Biomedical Engineering Reference
at the hot-spot again and rotated the coil to different orientations where we perform
the motor threshold estimation. Again, to reduce the effects of stress and varying
vigilance, we use steps of 20 and 10 to not unnecessarily prolong the session,
although we are able to use very small coil rotation steps with the robotized TMS
system. With our setup, each session lasts approximately 1.5 h. For session 1, we
rotate the coil clockwise from 0 to 80 in mixed steps of 20 and 10 , resulting in
orientations of 0 ,20 ,30 ,40 ,50 ,60 and 80 . In this case, 0 denotes the
reference (left-to-right) coil orientation. For session 2, we use orientations of 160 ,
180 , 200 , 210 , 220 , 230 , 240 and 260 , where 180 denotes the right-to-left
coil orientation, used as reference for session 2. The coil orientations are ran-
domized for each stimulation.
For realization of the experiment, an robot operator is responsible for accurate
coil placement with the robotized TMS system and an investigator performs the
actual stimulation with MEP recordings. A double-blind experiment is performed
for MT estimation with only the robot operator knowing the actual coil orientation.
Subject and investigator have no knowledge of the orientation. Therefore, the
investigator sits in reverse to the TMS robot.
220.127.116.11 Further Analysis
Due to biphasic stimulation [ 24 ], we can expect having a sinusoidal relation with two
minima between coil orientation a and motor threshold. This sinusoid should have
period p (as opposed to 2p which would be trivial) with the global minimum roughly
at p (standard orientation) and the second minimum approximately at 0 (left-to-right
orientation). Due to different slopes of the coil current pulse, the MT at p should be
smaller than the MT at 0. Therefore, a second sinusoid with period 2p should be
added to express the change of the amplitude which is orientation dependent.
Therefore, the sinusoidal relation should have the form
MT ð a Þ¼ a þ b cos ð 2 a þ c Þþ d cos ð a þ e Þ ;
ð 3 : 1 Þ
where a ; b ; c ; d ; e are constant factors. We therefore fit the experimental data to this
sinusoidal relation with nonlinear regression and estimate the error of the fit. The
fitting is performed using MATLAB (The MathWorks, Inc., Natick, MA, USA).
As a quantitative measure for the goodness of the sinusoidal fitting, we use the
coefficient of determination R 2 . It is defined as:
R 2 ¼ 1 P y i f i
ð 3 : 2 Þ
P y i y
Þ 2 ;
where y i represents the es ti mated MTs for a given coil rotation i, f i is the value of
the sinusoidal fit at i, and y symbolizes the arithmetic mean of the estimated MTs.
Minimal thresholds and thresholds at standard orientation were compared with
a repeated measures t-test using MATLAB. Similarly, the optimal coil orientation
is compared to the standard coil orientation.