Supplemental Document
Correlating stroke risk with non-invasive
cerebrovascular perfusion dynamics using a
portable speckle contrast optical spectroscopy
laser device: supplement
Y
U
X
I
H
UANG
,
1,†
S
IMON
M
AHLER
,
1,8,9,†,
A
IDIN
A
BEDI
,
2,3,4
J
ULIAN
M
ICHAEL
T
YSZKA
,
5
Y
U
T
UNG
L
O
,
2,6
P
ATRICK
D. L
YDEN
,
7
J
ONATHAN
R
USSIN
,
2,3
C
HARLES
L
IU
,
2,3,10,‡
AND
C
HANGHUEI
Y
ANG
1,‡
1
Department of Electrical Engineering, California Institute of Technology, Pasadena, CA 91125, USA
2
USC Neurorestoration Center, Department of Neurological Surgery, Keck School of Medicine, University
of Southern California, Los Angeles, CA 90033, USA
3
Rancho Research Institute, Rancho Los Amigos National Rehabilitation Center, Downey, CA 90242, USA
4
Department of Urology, University of Toledo College of Medicine and Life Sciences, Toledo, OH 43614,
USA
5
Division of Humanities and Social Sciences, California Institute of Technology, Pasadena, CA 91125, USA
6
Department of Neurosurgery, National Neuroscience Institute, Singapore 308433, Singapore
7
Department of Physiology and Neuroscience, Zilkha Neurogenetic Institute, and Department of Neurology,
Keck School of Medicine, University of Southern California, Los Angeles, CA 90033, USA
8
mahler@caltech.edu
9
sim.mahler@gmail.com
10
cliu@usc.edu
†
These authors contributed equally to this work.
‡
These authors jointly supervised this work.
This supplement published with Optica Publishing Group on 30 September 2024 by The Authors
under the terms of the Creative Commons Attribution 4.0 License in the format provided by the
authors and unedited. Further distribution of this work must maintain attribution to the author(s)
and the published article’s title, journal citation, and DOI.
Supplement DOI: https://doi.org/10.6084/m9.figshare.26984329
Parent Article DOI: https://doi.org/10.1364/BOE.534796
S1
Supplementary Material:
Correlating Stroke Risk with Non-Invasive Cerebrovascular Perfusion
Dynamics using a Portable Speckle Contrast Optical Spectroscopy Laser
Device
Yu Xi Huang,
a,†
Simon Mahler,
a,†, **
Aidin Abedi,
b,c,d
Julian Michael Tyszka,
e
Yu Tung Lo,
b,f
Patrick D.
Lyden
g
, Jonathan Russin,
b,c
Charles Liu,
b,c,*,***
Changhuei Yang
a,***
a
Department of Electrical Engineering, California Institute of Technology; Pasadena, CA 91125, USA.
b
USC Neurorestoration Center, Department of Neurological Surgery, Keck School of Medicine, University of Southern California;
Los Angeles, CA 90033, USA.
c
Rancho Research Institute, Rancho Los Amigos National Rehabilitation Center; Downey, CA 90242, USA.
d
Department of Urology, University of Toledo College of Medicine and Life Sciences; Toledo, OH 43614, USA.
e
Division of Humanities and Social Sciences, California Institute of Technology; Pasadena, CA 91125, USA.
f
Department of Neurosurgery, National Neuroscience Institute, Singapore 308433
g
Department of Physiology and Neuroscience, Zilkha Neurogenetic Institute, and Department of Neurology, Keck School of
Medicine, University of Southern California, Los Angeles, CA 90033 USA
†
These authors contributed equally to this work.
***
These authors co-supervised this work.
Corresponding authors:
*Email:
cliu@usc.edu
**Emails:
mahler@caltech.edu,
sim.mahler@gmail.com
S1. Heart rate changes and other features comparison during breath-holding
In this section, we show the resting heart rate and the maximal heart rate during breath-holding of
the two risk groups (Fig. S1). The heart rate was calculated by Fourier transforming the CBFI signal and
measuring the frequency of the heart rate Fourier peak (Fig. 1b). The resting heart rate was calculated by
selecting the first 60 seconds of the CBFI time trace, where the participant is at rest (sitting in a chair) and
breathing normally. The maximal heart rate was measured from the calculated heart rate time trace during
the breath-holding exercise, obtained by Fourier transforming a sliding window of 20 seconds in the CBFI
time trace (Fig. 1d). As shown in Fig. S1a, the distributions of resting heart rates of the two risk groups have
similar median and range of fluctuations, where the higher risk group has a slightly broader distribution.
However, in Fig. S1b, the low-risk group has a higher maximal heart rate than the higher-risk group. Those
results agree with previous studies on resting and exercising heart rates between young and older
populations: younger and older individual tends to have similar resting heart rates. In contrast, maximum
heart rates are typically higher in younger individuals and decline with age due to age-related factors such
as cardiovascular function, cardiac output, and blood vessel elasticity. Note that individual variations exist
within each age group, influenced by factors such as fitness level, overall health, genetics, and medications.
S2
Fig. S1
a. Resting heart rate and b. Maximal heart rate during breath-holding exercise for the two risk groups (low
and higher).
Next, we present the distributions of BHI
CBF
(Fig. S2a) and BHI
CBV
(Fig. S2b) also shown in the
main manuscript, now displayed in larger size for better visibility. As shown, both BHI
CBF
and BHI
CBV
distributions exhibit significant differences among the low-risk and higher-risk groups. Figure S2c show the
distribution of the duration of breath-holding
푇
퐵퐻
, where both risk groups have similar distributions. Finally,
Figure S2d shows the ratio between
푝푒푎푘푠
푟푎푡푖표
퐵퐻
/푝푒푎푘푠
푟푎푡푖표
푟푒푠푡푖푛푔
, where not much statistically
significant can be observed between the two risk groups.
Fig. S2
Distribution of a. BHI
CBF
, b. BHI
CBV
, c. breath hold duration, and d. ratio of the peaks ratio changes during
breath-holding exercise for the two risk groups (low and higher).
S3
Figure S3 shows the distributions of
푝푒푎푘푠
푟푎푡푖표
푟푒푠푡푖푛푔
(Fig. S3a) and
푝푒푎푘푠
푟푎푡푖표
퐵퐻
(Fig. S3b), where
statistically significant can be observed between the two risk groups.
Fig. S3
a. Distributions of
푝푒푎푘푠
푟푎푡푖표
푟푒푠푡푖푛푔
changes between the two risk groups during breath-holding. b.
Distribution of
푝푒푎푘푠
푟푎푡푖표
퐵퐻
changes between the two risk groups during breath-holding.
S2. Comparative analysis of compact SCOS forehead measurements and upper arm blood
pressure cuff readings
In this section, we compare the blood pressure ratio and resting heart rate results obtained with the
compact SCOS device to those from a blood pressure cuff. Figure S4a shows the distribution of resting
heart rates measured using the compact SCOS device on the forehead (Fig. 1 of main text). Figure S4b
shows the distribution of resting heart rates measured using a blood pressure cuff positioned on the upper
arm. The blood pressure cuff machine used was a digital automatic blood pressure monitor model Paramed
B22S, which provides systolic and diastolic blood pressure in mm Hg and heart rate in bpm. The blood
pressure readings from the cuff machine were taken at rest and approximately an hour apart from the
compact SCOS experiments. To ensure accuracy, each blood pressure reading was taken twice. Note that
blood pressure readings using a cuff were measured on 22 subjects in the low-risk group and 24 subjects
in the higher-risk group.
As shown in Figs. S4a and S4b, the distributions of resting heart rates among the two risk groups
measured from the compact SCOS on the forehead and from the blood pressure cuff on the upper arm are
similar. Note that since the two measurements were taken about an hour apart from each other and in
different environments, we did expect both distributions to be slightly different.
Figures S4c and S4d present the systolic and diastolic blood pressures measured from the cuff
machine. The distributions of blood pressure of the two risk groups are similar to those shown in Fig. 3 of
the main text using the compact SCOS device. However, the p-values in Figs. S4c and S4d are one-star
significance (
푝
푆푦푠푡표푙푒
< 0.05 and
푝
퐷푖푎푠푡표푙푒
< 0.05), compared to with the four-star significance (
푝
푆퐶푂푆
=
0.000001) obtained with the compact SCOS device. These results suggest that while the blood pressure
cuff can help distinguish between the two risk groups, the compact SCOS device is more effective. One
explanation is that the blood pressure cuff measures blood pressure at rest and on the upper arm, whereas
the compact SCOS device measures blood dynamics changes during breath-holding on the forehead,
providing a proxy for brain vascular health.
S4
Fig. S4
Comparison between SCOS and blood pressure cuff measurements. Purple-framed graphs indicate
measurements from blood pressure cuff. a. Resting heart rate measured by compact SCOS on the forehead, b.
Resting heart rate measured by a blood pressure cuff on the upper arm, c. Systolic and d. Diastolic blood pressure
measured by a blood pressure cuff on the upper arm.
To summarize, the measurements of resting heart rates and blood dynamics changes obtained
with the compact SCOS device on the forehead correlate well with the measurements of resting heart rates
and blood pressures obtained with the cuff on the upper arm. It has also recently shown that SCOS
measurements correlate well with transcranial Doppler (TCD) ultrasound measurements (Ref. [20] of main
text). These correlations indicate that compact SCOS can be used as a stroke risk assessment tool. Note
that compact SCOS only measures relative unitless analog of blood pressure.
S5
S3. Cleveland Stroke Risk Questionnaire
In Fig. 2a, the traditional stroke risk assessment was performed using the Cleveland Stroke Risk
Calculator. Table S1 shows the questionnaire filled out by each participant. In addition to the answers
provided by each participant, the investigator assessed the overall health of the participant to be very good,
good, fair, or poor.
Subject Stroke Risk Questionnaire
What is your age?
years old
What is your sex?
Female
Male
Were you ever told by a physician that you had a stroke?
Yes
No
Were you ever told by a physician that you had a mini-stroke or TIA, also known
as a transient ischemic attack?
Yes
No
Has a doctor or other health professional ever told you that you have atrial
fibrillation?
Yes
No
Has a doctor or other heath professional ever told you that you have diabetes or
high blood sugar or pre-diabetes?
Yes
No
Has a doctor or health professional ever told you that you have hypertension or
high blood pressure?
Yes
No
Has a doctor or other health professional ever told you that you had a myocardial
infarction or heart attack?
Yes
No
Have you smoked (cigarette or others) at least 100 times in your lifetime?
Yes
No
Do you smoke (cigarette or others) now, even occasionally?
Yes
No
What is your race?
Table S1.
Subject Stroke Risk Questionnaire completed by each participant for stroke risk assessment via the
Cleveland Clinic stroke risk calculator.
S6
S4. Peak Ratio Extraction Algorithm
In some situations, the peaks P1, P2, and P3 are not always clearly discernible as local maxima in
the CBFI traces, which can complicate their identification using conventional peak-finding algorithms. To
address this, we develop a method to detect peaks P1, P2, and P3 using a feature extraction approach
outlined as follows:
Step 1. Cardiac Period Identification: Each cardiac period within the CBFI signal is identified using local
minima peak detections on the CBFI time trace. The time-period between two local minima was
enforced to be at least 60% of the cardiac period measured from the Fourier method (Figs. 1b and
2b).
Step 2. Normalization: After identifying all the cardiac periods, we normalize the CBFI values of each
cardiac period from 0 to 1, standardizing the CBFI amplitude.
Step 3. Segmentation: Each normalized cardiac period is divided into 12 segments, shown in the figure
below.
Fig. S5
Example of segmented cardiac period using our algorithm.
Step 4. Segment Areas Calculation: We calculate the area under each segment.
Step 5. Peak and Dicrotic Notch Identifications: As shown in Fig. S5, segment 1 typically corresponds
to peak P1, segment 3 to peak P2, segment 4 to the dicrotic notch, and segment 6 to peak P3.
Step 6.
Feature Calculation: The peaks ratio at rest is determined by calculating the averaged ratio
between segment 3 and segment 1 over the initial 60 seconds of a breath-holding time trace. The
peak ratio during breath-holding is determined by calculating the averaged ratio between segment
3 and segment 1 over the final seconds of the breath hold period.
Figure S6 provides an example of the segment 3 to segment 1 ratio during an entire recording
period. As shown, the ratio significantly increases during breath-holding. The peak ratio at rest
correspond to the average ratio during the first 60 seconds, while the peak ratio during breath holding
corresponds to the average ratio during the final seconds of the breath hold period.
S7
Fig. S6
Example of ratio of segment 3 over segment 1 during a breath-holding recording time trace. As shown, the
ratio significantly increases during breath-holding.
S5. Detailed metrics analysis at various risk levels
We first present in Table S2 the mean and standard deviation of the different extracted features from Table
1, but now breaking down to specific risk groups similar to Fig. 4. We also show compare the results
obtained from SCOS BHI ratio and blood pressure cuff for each risk subgroup.
Table S2.
Mean and standard deviations of all the different features of each risk subgroup.
Feature
Risk 1
mean (std)
Risk 4
mean (std)
Risk 5
mean (std)
Risk 6-7
mean (std)
T
BH
(s)
33.49 (10.21)
36.58 (13.55)
41.06 (12.75)
23.32 (5.30)
τ
growth
(s)
14.89 (10.48)
17.26 (9.60)
19.87 (11.27)
12.96 (10.53)
τ
decay
(s)
7.78 (8.55)
7.40 (6.00)
20.18 (34.03)
4.45 (6.15)
CBFI
change
(%)
42.00 (20.35)
48.09 (18.71)
61.54 (34.56)
68.13 (31.01)
CBVI
change
(%)
16.31 (14.68)
8.27 (6.46)
7.08 (9.52)
7.47 (5.37)
BHI
CBF
(%/s)
1.34 (0.71)
1.53 (0.96)
1.53 (0.72)
3.18 (1.91)
BHI
CBV
(%/s)
0.54 (0.57)
0.27 (0.20)
0.17 (0.22)
0.36 (0.29)
BHI
CBF
/BHI
CBV
(arb. u.)
1.23 (0.15)
1.37 (0.17)
1.51 (0.28)
1.57 (0.28)
Resting heart rate (bpm)
67.51 (11.04)
68.82 (9.18)
64.00 (9.10)
71.08 (4.99)
Maximum heart rate (bpm)
83.68 (12.58)
79.74 (10.99)
72.59 (11.13)
77.95 (6.11)
푇
푚푎푥
퐵푉퐼
-
푇
푚푎푥
퐵퐹퐼
(s)
-0.18 (3.89)
-0.17 (3.72)
-0.31 (3.40)
0.88 (4.69)
Peaks ratio
resting
(arb. u.)
0.85 (0.14)
1.25 (0.46)
1.05(0.17)
0.95 (0.17)
Peaks ratio
BH
(arb. u.)
1.05 (0.16)
1.40 (0.44)
1.25 (0.24)
1.15 (0.18)
Peaks ratio
BH
/
Peaks ratio
resting
1.24 (0.11)
1.16 (0.14)
1.19 (0.14)
1.22(0.15)
S8
Fig. S7
Comparison between SCOS BHI ratio and blood pressure cuff measurements for each risk subgroup. Purple-
framed graphs indicate measurements from blood pressure cuff. a. BHI ratio measured by compact SCOS on the
forehead, c. Systolic and d. Diastolic blood pressure measured by a blood pressure cuff on the upper arm.
S9
To summarize the results in Table S2, we also conducted ANOVA test and Spearman correlation tests
between the different risk subgroups, shown in Table S3.
Table S3.
ANOVA test and Spearman rank correlation results for each feature.
Feature
ANOVA F
ANOVA PR
Spearman Rank
Correlation
Spearman
p-value
T
BH
(s)
4.68
0.01
-0.13
1.11E-01
τ
growth
(s)
3.81
0.02
0.06
4.53E-01
τ
decay
(s)
2.37
0.08
-0.10
2.07E-01
CBFI
change
(%)
5.32
0.00
0.36
2.64E-06
CBVI
change
(%)
3.84
0.02
-0.37
1.37E-06
BHI
CBF
(%/s)
9.75
0.00
0.35
7.15E-06
BHI
CBV
(%/s)
3.12
0.03
-0.28
4.09E-04
BHI
CBF
/BHI
CBV
(arb. u.)
14.56
0.00
0.55
5.57E-14
Resting heart rate (bpm)
1.17
0.33
0.07
4.10E-01
Maximum heart rate (bpm)
3.43
0.02
-0.26
8.76E-04
푇
푚푎푥
퐵푉퐼
-
푇
푚푎푥
퐵퐹퐼
(s)
0.97
0.42
0.03
7.48E-01
Peaks ratio
resting
(arb. u.)
13.67
0.00
0.28
7.18E-04
Peaks ratio
BH
(arb. u.)
9.04
0.00
0.26
1.97E-03
Peaks ratio
BH
/
Peaks ratio
resting
8.14
0.00
-0.05
5.26E-01