PreprintPDF Available

Accuracy of robotic coil positioning during transcranial magnetic stimulation


Abstract and Figures

Objective: Robotic positioning systems for transcranial magnetic stimulation (TMS) promise improved accuracy and stability of coil placement, but there is limited data on their performance. This text investigates the usability, accuracy, and limitations of robotic coil placement with a commercial system, ANT Neuro, in a TMS study. Approach: 21 subjects underwent a total of 79 TMS sessions corresponding to 160 hours under robotic coil control. Coil position and orientation were monitored concurrently through an additional neuronavigation system. Main Results: Robot setup took on average 14.5 min. The robot achieved low position and orientation error with median 1.34 mm and 3.48 deg. The error increased over time at a rate of 0.4%/minute for both position and orientation. Significance: After the elimination of several limitations, robotic TMS systems promise to substantially improve the accuracy and stability of manual coil position and orientation. Lack of pressure feedback and of manual adjustment of all coil degrees of freedom were limitations of this robotic system.
Content may be subject to copyright.
Coil placement has to be maintained during long
stimulation sessions in the majority of research and
clinical applications of transcranial magnetic stimu-
lation (TMS). Conventional holders to stabilize the
heavy stimulation coil are cumbersome, show elas-
ticity of their arms as well as joints, render adjust-
ments with millimeter accuracy difcult, and do
not compensate for subject movement [1, 2]. Al-
ternatively, manually holding the coil throughout
long sessions requires continuous attention, causes
operator fatigue, depends on operator skill, and is
costly [3].
Several navigated robotic positioning systems for
TMS have been introduced to address these short-
comings [2, 4-9]. These systems include a robotic
arm that holds the coil in a specied position and
orientation relative to the subjects head. With the
aid of sensors, such as stereocameras, the coil
placement is maintained automatically even if the
subject moves [7]. Given the growing need for
precision in coil placement to match the develop-
ment of fMRI-guided targeting, robotic systems
show great promise. However, there is a lack of
information about the performance of such sys-
tems in routine TMS studies.
This paper reports on the accuracy and user expe-
rience with an ANT Neuro robotic positioning sys-
tem during real-world conditions of a TMS study.
We used a robotic system with adaptive positioning
(ANT Neuro, Enschede, Netherlands) including a
six-axis industrial robot (Omron Adept Viper s650,
San Ramon, CA) and a tracking camera (NDI Polaris,
Waterloo, Canada). We monitored and recorded
the position and orientation of the coil at the time
of each stimulus with six degrees of freedom (three
spatial coordinates and three angles) using a sec-
Accuracy of robotic coil positioning during transcranial magnetic
Stefan M. Goetz
a,b,c, I. Cassie Kozyrkova, Bruce Lubera,f, Sarah H. Lisanbya,f, David L. K. Murphya, Warren M.
b,c,d,e, and Angel V. Petercheva,b,c,d
Department of Psychiatry & Behavioral Sciences, Duke University, Durham, NC 27710, USA
Department of Neurosurgery, Duke University, Durham, NC 27710, USA
Department of Electrical & Computer Engineering, Duke University, Durham, NC 27708, USA
Department of Biomedical Engineering, Duke University, Durham, NC 27708, USA
Department of Neurobiology, Duke University, Durham, NC 27710, USA
Noninvasive Neuromodulation Unit, Experimental Therapeutics & Pathophysiology Branch,
National Institute of Mental Health, Maryland, USA
Robotic positioning systems for transcranial magnetic stimulation (TMS) promise improved accuracy
and stability of coil placement, but there is limited data on their performance. This text investigates the usability,
accuracy, and limitations of robotic c
oil placement with a commercial system, ANT Neuro, in a TMS study.
21 subjects underwent a total of 79 TMS sessions corresponding to 160 hours under robotic coil co
trol. Coil position and orientation were monitored concurrently through an additi
onal neuronavigation system.
Main Results:
Robot setup took on average 14.5 min. The robot achieved low position and orientation error with
median 1.34
mm and 3.48°. The error increased over time at a rate of 0.4%/minute for both position and orie
After the elimination of several limitations, robotic TMS systems promise to substantially improve
the accuracy and stability of manual coil position and orientation. Lack of pressure feedback and of manual
adjustment of all coil degrees of fr
eedom were limitations of this robotic system.
Transcranial magnetic stimulation (TMS); brain stimulation; robotic coil positioning; motion compensation;
targeting accuracy; positioning variability; real
-world performance evaluation.
Dr. Bruce Luber and Dr. Sarah H. Lisanby are currently with the National Institute of Mental Health, Bethesda, MD 20852, USA. They con-
tributed to this article while at Duke University, prior to joining NIMH. The views expressed are their own (the authors) and
do not neces-
sarily represent
the views of the National Institutes of Health or the United States Government.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
ond stereotactic system (Brainsight, Rogue Research,
Montreal, Canada). The second stereotactic system
received a trigger from the stimulator to synchro-
nize recording of position and orientation with the
pulse (Figure 1).
The coil positioning data were obtained during ex-
periments for a previously published repetitive
TMS study [11]. TMS was performed on 21 subjects
(age 18 48 years, median 21, 14 females, 7 males,
all right handed) in 17 sessions (3.76 sessions per
subject on average). Sessions lasted 48194 min
(mean 124 min), and contained both single-pulse
TMS with jittered inter-stimulus intervals between
8 s and 12 s and 1 Hz repetitive TMS over the left
primary motor cortex. In total, 106,904 coil position
samples synchronized to stimulation pulses were re-
Subjects were seated in a comfortable chair and
told to move their head as little as possible and
only very slowly so that the robot could compen-
sate for movement. The two stereotactic systems
each used its own camera but the same reflective
ducial markers on the subject’s head, also called
optical trackers in the following (Figure 1). For both
stereotactic systems, each subjects head was regis-
tered using characteristic anatomic landmarks (na-
sion, right and left intertragic notches) identied by
the same operator throughout the study. The same
operator also performed the coil and robot setup in
every session. For the orientation control and colli-
sion prevention of the robot, the surface of the
head was sampled with at least 800 points covering
the entire surface formed by the frontal, occipital,
parietal, and temporal bones, including at least 300
points in the target area. For tracking of head posi-
tion, we used glasses with reflective trackers and
double-sided tape on the nasal bridge to stabilize
the position of the glasses throughout the session.
For safety, we avoided any head rest or xation as
the robot could press the subject’s head against it.
Figure 1.
Picture of the robotic TMS setup, showing a stereocamera for the robot control system (A), an indu
trial robot (B), a
stereocamera for monitoring the coil placement (C), a TMS coil with tracker (D), and a ne
ronavigation system for monitoring the coil placement (E). All position recordings are relative to the head (Ta-
lairach coordinates); therefore the x direction refers to the axis pointing from right to left, y from anterior to
posterior, and z from inferior to superior, while
α denotes the angle of rotation around the x axis (pitch),
around the y axis (yaw), and
γ around the z axis (roll).
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
The center of the gure-of-eight coil (Magventure
Cool-B65, Farum, Denmark) was covered with a
layer of compressible gauze, generating light pres-
sure and friction for comfort and stabilization of
the coilhead contact.
Statistical analysis was performed in JMP (SAS Insti-
tute, Cary, NC, USA). Prior to further effect screen-
ing, all data underwent analysis for statistical dis-
tribution; exclusively nonnegative data was also
subjected to Box-Cox distribution analysis based on
Akaike’s Information Criterion (AIC). With appropri-
ate data transformation to incorporate the specic
statistical distribution, we analyzed the absolute
position and orientation and the deviation from
the position and orientation of the location found
to generate largest MEPs (“hotspot”) with mixed-
effects models.
From the time course of the errors of position and
orientation, we extracted the time the robot re-
quired after major deflections to reposition the coil
again and correct 80% of the deviation in position
or orientation. Major deflections were dened as
exceeding the error of the previous sample by at
least 20% and 2 mm or 5° for the position and
orientation, respectively.
The net setup time of the robot for a subject at the
beginning of each session was (14.48 ± 4.22) min
(mean ± standard deviation), dened as the time
from the last neuronavigated handheld hotspotting
pulse to the rst robot-controlled pulse.
The use of the robot led to low targeting errors
throughout the entire study with a median of
1.34 mm and 3.48° for position and orientation,
respectively. The various effects contributing to the
error are summarized in Figure 2 and Ta bl e 1.
The absolute position depended on Sex and Age,
which may have influenced height as well as the
seating position. There was also an interaction be-
tween Session number and Session time. Impor-
tantly, the effect of Session time was insignicant,
indicating that the coil placement did not drift sys-
tematically in any direction.
The error of the coil position with respect to the
target identied at the beginning of each session,
was on average (3.9 ± 28.0) mm (mean ± standard
deviation) (see Figure 2). The orientation error was
on average (4.04 ± 3.13) ° (mean ± standard devia-
tion). The correlation coefcient between position
and orientation error was linearly 0.594 (p < 0.001)
and logarithmically 0.179 (p < 0.001).
The deviation of the position from the target was
dependent on Sex, Age, Session time, Session num-
ber, and the interactions of Sex and Session time,
of Age and Session time, and of Session number
and Session time. The overall position error de-
creased with age, while older subjects showed a
slower increase of the error.
Likewise, the orientation error depended on Sex,
Age, Session time, Session number, and the inter-
actions of Age and Session time and of Session
number and Session time (see Table 1). The error
grew over time with 0.387% per minute. Further,
Figure 2.
Dependence of the position error (top) and the orientation error (bottom) on key effects. Red lines
indicate regression trends on the logarithmic scale.
Position error (mm)
Orientation error
Session time (min)
040 80 120 160 200 Female Male
Sex Session numberAge (years)
1 2 3 4 5 6 720 30 4025 35 45
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
both the overall orientation error as well as the
speed of increase of the orientation error de-
creased with age.
Large brief targeting errors up to 860 mm occurred
after subjects rapidly moved their head away from
the coil. After such major deflections, the time for
the robot to return to baseline was (10.7 ± 11.4) s
(mean ± standard deviation). The duration for ori-
entation errors to return to baseline was (8.14 ±
7.05) s (mean ± standard deviation). The durations
for both the return of the position and orientation
exhibited approximately lognormal distributions.
For both position and orientation errors, age and
male sex tended to increase the duration for cor-
rection, and the effect of age was signicant for
orientation errors.
In this robotic rTMS study, we detected coil posi-
tion errors consistent with the maximum values of
2 mm reported in previous benchtop measure-
ments of similar systems [5, 8]. The identied repo-
sitioning error of 3.64 mm between sessions of the
same subject is likely less an indicator of robot
error than of different head registration errors
between the sessions. The robot is comparable or
may outperform the accuracy of manual placement
with stereotactic support, which was reported to
range from 2 mm to more than 5 mm, although our
study did not intend to compare robotic and man-
ual coil placement [12, 13]. The practicality of ro-
botic placement appears superior for long sessions,
in which the additional setup time of about 15 min
is acceptable. A conclusive quantitative comparison
with manual placement concerning positioning
accuracy, on the other hand, will require a one-to-
one comparison with a repeated measures setup,
which appears justied as soon as the revealed
technical limitations are solved.
The overall error and to a lesser extent the rate of
increase of the position and angular errors were
smaller for older individuals. These trends may be
related to the anecdotally observed increased com-
pliance of older subjects with the guideline to keep
the head as still as possible. Frequent head move-
ments can lead to constant errors at the time of
pulses as the robot repositioning lags behind and
to growing errors because of movement-mediated
shift of the head trackers.
While the absolute position was not dependent on
the session time, the deviation from the target
was. Accordingly, the error grew over time, but the
robot did not show a preferential absolute direc-
tion. Thus, all three directions contributed approx-
imately similarly to deviations.
The positioning errors can be attributed to several
factors. The repeatability of positioning of industri-
al robots is typically excellent, even though coil
position and orientation are estimated exclusively
from the states of the six joints along the kinematic
chain [14]. Therefore, the largest portion of the
error may result from control. The controller esti-
mates the local head curvature at the target from
an idealized head scaled to the sampled outline of
the subject. This procedure typically does not con-
sider individual differences of the head shape and
may misrepresent the surface curvature and the
normal direction [15].
With a misestimated local curvature, the coil does
not contact the head with its focus point, which is
approximately in the center of the bottom face of a
gure-of-eight coil. Thus, the coil is not perfectly
tangential to the head at the focus point. This issue
is often underestimated in manual coil positioning
as well, where it requires experienced operators to
appreciate and manage the impact of imperfect
tangential orientation on the head surface. For
example, on a sphere with a radius of 85 mm,
which represents a typical head curvature at the
motor cortex, rolling the coil by only 5° shifts the
point on the scalp where the coil touches the head
by 7.4 mm and lifts the focal point of the coil by
0.65 mm, further increasing the distance to the
target [16-18].
The ANT robot control software allowed correc-
tions of the three translational degrees of freedom
as well as the orientation of the coil relative to the
central gyrus. However, it did not allow manual
adjustment of the two angular degrees of freedom
that control the position of the coil surface relative
to the assumed local head surface normal. Where-
as one of the two degrees of freedom can be and
was corrected by rolling the coil handle in the
clamp, the other one did not allow a simple me-
chanical adjustment. It is expected that corrections
of less than 5° could control a large portion of the
constant part of the observed positioning err or.
The current study also revealed that the coil posi-
tion constantly drifts relative to the target without
any preferential direction so that targeting accura-
cy decreases over time. On average, however, the
rate of drift appears moderate on the order of 0.4%
per minute (i.e., a doubling time of about 175 min)
for both position and orientation so that an error
of 1 mm and 5° increases to 1.27 mm and 6.3° over
an hour. This growing offset may arise from a num-
ber of contributions. Dominant may be minor move-
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
ments and position drifts of the optical trackers on
the coil and the subject, which are a key limitation
of all frameless stereotaxy systems in TMS [13, 19].
These drifts accumulate over time and are am-
plied by movements of the subject. Tracker move-
ments can deteriorate the tight coilhead contact,
without which subjects tend to constantly reposi-
tion their head and the robot lags behind causing a
deviation. In contrast to widely-used expandable
headbands, which can easily move and allow the
lever that accommodates the trackers to swing, we
used trackers on glasses worn by the subject. The
position of the glasses is comparably well dened
as they rest on the rigid nasal bone and their orien-
tation is stabilized by the temples which provide a
long lever to the ears. However, despite our use of
double-sided tape to x the eyeglass bridge to the
skin, movement of the skin underneath the bridge
can still contribute to the error.
Whereas this study used a commercial system based
on an industrial robot with a relatively moderate
cost, other purpose-designed robots have been
presented or are commercially available for TMS [5,
7, 9, 20]. The alternative robot mechanics are un-
likely to improve positioning accuracy, as the accu-
racy is limited by other factors, and indeed bench
measurements showed deviations that are compa-
rable with industrial robots [9]. Nevertheless, kin-
ematics that are adjusted to the typical movement
range of the human head can improve reposition-
ing speed and range, which is limited with small
industrial robots. During larger rotation or move-
ment of the head that would require the robot to
leave its range, the robot risks losing the target
entirely, which we observed several times during
training sessions.
A major issue was identied prior to the study. The
coil surface abutting the head is smooth and rigid.
Without some pressure against the coil, subjects
tended to perform constant minor position ad-
justments of the head, which the robot compen-
sated with time-lagged millimeter-scale oscillatory
movements. Accordingly, the robotic system would
require a highly accurate adjustment of the coil
head contact in 0.1 mm steps to nd an equilibri-
um between the following two cases as observed
in our lab: In one scenario, the robotically actuated
coil slowly pushed the subject’s head into uncom-
fortable positions and potentially out of range
throughout the session, which can happen with
speeds as slow as a few millimeters per minute. In
the opposite scenario, there was a loose (no pres-
sure) mechanical contact between the coil and the
head. In this case the subject may lean into the
coil, to which the robot responds by moving the
coil away, again resulting in a slow drift.
This contact issue is a direct consequence of the
robot control approach, which is exclusively based
on position. For this study, we introduced passive
pressure control as soon as the focal point of the
coil touched the head by locally adding compressi-
ble gauze between the coil and the head. For small
coilscalp distances, the gauze was compressed to
a thin layer and the gauze elasticity converted the
coilscalp distance into contact pressure. Thus, if
the target location is set for a position where the
coil compresses the gauze tightly but not entirely,
an increase of the pressure by the subject moves
the coil away, whereas a release attracts the coils,
keeping the relative position to the target and the
pressure practically constant. In addition, the in-
creased friction between the scalp and the coil
reduces the small continuous oscillatory move-
ments. The pressure control and the additional
friction turn out to be better controllable for the
application of TMS coil positioning.
Robot systems that provide force control based on
pressure sensors promise better control over the
coilhead interface [9]. Multi-cell pressure sensors
can further ensure that the coil touches the target
with its focus point and is tangential to the head
surface at the target [10].
Research reported in this publication was support-
ed by the Duke-Coulter Translational Partnership,
Brain & Behavior Research Foundation under NAR-
SAD Young Investigator Award 22796, and National
Institutes of Health under award numbers
RF1MH114268 and U01AG050618, and in part by
the Intramural Research Program of the National
Institute of Mental Health (ZIAMH00295). The con-
tent is solely the responsibility of the authors and
does not necessarily represent the ofcial views of
the National Institutes of Health. We thank Dr. L.
Gregory Appelbaum and Dr. Lysianne Beynel for their
valuable comments on the manuscript.
S. M. Goetz, S. H. Lisanby, D. L. K. Murphy, and A. V.
Peterchev are inventors on patents and patent ap-
plications on TMS technology. S. M. Goetz has re-
ceived research funding from Magstim Inc. A. V.
Peterchev has received research and travel support
as well as patent royalties from Rogue Research;
research and travel support, consulting fees, as
well as equipment donation from Tal Medical/Neu-
rex; patent application and research support from
Magstim; as well as equipment loans from Mag-
Venture, all related to TMS technology. I. C. Kozyr-
kov, B. Luber, and W. M. Grill report no relevant dis-
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
1. Traad, M., A Quantitative Positioning Device For Trans-
cranial Magnetic Stimulation. Proc. IEEE Eng. Med. Biol.
Soc. (EMBC), 1990. 12: p. 2246-2246.
2. Richter, L., et al., Stimulus Intensity for Hand Held and
Robotic Transcranial Magnetic Stimulation. Brain Stimula-
tion, 2013. 6(3): p. 315-321.
3. Sollmann, N., et al., Clinical Factors Underlying the Inter-
individual Variability of the Resting Motor Threshold in
Navigated Transcranial Magnetic Stimulation Motor Map-
ping. Brain Topography, 2017. 30(1): p. 98-121.
4. Richter, L., et al., Optimal Coil Orientation for Transcranial
Magnetic Stimulation. PLOS ONE, 2013. 8(4): p. e60358.
5. Lancaster, J.L., et al., Evaluation of an image-guided, robot-
ically positioned transcranial magnetic stimulation system.
Human Brain Mapping, 2004. 22(4): p. 329-340.
6. Kantelhardt, S.R., et al., Robot-assisted image-guided trans-
cranial magnetic stimulation for somatotopic mapping of
the motor cortex: a clinical pilot study. Acta Neuro-
chirurgica, 2010. 152(2): p. 333-343.
7. Lebosse, C., et al. A robotic system for automated image-
guided transcranial magnetic stimulation. in 2007 IEEE
/NIH Life Science Systems and Applications Workshop.
8. Richter, L., et al. Towards direct head navigation for robot-
guided Transcranial Magnetic Stimulation using 3D laser-
scans: Idea, setup and feasibility. in 2010 Annual Inter-
national Conference of the IEEE Engineering in Medicine
and Biology. 2010.
9. Zorn, L., et al., Design and Evaluation of a Robotic System
for Transcranial Magnetic Stimulation. IEEE Transactions
on Biomedical Engineering, 2012. 59(3): p. 805-815.
10. Lebosse, C., et al., Modeling and Evaluation of Low-Cost
Force Sensors. IEEE Transactions on Robotics, 2011. 27(4):
p. 815-822.
11. Goetz, S.M., et al., Enhancement of Neuromodulation with
Novel Pulse Shapes Generated by Controllable Pulse Pa-
rameter Transcranial Magnetic Stimulation. Brain Stimula-
tion, 2016. 9(1): p. 39-47.
12. Schönfeldt-Lecuona, C., et al., Accuracy of Stereotaxic
Positioning of Transcranial Magnetic Stimulation. Brain
Topography, 2005. 17(4): p. 253-259.
13. Ruohonen, J. and J. Karhu, Navigated transcranial magnet-
ic stimulation. Neurophysiologie Clinique/Clinical Neuro-
physiology, 2010. 40(1): p. 7-17.
14. Shiakolas, P.S., K.L. Conrad, and T.C. Yih, On the Accuracy,
Repeatability, and Degree of Influence of Kinematics
Parameters for Industrial Robots. International Journal of
Modelling and Simulation, 2002. 22(4): p. 245-254.
15. Konigsberg, L.W., B.F.B. Algee-Hewitt, and D.W. Steadman,
Estimation and evidence in forensic anthropology: Sex and
race. American Journal of Physical Anthropology, 2009.
139(1): p. 77-90.
16. Örmeci, A.R., et al., Adult head circumferences and cen-
tiles. Journal of Turgut Ozal Medical Center, 1997. 4(3): p.
17. Deng, Z.D., S.H. Lisanby, and A.V. Peterchev, Effect of
Anatomical Variability on Electric Field Characteristics of
Electroconvulsive Therapy and Magnetic Seizure Therapy:
A Parametric Modeling Study. IEEE Transactions on Neural
Systems and Rehabilitation Engineering, 2015. 23(1): p. 22-
18. Kristeva, R., D. Cheyne, and L. Deecke, Neuromagnetic
elds accompanying unilateral and bilateral voluntary
movements: topography and analysis of cortical sources.
Electroencephalography and Clinical Neurophysiology
/Evoked Potentials Section, 1991. 81(4): p. 284-298.
19. Comeau, R., Neuronavigation for Transcranial Magnetic
Stimulation, in Transcranial Magnetic Stimulation, A.
Rotenberg, C.J. Horvath, and A. Pascual-Leone, Editors.
2014, Springer New York: New York, NY. p. 31-56.
20. Bohning, D.E., et al., A TMS coil positioning/holding system
for MR image-guided TMS interleaved with fMRI. Clinical
Neurophysiology, 2003. 114(11): p. 2210-2219.
Tab le 1. Results of mixed effects model of the robot’s absolute position and deviation from the target.
Signicant effects (p < 0.05) are highlighted in bold.
Val ue
Absolute position & orien-
F(1,106904) = 528, p <
for females position
shifted by
x: 2.17 mm
y: 5.15 mm
z: 2.83 mm
F(1,106904) = 3290, p <
shift per year of age
x: +0.149 mm
y: 0.382 mm
z: 0.387 mm
Session time
F(1,106904) = 0.0031, p
= 0.956
27.1 µm/min in x,
141 µm/min in y,
+6.88 µm/min in z,
0.0233°/min in α,
0.0400°/min in β,
+0.0331°/min in γ
Session number
F(6, 106904) = 1.91, p =
Sex × Session time
F(1,106904) = 0.159, p =
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
Age × Session time
F(1,106904) = 2.96, p =
Session time × Session
F(6,106904) = 5.32, p <
Position error
F(1,106904) = 96.8, p <
6.82% higher for
F(1,106904) = 2170, p <
1.76% per year of
Session time
F(1,106904) = 577, p <
0.404% per minute
Session number
F(6,106904) = 4811, p <
Sex × Session time
F(1,106904) = 467, p <
0.172% per minute
for females,
0.643% per minute
for males
Age × Session time
F(1,106904) = 102, p <
0.119% rate de-
crease per year of age
Session number × Session
F(6,106904) = 258, p <
Orientation error††
F(1,106904) = 5634, p <
65.2% larger for
F(1,106904) = 8814, p <
3.52% per year
Session time
F(1,106904) = 526, p <
0.387% per minute
Session number
F(6,106904) = 2523, p <
Age × Session time
F(1,106904) = 66.4, p <
0.0964% rate de-
crease per year of age
Session number × Session
F(6,106904) = 769, p <
Sex × Session time
F(1,106904) = 1.32, p =
0.376% per minute
for males,
0.401% per minute
for females
Duration of position cor-
rection after major deflec-
F(1,79) = 1.09, p = 0.301
+17.8% for males
F(1,79) = 1.75, p = 0.190
+1.2% per year of age
Session number
F(6,79) = 1.53, p = 0.192
Duration of orientation
correction after major
F(1,79) = 1.39, p = 0.238
+18.9% for males
F(1,79) = 5.10, p = 0.027
1.9% per year of age
Session number
F(6,79) = 1.81, p = 0.102
* normal distribution of position (AIC(normal) = 9.21·105, AIC(Weibull) = 9.60·105, AIC(lognormal) > 1.5·106) and orientation (AIC(normal) =
3.26·105, AIC(Weibull) = 4.80·105, AIC(lognormal) > 9.0·105)
lognormal distribution (AIC(normal) = 1.02·106, AIC(Weibull) = 4.23·105, AIC(lognormal) = 3.39·105, Box-Cox Coefcient = 0.1)
†† lognormal distribution (AIC(normal) = 5.44·105, AIC(Weibull) = 4.79·105, AIC(lognormal) = 4.70·105, Box-Cox Coefcient = 0.2)
lognormal distribution (AIC(normal) = 612, AIC(Weibull) = 528, AIC(lognormal) = 494, Box-Cox Coefcient = 0.25)
‡‡ lognormal distribution (AIC(normal) = 487, AIC(Weibull) = 436, AIC(lognormal) = 417, Box-Cox Coefcient = 0.051)
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted May 19, 2019. ; bioRxiv preprint
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Correctly determining individual's resting motor threshold (rMT) is crucial for accurate and reliable mapping by navigated transcranial magnetic stimulation (nTMS), which is especially true for preoperative motor mapping in brain tumor patients. However, systematic data analysis on clinical factors underlying inter-individual rMT variability in neurosurgical motor mapping is sparse. The present study examined 14 preselected clinical factors that may underlie inter-individual rMT variability by performing multiple regression analysis (backward, followed by forward model comparisons) on the nTMS motor mapping data of 100 brain tumor patients. Data were collected from preoperative motor mapping of abductor pollicis brevis (APB), abductor digiti minimi (ADM), and flexor carpi radialis (FCR) muscle representations among these patients. While edema and age at exam in the ADM model only jointly reduced the unexplained variance significantly, the other factors kept in the ADM model (gender, antiepileptic drug intake, and motor deficit) and each of the factors kept in the APB and FCR models independently significantly reduced the unexplained variance. Hence, several clinical parameters contribute to inter-individual rMT variability and should be taken into account during initial and follow-up motor mappings. Thus, the present study adds basic evidence on inter-individual rMT variability, whereby some of the parameters are specific to brain tumor patients.
Full-text available
In this study the head circumference (HC), height and weight of 408 adults were measured. An adult’s HC was been shown to be affected by his/her height and weight. Our data suggest that absolute measurements of HC without regard to stature and weight are inadequate for demonstrating clinically significant abnormalities of head size. On average a male’s head is 1.33 cm larger than that of a female. We produced centile charts for adult HCs to accurately detect relative abnormalities. The pediatric charts are also appropriate for use in adults of either sex. [Journal of Turgut Özal Medical Center 1997;4(3):261-264]
Full-text available
We study the impact of coil orientation on the motor threshold (MT) and present an optimal coil orientation for stimulation of the foot. The result can be compared to results of models that predict this orientation from electrodynamic properties of the media in the skull and from orientations of cells, respectively. We used a robotized TMS system for precise coil placement and recorded motor-evoked potentials with surface electrodes on the abductor hallucis muscle of the right foot in 8 healthy control subjects. First, we performed a hot-spot search in standard (lateral) orientation and then rotated the coil in steps of 10° or 20°. At each step we estimated the MT. For navigated stimulation and for correlation with the underlying anatomy a structural MRI scan was obtained. Optimal coil orientation was 33.1±18.3° anteriorly in relation to the standard lateral orientation. In this orientation the threshold was 54±18% in units of maximum stimulator output. There was a significant difference of 8.0±5.9% between the MTs at optimal and at standard orientation. The optimal coil orientations were significantly correlated with the direction perpendicular to the postcentral gyrus ([Formula: see text]). Robotized TMS facilitates sufficiently precise coil positioning and orientation to study even small variations of the MT with coil orientation. The deviations from standard orientation are more closely matched by models based on field propagation in media than by models based on orientations of pyramidal cells.
Full-text available
Low-cost piezoresistive sensors can be of great interest in robotic applications due not only to their advantageous cost but to their dimension as well, which enables an advanced mechanical integration. In this paper, a comparison of two commercial piezoresistive sensors based on different technologies is performed in the case of a medical robotics application. The existence of significant nonlinearities in their dynamic behavior is demonstrated, and a nonlinear modeling is proposed. A compensation scheme is developed for the sensor with the largest nonlinearities before discussing the selection of a sensor for dynamic applications. It is shown that force control is achievable with these kinds of sensors, in spite of their drawbacks. Experiments with both types of sensors are presented, including force control with a medical robot.
Full-text available
Transcranial magnetic stimulation is a noninvasive brain stimulation technique. It is based on current induction in the brain with a stimulation coil emitting a strong varying magnetic field. Its development is currently limited by the lack of accuracy and repeatability of manual coil positioning. A dedicated robotic system is proposed in this paper. Contrary to previous approaches in the field, a custom design is introduced to maximize the safety of the subject. Furthermore, the control of the force applied by the coil on the subject's head is implemented. The architecture is original and its experimental evaluation demonstrates its interest: the compensation of the head motion is combined with the force control to ensure accuracy and safety during the stimulation.
In order to ensure successful application of transcranial magnetic stimulation (TMS), practitioners must be certain that they are targeting the correct cortical location. To aid in this, a number of clinicians and practitioners have begun utilizing various neuronavigation systems to track coil and participant-head position in space for the duration of a stimulation session. In this chapter, I explore the history of neuronavigation and the developments that made combining this technology with TMS possible. Following this, I discuss the practical aspects of properly utilizing a neuronavigation system: including MRI acquisition, 3D-reconstruction, head and coil co-registration, cortical targeting, motion tracking, and electric field modeling. I conclude with a brief discussion of incorporating robotic assistance in coil positioning and tracking.
In this article, we discuss industrial robot characteristics of accuracy and repeatability. The factors that affect these characteristics are identified, and an error tree is developed. Subsequently, the accuracy and repeatability are investigated utilizing the Denavit-Hartenberg kinematics parameters, the homogeneous transformation matrix, and the differential transformation matrix theory, and corresponding measures are developed. The formulation indicates that the influence matrices associated with joint variables are constant. A new measure called degree of influence is established that qualitatively assesses the relative contribution of each kinematic parameter variation to the accuracy and repeatability of rigid manipulators. The developed formulation provides for easy evaluation of the degree of influence measures for rigid manipulators in either numerical or symbolic form. A numerical example is included in which the degree of influence of the kinematics parameters for an articulated manipulator, PUMA 560, are evaluated and analysed.
Background: Transcranial Magnetic Stimulation (TMS) is based on a changing magnetic field inducing an electric field in the brain. Conventionally, the TMS coil is mounted to a static holder and the subject is asked to avoid head motion. Additionally, head resting frames have been used. In contrast, our robotized TMS system employs active motion compensation (MC) to maintain the correct coil position. Objective/hypothesis: We study the effect of patient motion on TMS. In particular, we compare different coil positioning techniques with respect to the induced electric field. Methods: We recorded head motion for six subjects in three scenarios: (a) avoiding head motion, (b) using a head rest, and (c) moving the head freely. Subsequently, the motion traces were replayed using a second robot to move a sensor to measure the electric field in the target region. These head movements were combined with 2 types of coil positioning: (1) using a coil holder and (2) using robotized TMS with MC. Results: After 30 min the induced electric field was reduced by 32.0% and 19.7% for scenarios (1a) and (1b), respectively. For scenarios (2a)-(2c) it was reduced by only 4.9%, 1.4% and 2.0%, respectively, which is a significant improvement (P < 0.05). Furthermore, the orientation of the induced field changed by 5.5°, 7.6°, 0.4°, 0.2°, 0.2° for scenarios (1a)-(2c). Conclusion: While none of the scenarios required rigid head fixation, using a simple holder to position a coil during TMS can lead to substantial deviations in the induced electric field. In contrast, robotic motion compensation results in clinically acceptable positioning throughout treatment.
Movement-related magnetic fields (MRMFs) accompanying left and right unilateral and bilateral finger flexions were studied in 6 right-handed subjects. Six different MRMF components occurring prior to, and during both unilateral and bilateral movements are described: a slow pre-movement readiness field (RF, 1–0.5 sec prior to movement onset); a motor field (MF) starting shortly before EMG onset; 3 separate “movement-evoked” fields following EMG onset (MEFI at 100 msec; MEFII at 225 msec; and MEFIII at 320 msec); and a “post-movement” field (PMF) following the movement itself. The bilateral topography of the RF and MF for both unilateral and bilateral movements suggested bilateral generators for both conditions. Least-squares fitting of equivalent current dipole sources also indicated bilateral sources for MF prior to both unilateral and bilateral movements with significantly greater strength of contralateral sources in the case of unilateral movements. Differences in pre-movement field patterns for left versus right unilateral movements indicated possible cerebral dominance effects as well. A single current dipole in the contralateral sensorimotor cortex could account for the MEFI for unilateral movements and bilateral sensorimotor sources for bilateral movements. Other MRMF components following EMG onset indicated similar sources in sensorimotor cortex related to sensory feedback or internal monitoring of the movement. The results are discussed with respect to the possible generators active in sensorimotor cortex during unilateral and bilateral movement preparation and execution and their significance for the study of cortical organization of voluntary movement.