, Espoo, Finland,
4 Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General
Hospital/Harvard Medical School, Charlestown, MA, United States of America, 5 Harvard-MIT Division of
a1111111111 Health Sciences and Technology, Cambridge, MA, United States of America, 6 Institute of Neuroscience and
a1111111111 Physiology, University of Gothenburg and MedTech West, Gothenburg, Sweden, 7 Donders Institute for
a1111111111 Brain, Cognition and Behaviour, Radboud University, Nijmegen, Netherlands
a1111111111
* [email protected]
a1111111111
Abstract
OPEN ACCESS Accurate estimation of the neural activity underlying magnetoencephalography (MEG) sig-
Citation: Pfeiffer C, Andersen LM, Lundqvist D, nals requires co-registration i.e., determination of the position and orientation of the sensors
Hämäläinen M, Schneiderman JF, Oostenveld R with respect to the head. In modern MEG systems, an array of hundreds of low-Tc SQUID
(2018) Localizing on-scalp MEG sensors using an sensors is used to localize a set of small, magnetic dipole-like (head-position indicator, HPI)
array of magnetic dipole coils. PLoS ONE 13(5):
coils that are attached to the subject’s head. With accurate prior knowledge of the positions
e0191111. https://doi.org/10.1371/journal.
pone.0191111 and orientations of the sensors with respect to one another, the HPI coils can be localized
with high precision, and thereby the positions of the sensors in relation to the head. With
Editor: Blake Johnson, Australian Research Council
Centre of Excellence in Cognition and its Disorders, advances in magnetic field sensing technologies, e.g., high-Tc SQUIDs and optically
AUSTRALIA pumped magnetometers (OPM), that require less extreme operating temperatures than
Received: July 14, 2017 low-Tc SQUID sensors, on-scalp MEG is on the horizon. To utilize the full potential of on-
scalp MEG, flexible sensor arrays are preferable. Conventional co-registration is impractical
Accepted: April 18, 2018
for such systems as the relative positions and orientations of the sensors to each other are
Published: May 10, 2018
subject-specific and hence not known a priori. Herein, we present a method for co-registra-
Copyright: © 2018 Pfeiffer et al. This is an open tion of on-scalp MEG sensors. We propose to invert the conventional co-registration
access article distributed under the terms of the
approach and localize the sensors relative to an array of HPI coils on the subject’s head. We
Creative Commons Attribution License, which
permits unrestricted use, distribution, and show that given accurate prior knowledge of the positions of the HPI coils with respect to
reproduction in any medium, provided the original one another, the sensors can be localized with high precision. We simulated our method
author and source are credited. with realistic parameters and layouts for sensor and coil arrays. Results indicate co-registra-
Data Availability Statement: All data published on tion is possible with sub-millimeter accuracy, but the performance strongly depends upon a
Zenodo: https://doi.org/10.5281/zenodo.1044370. number of factors. Accurate calibration of the coils and precise determination of the posi-
Funding: The NatMEG facility was supported by tions and orientations of the coils with respect to one another are crucial. Finally, we propose
the Knut and Alice Wallenberg foundation and the methods to tackle practical challenges to further improve the method.
Swedish Research Council. This work has further
been supported by the Knut and Alice Wallenberg
foundation (KAW2014.0102), the Swedish
Research Council (621-2012-3673), the Karolinska
Institutet Strategic Neuroscience Program -
StratNeuro, the Swedish Childhood Cancer
PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 1 / 21

Localizing on-scalp MEG sensors using an array of magnetic dipole coils
Foundation (MT2014-0007), the National Institute Introduction

of Biomedical Imaging and Bioengineering
(P41EB015896; 5R01EB009048), and the Magnetoencephalography (MEG), which measures magnetic fields generated by neural cur-
Innovative Training Network, ChildBrain, funded by rents, is a tool for non-invasive studies of human brain function. Via sampling of the magnetic
the Marie Curie Actions of the European fields around the head surface, one can determine neural activity in the brain with high tempo-
Commission (H2020-MSCA-ITN-2014). The ral and spatial accuracy. Detection of these weak fields (on the order of 10−14 to 10−15 T)
funders had no role in study design, data collection
requires very sensitive magnetometers. State-of-the-art MEG systems employ hundreds of low
and analysis, decision to publish, or preparation of
the manuscript. critical temperature (low-Tc) superconducting quantum interference devices (SQUIDs). [1, 2]
Conventional low-Tc SQUIDs have to be cooled with liquid helium (4.2 K), which necessi-
Competing interests: The authors have declared
tates a well-insulated dewar. This results in a distance between sensor (cryogenic temperature)
that no competing interests exist.
and head (room temperature) that is on the order of ca. 2–3 centimeters. As magnetic fields
decay rapidly with distance, significant gains in signal-to-noise ratio (SNR) would be achieved
if the distances between the sensors and the head were reduced.
Fixed-helmet sensor arrays, like the ones used in conventional state-of-the-art MEG sys-
tems, allow a high sensor density but are restricted to a single, fixed sensor arrangement. The
helmet design and sensor arrangement are chosen to fit the majority of head sizes of a target
population, in most cases adults. This excludes some subjects, e.g., with large heads. On the
other hand, for subjects with heads significantly smaller than the helmet, a large fraction of the
sensors is far away from the head. This is especially problematic for research on young chil-
dren–a subject group of high interest in neuroscience (e.g. in studies of brain development).
New technologies, such as optically pumped magnetometers (OPMs) and high-Tc SQUIDs,
allow us to get much closer (within a few millimeters) to the head; we refer to these techniques
jointly as “on-scalp MEG”. Reducing the distance between sensor and source results in stron-
ger signals and improved spatial resolution, allowing for an improved view on brain activity.
Despite their typically higher sensor noise levels, several groups have shown such MEG sys-
tems can outperform state-of-the-art systems in terms of extracted information and localiza-
tion accuracy [3–5] and several measurements have demonstrated their potential [6–11].
On-scalp MEG allows the construction of flexible arrays where the magnetic field sensors
are positioned individually or in small modules on the head of the subject. This allows the
head size variability in the population to be dealt with by adjusting the sensor array to the indi-
vidual head–ensuring high SNR independent of head size.
While offering good coverage and improved sensor-to-head distance, adjustable sensor
arrays have a practical issue that does not apply to fixed-helmet sensor arrays: how to deter-
mine the positions and orientations of all individual magnetometers with respect to the sub-
ject’s head. Accurate reconstruction of neural activity from a set of magnetic field
measurements from outside the head requires a precise assessment of the location and orienta-
tion at which the field is sampled. This is especially important when recording closer to the
head [4].
In a flexible system the location and orientation of each individual sensor has to be deter-
mined—in contrast to current systems, where the sensor array is fixed and only the whole
array needs to be localized relative to the head. We propose to do this using localization coils
similar to the ones used in commercial state-of-the-art MEG systems. In commercial systems,
three to five localization coils are attached to the head of the subject. These are driven with
well-defined, periodic currents at unique frequencies, generating a magnetic dipole field distri-
bution at these frequencies for the different coils. The field strength of each coil detected at
each sensor location is extracted from the sensor signals, e.g. by Fourier transform. When the
sensor locations inside the helmet and the magnetic moments of the coils are known, the mea-
sured field distribution can be used to determine the positions of the coils relative to the sen-
sors and, thereby, the location and orientation of the head with respect to the helmet.

Fig 1. Head localization vs. sensor localization. Left: classical head localization where a large array of sensors (blue;
here: the low-Tc SQUID magnetometer locations of the Elekta Neuromag TRIUX helmet) localizes a set of small,
dipolar coils (green) that are attached to the subject’s head. Right: example of the proposed magnetometer localization
where an array of dipolar coils (green) is used to localize individual (or small arrays of) on-scalp magnetometers (red).
https://doi.org/10.1371/journal.pone.0191111.g001
Instead of localizing the dipolar coils attached to the head with a set of well-known (in
respect to position and orientation) sensors, we propose to invert the procedure. We use a
well-defined array of dipolar coils to localize the positions and orientations of individual sen-
sors (see Fig 1).
The aim of this study was to simulate the proposed sensor localization procedure with dipo-
lar coils for flexible on-scalp MEG systems and–in order to guide future work—investigate the
influence of different parameters on the localization accuracy. The parameters we varied are
(i) accuracy of a priori knowledge of calibration, position, and orientation of the localization
coils, (ii) magnetic moment of the localization coils, (iii) magnetometer noise, (iv) localization
time, and (v) number of localization coils.
The sensor localization procedure proposed here is not needed for commercial state-of-the-
art MEG systems, which contain a large number of magnetometers and/or gradiometers
whose position is fixed and well calibrated. Such systems, however, provide a good reference
and can hence be used to test the performance of the proposed procedure in practical experi-
ments. Besides exploring the parameters of the localization coil configuration and the record-
ing details, we investigated how accurately small magnetometer arrays (1 to 10 channels) with
a known relative position and orientation can be localized using this approach.
Materials and methods

All simulations were done on the basis of geometrical data that was acquired from a MEG
measurement in a previous study [11]. The involvement of the human volunteer was per-
formed in accordance with the technical development prerogative of the Swedish law for ethi-
cal approval of research. The subject was informed about the purpose of the study and
consented to participating in the previous as well as in this study.
We performed a number of localizations of individual magnetometers to assess statistical
variability. The magnetometer positions were based on the layout of the commercial state-of-
the-art low-Tc SQUID-based Elekta Neuromag “TRIUX” system (hereafter called low-Tc
system). A corresponding set of on-scalp sensor positions were obtained by projecting the
TRIUX sensor positions to a distance of 1mm from the subject’s scalp by finding the closest

Fig 2. Low-Tc vs. on-scalp magnetometer positions. Left: Magnetometer positions used for the low-Tc system (blue;
based on the Elekta Neuromag TRIUX). Right: Magnetometer positions used for the on-scalp system (red).
Localization is performed for each magnetometer individually.
point on the scalp surface and moving it 1 mm in direction of the surface normal at that point.
This distance was chosen to account for the minimum separation that can reasonably be
achieved with on-scalp sensors [9]. Our particular choice of on-scalp sensor locations was
motivated by it being evenly distributed over the head and its similarity to the low-Tc channel
distribution, as shown in Fig 2. Each system has 102 possible magnetometer locations.
We approximated the localization coils as magnetic dipoles to simplify the simulations.
This is a valid assumption for coils that are flat and small in size compared to the distance to
the magnetometers, like the ones that are typically used in commercial systems (e.g. CTF and
Elekta Neuromag TRIUX, see Fig 3).
The low-Tc system allows for 12 localization coils to be configured in software and hard-
ware and comes with localization coils in sets of 10 [11]. Consequently, we have been using 10
localization coils in previous measurements. For most simulations we used 10 locations based
Fig 3. A localization coil used in the Elekta Neuromag TRIUX system (ruler scale is in cm).

on the actual locations from a previous measurement. The coil locations on the head were
obtained using a Polhemus Fastrak 3D digitization device (which is part of the standard co-
registration procedure when using the low-Tc system). In optimal conditions, the Fastrak
achieves accuracies of <1 mm. For practical measurements, however, position errors of
around 2 mm are more common [12]. Fig 4 shows the standard localization coil positions on
the subject’s head we used.
Magnetometer localization
The field from a magnetic dipole that is picked up by a magnetometer, i.e., the forward solu-
tion, is given by:
!
m0 3rjk ðmj rjk Þ mj
S rjk ; mj ; n ^k ¼
^ k ¼ B rjk ; mj n 5 3
^k
n ðEq 1Þ
4p jrjk j jrjk j
where rjk is a vector describing the relative position of magnetometer k to the magnetic dipole
j, m is the magnetic dipole moment vector and n ^ is a unit vector describing the sensitive direc-
tion of the magnetometer. For convenience the forward solution is typically defined by the
sensor’s spatial sensitivity to (here: magnetic) dipoles, the so-called lead field Lm.
!
mj
^ k ¼ B rjk ;
S rjk ; mj ; n ^ k ¼ Lm rjk ; m
jmj j n ^ j; n
^ k jmj j ðEq 2Þ
jmj j
Localization of a magnetic dipole coil is achieved by fitting a magnetic dipole to the measured/
simulated data. The error between the data and the forward solution for the magnetic dipole is
Fig 4. Standard localization coil array (green).


the minimized with a non-linear optimization procedure by varying the position, orientation,
and moment of the dipole. The error function e used for the optimization is:
eðmfit Þ ¼ Sdata Lm mfit ðEq 3Þ
with Sdata a vector containing the measured (or simulated) magnetometer signals, Lm a magne-
tometers-by-localization-coils lead field matrix, and mfit a vector containing the dipole
moment strengths of the localization coil(s).
For our approach we want to localize individual magnetometers with a set of localization
coils instead of localizing coils with a sensor array as in current commercial low-Tc systems.
We therefore exchange the roles of magnetometer and localization coils:
m ¼ Lm þ S ðEq 4Þ
From symmetry follows that ½Lm ðr; m; ^ n^ Þþ ¼ Lm ð r; n

^ ; mÞ
^ i.e., the pseudo inverse of the
lead field equals the lead field that would result if the magnetometers were generating the mag-
netic field and the localization coils sensing it.
By swapping in- and outputs the same dipole fitting functions used in classical head locali-
zation can therefore be used to localize a magnetometer with a set of localization coils. A requi-
site hereto is that positions, orientations, and amplitudes of the localization coils are known. In
comparison: classical head localization in commercial full-head systems requires knowledge of
position, orientation, and (input) sensitivity of the magnetometers.
The magnetic dipole fits were performed in MATLAB version R2015a (Natick, MA) using
the FieldTrip toolbox [13].
Data generation
The data was generated by applying the forward solution to the signals driving the individual
localization coils and summing the resulting magnetic fields that a magnetometer picks up.
The localization coils were driven with sine wave signals at different frequencies from 218 Hz
up in steps of 7 Hz, in line with the TRIUX system. The driving signal m of coil j is given by:
mj ðtÞ ¼ mj;0 sinð2pfj tÞ ðEq 5Þ
where f is the frequency and m0 the amplitude of the magnetic moment.

In addition to the signals from the localization coils, Gaussian noise N N ð0; s2 Þ was
added to the magnetometer signals.
P
SðtÞ ¼ j ðLm mj ðtÞÞ þ N ðEq 6Þ
Due to the use of different frequencies, the magnetic moment amplitudes of the individual
localization coils can be extracted from the magnetometer signal through Fourier transform or
phase-locked detection [14].
To reduce computation time, we chose the real positions of the localization coils as starting
points for the optimization. This represents a fitting with an initial guess (i.e., from the digiti-
zation point) similar to the way the digitized locations of the coils are used in conventional
head localization. Alternatively, a grid-search could be performed to find an initial starting
point for the non-linear fit.
We localized the magnetometers individually and defined the localization error as the dif-
ference between the actual magnetometer location and the location obtained from the fit. The
localization errors were then averaged over all magnetometers; this entire process was executed
for each of the two systems.

Signal-to-noise ratio. To investigate the influence of the signal-to-noise ratio (SNR) on

the localization accuracy, we performed simulations with different magnetometer noise levels
and magnetic moment amplitudes. The noise was varied from σ = 0.1 to 500 fT/Hz1/2, the
magnetic moment amplitudes from mj,0 = 0.25 nAm2 to 250 nAm2.
Since the Fourier transform is used for localization, the length of the recorded data also
determines the SNR. Random noise reduces by averaging with the square root of the data
length. If the signal does not change, the SNR therefore increases with increasing data length.
Head localization with commercial systems is limited by practical considerations: averaging
over time acts as a low pass filter, thereby limiting the speed with which movements of the
head relative to the sensor array can be detected. Head localization is therefore typically per-
formed with a sliding window of about one second.
Recently experimental procedures for low-Tc MEG were proposed where the relative posi-
tions of the head of the subject with respect to the sensors are restricted by the use of a 3-D
printed or molded head cast [15, 16]. Since head movements are minimized by such an
approach, the data length used for localization can be increased significantly to achieve higher
SNR. Although on-scalp MEG systems are still at an experimental stage, we anticipate that
these systems will have the sensors fixed relative to the head, as with EEG electrode caps or
with the 3-D printed MEG head cast procedure (as in [16]). Therefore we expect that the head
and sensors will not move relative to each other and that long stretches of data can be acquired
with a constant localization coil signal.
To compare our simulations to present-day commercial MEG systems, we used data seg-
ments of one second with a sampling frequency of 1 kHz. Additionally, we performed a
simulation with 10-minute data length to test how much the localization accuracy could be
improved in systems where head movements are not a concern.
In the following simulations we used typical noise values for low-Tc and on-scalp magne-
tometers. While low-Tc SQUID magnetometers with noise levels around 0.15 fT/Hz1/2 have
been reported [17], these are, to date, not standard in MEG. Magnetometers in commercial
systems like the TRIUX have typical noise levels of around 3 fT/Hz1/2. While there have been
several reports of OPMs and high-Tc SQUIDs with noise levels as low as a few fT/Hz1/2 [18,
19], such high-performance sensors have only been produced in small quantities. We therefore
employed a more conservative noise level of 20 fT/Hz1/2 that may be more typical of commer-
cially-produced on-scalp sensors. We furthermore used 10 nAm2 for the magnetic dipole
moment amplitude, consistent with the TRIUX system.
Calibration, position and orientation. As mentioned above, localizing the magnetome-
ters requires knowledge of the magnetic moment, position and orientation of the localization
coils. To investigate the influence of these parameters on the localization accuracy we simu-
lated different types of inaccuracies in these parameters.
Instead of varying the magnetic moment we applied an error to the calibration values of
the localization coils. In practice, the operator of a MEG system has control over the current
applied to the localization coils, but has to rely on the accuracy of the calibration value for the
conversion from current to magnetic moment. Errors in the calibration value thus translate to
the magnetic moment. We therefore performed simulations with random calibration errors in
a range of ecalib = ±0.1 to ±10%.
We furthermore simulated errors in the position and orientation of the localization coils.
First, we added random position errors of epos = ±0.5 to ±5 mm in three orthogonal directions.
Subsequently, we performed simulations with randomly applied orientation errors by rotating
the localization coils eori = ±1 to ±10 degrees around the X-, Y-, and Z-axes.

Fig 5. Local localization coil arrays. Placement of localization coils (green) in a small (here: 7-coil) array around an
on-scalp magnetometer position (red).
We simulated each of the aforementioned error cases (calibration, position and orientation)
ten times. The errors were then averaged over those repetitions as well as over the magnetome-
ter locations.
Localization coil arrays. We also simulated the effects of using different numbers of local-
ization coils. Here we followed two approaches corresponding to different experimental proce-
dures for MEG recordings.
First, we assumed the case of recording the neuromagnetic field only on a small area of the
head, e.g., to measure the dipolar pattern of a single source. In such a case, the localization
coils can be placed close to that area to ensure good signals from all coils. We realized this by
constructing a ring of localization coils with a 5 cm radius around each of the magnetometer
positions (see Fig 5) and localizing all the magnetometers falling inside that ring. We simulated
such ring-shaped localization arrays at all 94 sensor locations where such a ring fit on the
head. This test was furthermore carried out with 4 to 10 localization coils.
Second, we assumed that MEG signals on the whole head are to be recorded in a single ses-
sion. In such a measurement, a single localization array is required that should provide good
coverage over the whole head. We simulated whole head arrays of 10, 12, 21 and 32 localization
coils (see Fig 6). The layouts are based on the 10–10 EEG system locations and chosen with the
aim of uniform coverage. With such whole-head localization arrays, we localized all magne-
tometer positions and averaged the localization error.

Fig 6. Full-head localization coil arrays. Localization coil placement in 10- (top-left), 12- (top-right), 21- (bottom-
left) and 32- (bottom-right) coil full-head arrays.
Small magnetometer arrays. Finally, we investigated how a small array of magnetometers

would compare in terms of localization accuracy. OPMs can be designed such that a single vapor
cell provides 4 measurement points on the corners [20] and high-Tc systems can be composed of
several small cryostats that each contain multiple, tightly packed sensors. If the relative positions
and orientations of a set of magnetometers are known, they can be fitted as a group, potentially
improving accuracy. We simulated this for arrays of 1 to 10 magnetometers, where we used the
nearest neighbors of each magnetometer to form a mini-array (see Fig 7). We determined the
average localization error as a function of the numbers of magnetometers in the mini-array.
Results
Signal-to-noise ratio
The results of our simulations with varying noise levels and magnetic moment for the low-Tc
and on-scalp systems are presented in Fig 8. Both magnetic moment and magnetometer noise
have a significant impact on the localization error. The accuracy decreases rapidly with higher
noise levels, while the magnetic moment shows the opposite effect. This shows that the SNR is
the defining characteristic for this localization approach. At identical noise levels and magnetic
moments, the on-scalp system shows better accuracy.

Fig 7. Small magnetometer array. Example of a small array of on-scalp magnetometers (red) that are fit as a group
with a 32-coil localization array (green).
Considering the typical sensor noise levels, a typical low-Tc system with 3 fT/Hz1/2 noise
level and 10 nAm2 magnetic moment would achieve a magnetometer localization accuracy of
around 0.01 mm, while a typical on-scalp system with the same magnetic moment and 20 fT/
Hz1/2 noise level would reach approximately 0.1 mm.
The results for longer measurement time are presented in Fig 9. Computing the localization
coil topographies over 10 minutes of data reduces the localization error compared to one sec-
ond of data. As expected, this demonstrates that longer localization measurements are prefera-
ble in terms of accuracy.
Calibration, position and orientation

Fig 10 shows the magnetometer localization errors that result from imprecise assumptions on
the localization coil calibration, position, and orientation.
For typical calibration errors of less than ±1%, the simulations show that both systems local-
ize with an accuracy better than 1 mm. Higher calibration errors can lead to localization errors
of >3 mm.
The magnetometer localization error shows a strong, approximately linear dependence on
the localization coil position inaccuracy. With position errors on the order of 2 mm (i.e., the
typical precision of the digitization device used [12]), mean localization errors of around 4
mm can be expected.

Fig 8. Localization error vs. SNR. Mean localization error with low-Tc (solid blue) and on-scalp (solid red) systems for different magnetometer noise levels
and localization coil magnetic moments (both axes plotted logarithmic). Typical noise levels are marked by vertical blue (low-Tc) and red (on-scalp) dotted
lines. The horizontal black dotted line indicates 1 mm localization error.
The magnetometer localization accuracy as a function of the localization coil orientation

inaccuracies shows, in contrast to calibration and position inaccuracies, a clearly different
behavior in the low-Tc and on-scalp systems. Especially for higher orientation inaccuracies the
low-Tc system shows a stronger increase in the localization error compared to the on-scalp sys-
tem. For small orientation errors the localization accuracies converge. The difference results
from the larger distance between magnetometers and localization coils in the low-Tc system as
compared to the on-scalp system.
Localization coil arrays

Fig 11, panel a) shows how the localization errors change with an increasing number of locali-
zation coils in a small, ring-shaped localization coil array. The on-scalp system shows slightly
higher localization errors and both systems show decreasing localization errors when using
more than five coils. With localization errors of around 10−4 mm (even with four localization
coils) both systems show more than sufficiently high accuracy.

Fig 9. Localization error vs. SNR for different measurement time. Mean localization error with low-Tc (solid) and on-scalp (dashed) systems for different
data lengths. Magnetic moment = 1 nAm2. Both axes are plotted logarithmic.
The localization error as a function of the number of localization coils arranged as full-
head arrays can be seen in Fig 11, panel b). Both systems show small localizations errors
(<0.1 mm) even with just 10 localization coils. This improves further with increasing
number of coils. The on-scalp system benefits especially from increasing the number of
localization coils.
Magnetometer arrays
As expected, fitting multiple magnetometers with known relative positions and orienta-
tions as a group results in higher accuracy (see Fig 12). The reduction in localization error
is strongest for the first few additional magnetometers and slowly flattens out. The on-
scalp system benefits more from the use of small magnetometer arrays than the low-Tc
system. When fitting 10 magnetometers combined they show a reduction of the localiza-
tion error by a factor of approximately 27, and 9 respectively, as compared to localizing
individual sensors.

Fig 10. Localization error vs. a priori errors. Mean localization error with low-Tc (blue) and on-scalp (red) systems
for different a) calibration error ranges, b) position inaccuracies ranges, and c) orientation error ranges. Magnetic
moment = 10 nAm2, on-scalp magnetometer noise = 20 fT/Hz1/2, and low-Tc magnetometer noise = 3 fT/Hz1/2. Error
bars indicate one standard deviation.
Discussion
Our results demonstrate the feasibility of the proposed approach to localize individual magne-
tometers with an array of localization coils. Our simulations show that for on-scalp systems we
expect better accuracy than for low-Tc systems due to the closer proximity between the sensors


Fig 11. Localization error vs. number of coils. Mean localization error with low-Tc (blue) and on-scalp (red) systems for different numbers of
localization coils a) in small, local localization arrays b) in full-head localization arrays. Magnetic moment = 10 nAm2, on-scalp magnetometer
noise = 20 fT/Hz1/2 and low-Tc magnetometer noise = 3 fT/Hz1/2. Error bars indicate one standard deviation.
and the localization coils. At identical noise levels they show smaller magnetometer localiza-
tion errors, less sensitivity to inaccuracies in the assumed localization coil orientations, and a
stronger improvement with increasing number of localization coils in small, local ring-arrays.
Practically though, currently available on-scalp magnetometers have higher noise levels than
low-Tc SQUIDs. With typical noise levels (low-Tc: 3 fT/Hz1/2; on-scalp: 20 fT/Hz1/2), the locali-
zation of magnetometers in low-Tc systems is better than the localization of on-scalp magne-
tometers in most cases.
In the following we discuss the individual results from above and a limitation of the simula-
tions due to the use of dipole approximations.
Fig 12. Localization error vs. number of magnetometers. Mean localization error with low-Tc (blue) and on-scalp (red) systems for different numbers of
magnetometers fitted as a group. Magnetic moment = 10 nAm2, on-scalp magnetometer noise = 20 fT/Hz1/2 and low-Tc magnetometer noise = 3 fT/Hz1/2.
Error bars indicate one standard deviation.

Signal-to-noise ratio
The Signal-to-Noise ratio has shown to be a key factor. Since it is hard to significantly decrease
the sensor noise for on-scalp sensors, a sensible step is to increase the strength of the localiza-
tion signal. However, the localization coil field strength is limited by the dynamic range of the
sensors. To optimize the SNR, the localization signal should therefore be chosen such that the
magnetic field is maximal at the magnetometers without saturating them. Imagining an ideal
system, one could consider implementing active control of the localization signal (with sensor
signal as an input) to maximize the SNR while avoiding sensor saturation.
Another way to improve the SNR is to increase the recording and averaging time. Here it is
important to remember the trade-off between recording time for localization and the ability to
register head movements: longer averaging improves SNR but limit the speed with which
movements between the head and sensors can be detected. This might especially be an issue
when trying to localize epileptic foci in patients that experience strong seizures [21]. However,
there are ways to avoid the head movement problem. With custom-designed head casts, for
example, the head of a subject can be fixed inside a conventional MEG dewar [14]. While this
improves the SNR, the physical constraint of the subject’s head might cause other problems
such as claustrophobia, discomfort and risk of injury (e.g. in case of seizure). An alternative
approach is to fix the system to the head, instead of the head to the system. This requires the
system to be light enough for the head of a subject to hold it, as for example with an EEG cap.
For low-Tc systems that is (with current technology) impossible, but for on-scalp sensors it is
conceivable that this can be implemented. While the sensor locations can be determined rela-
tively simple in such cap systems (e.g. with measuring tape as in EEG systems) a more sophisti-
cated co-registration technique would still be necessary to provide the sensor orientations
which are–in contrast to EEG–crucial for reconstructing neural activity in MEG.
Calibration, position and orientation

The strong dependence of the localization accuracy on the calibration of the localization coils
shows the importance of accurate calibration of the localization coils. As mentioned before
(see Methods), head localization requires exact knowledge of the position, orientation, and cal-
ibration factor of the magnetometers, but not the localization coils. As commercial systems
employ hundreds of magnetometers to localize a few coils, it is not necessary to know the mag-
netic dipole strength of the localization coils as these are fitted along with the position and ori-
entation. For this reason, the localization coils used in commercial systems are typically not
calibrated and can easily be swapped out for new ones in case they break due to daily handling.
For on-scalp magnetometer localization our results suggest that they should be carefully and
consistently calibrated.
Like the calibration, the position of the localization coils has a strong influence on the accu-
racy of the localization. As such, the more accurately the localization coils positions relative to
the head can be determined, the better. The standard technology used for localizing localiza-
tion coils on the head–the Polhemus Fastrak–offers relatively good accuracy. In the last years
optical localization and 3-D scanning technologies have emerged that might provide better
accuracy and faster recording procedures, while avoiding the potential issue of localization coil
movements due to physical contact with a digitizer stylus [22–24].
An issue to be addressed is the accuracy of determining the localization coil orientation.
The small localization coils like the ones used in commercial systems (e.g., Elekta Neuromag
TRIUX, which is approximately the size of an EEG cup electrode) offer no simple way to deter-
mine their orientation. To improve on this problem, the design of the localization coils could
therefore be revised. By fixing a coil to a larger plate, for example, it would be possible to

Fig 13. An orientation-digitizable dipolar coil prototype. Photograph of a localization coil mounted on a plate for
accurate determination of the coil orientation. At the corners are small indentions for the tip of the Polhemus stylus to
assist in the digitization.
determine the coil orientation by digitizing the plate with several points (given that the orien-
tation of the coil relative to the plate is known). A prototype for such a coil setup can be seen
in Fig 13. The setup could be further improved by, e.g., using fabrication techniques such as
multilayer printed circuit boards to develop smaller, more reproducible coils where the plate
for determining the orientation is integrated.
Another way to determine the orientation could be to use multiple coils with fixed relative
orientations and positions. Knuutila et al., for example, mounted three coils together onto a
fiberglass plate [25] and Adjamian et al., used five coils attached to a bite-bar for improved co-
registration [26]. By reducing the number of independent variables a fixed localization coil
arrangement can be used to localize sensors even without accurate knowledge of the coil
orientations.
To ensure accurate and maximally stable coil positions and orientations the coil array could
further be supported by a mechanical support structure. A simple, thin head-mold made out
of, e.g., plaster could be made for a specific subject. Coils could then be glued to it to improve
stability.
Localization coil arrays

Both local and full-head localization coil arrays demonstrated good localization accuracy for
the magnetometers, even with few localization coils. The design of the arrays that we simulated
was based on simple, intuitive ideas; these could be further optimized to achieve better results.
Increasing the number of localization coils clearly improves the magnetometer localization
accuracy. For on-scalp systems a large number of coils could, however, bring new problems,
especially in case of full-head arrays. The localization coils used in commercial systems are typ-
ically around 1–2 mm thick. That means the magnetometers have to be placed either at loca-
tions where there is no coil, or have to be placed further away from the head. Since the
reduced sensor-to-scalp distance is one of the main benefits of on-scalp MEG, as compared to
conventional, MEG, it is preferable to avoid moving the sensors away from the head. More
coils would thus result in fewer possibilities for magnetometer placement. When designing a
full-head localization coil array, this should be kept in mind.

Fig 14. Accuracy of magnetic dipole approximation. Ratio between magnetic dipole approximation and exact solution of the magnetic field for coils with 2
mm (red, dotted) and 4 mm (blue) radii as a function of distance between sensor and coil.
Magnetometer arrays
We observed that fitting small arrays of magnetometers results in improved localization accu-
racy compared to individually localizing each magnetometer. In the extreme case this can be
compared to commercial low-Tc MEG systems, where a single, fixed 100+–channel sensor
array can be fitted using a few localization coils. Chella et al., for example, used a similar
method to calibrate a 153-channel low-Tc MEG system with a 31-coil phantom [27]. For flexi-
ble on-scalp systems this is not possible as the position of each channel will be different in each
measurement/subject. For several small arrays with a few sensors each, on the other hand, this
might be an option for improving the localization of the arrays relative to the head.
Fitting magnetometer arrays in unison should only be done if there is high confidence in
the accurate relative positions and orientations of the sensors to each other. If the geometry of
the array were inaccurate, the errors would severely affect the localization accuracy, similar to
position/orientation errors of the localization coils. This could possibly lead to joint localiza-
tion that is worse than the individual sensor case.

Dipole approximation
We used magnetic dipoles to approximate the localization coils for faster computation. Since
the coils were approximated both in data generation and in localization they match exactly. In
practical measurements, however, this would not hold true. It is therefore necessary to mind
the limitations of approximating coils with magnetic dipoles.
Comparing the magnetic fields of a coil and the magnetic dipole approximation of the same
(Fig 14) shows how the accuracy of the approximation depends on the distance between the
coil and the sensor as well as on the size of the coil. The difference between actual and approxi-
mated magnetic fields is high as sensor and coil are close (up to 30% at 5 mm separation) and
decreases significantly with distance. Placing the coils very close to the sensors could therefore
lead to localization errors and should generally be avoided.
Since co-registration in commercial MEG systems is an over-determined problem, require-
ments on coil calibration and approximations of the fields they generate are not strict. As a
result, the wire wound coils used typically (like the one in Fig 3) have high construction inac-
curacies, making it hard to accurately model them. It would therefore be advisable to develop
new coils that are optimized for this application. With printed circuit boards, for example, it
should be possible to make smaller, more reliable coils. With more accurate coils it would be
possible to use more sophisticated forward modeling of localization coils (e.g. elliptic integrals)
to obtain better approximations of the magnetic fields.
Conclusion
We introduced and evaluated a novel approach for localizing magnetometers for MEG using
multiple localization coils. Our method is potentially beneficial for on-scalp MEG with flexible
sensor arrays. We simulated the localization, showing that the approach works well in theory,
and investigated the influence of different parameters on its accuracy. We found that, in order
to provide high localization accuracy, position, orientation and calibration of the localization
coils need to be determined accurately. SNR also has a strong impact on the localization accu-
racy and can be affected practically by magnetometer noise, the magnetic moment strengths of
the localization coils (through the driving current) and the averaging time. Furthermore, the
number of coils as well as their placement (i.e., their distance to the sensors) plays a role in the
performance of the localization.
Of the parameters investigated the a priori knowledge about the localization coils is the
main determining factor for the overall accuracy. With typical noise levels and even small
localization coil arrays (4 coils in the local case and 10 in the full-head) the localization errors
are well below 1 mm. Position errors on the order of the accuracy of a standard digitization
device like the Polhemus Fastrak (1 mm) would, however, result in localization errors on the
order of 2 mm.
We have proposed ways to tackle some of the challenges that we anticipate with the
approach and improve its accuracy. These ideas should be further investigated and/or tested
experimentally. Using the knowledge gained here we will be able to verify the technique in
practical measurements and can try to optimize the localization of sensors for the accurate
measurement and analysis of neuromagnetic activity in on-scalp MEG.
Acknowledgments
Data for this study was collected at NatMEG (www.natmeg.se), the National infrastructure for
Magnetoencephalography, Karolinska Institutet, Sweden.

Author Contributions
Conceptualization: Christoph Pfeiffer, Lau M. Andersen, Matti Hämäläinen, Justin F. Schnei-
derman, Robert Oostenveld.
Formal analysis: Christoph Pfeiffer.
Funding acquisition: Daniel Lundqvist, Justin F. Schneiderman.
Investigation: Christoph Pfeiffer, Daniel Lundqvist, Matti Hämäläinen, Robert Oostenveld.
Methodology: Christoph Pfeiffer, Daniel Lundqvist, Matti Hämäläinen, Justin F. Schneider-
man, Robert Oostenveld.
Project administration: Daniel Lundqvist.
Resources: Daniel Lundqvist, Justin F. Schneiderman, Robert Oostenveld.
Software: Christoph Pfeiffer, Robert Oostenveld.
Supervision: Daniel Lundqvist, Justin F. Schneiderman, Robert Oostenveld.
Validation: Christoph Pfeiffer.
Visualization: Christoph Pfeiffer.
Writing – original draft: Christoph Pfeiffer.
Writing – review & editing: Christoph Pfeiffer, Lau M. Andersen, Daniel Lundqvist, Matti
Hämäläinen, Justin F. Schneiderman, Robert Oostenveld.
References
1. Hämäläinen MS, Hari R, Ilmoniemi RJ, Knuutila J, Lounasmaa OV. Magnetoencephalography–theory,
instrumentation, and applications to noninvasive studies of the working human brain. Rev Mod Phys.
1993; 65(2):413–497.
2. Supek S, Aine CJ, editors. Magnetoencephalography–From signals to dynamic cortical networks. 1st
ed. Heidelberg: Springer. 2014; ISBN: 978-3-642-33045-2.
3. Schneiderman JF. Information content with low- vs. high-Tc SQUID Arrays in MEG Recordings: The
Case for High-Tc SQUID- based MEG. J Neurosci Methods. 2014; 22:42–46.
4. Boto E, Bowtell R, Krüger P, Fromhold TM, Morris PG, Meyer SS, et al. On the potential of a new gener-
ation of magnetometers for MEG: A beamformer simulation study. PLOS ONE. 2016; https://doi.org/10.
1371/journal.pone.0157655 PMID: 27564416
5. Iivanainen J, Stenroos M, Parkkonen L. Measuring MEG closer to the brain: Performance of on-scalp
sensor arrays. bioRxiv. 2016; https://doi.org/10.1101/073585
6. Öisjöen F, Schneiderman JF, Figueras GA, Chukharkin ML, Kalabukhov A, Hedström A, et al. High-Tc
superconducting quantum interference device recordings of spontaneous brain activity: Towards high-
Tc magnetoencephalography. Appl Phys Lett. 2012; 100:132601.
7. Sander TH, Preusser J, Mhaskar R, Kitching J, Trahms L, and Knappe S. Magnetoencephalography
with a chip-scale atomic magnetometer. Biomed Opt Express 2012; 3(5):981–990. https://doi.org/10.
1364/BOE.3.000981 PMID: 22567591
8. Dammers J, Chocholacs H, Eich E, Boers F, Faley M, Dunin-Borkowski RE, et al. Source localization of
brain activity using helium-free interferometer. Appl Phys Lett. 2014; 104:213705.
9. Xie M, Schneiderman JF, Chukharkin ML, Kalabukhov A, Whitmarsh S, Lundqvist D, et al. High-Tc
SQUID vs. low-Tc SQUID- based recordings on a head phantom: Benchmarking for magnetoencepha-
lography. IEEE Trans Appl Supercond. 2015; 25:1601905.
10. Boto E, Meyer SS, Shah V, Alem O, Knappe S, Kruger P, et al. A new generation of magnetoencepha-
lography: Room temperature measurements using optically-pumped magnetometers. NeuroImage.
2017; 149:404–414. https://doi.org/10.1016/j.neuroimage.2017.01.034 PMID: 28131890
11. Andersen LM, Oostenveld R, Pfeiffer C, Ruffieux S, Jousmäki V, Hämäläinen M, et al. Similarities and
differences between on-scalp and conventional in-helmet magnetoencephalography recordings. PLOS
ONE 2017; https://doi.org/10.1371/journal.pone.1712108

12. Engels L, De Tiege X, Op de Beeck M, Warzée N. Factors influencing the spatial precision of electro-
magnetic tracking systems used for MEG/EEG source imaging. Clin Neurophysiol. 2010; 40(1):19–25.
13. Oostenveld R, Fries P, Maris E, Schoffelen JM. FieldTrip: Open source software for advanced analysis
of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosc. 2011; https://doi.org/10.
1155/2011/156869 PMID: 21253357
14. Uutela K, Taulu S, Hämäläinen M. Detecting and correcting for head movements in neuromagnetic
measurements. NeuroImage 2001; 14:1424–1431. https://doi.org/10.1006/nimg.2001.0915 PMID:
11707098
15. Troebinger L, López JD, Lutti A, Bradbury D, Bestmann S, Barnes G. High precision anatomy for MEG.
NeuroImage. 2014; 86:583–591. https://doi.org/10.1016/j.neuroimage.2013.07.065 PMID: 23911673
16. Meyer SS, Bonaiuto J, Lim M, Rossiter H, Waters S, Bradbury D, et al. Flexible head-casts for high spa-
tial precision MEG. J Neurosci Methods. 2017; 276: 38–45. https://doi.org/10.1016/j.jneumeth.2016.
11.009 PMID: 27887969
17. Strom JH, Hömmen P, Drung D, Körber R. An ultra-sensitive and wideband magnetometer based on a
superconducting quantum interference device. Appl. Phys. Lett. 2017; 110:072603.
18. Faley MI, Poppe U, Urban K, Paulson DN, Fagaly RL. A new generation of the HTS multilayer DC-
SQUID magnetometers and gradiometers. J Phys Conf Ser. 2006; 43:1199–1202.
19. Kominis IK, Kornack TW, Allred JC, Romalis MV. A subfemtotesla multichannel atomic magnetometer.
Nature. 2003; 422:596–599. https://doi.org/10.1038/nature01484 PMID: 12686995
20. Colombo AP, Carter TR, Borna A, Jau YY, Johnson CN, Dagel AL, et al. Four-channel optically pumped
atomic magnetometer for magnetoencephalography. Optics Express. 2016; 24(14):15403–15416.
https://doi.org/10.1364/OE.24.015403 PMID: 27410816
21. Kakisaka Y, Wang ZI, Mosher JC, Dubarry AS, Alexopoulos AV, Enatsu R, et al. Clinical evidence for
the utility of movement compensation algorithm in magnetoencephalography: Successful localization
during focal seizure. Epilepsy Res. 2012; 101(0):191–196. https://doi.org/10.1016/j.eplepsyres.2012.
03.014 PMID: 22503605
22. Urban E, Wakai RT. Optical sensor position indicator for neonatal MEG. IEEE Trans Biomed Eng.
2012; 59(1):255–262. https://doi.org/10.1109/TBME.2011.2171960 PMID: 22010142
23. Bardouille T, Krishnamurthy SV, Ghosh Hajra S, D’Arcy RCN. Improved localization accuracy in mag-
netic source imaging using a 3-D laser scanner. IEEE Trans Biomed Eng. 2012; 59(12):3491–3487.
https://doi.org/10.1109/TBME.2012.2220356 PMID: 23033325
24. Krishna Murthy SV, MacLellan M, Beyea S, Bardouille T. Faster and improved 3-D head digitization in
MEG using Kinect. Front Neurosci. 2014; https://doi.org/10.3389/fnins.2014.00326 PMID: 25389382
25. Knuutila J, Ahlfors S, Ahonen A, Hällström J, Kajola M, Lounasmaa OV, et al. Large-area low-noise
seven-channel dc SQUID magnetometer for brain research. Rev Sci Instrum. 1987; 58(11):2145–
2156.
26. Adjamian P, Barnes GR, Hillebrand A, Holliday IE, Singh KD, Furlong PL, et al. Co-registration of
magnetoencephalography with magnetic resonance imaging using bite-bar-based fiducials and sur-
face-matching. Clin. Neurophysiol. 2004; 115:691–698. https://doi.org/10.1016/j.clinph.2003.10.023
PMID: 15036065
27. Chella F, Zappasodi F, Marzetti L, Della Penna S, Pizzella V. Calibration of a multichannel MEG system
based on the Signal Space Separation method. Phys. Med. Biol. 2012; 57:4855–4870. https://doi.org/
10.1088/0031-9155/57/15/4855 PMID: 22797687

Pone 0191111

Uploaded by

Copyright:

Available Formats

Pone 0191111

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pone 0191111

Uploaded by

Copyright:

Available Formats

RESEARCH ARTICLE

Localizing on-scalp MEG sensors using an

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 1 / 21

Foundation (MT2014-0007), the National Institute Introduction

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 2 / 21

Materials and methods

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 3 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 4 / 21

Fig 4. Standard localization coil array (green).

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 5 / 21

From symmetry follows that ½Lm ðr; m; ^ n^ Þþ ¼ Lm ð r; n

where f is the frequency and m0 the amplitude of the magnetic moment.

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 6 / 21

Signal-to-noise ratio. To investigate the influence of the signal-to-noise ratio (SNR) on

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 7 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 8 / 21

Small magnetometer arrays. Finally, we investigated how a small array of magnetometers

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 9 / 21

Calibration, position and orientation

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 10 / 21

The magnetometer localization accuracy as a function of the localization coil orientation

Localization coil arrays

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 11 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 12 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 13 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 14 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 15 / 21

Calibration, position and orientation

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 16 / 21

Localization coil arrays

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 17 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 18 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 19 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 20 / 21

PLOS ONE | https://doi.org/10.1371/journal.pone.0191111 May 10, 2018 21 / 21

You might also like

From symmetry follows that ½Lm ðr; m; ^ n^ Þþ ¼ Lm ð r; n