1
A compact
and cost
-
effective
laser
-
powered
speckle visibility
spectroscopy
(SVS)
device for measuring cerebral blood flow
Yu Xi Huang,
a
,†
Simon Mahler,
a
,†,*
Maya Dickson,
a
Aidin Abedi,
b
Julian M. Tyszka,
c
Yu
Tung Lo
,
b
Jonathan Russin,
b
Charles Liu,
b
,*
Changhuei Yang
a
,**
a
Department of Electrical Engineering, California Institute of Technology, Pasadena, California 91125, USA
b
USC Neurorestoration
Center and the Departments of Neurosurgery and Neurology, University of Southern
California; Los Angeles, CA 90033, USA
c
Division of Humanities and Social Sciences, California Institute of Technology, Pasadena, California 91125, USA
d
Rancho Los Amigos National Rehabilitation Center, Downey CA, 90242, USA
†
These authors contributed equally to this work.
*
Email:
cliu@usc.edu
**
Email:
chyang@caltech.edu
Abstract
In the realm of cerebrovascular monitoring, primary metrics typically include blood pressure, which
influences cerebral blood flow (CBF) and is contingent upon vessel radius.
Measuring CBF non
-
invasively
poses a persistent challenge, primarily attributed to the difficulty of accessing and obtaining signal from the
brain.
This study aims to
introduce a compact speckle visibility spectroscopy (SVS) device designed for
non
-
invasive CBF measurements
, offering cost
-
effectiveness and scalability while tracking CBF with
remarkable sensitivity
and temporal resolution.
The
wearable hardware
has a modular design
approach
consisting
solely
of
a laser diode as the source and a
meticulously selected
board camera as the detector.
They both
can be easily placed on a subject’s head to measure CBF with no
additional
optical elements
.
The SVS device
can
achie
ve a sampling rate of
80
Hz with minimal susceptibility to external disturbances.
The device also achieves better SNR compared with traditional fiber
-
based SVS devices, capturing about
70 times more signal
and
showing
superior stability
and reproducibility
. It is designed to be paired and
distributed in multiple configurations around the head, and measure signals that exceed the quality of prior
optical CBF measurement techniques.
Given its cost
-
effectiveness, scalability, and simplicity, this
laser
-
centric tool offers significant potential in advancing non
-
invasive cerebral monitoring technologies.
1
Introduction
The brain stands as the most complicated and indispensable organ within the
human body.
M
onitoring the cerebral blood flow
(CBF) non
-
invasively bears significance
in both clinical settings and cognitive neuroscience research
1
. Measuring CBF non
-
invasively poses a persistent challenge, primarily attributed to the difficulty of accessing
and obtaining signal from the brain, especially in biomedical context where the exposure
levels are restricted for the safety of the subjects
2
. As a result, efforts have been devoted
in diverse methods for measuring CBF. Some notable techniques include transcranial
Doppler ultrasound
3,4
, magnetic resonance imaging (MRI)
5,6
, near
-
infrared
spectroscopy
7,8
, electroencephalography
9
, or cerebral oximetry. Optical monitoring of
CBF stands to be expected more sensitive than other techniques
10
, as
it
can better
penetrate through skulls and tissues while providing
high
er
temporal resolution.
Diffusing wave spectroscopy utilizing laser light transmitted through a scattering
medium to extract the
dynamic information has recently garnered attention as a promising
tool for CBF monitoring
11
–
17
. One advantage of
diffusing wave spectroscopy
is the
2
ca
pability
to collect a substantial number of photons that
have
interacted with the brain.
It also presents numerous operational benefits, including its non
-
ionizing, safe radiation,
straightforward methodology, use of relatively lightweight and cost
-
effective equipment,
and compatibility with advanced commercial optic
al systems that can be readily adapted.
In
diffusing wave spectroscopy
scheme, laser light is injected into the head using a laser
source, and the emerging light is collected by a detector positioned at a source
-
to
-
detector (S
-
D) separation distance from the injection spot. The movements of blood cells
within the travelling l
ight’s path will scatter and change the effective optical path lengths,
resulting in a fluctuating laser speckle field.
There exist two types of sampling techniques to infer the blood flow: temporal and
spatial. The temporal sampling technique, called time
-
domain diffuse correlation
spectroscopy, is based on the use of the temporal ensemble of the speckle field and uses
a
photodetector
working at a high frame rate (typically above 100 kHz) on a single (or on
a small group of) speckle
11,12
. The spatial sampling technique is an off
-
shoot of laser
speckle contrast imaging (LSCI) and is based on the use of spatial ensemble of the
speckle field, usually referred as speckle visibility spectroscopy (SVS)
15,16,18
–
20
or as
speckle contrast optical spectroscopy (SCOS)
17,21,22
.
In SVS, instead of a high frame rate detecting device, a camera with
a larger
detecting area and
a large number of
pixels
is
used to collect more photons within the
same frame
15,16,18
. The camera is typically working at an exposure time longer than the
decorrelation time of the speckle field. This results in multiple different speckle patterns
summing up onto a single camera frame. As the speckle field fluctuates, the speckle
pattern r
ecorded by the camera is smeared and washed out within the exposure time.
Because the
smearing or
the
washing out
effect
is due to the dynamics of the blood cells,
the decorrelation time can be calculated from the degree of blurring of the captured frame,
typically by calculating the speckle contrast
.
SVS was applied on the human head to
monitor cerebral blood flow non
-
invasively, allowing for the detection of a larger number
of speckles and an increased proportion of detected light from the brain
15,16
.
This paper
reports
a compact SVS device designed for monitoring CBF
.
This
wearable hardware consists solely of a continuous
-
wave laser diode and a
high
-
resolution
CMOS
-
based board camera that can be easily placed on a subject’s head to measure
CBF with no external optical elements. It offers real
-
time CBF monitoring at
80
Hz
sampling rate
while maintaining a lightweight and budget
-
friendly design.
While similar
wearable optical system
was recently used to
measure
the
change
s
of
CBF
during breath
hold maneuver
23
, this paper
presents
the design
and processing of compact SVS and
compares its
gain in stability and SNR
over fiber
-
based SVS
.
We
expect
t
he
compact SVS
device
to have
certain advantages over
fiber
-
based
SVS devices
.
First
,
we show that compact SVS
achieve
s
better SNR
compared with fiber
-
based SVS devices
by collecting
a
large
r
amount of photons
due to a significant increase
in
the detecting
area and
numerical aperture
.
Specifically, we measured the compact
SVS
version to capture about 70 times more signal relative to a comparable fiber
-
based SVS
device, improving detectability of CBF at extended S
-
D distances.
Second, it
eliminates
the motional artifacts associated with the light guide running from the head to the camera
,
showing
a superior stability
and reproducibility
. Typically,
when
a large
-
diameter
multimode fibers is used to collect the photons from the head to the camer
a, s
light
movements of the fiber can cause significant speckle changes, disrupting the results
17,19
.
3
This issue is currently mitigated by minimizing fiber perturbations with extraordinary
measures, which is not ideal.
The paper is
organized
as follows.
First, we detail
the design
and experimental
arrangement
of the
proposed
compact SVS system
and describe the data
processing
for
calculating blood flow from
the
recorded camera images
.
Second
, w
e
compare
the
performance of
compact SVS and fiber SVS
by
using static
and moving
phantom
s
.
Finally, we experimentally
compare the
CBF
measured
from
compact SVS
and fiber SVS
device
s
at different S
-
D distances
from a cohort of five subjects
. Our results
show
significant improvements
in
CBF measurement
with the
compact SVS
over
the
fiber SVS
device.
2
Methods
The arrangement of our
compact
SVS
device is shown in Fig. 1.
The
system
design
is shown in Fig. 1(a) with the schematics shown on the left and a photograph of the 3D
printed device shown on the right. The system includes a laser source for illumination and
a board camera for detection. A dime is included in the photograph for size co
mparison.
For this study, we used a single
-
mode continuous wave 785 nm laser diode
as a
source
[Thorlabs L785H1] which can deliver up to 200 mW. To ensure control over the
illumination spot size and prevent undesirable laser light reflections
or stray light
, we
housed the laser diode within a 3D
-
printed mount. The mounts were printed using black
resin which absorbs light, minimizing back reflection and stray light. The laser diode was
set several millimeters away from the skin of participants such that the i
llumination spot
diameter was 5 mm
16
. The total illumination power was limited to 45 mW to ensure that
the laser light intensity level of the area of illumination is well within the American National
Standards Institute (ANSI) laser safety standards for maximum permissible exposure
(2.95 mW/
mm
2
) for skin exposure to a
785nm
laser beam
2
.
At a
specific
S
-
D distance from the illumination spot, the detector was positioned
on the head of the subject to
collect
the emerging light away from the laser illumination
spot, Figs. 1(b) and 1(c). The collected laser light was directed onto a carefully selected
camera equipped with a large sensor area and small pixel size, maximizing the number
of speckles captured.
We used a USB
-
board camera [Basler daA1920
-
160um (Sony
IMX392
sensor
)]
as
the
detector
. For optimal performance and stability, we typically
operated the camera at a framerate of
80
frame
-
per
-
second (fps).
The compact SVS
system has the potential to achieve a sampling rate of up to 160 fps. However, it is capped
at
80
fps to provide a balance between storage space
and temporal resolution.
To ensure
time
-
synchronization among all pixels, the camera was configured with a global shutter
setting. This camera features a pixel pitch of 3.4 μm, which offers a balance between the
average intensity per pixel and the number of speckles per pixel
which was measured to
be
about 10
speckles per pixels
.
The depth to which the photons have travelled deep into the head is related to the
S
-
D distance
13,16
. By tuning the S
-
D
distance, one can tune the depth of penetration into
the head, where a banana
-
shaped spatial sensitivity of the light path is usually observed
as shown in Fig.
1(b)
13,16
. As the S
-
D distance increases, the banana
-
shape extends
deeper into the brain, with deeper brain regions being more challenging to access.
The
4
spatial distribution of the exiting photons collected by a camera exhibit a granular pattern
with areas of high and low intensity called speckles.
The motions within the light paths,
primarily due to the movement of red blood cells, will scatter and change the effective
optical path lengths resulting in a fluctuating speckle field that varies in time
.
Speckles arise
from
interference between the numerous random scatterings with
the coherent light field
and constitute a vast
area
of research
24
. We image these speckles
onto a camera with a finite
exposure time
, Fig.
2(a)
.
T
he camera must operate at a high
enough frame rate to temporally resolve the dynamics,
typically
above 20 fps for blood
flow measurements.
Speckles
undergo
dynamic changes with a specific temporal
evolution
25
–
27
, characterized by the decorrelation time
휏
!
of the speckle field
28,29
.
Typically, the camera is configured with an exposure time
푇
that is
significantly larger than
the decorrelation time
휏
!
.
As the speckle field fluctuates, the
recorded speckled image
would be smeared and washed out
:
the shorter the speckle decorrelation time, the more
washed out the image.
T
he
dynamics
of
the
speckle
s
can be
quantif
ied
by calculating the
speckle contrast of the recorded image.
Fig. 1
Compact speckle visibility spectroscopy (SVS) setup. (a) Design of the SVS device, consisting of a laser diode
(source) and a CMOS
-
based board camera (detector) both housed in a 3D
-
printed mount. (left panel) 3D schematics
breakdown. (right panel) Photogr
aph of the actual SVS device. (b) Top
-
view and (c) front
-
view schematics illustrating
5
the SVS device in use on a subject’s forehead. When set at a specific S
-
D distance, the SVS device can effectively
measure
cerebral blood flow
.
The experimental SVS processing analysis flowchart for deriving the CBF from
recorded camera images is shown in Fig. 2.
T
he squared speckle contrast
퐾
"#$
%
(
퐼
)
of
a
recorded camera image
퐼
, Fig. 2(a)
is calculated
as:
퐾
"#$
%
(
퐼
)
=
휎
%
(
퐼
)
)
휇
%
(
퐼
)
)
,
(
1
)
where in Eq. (1),
퐼
)
=
퐼
−
퐼
&''()*
, w
ith
퐼
the recorded camera image
and
퐼
&''()*
the camera
offset
.
To experimentally measure the camera offset,
we
capture a series of images
without any source illumination, and then calculate the mean offset image
퐼
&''()*
.
The
variance of
퐼
)
is
휎
%
(
퐼
)
)
and
the mean is
휇
(
퐼
)
)
.
This calculation does not account for noises
that contribute to the variance of the images. To account for these noises,
we use
an
adjusted squared speckle contrast
퐾
#+,-(*)+
%
,
which is commonly calculated
as
17,30
–
32
:
퐾
#+,-(*)+
%
=
퐾
"#$
%
−
퐾
(
.
&*
%
−
퐾
/-#0*
%
−
퐾
!#1
%
−
퐾
(2
%
,
(
2
)
with
퐾
(
.
&*
%
accounting for variance contributions from the shot noise,
퐾
/-#0*
%
for the
variance inherited from quantization,
퐾
!#1
%
for the variance contributions of the camera’s
readout noise and dark noise, and
퐾
(2
%
for the spatial inhomogeneities.
See Fig. 2(b) for
examples of raw and noise speckle cont
r
ast
measurement
s
.
For
each of the image
퐼
)
, t
hey
can be calculated as the following
17,30
–
32
:
퐾
(
.
&*
%
(
퐼
)
=
0
훾
휇
(
퐼
)
)
2
,
(
3
푎
)
퐾
/-#0*
%
(
퐼
)
=
0
1
12
휇
(
퐼
)
)
%
2
,
(
3
푏
)
퐾
!#1
%
(
퐼
)
=
0
휎
!#1
%
휇
(
퐼
)
)
%
2
,
(
3
푐
)
퐾
(2
%
(
퐼
)
=
0
휎
(2
%
휇
(
퐼
)
)
%
2
.
(
3
푑
)
In Eq. (3a),
훾
is the analog to digital conversion ratio associated to the camera,
which depends on the gain setting and the conversion factor
퐶퐹
of the camera, as
훾
=
3#40
56
. In our investigations, the gain was set within a range of 1 to 72, corresponding to a
0 to 37 dB setting. The gain was tuned depending on the signal intensity.
In
8
-
bit mode,
the Basler camera we used
had
a conversion factor of
퐶퐹
=
40
.
7
. To reduce quantization
noise, the gain of the camera was adjusted such that the camera readout grayscale
values fell within the range of 40 to 255 at 8
-
bit recording
unless the signal is too low
. The
camera noise
휎
!#1
%
was estimated by calculating the variance of a series of 500 camera
images recorded in the absence of any illumination sources. The measured camera noise
may introduce an offset bias, which is rectified by subtracting a bias term. The spatial
6
variations in photon flux across the sensor area is accounted by
휎
(2
%
.
With
these
calibration
s
, we can acquire the
adjusted speckle contrast
퐾
#+,-(*)+
%
for each of the
recorded image.
After obtaining
퐾
#+,-(*)+
%
, one can calculate the decorrelation time
휏
as
15,17,33,34
:
퐾
#+,-(*)+
%
=
β
?
휏
푇
@
1
+
휏
2
푇
B
exp
B
−
2
푇
휏
F
−
1
F
G
,
(
4
)
where
푇
is the exposure time of the camera, and
훽
?
=
훽
−
훽
&''()*
is a constant that
accounts for the loss of correlation associated with the ratio of the detector size to the
speckle size and polarization
33
.
At low signal,
훽
may deviate from typical calibration due
to high sensitivity to noise. To mitigate the issue that
퐾
#+,-(*)+
%
<
0
in low signal situations,
a correction term
훽
&''()*
enforces a positive speckle contrast.
With our proposed compact
SVS setup, we employ a lensless imaging configuration to enhance the numerical
aperture, enabling the recording of multiple speckles within a single pixel, leading to
훽
≈
0
.
05
, measured with a static sample. The measured
훽
value is relatively low because the
average speckle size is smaller than the pixel size, resulting in multiple speckles per pixel.
In SVS, the detecting device operates with an exposure time significantly greater than the
decorrelation time of the sample, i.e.
푇
≫
휏
. In our case,
푇
=
6
ms.
Consequently, Eq.
(4) simplifies to:
퐾
#+,-(*)+
%
≈
(
훽
−
훽
&''()*
)
휏
푇
.
(
5
)
The cerebral blood flow (CBF) can be related to
퐾
#+,-(*)+
%
(and
휏
) as
35,36
:
퐶퐵퐹
=
1
퐾
#+,-(*)+
%
≈
푇
(
훽
−
훽
&''()*
)
휏
.
(
6
)
See Fig. 2(c) for typical example of measured CBF dynamics with
our
SVS
device
.
The CBF accounts for the total volume of blood moved
in each
time period. It can also be
measured in blood flow index (BFI) metric
1
.
The CBF metric accounts for the total volume of blood moved in a given time
period. According to classical fluid mechanics and Poiseuille’s law, blood flow can be
expressed as
퐵퐹
=
789
"
!
:
;<
where
훿푃
represents the difference in blood pressure,
푟
denotes the radius of the blood vessel,
휂
is the dynamic viscosity of the blood, and L
signifies the length of the blood vessel
1
. Thus, any alteration in the blood flow means that
there is a change in either the blood pressure or a change in the diameter of the blood
vessel. It is worth noting that even a slight adjustment in the blood vessels' radius can
have a profound impact on
blood flow due to the fourth power relationship with
푟
. Such
variations are especially significant as they accompany the modulation and regulation of
CBF. In results shown later, we utilize the relative cerebral blood flow (rCBF) metric to
provide normalized blood flow information for enhanced comparability a
cross
measurements.
Note that the computational requirements needed to calculate the speckle contrast
in Eqs. (1)
-
(3) from
the recorded camera images can be handled by a standard consumer
-
7
grade computer [e.g., AMD 7950X CPU]. The most resource
-
demanding step is the
calculation of
퐾
"#$
%
in Eq. (1), as the noise terms in Eq. (3) only need to be calculated
once (pre
-
calibration) or were already calculated in Eq. (1). Therefore, the data
recorded
from our SVS compact device can be processed and stored in real
-
time by using
a
dedicated Basler USB
-
PCIE card and SSD. It is also possible to expand the device to
multiple channels.
Fig.
2
Compact
SVS
processing analysis flowchart for deriving the cerebral blood flow (CBF) from recorded camera
images. (a) Recording and
storing of SVS camera images. (b) Measured raw speckle contrast calculated from the
images in (a). (c) Calculated CBF after calibrating the raw speckle contrast
i
n
(b)
.
.
3
Results
and Discussion
Relative to the
traditional
fiber
-
based SVS systems
15
–
17
, the compact SVS
arrangement, where the sensor is directly positioned atop the region of interest offers the
larger collection area and numerical aperture of the sensor allow for two orders of
magnitude increase in the number of photons collected.
To
demonstrate the superior
signal strength and stability of the compact SVS over traditional fiber
-
based SVS systems
,
we compared the two systems
. The experimental configuration
is shown in Fig. 3(a) and
featur
e
s
a continuous
-
wave 785 nm laser diode, acting as
a
common
light source, and
two SVS detection modules symmetrically placed on each side of the laser source at
the
same S
-
D separation distance. On one detection side, the compact SVS was composed
of a board camera
[Basler daA1920
-
160um
],
directly positioned on the sample. On the
opposing detection side, the
fiber SVS was composed of a 600
-
um diameter multimode
8
optical fiber [Thorlabs FT600UMT], positioned on the sample. The other end of the fiber
was coupled onto a
n identical
camera to the one used in the compact SVS
16
.
Theoretically, we expect
the compact version
to yield a signal gain of
about
70
times
compared to the fiber,
as a result of
the
increased
collection area.
The
camera
sensor
’s
dimension
is
6.6
mm by 4.
1
mm,
resulting in an approximate sensor area of
2
7
mm
2
compared to the
0.28 mm
2
area
of the 600
-
um diameter multimode optical fiber
,
leading to about 95 times gain in detecting area
.
However, the camera sensor is
positioned with a
5 to
7
mm gap from the sample
,
while the fiber is directly placed in
contact with the sample.
In this configuration, we calculated the numerical aperture of the
camera to be
푁
퐴
=
=
0.
28
on
one dimension
and
푁
퐴
>
=
0.4
2
on the
other dimension
.
The
fiber has a numerical aperture of
푁
퐴
'4?)"
=
0.39.
By taking into account the NA difference
between the camera and fiber,
we expect the
collected
signal
gain between the compact
SVS over fiber SVS
to be
@
A
"
⋅
@
A
#
C
@
A
$%&'(
D
)
⋅
A")
#
*+,-
A")
#
$%&'(
≈
75
times
.
To
experimentally validate this gain, we used
a static sample (a thick slice of meat)
and measured
the camera readout signal at different S
-
D distances for the two detection
units.
The S
-
D distances range
d
from 1.5 cm (2.5 cm) for the fiber (compact) SVS to 10.5
cm.
The mounting encasing units prevent smaller S
-
D
distances
.
The results are
presented in Fig.
3(b)
and were averaged over six different realizations at different
locations
.
The error bars were estimated by calculating the standard deviation over
the
six different realizations.
As shown, a consistent gain in the number of photons is captured
for the compact SVS over the fiber SVS system across the multiple S
-
D distances. By
calculating the signal ratio of the two, we determined that the compact version capture
about
70
times more signal than its fiber
-
based
SVS counterpart. This significant
improvement leads to an enhanced detectability at extended S
-
D distances, up to an S
-
D distance increase of 2.5 cm for the same signal readout
in this case
. Note that both
devices reach the noise floor of the camera, although at different S
-
D distances.
The fiber
SVS reaches the noise floor level at an S
-
D distance of approximately 5.5 cm, whereas
the compact SVS maintains a robust signal even at a S
-
D distance of 8.
0
cm
on a static
sample
.
These results showcase the superiority of compact SVS over fiber SVS in the
ability to collect
more
signal, enabling
the
detectability of CBF at extended S
-
D distances.
Note that this increase has not yet consider the
increase in stability
by eliminating the
potential fiber movement during the camera exposure time.
9
Fig. 3
Experimental comparison between fiber SVS and compact SVS: (a) Overview of the experimental setup. (b)
Camera readout signal intensity for fiber and compact SVS measured at various S
-
D distances on a static sample.
Notably, the compact SVS demonstrates a s
ignificantly higher readout signal, averaging approximately 70 times more
than its fiber counterpart. The
f
iber SVS reaches the noise floor level at an S
-
D distance of approximately 5.5 cm,
whereas the
c
ompact SVS maintains a robust signal even at a S
-
D distance of 8.5 cm.
Next, we characterized the stability of the compact SVS over the fiber SVS
.
For
that, we
designed
two
distinct
experiment
s
.
In the first experiment,
presented in Fig. 4
(a)
,
both the compact and fiber SVS systems were affixed
on top of
a
n
one
-
layer phantom,
which comprised of a sealed
container
filled with a liquid mixture (3D printing resin)
16
. The
liquid mixture was
positioned
on an orbital shaker
set at a
rotating speed of
9
0 rotations
per minute
16
.
Both compact SVS and fiber SVS
systems
rotat
ed synchronously
with
the
liquid mixture. During
each
rotation, the SVS system
s
are
measuring the change of
decorrelation time
within
the liquid mixture
16
.
In this context
, the
SVS system
s
measur
e
a
liquid flow dynamic similar to the blood flow dynamic for humans, see
Appendix A of Ref
16
.
T
he compact SVS and fiber SVS were set at S
-
D distances that
yielded an equivalent
photon count
. In this configuration, the light power collected by the compact SVS and
fiber SVS detection
systems
are
equivalent
.
Consequently, we anticipate assessing the stability of the compact SVS and fiber
SVS by comparing the recorded
liquid flow
from the rotating phantom
.
The results are
shown in Fig. 4(
b
). As expected, the liquid flow measured by the compact SVS
exhibits
superior signal quality compared to that obtained by the fiber SVS.
This observation is
further validated when examining the frequencies present in the Fourier spectrum of the
flow signal
intensity
퐼
, Fig. 4(c)
. The Fourier amplitude peak, centered around 1.5 Hz,
corresponds to the rotational frequency of the orbital shaker (90 rotations per minute,
translating to 1.5 rotations per second).
The peak at
around
3 Hz
(4.5 Hz)
,
represent
ing
the second
(third)
harmonic of the
flow pulsation
37
,
is
observable
in both the compact
SVS
and fiber SVS spectra.
However, the noise level is
slightly
higher in the fiber SVS Fourier
spectrum
, indicating that the
measured
flow intensity from the
fiber SVS is less
reproducible
than
that from
compact SVS
.
To quantitatively evaluate the reproducibility
of the measured liquid flow signal, we computed the Pearson correlation factor for each
SVS system as:
휌
W
퐼
(
푡
)
,
퐼
(
푡
+
푑푡
)
Y
=
∑
(
퐼
(
푡
)
−
퐼
̅
)
(
퐼
(
푡
+
푑푡
)
−
퐼
̅
)
E
*
F
G
\
∑
(
퐼
(
푡
)
−
퐼
̅
)
%
E
*
F
G
∑
(
퐼
(
푡
+
푑푡
)
−
퐼
̅
)
E
*
F
G
%
,
(
5
)
10
w
here
퐼
(
푡
)
i
s
the measured flow signal in Fig. 4(a)
,
퐼
̅
is the mean flow intensity,
and
퐼
(
푡
+
푑푡
)
is the signal shifted
by one period of
푑푡
=
1
1
.
5
퐻푧
_
=
0
.
66
sec.
The compact SVS
correlation factor was
휌
!&12#!*
=
0
.
94
and
the fiber SVS was
휌
'4?)"
=
0
.
67
,
demonstrating
that the measured
periodic
signal from the compact SVS is
significantly
more stable than its fiber counterpart.
We
further investigated the compact SVS’s
robustness against human head movements.
Fig. 4
Experimental comparison of flow reproducibility between fiber SVS and compact SVS by using a one
-
layer
phantom rotating on an orbital shaker. (a) Experimental arrangement. (b) Measured flow intensity. (c) Rotating
frequencies obtained by Fourier transfor
ming the flow intensity signal in (b).
The second experiment
, presented in Fig.
5
(a)
,
entailed an evaluation of head
movement instabilities. To conduct this assessment, both SVS systems were positioned
on the forehead of a subject, with a static scattering block interposed between the SVS
setups and the subject’s forehead. The static scatt
ering block comprised a rigid block of
packaging foam, complemented by a thick layer of black tape on its backside to prevent
any laser light from entering the subject’s head. Consequently, the SVS systems
ex
clusively detected light interacting with the static scattering block.
The S
-
D distances
of
both the compact SVS
and fiber SVS are equivalent
, in order to replicate
CBF
data
acquisition scenarios. As a result, the signal intensity on the fiber SVS is
about
70 times
lower than that of the compact SVS.
The measurement
was performed
over a 30
-
second
interval, following a specific protocol: from 0 to 10 seconds, the subject maintained a still
position; at the 10
-
second mark, the subject was asked to laterally move their head (left
to right and right to left) for the subsequent 10 secon
ds; and finally, the subject resumed
a stationary position for the remaining 10 seconds. The results are presented in Fig.
5
(b)
and show that the flow measured by the compact SVS exhibits less noise movement that
with the fiber SVS during head movements. I
n addition, the overall flow intensity notably
rises due to the SVS systems’ movements accompanying head motions. As shown in Fig.
5
(b), this increase in flow intensity is more pronounced for the fiber SVS than the compact
SVS, indicating than the compact SVS experiences less movement vibrations than the
fiber SVS during head motions. The more prominent shift in the fiber SVS is due to
additional decorrelation resulting from fiber movement, which creates blurrier speckle
images. These blurrier images corres
pond to lower contrast, leading to higher flow
intensity as explained in Eq. 5 and Eq. 6. By removing the fiber, this source of instability