#4 recognizing_and_correcting_meg_artifacts.5

INVITED REVIEW
Recognizing and Correcting MEG Artifacts

Richard C. Burgess
Cleveland Clinic Epilepsy Center, Cleveland, Ohio, U.S.A.
Downloaded from http://journals.lww.com/clinicalneurophys by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4
XMi0hCywCX1AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 11/17/2024
Summary: Noise sources in magnetoencephalography (MEG) fields, real-time active compensation using reference sensors,
include: (1) interference from outside the shielded room, (2) and postprocessing with advanced spatio-temporal filters.
other people and devices inside the shielded room, (3) Many of the artifacts that plague MEG are also seen in EEG, so
physiologic or nonphysiologic sources inside the patient, (4) an experienced electroencephalographer will have the
activity from inside the head that is unrelated to the signal of advantage of being able to transfer his knowledge about
interest, (5) intrinsic sensor and recording electronics noise, artifacts to MEG. However, many of the procedures and
and (6) artifacts from other apparatus used during recording software used during acquisition and analysis may themselves
such as evoked response stimulators. There are other factors contribute artifact or distortion that must be recognized or
which corrupt MEG recording and interpretation and should prevented. In summary, MEG artifacts are not worse than EEG
also be considered “artifacts”: (7) inadequate positioning of the artifacts, but many are different, anddas with EEGdmust be
patient, (8) changes in the head position during the recording, attended to.
(9) incorrect co-registration, (10) spurious signals introduced Key Words: Magnetoencephalography, Artifacts, Inteference,
during postprocessing, and (11) errors in fitting. The major Noise, Active compensation, Temporal signal space separation,
means whereby magnetic interference can be reduced or Magnetic shielding, Gradiometers, Degaussing.
eliminated are by recording inside a magnetically shielded
room, using gradiometers that measure differential magnetic (J Clin Neurophysiol 2020;37: 508–517)
M agnetoencephalography (MEG) has significant advantages

for localization compared with EEG because of its higher
temporal and spatial resolution,1–9 along with its insensitivity
This article, accompanied by pictorial examination, will
focus on both classifying/explaining MEG artifacts and on the
steps that can be taken to eliminate or mitigate their effects.
to tissue inhomogeneities (e.g., skull, scalp).10 Magneto-
encephalography is used primarily for precise localization,
and the accuracy of this source localization task is highly
dependent on the signal to noise ratio. Like almost all ORIGIN OF MEG ARTIFACTS
bioelectric signals, the magnetic signals picked up from the Sources that interfere with the MEG recording can come
brain are quite tiny when compared with other magnetic from the environment, the patient, and the instrumentation itself.
activity in the environment. Interfering signals can be conveniently divided into the following
Many sources of interference familiar to electroence- categories:
phalographers plague MEG as well, although there are some
(such as EMG) that are less disruptive to MEG than EEG. 1. Interference from sources outside the shielded room
Disturbances in MEG like eye blinks, pacemaker pulses, and • Large moving magnetized objects such as elevators, vehicles,
tremors will all be recognizable to electroncephalgraphers, metal doors
and others, like heart-related artifacts, may show an exag- • Frequent transportation of patients on gurneys or equipment on
gerated ballistocardiogram in MEG. Because MEG is carts
a recording of magnetic fields, there are also several artifacts • Motors and power lines
not seen on EEG; magnetoencephalographers need to • Construction activities
recognize and minimize these interfering signals. Further-
more, because MEG source localization is inherently depen- 2. Noise sources inside the shielded room
dent on computerized mathematical calculations, there are • Other people (parent, nurse, neuropsychologist)
errors that can arise from subsequent software processing • Devices (projectors, cameras, stimulators, EEG cables)
steps.
3. Physiologic or nonphysiologic sources inside the patient
• Electyrocardiogram
• Orthodontia, other dental materials
• Implants (vagal nerve stimulators [VNS], artificial joints)
The author has no funding or conflicts of interest to disclose.
Address correspondence and reprint requests to Richard C. Burgess, MD, PhD,
4. Noise from inside the head
Cleveland Clinic Epilepsy Center, S-51, 9500 Euclid Avenue, Cleveland, OH • Shunts
44195, U.S.A.; e-mail: burgesr@ccf.org. • Ferrous particles left in the head after craniotomy
Copyright 2020 by the American Clinical Neurophysiology Society
ISSN: 0736-0258/20/3706-0508 • Background brain noise unrelated to the signal of interest
DOI 10.1097/WNP.0000000000000699 • Activity from brain regions not of interest
508 Journal of Clinical Neurophysiology Volume 37, Number 6, November 2020 clinicalneurophys.com
Copyright © by the American Clinical Neurophysiology Society. Unauthorized reproduction of this article is prohibited.
MEG Artifacts R. C. Burgess
5. Materials on the patient

• Jewelry OTHER SOURCES OF ERROR
• Clothing (such as belts, bras, snaps, zippers) In addition to the classic sorts of interfering signals,
• Contact lenses problems can occur during the source localization and other
• Mascara and other cosmetics or hair products parts of the process that lead to localization errors. These
other factors that corrupt MEG recording and interpretation

should also be considered “artifacts.” The vulnerable steps
6. Intrinsic sensor noise

7. Artifacts from other apparatus include:
• Crosstalk from other recording devices (EEG)
• Evoked response stimulators 1. Inadequate positioning of the patient
• Monitoring instruments 2. Changes in the head position during the recording
These may all be encountered in the process of MEG 3. Incorrect co-registration with anatomic images
recording. Examples of many of them are discussed in this article. 4. Introduction of spurious artifactual signals during postprocess-
ing to remove noise
5. Creation of erroneous signals during postprocessing to correct
for changes in head position
METHODS FOR ELIMINATION OF 6. Errors during management of the source localization
MAGNETIC INTERFERENCE algorithm
To focus on the brain magnetic field and exclude extraneous
magnetic noise, several strategies are used. Of course, prevention of Because the interpretation of a MEG study results in a series
artifact during recording is the first and, usually, the most important of pictures thatdfrom a signal processing standpointdare
step. This can be accomplished by attention to the details of good considerably removed from the raw data, the errors produced
practice and careful instructions to the patient. The following are some by the various components of the evaluation process are often
of the techniques used to mitigate the effects of corrupting signals: hidden within the intermediate results. These “artifacts” demand
vigilance to identify and remove them.
1. Magnetoencephalography recordings are always carried out The sections that follow, by focusing on some specific
within a magnetically shielded room examples, provide some guidance on recognizing artifacts and
2. Gradiometers measure differential magnetic fields (somewhat recommendations on handling their potentially deleterious
analogous to bipolar EEG montages), thereby canceling effects.
magnetic signals that appear on both the sensors, making
them blind to distant sources of interference
3. Real-time active compensation using reference sensors and
external feedback coils11
4. Postacquisition processing to remove noise SENSOR INSTRUMENTATION NOISE
• Signal space projection12 Superconducting quantum interference device sensor ther-
• Signal space separation (SSS)13; spatiotemporal SSS (tSSS)14 mal noise, flux jumps or slow drifts, improperly tuned sensors,
• Averaging or other digital subtraction techniques. and 50 or 60 Hz mains power interference are all
instrumentation-related causes of noise in the signal. Sudden
changes in the quantum state of the superconducting quantum
In most mapping and cognitive-related protocols, averaging interference device sensors result in instrumentation artifacts
is used to improve signal to noise ratio. During spontaneous referred to as “flux jumps.” An example of a flux jump is shown
MEG recordings in patients with epilepsy, however, averaging of in Supplemental Digital Content 1 (see Figure, http://links.
interictal activity may not be desirable, and alternative methods lww.com/JCNP/A120).
of noise reduction are sought. Instrumentation-related noise will generally be detected
The most recent technique, the tSSS method, recognizes and before the patient arrives, during calibration, tuning, or
removes both external interference and artifacts produced by the setupddepending on the procedures for a given manufac-
nearby sources, even those on the scalp. The basic separation into turer’s MEG system.
brain-related and external interferences signals is accomplished with
signal space separation based on sensor geometry and Maxwell
equations only. The artifacts from nearby sources are extracted by
a simple statistical analysis in the time domain and projected out. FERROMAGNETIC SOURCES ON THE PATIENT
The tSSS method generally works well, with relatively little user Occasionally encountered are cosmetics or hair-care
intervention, and the beneficial (and frequently essential) effects of products that contain iron. Figure 2A (see Supplemental
tSSS filtering are shown in many of the examples that follow. Digital Content 2, http://links.lww.com/JCNP/A121) shows
Alternatively, subsets of the tSSS methods provided as part of the large amplitude artifacts produced from iron-containing
MEGIN (Helsinki, Finland) product are also available in open- mascara. The top view of the sensors shown in Fig. 2B
source software packages from FieldTrip,15 BrainStorm,16 (see Supplemental Digital Content 2, http://links.lww.com/
SPM12,17 and MNE-Python.18 JCNP/A121) demonstrates that the artifact is maximal in the
clinicalneurophys.com Journal of Clinical Neurophysiology Volume 37, Number 6, November 2020 509
R. C. Burgess MEG Artifacts
FIG. 1. Artifact from outside the

patient. The raw MEG data in panel
(A) shows a large amplitude discharge in

a nonphysiologic distribution, maximal in
the left temporal sensors. The top view of
all the gradiometers in panel
(B) demonstrates that the activity is
distributed over all the sensors,
suggesting that it is not from the head.
The same data after tSSS filtering is
shown in panel (C). This was an artifact
from a cell phone carried surreptitiously
by the patient into the magnetically
shielded room. MEG,
magnetoencephalography; tSSS,
spatiotemporal signal space separation.
frontal sensors, and that after the patient removed her snaps to avoid the possibility of artifact from belt buckles, jewelry,
mascara, the artifact disappeared. metallic hook and loop clips, brassiere underwiresdand even the
occasional contraband. The artifact in Fig. 1A was produced by the
patient’s cell phone. Despite careful instructions regarding items
proscribed in the magnetically shielded room, this patient managed
ELECTRONIC DEVICES NEAR OR ON THE PATIENT to smuggle the device into the magnetically shielded room inside
Some of the most problematic sources are caused by sources his clothing. Once in the magnetically shielded room where
in or on the patient and thereby close to the sensors. Patients naturally no cell service is available, the phone came alive with
should normally change into hospital gowns with nonmetallic attempts to find a tower, producing sporadic large amplitude
FIG. 2. Artifact from metal surgical

plate in right side of skull. A, The MEG
waveforms before degaussing by hand-
degausser, with the very prominent
artifact in the right temporal sensors
displayed in the top view in (B). In the
waveforms shown in (C) and the
corresponding top view in (D), there is no
artifact after degaussing with the hand-
degausser. MEG,
magnetoencephalography.
FIG. 3. All four panels show the same data segment, during jaw
movement in a patient with dental braces. In the upper panels, the
frontal MEG sensors closest to the teeth are most affected by the
magnetic artifact, but even the posterior sensors also show
substantial artifact, despite degaussing. After tSSS filtering, shown
in the lower panels, the artifact from outside the brain has been
removed, and the data are clearly interpretable. MEG,
magnetoencephalography; tSSS, spatiotemporal signal space
separation.
discharges. Initially mysterious to the technologists, the distribu-

tion of the interference over all of the sensors, as shown in Fig. 1B,
suggested that this activity was not from the head. Confrontation
and a search revealed the device. Figure 1C demonstrates that tSSS
processing was able to eliminate the cell phone artifact, despite its
large amplitude.
MAGNETIC ARTIFACTS PRODUCED BY METAL

INSIDE THE PATIENT
Metallic foreign bodies (bullet fragments, shrapnel, or
orthopedic implants) anywhere in the patient’s body can produce FIG. 4. Although VP shunts containing permanent magnets (such as
interference and should be degaussed before recording. These the “magnetic rotors” found in most programmable shunts) make
magnetized objects and implants produce artifact because MEG recording impossible because of the excessive artifact,
movement generates a dynamic magnetic field, either volitionally nonmagnetic shunts that contain metal but no permanent magnets do
by the patient or as a consequence of automatic internal not overwhelm the SQUIDs. Before postprocessing (panel A), the shunt
movement, such as respiration or cardiac contractions. Objects artifact obliterates the MEG data, especially in the right hemisphere
inside the head, such as neurosurgical clips or plates, are likely to where the shunt was implanted. After tSSS filtering (panel B), the MEG
produce substantial interference. Figure 2 demonstrates the data are clear, and source localization on the right frontal sharp wave
can be carried out. MEG, magnetoencephalography; tSSS,
beneficial effect of degaussing on the artifact from a metal skull
spatiotemporal signal space separation; VP, ventriculo-peritoneal.
plate. Despite the fact that components designed for implantation
are not composed of ferrous materials, alloy purity is not perfect,
and they can easily retain residual magnetism. Figure 3 (see Perhaps, the greatest difficulty is produced by dental fillings
Supplemental Digital Content 3, http://links.lww.com/JCNP/ and orthodontia because of movement of the jaw. Impurities in the
A122) demonstrates that the retained bullet fragment in this amalgam material and accidental magnetization create magnetic
patient’s head produces no artifact on EEG (abnormal cortical sources inside the head that are both common and difficult to
slowing is seen, but no artifact) but substantially affects the MEG remove. In Fig. 3, artifact from the patient’s braces obscure much
recording. Even for sources coming from inside the head, tSSS of the MEG, despite degaussing. Spatiotemporal signal space
filtering is of significant value in reducing the artifact, as shown separation filtering provides artifact reduction sufficient to enable
in the last panel. interpretation (also shown by Hillebrand19).
FIG. 5. Effect of tSSS noise cancellation in a patient with left temporal spikes and a VNS. (All three columns display the same time
segment from the left temporal sensors, with the magnetic field contours of the same 400 ms epoch in the bottom row.) The MEG
artifact seen in patients with a VNS is not because of the operation of the electronics, which can, of course, be turned off during MEG
recording. Movement of the metal case of the VNS in the chest, although it is almost devoid of magnetizable ferrous material,
produces an artifact with each heartbeat, as shown prominently in column (A). Even with SSS filtering (column B), often used in MEG,
the heart pulsation artifact still hides the important MEG features. With the development of adaptive tSSS filtering, extracranial
signals can be eliminated, so that the sharp wave indicated by the arrow becomes clear for both manual visualization and accurate
source localization (column C). MEG, magnetoencephalography; SSS, signal space separation; tSSS, spatiotemporal signal space
separation; VNS, vagal nerve stimulator.
mechanism contains a permanent magnet, and the artifact

MAGNETIC ARTIFACTS FROM IMPLANTED ITEMS produced generally obliterates the MEG signal and cannot be
Implanted electronic devices are encountered frequently in filtered out. Cochlear implants also include permanent magnets
the MEG laboratory. These include pacemakers, automatic used to affix the external part of the stimulator to the head. These
implantable defibrillators, pumps, VNS, etc. Figure 4 (see strong magnets produce artifact that make it impossible to record
Supplemental Digital Content 4, http://links.lww.com/JCNP/ MEG signals. Patients should always be asked about the presence
A123) illustrates the MEG recording on a patient who had both of implanted devices before coming to the MEG laboratory.
an automatic implantable defribillator and a VNS. Note that tSSS
filtering does a good job of eliminating the artifacts from these
two extracranial sources. In Fig. 5 (see Supplemental Digital POSTPROCESSING OF MEG DATA TO
Content 5, http://links.lww.com/JCNP/A124) however, the arti- REMOVE ARTIFACT
facts in a patient with a spinal cord stimulator and an analgesic Many patients with epilepsy are either too young or too
pump could not be overcome by tSSS at the moment of the left impaired to cooperate fully with rigid recording requirements.
temporal spike recorded on the simultaneous EEG. Furthermore, the frequent presence of devices and implants,
Devices that are implanted intracranially can be especially whether for treatment of their epilepsy or for other disorders,
problematic, such as shunts, responsive neural stimulators, and makes it difficult to obtain a noise-free recording. Advances in
cochlear implants. Ventriculo-peritoneal shunts are especially postprocessing methods have been of proven benefit and have
tricky as certain programmable types contain permanent mag- revolutionized MEG.21 As stated by Hillebrand et al.,19 “We
nets.20 Figure 4 shows the artifact from a programmable, non- have further shown that tSSS is a required pre-processing step
magnetic ventriculo-peritoneal shunt, before and after tSSS for data recorded .. these techniques enable the use of MEG
filtering. The shunt contains metallic parts and is therefore for pre-surgical evaluation in a much larger clinical population
somewhat magnetizable, but it does not contain a permanent than previously thought possible.” tSSS, an extension of SSS,
magnet. The second panel illustrates that this small amount of assumes that waveforms that are strongly correlated are artifacts
magnetization, while producing a large artifact in the unfiltered (because brain signals typically are not correlated) and projects
data, can be cleaned up with postprocessing. There are, however, them out from the data in the time domain. As such, tSSS is
ventriculo-peritoneal shunts that allow control of the flow by particularly effective in reducing signals coming from outside
adjusting the tension using a “magnetic rotor.” This valve the head.
FIG. 6. Neuropace artifact in MEG. Even electronic

stimulation devices implanted into the skull, such as
the responsive neural stimulator, do not preclude
satisfactory MEG recording. This patient’s device
was implanted in the right parietal skull region, and
the raw MEG signal in (A) shows an overwhelming

magnetic artifact in that region, whereas the

opposite hemisphere (B) is considerably less
affected. Data after tSSS postprocessing is shown in
the right-hand column. Even with the gain increased
to the normal setting (5· greater than in A), the
artifact is substantially gone (C), affecting only a few
channels in the immediate area (D), allowing SECD
source localization (co-registered to a pre-responsive
neural stimulator-implant MRI in E). MEG,
magnetoencephalography; SECD, single equivalent
current dipole; tSSS, spatiotemporal signal space
separation.
In our laboratory, 15% of the patients have implanted VNS,

which naturally produce considerable artifact. These program- PROBLEMS BECAUSE OF HEAD POSITION IN
mable devices can be turned off for the duration of the MEG THE ARRAY
recording with a simple programming wand, so it is not the active The fall-off in magnetic field strength is proportional to the cube
stimulation of the VNS that produces most of the artifact. Rather of the distance. It is important to make sure that the patient’s head is
it is the metal enclosure of the device that cannot be completely adequately inserted into the sensor helmet. Patients who are unable to
demagnetized. Thus, with each respiration and heartbeat, this cooperate may continuously change their position during recording (an
magnetic source pulsates and produces artifact, as shown in Fig. example of position changes during a single hour of recording is
5. Patients whose epilepsy failed to respond to VNS treatment are shown in Fig. 7 (see Supplemental Digital Content 7, http://links.
an important population to be considered for epilepsy surgery, lww.com/JCNP/A126) resulting in varying attenuation, worsening as
and therefore, it is imperative to achieve a satisfactory signal to the source-to-sensor distance increases. (All the illustrations of head
noise ratio during MEG recording. Filtering with tSSS enables position were generated using the free research software Brainstorm.16)
high fidelity recordings in these patients.22 The placement of Unlike EEG where the sensor electrodes remain affixed to the
vagal nerve stimulators is in the chest, but this postprocessing scalp, the relationship of the MEG sensors is not fixed with regard
technique even allows MEG recording from patients with to the brain. The waveform appearance in sensor space may
responsive neural stimulators, which are implanted in the skull,23 therefore be misleading. In children especially, there can be
discordance of the location in sensor space versus head space
as shown in Fig. 6.
because of substantial head rotation inside the array. Figure 9
Despite the sophistication and power of noise reduction
demonstrates that despite the left hemisphere epileptic discharges
postprocessing methods, they must be used with caution and an
paradoxically appearing in the right-sided sensors, the continuous
appreciation for their potential for distortion. Although very
head position monitoring and movement compensation system
rare, there are artifacts that occur as a direct result of these
insures that the dipole fit localizes properly to the left hemisphere.
methods. Such errors are most often encountered as a result of Because considerable patient head displacement can happen
processing an enormous transient. Either SSS or tSSS process- during MEG recording, mislocalization may occur if not corrected.
ing may “spread” an artifact that originated in only a restricted Fortunately, continuous (100 times per second) head position
number of channels, as illustrated in Fig. 7. These artifacts can monitoring and compensation is available on modern MEG
be identified because changes in the statistical properties of the systems.24,25 This enhancement has been validated24,26 and its use
signal between one epoch and the next may result in greatly simplifies MEG recording and improves the accuracy of
a “boundary artifact” as illustrated in Fig. 6 (see Supplemental source localization. Use of movement compensation should,
Digital Content 6, http://links.lww.com/JCNP/A125). In Fig. however, not be carried to the extreme. When the head position
8, the artifact is not just a transient at the boundary of the must be corrected too far from its starting position, the signal may
processing epochs, rather the statistics of the background become significantly more noisy, as the algorithm attempts to re-
activity are so altered in the succeeding epoch that it appeared create the signal that would have occurred if the head had remained
to represent the onset of seizure activity. As with electroen- in the original position. An example of the noisy data that results is
cephalography, using complex and sophisticated computerized shown in Fig. 8 (see Supplemental Digital Content 8, http://links.
tools to postprocess the data requires an in-depth understand- lww.com/JCNP/A127). The solution to this problem is for the
ing of the underlying physiology, the physics of the method, technologist to carefully watch the patient’s position on the video
and the technical details of the implementation in order not to monitor and to start a new file (with an accompanying new initial
be fooled by artifacts. position) every time that the patient changes position excessively.
FIG. 7. Postprocessing itself can produce its own artifact under

certain circumstances. Not only must the postprocessed MEG be FIG. 8. The tSSS processing artifact. On initial review of the tSSS-
looked at with the simultaneous EEG but also with the raw MEG data filtered data shown in (A), our impression was that there was a sudden
as needed. Panel (A) at the top shows the MEG data as it is ordinarily change in the activity, beginning at the time indicated by the arrow,
viewed to hunt for spikes, i.e., after postprocessing. A first glance at perhaps indicative of a seizure. However, review of the raw data
these data seems to suggest a spike just before the QRS on displayed in (B) shows some change, but not nearly as dramatically as
Electyrocardiogram, which is seen on many of the right frontal MEG in (A), and much less consistent with seizure activity. The position of
channels. Panel (B), however, demonstrates that a high-amplitude the arrow coincides with the borderline between the 5-second analysis
artifact on a single channel is the culprit, but which has been spread in windows of Maxfilter. The statistical properties of the second half of
a “physiologically appearing” pattern over this area by tSSS processing. the page are different enough from the first half to produce a very
This MEG study was entirely normal. MEG, magnetoencephalography; exaggerated change of the entire 5-second epoch, leading to this false
tSSS, spatiotemporal signal space separation. impression. tSSS, spatiotemporal signal space separation.
model correctly, is dependent on the activity being modeled, but is

ERRORS OCCURRING DURING THE still part of the “art” of MEG. The single equivalent current dipole
FITTING PROCESS (SECD) is the most common source model used in clinical MEG,
The primary advantage of MEG over EEG is the precision of but it can be abused. The SECD will always provide an “answer,”
source localization.1,27–31 Because the forward model is so much i.e., a “fit” can be obtained for any time point of any waveform, and
simpler, solution to the inverse problem is considerably more the SECD assumes that the activity source is a single point in space.
accurate.32–34 Selecting the appropriate source model, and using the The more spread out from a theoretical single point that the actual
FIG. 9. The simultaneously recorded EEG

(panel A) clearly shows bursts of slow/sharp
activity arising from the left frontal region.
However, panel (B) shows that in the MEG
sensors the epileptic activity is larger in the
right sensors! Nevertheless, SECD localization
places the source properly, in the left
hemisphere adjacent to the lesion, as shown
in panel (C). The answer to this riddle is
shown in (D); the patient’s head is rotated so
far to the right that the left frontal lobe is
actually closer to the right anterior MEG
sensors. MEG, magnetoencephalography;
SECD, single equivalent current dipole.
considered an “artifact.” Figure 10 shows an example of inappro-

priate application of the SECD model. Although the statistics and
magnetic field pattern can help the magnetoencephalographer
decide when to reject a fit, these results are dependent on the
choice of time point, baseline, and region of interest.
CO-REGISTRATION WITH THE MRI VOLUME AS A

POTENTIAL SOURCE OF ERROR
Magnetoencephalography is capable of highly accurate
source localization within the volume enclosed by the sensor
dewar to within 2 to 3 mm.27229 But this accuracy is wasted if
not precisely co-registered to the patient’s brain volume.
Therefore, careful co-registration of the MEG results with MRI
is of crucial importance for localization, especially if the results
are to be used for surgical decision making. The examples shown
in Fig. 9 (see Supplemental Digital Content 9, http://links.lww.
com/JCNP/A128) are but three ways that the co-registration
process may go awry, thereby producing artifacts in localization.
CONCLUSIONS
Because the signals picked up by recording the magnetic
fields of the brain are so small compared with other sources of
magnetic activity (both physiologic and nonphysiologic), miti-
gating the effects of these interfering fields by prevention or
postprocessing is very important. For those artifacts that do make
their way into the recording, recognition of them and their effects
on the overall recording is an important magnetoencephalogra-
pher skill. Those experienced in electroencephalography will
have the advantage of being able to transfer their EEG
knowledge about artifacts to MEG. However, there are other
MEG artifacts not seen in EEG, and because MEG is more
heavily dependent on computer processing, awareness of and
attention to all potential sources of error is crucial.
REFERENCES
1. Leahy RM, Mosher JC, Spencer ME, Huang MX, Lewine JD. A study
of dipole localization accuracy for MEG and EEG using a human
skull phantom. Electroencephalogr Clin Neurophysiol
1998;107:159–173.
FIG. 10. Using head and source models correctly is the essence of 2. Baillet S, Mosher JC, Leahy RM. Electromagnetic brain mapping. IEEE
Signal Process Mag 2001;18:14–30.
source localization. Application of the SECD model will always result in 3. Oishi M, Otsubo H, Holland-Bouley K, Greiner HM, Arthur T, Mangano
an answer. However, inappropriate use of any model will produce an FT. Epileptic spikes: magnetoencephalography vs simultaneous electro-
artifactual location for the source. Panel (A) displays an example of corticography. Epilepsia 2002;43:1290–1295.
repetitive spikes, which were widespread, maximum in the left frontal 4. Barkley GL, Baumgartner C. MEG and EEG in epilepsy. J Clin
sensors. The SECD analysis of the discharge just after the time bar Neurophysiol 2003;20:163–178.
produced an erroneously deep source, shown in panel (B). The 5. Barkley GL. Controversies in neurophysiology. MEG is superior to EEG
in localization of interictal epileptiform activity: pro. Clin Neurophysiol
impossible location (inside the anterior part of the left lateral ventricle 2004;115:1001–1009.
where no cortex exists) and excessive amplitude (620 nAm) of the 6. Tao JX, Baldwin M, Hawes-Ebersole S, Ebersole JS. Cortical substrates
modeled dipole indicates that the source is too widely distributed to use of scalp EEG epileptiform discharges. J Clin Neurophysiol 2007;24:96–
the SECD model. SECD, single equivalent current dipole. 100.
7. Agirre-Arrizubieta Z, Huiskamp GJ, Ferrier CH, van Huffelen AC,
activity is, the more this assumption is violated, and the less Leijten FS. Interictal magnetoencephalography and the irritative zone in
the electrocorticogram. Brain 2009;132:3060–3071.
appropriate is the SECD model. Thus, the inappropriate use of 8. Ray A, Bowyer SM. Clinical applications of magnetoencephalography in
a source model will lead to localization error, and this should be epilepsy. Ann Indian Acad Neurol 2010;13:14–22.
9. Gavaret M, Badier JM, Bartolomei F, Bénar CG, Chauvel P. MEG and 23. Wang ZI, Alexopoulos AV, Nair D, et al. Feasibility of magneto-
EEG sensitivity in a case of medial occipital epilepsy. Brain Topogr encephalography recording in an epilepsy patient with implanted
2014;27:192–196. responsive cortical stimulation device. Clin Neurophysiol
10. Nakasato N, Levesque MF, Barth DS, Baumgartner C, Rogers RL, 2013;124:1705–1706.
Sutherling WW. Comparisons of MEG, EEG, and ECOG source 24. Nenonen J, Nurminen J, Kicic D, et al. Validation of head movement
localization in neocortical partial epilepsy in humans. Electroencepha- correction and spatiotemporal signal space separation in magneto-
logr Clin Neurophysiol 1994;91:171–178.
encephalography, Clin Neurophysiol 2012;123:2180–2191.

11. Vrba J, Robinson SE. Signal processing in magnetoencephalography. 25. Taulu S, Simola J, Nenonen J, Parkkonen L. Novel noise reduction
Methods 2001;25:249–271. methods. In: Spuek S, Aine C, eds. Magnetoencephalography from

12. Uusitalo MA, Ilmoniemi RJ. Signal-space projection method for
Signals to Dynamic Cortical Networks, Berlin, Heidelberg: Springer-
separating MEG or EEG into components. Med Biol Eng Comput
1997;35:135–140. Verlag, 2014; 35–71.
13. Taulu S, Simola J, Kajola M, Applications of the signal space separation 26. Kakisaka Y; Wang ZI; Mosher JC, et al. Clinical evidence for the utility
method, IEEE Trans Sign Proc 2005;53:3359–3372. of movement compensation algorithm in magnetoencephalography:
14. Taulu S, Simola J. Spatiotemporal signal space separation method for successful localization during focal seizure. Epilepsy Res
rejecting nearby interference in MEG measurements, Phys Med Biol 2012;101:191–196.
2006;51:1759–1768. 27. Barth DS, Sutherling W, Broffman J, Beatty J. Magnetic localization of
15. Oostenveld R, Fries P, Maris E, Schoffelen JM. FieldTrip: open a dipolar current source implanted in a sphere and a human cranium.
source software for advanced analysis of MEG, EEG, and invasive Electroencephalogr Clin Neurophysiol 1986;63:260–273.
electrophysiological data. Comput Intelligence Neurosci 28. Weinberg H, Brickett P, Robertson A, et al. The magnetoencephalo-
2011;2011:156869. graphic localisation of source-systems in the brain: early and late
16. Tadel F, Baillet S, Mosher JC, Pantazis D, Leahy RM. Brainstorm: components of event related potentials. Alcohol 1987;4:339–345.
a user-friendly application for MEG/EEG analysis. Comput Intel 29. Yamamoto T, Williamson SJ, Kaufman L, Nicholson C, Llinás R.
Neurosci 2011;2011:8. Magnetic localization of neuronal activity in the human brain. Proc Natl
17. Friston KJ, Ashburner JT, Kiebel SJ, Nichols TE, Penny WD. Statistical Acad Sci U S A 1988;85:8732–8736.
Parametric Mapping: The Analysis of Functional Brain Images. Cam- 30. Romani GL, Williamson SJ, Kaufman L. Tonotopic organization of the
bridge: Elsevier/Academic Press, 2007. human auditory cortex. Science 1982;216:1339–1340.
18. Gramfort A, Luessi M, Larson E, et al. MNE software for processing 31. Hari R, Hämäläinen M, Ilmoniemi R, et al. Responses of the primary
MEG and EEG data, Neuroimage 2014;86:446–460. auditory cortex to pitch changes in a sequence of tone pips: neuro-
19. Hillebrand A, Fazio P, de Munck JC, van Dijk BW. Feasibility of
magnetic recordings in man. Neurosci Lett 1984;50:127–132.
clinical magnetoencephalography (MEG) functional mapping in the
32. Gharib S, Sutherling WW, Nakasato N, et al. MEG and ECoG
presence of dental artefacts. Clin Neurophysiol 2013;124:107–113.
20. Soler GJ, Bao M, Jaiswal D, et al. A review of cerebral shunts, current localization accuracy test. Electroencephalogr Clin Neurophysiol
technologies, and future endeavors. Yale J Biol Med 2018;91:313–321. 1995;94:109–114.
21. Jin K, Alesopoulos AV, Mosher JC, Burgess RC. Implanted medical 33. Mikuni N, Nagamine T, Ikeda A, et al. Simultaneous recording of
devices or other strong sources of interference are not barriers to epileptiform discharges by MEG and subdural electrodes in temporal
magnetoencephalographic recordings in epilepsy patients. Clin Neuro- lobe epilepsy. Neuroimage 1997;5:298–306.
physiol 2013;124:1283–1289. 34. Scheler G, Fischer MJ, Genow A, et al. Spatial relationship of source
22. Kakisaka Y, Mosher JC, Wang ZI, et al. Utility of temporally-extended localizations in patients with focal epilepsy: comparison of MEG and
signal space separation algorithm for magnetic noise from vagal nerve EEG with a three spherical shells and a boundary element volume
stimulators. Clin Neurophysiol 2013;124:1277–1282. conductor model. Hum Brain Mapp 2007;28:315–322.

#4 recognizing_and_correcting_meg_artifacts.5

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

#4 recognizing_and_correcting_meg_artifacts.5

Uploaded by

Document Information

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

#4 recognizing_and_correcting_meg_artifacts.5

Uploaded by

Copyright:

Available Formats

INVITED REVIEW

Recognizing and Correcting MEG Artifacts

M agnetoencephalography (MEG) has signiﬁcant advantages

5. Materials on the patient

other factors that corrupt MEG recording and interpretation

6. Intrinsic sensor noise

FIG. 1. Artifact from outside the

(A) shows a large amplitude discharge in

FIG. 2. Artifact from metal surgical

discharges. Initially mysterious to the technologists, the distribu-

MAGNETIC ARTIFACTS PRODUCED BY METAL

mechanism contains a permanent magnet, and the artifact

FIG. 6. Neuropace artifact in MEG. Even electronic

the raw MEG signal in (A) shows an overwhelming

magnetic artifact in that region, whereas the

In our laboratory, 15% of the patients have implanted VNS,

FIG. 7. Postprocessing itself can produce its own artifact under

model correctly, is dependent on the activity being modeled, but is

FIG. 9. The simultaneously recorded EEG

considered an “artifact.” Figure 10 shows an example of inappro-

CO-REGISTRATION WITH THE MRI VOLUME AS A

encephalography, Clin Neurophysiol 2012;123:2180–2191.

Methods 2001;25:249–271. methods. In: Spuek S, Aine C, eds. Magnetoencephalography from

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.