Biomedical Engineering Reference
In-Depth Information
was applied to remove the power line interference. Continuous MEG recordings were
converted into multiple trials. A single trial had a time period between
−
+
6s
allowing an overlap of about 2 s to its adjacent trials. Here, the time origin was defined
as the time of starting the initial rise in the electromyogram waveform. We excluded
trials with the magnetic field strength exceeding 2.5 pT in any sensor channels.
6 and
9.5.2.2 Source Estimation
The watershed algorithm in FreeSurfer [
18
] was applied to the subject's MRI to gen-
erate triangulations on the inner skull surface. The dense triangulation on the folded
cortical surface was divided into a grid of 15,000 voxels. The source space includes
both hemispheres. The average distance between voxels is 2 mm. The sensor lead
field was computed using the boundary element model provided by OpenMEEG [
19
].
The source estimation was performed using the depth-weighted minimum L2 norm
method implemented in Brainstorm [
20
]. The anatomical labels were identified auto-
matically by using the FreeSurfer software with the Desikan-Killiany atlas [
21
].
Since in our study, we focused only on the contralateral sensorimotor area, we defined
the region of interest as one containing subareas such as precentral, postcentral, para-
central, caudal media frontal areas, and a part of the superior frontal area. The total
number of voxels in the ROI resulted in 1,258. We computed a single time course at
source orientation has a 180
◦
ambiguity, we use either the estimated orientation or
its opposite orientation that is closer to the direction of the cortical surface provided
by Freesurfer.
9.5.2.3 Modulation Index Analysis
The LF and HF signals are extracted from voxel time courses by applying a finite
impulse response band-pass filter. The frequency band of the LF signal ranges from
8.5 to 14.5 Hz, which was divided into 11 sub-bands with a bandwidth of 1 Hz; each
sub-band half-overlapped with adjacent sub-bands. The frequency band of the HF
signal ranges from 50 to 200 Hz; the band is divided into five half-overlapped sub-
bands with the bandwidth of 50 Hz. We computed the modulation index described
in Sect.
9.4.3
using all the combinations of the LF and HF sub-bands. The number
of the phase bins,
q
, was set to 12.
When computing the time course of the modulation index, we divided a 12-second-
long interval of trial data into 21 time windows in which the window width is 2 s
with a 1.5 s overlap. The modulation index
M
I
was computed using all 21 sets of
the time-windowed data to obtain the time course,
M
I
(
t
1
),...,M
I
(
t
21
)
, where
t
1
,...,
t
21
indicates the midpoints of the 21 time windows. We also computed the
maximum of the modulation index, such that
M
P
=
(
M
I
(
t
k
)
)
max
(9.6)
