Assessment of cerebral autoregulation using continuous-wave near-infrared spectroscopy during squat-stand maneuvers in subjects with symptoms of orthostatic intolerance

Orthostatic lightheadedness in healthy young adults often leads to syncope in severe cases. One suggested underlying mechanism of orthostatic lightheadedness is a drop in transient blood pressure (BP); however, a decrease in BP does not always lead to a drop in cerebral blood flow (CBF) due to cerebral autoregulation (CA). We present a direct assessment method of CA using a multichannel continuous-wave near-infrared spectroscopy (CW-NIRS) device that measures the temporal changes in oxy- and deoxy-hemoglobin concentrations in the prefrontal cortex. Twenty healthy young adults were recruited. During the experiment, continuous beat-to-beat BP and heart rate were simultaneously measured during repetitive squat-stand maneuvers. We introduce a new metric termed ‘time-derivative hemodynamic model (DHbT)’, which is the time-derivative of total-hemoglobin concentration change that reflects the changes of cerebral blood volume and CBF. Although the absolute levels and the variations of systolic and diastolic BPs and mean arterial pressure showed no significant difference between the two groups, the proposed model showed a distinct difference in slope variation and response time of DHbT between the subjects with frequent symptom of orthostatic intolerance and the healthy control subjects. Thus, these results clearly demonstrate the feasibility of using CW-NIRS devices as a CA performance assessment tool.

Experimental Design. Figure 1A shows the data acquisition setup. The changes in optical intensity were measured using a commercial wireless continuous-wave near-infrared spectroscopy (CW-NIRS) system (NIRSIT, OBELAB Inc., Republic of Korea) at a sampling rate of 8.13 Hz. The optical probes consist of 24 sources and 32 detectors. The sources of the CW-NIRS system were dual-wavelength vertical cavity surface emitting laser diodes operating at 780 nm and 850 nm and the detectors were implemented with silicon photodiodes. The output power of the laser diodes was automatically controlled to satisfy the FDA regulations of maximum permissible exposure standards (i.e. under 1 mW) to human subjects. The gain of the detectors was individually calibrated for each subject to maximize the signal intensity without saturation, taking different skin pigmentations, skin and skull thickness, and other biological factors into account. The optical probes were arranged at distances of 1.5 cm as shown in Fig. 1C, where a total of 48 pre-defined channels with a source-detector separation of 3 cm were measured. Such separation is adequate for the depth penetration of 20 mm from the scalp, which is required for the inclusion of the measurements from the microvasculature of the cerebral cortex 11 . The measurements were obtained from the prefrontal cortex (see Fig. 1D), where the center of the lowermost optical probes was aligned to the frontal pole zero (FPz) location of the 10-20 EEG system to remove positional uncertainty between subjects. The device features a rubber cap at the tip of the optical probe to ensure ambient-light-proof contact to the forehead; the device was tightly fastened to minimize additional motion-induced contamination from contact variations.
A continuous beat-to-beat finger photoplethysmography system (Finometer ® , Finapres Medical Systems, Amsterdam, Netherlands) was used to measure systolic BP (SBP), diastolic BP (DBP), mean arterial pressure (MAP), and HR non-invasively at a sampling frequency of 1.4 Hz. A pneumatic finger cuff was wrapped around the left middle finger for the measurements. The outputs of the CW-NIRS and the photoplethysmography devices were recorded simultaneously in a dark environment to maintain signal integrity by preventing ambient light during the experiment.
All subjects underwent an identical test protocol composed of series of repetitive maneuvers to trigger CA. Prior to the CW-NIRS and the photoplethysmography measurements, subjects were trained to stand upright and to squat in a uniform manner. The protocol consisted of a two minutes initialization for the CW-NIRS system Scientific REPORTS | (2018) 8:13257 | DOI:10.1038/s41598-018-31685-y and the pneumatic cuff for stabilization while the participants were in a standing-rest position, as shown in Fig. 1B. Then, the subjects rested again in the standing-rest position for one minute to record the baseline (BL) levels. Subsequently, they were instructed to gently sit into a squatting position and hold for one minute, during which they could hold onto a preinstalled grab-bar with their right hand to aid balance, so as to minimize motion artifacts caused by abrupt movements. During the squatting-rest posture, subjects could either stand on tiptoe or with their feet flat, depending on their preference. Lastly, the subjects were instructed to return to the standing-rest position and hold for 3 minutes. The entire maneuver (one minute squat followed by three minutes of standing) was repeated five times per subject (see Fig. 1B). The subjects were instructed to breathe normally during the test protocol. During the rest periods, the subjects focused on a black fixation cross that was positioned according to their head position (i.e. during standing or squatting), to minimize hemodynamic changes caused by both cognitive activities and gravitational blood shifting under head tilting 12 . The event markers were simultaneously recorded manually by the observer when the subjects change postures, by using embedded event marking functions in each system. Data Processing. Data analysis was performed using MATLAB (R2013b; MathWorks, Natick, USA). Prior to the analysis, the CW-NIRS channels with insufficient signal integrity, i.e., signal-to-noise ratio (SNR) less than 30 dB, were excluded for further analysis. We obtained the mean (S m ) and standard deviation (S std ) of the motion-less BL signal (i.e. 10 to 15 seconds) to calculate the SNR (=20·log 10 (S m /S std )). The signals were then converted into logarithmic changes in optical density. The optical density variations were then filtered using a low-pass filter with a cutoff frequency of 0.2 Hz to suppress the instrumental noise and physiological oscillations (i.e. Mayer's waves, respiration noise, and heartbeat). These filtered optical density variations were converted to changes in HbO 2 and HbR using the modified Beer-Lambert law 13 . In this study, the variation in the total hemoglobin concentration (HbT = HbO 2 + HbR) was the main focus during the squat-stand maneuver because HbT is regarded as a surrogate for cerebral blood volume (CBV) [14][15][16] .
The raw SBP, DBP, MAP, and HR signals acquired from the photoplethysmography device were filtered using a 4-tap moving average filter to suppress the non-physiological signal caused by instant contact variation during the maneuvers.
For reliability, all the measurements (i.e. HbT, SBP, DBP, MAP, and HR) from the CW-NIRS and the photoplethysmography devices were block-averaged across five maneuver trials per subject.

Time-Derivative Hemodynamic Model.
In this study, a new metric for the assessment of individual CA performance was introduced. The metric, referred to as "DHbT", the time-derivative of HbT where DHbT = dHbT/dt, was used to quantify HbT variations during the squat-stand maneuvers. For the numerical differentiation, a two-point difference algorithm was used 17 .
A typical change in HbT and DHbT under the squat-stand transition is shown in Fig. 2A. Since HbT variation is tightly coupled with that of CBV [16][17][18] , we assumed that the hemodynamic recovery response extracted from the HbT is caused by the CA in conjunction with the baroreflex mechanism of any given subject. Since CW-NIRS measures only relative hemodynamic changes from BL, assessment of individual CA performance using only measured HbT is difficult. In order to overcome this limitation, DHbT which translates relative change into absolute numbers, was proposed for assessment of individual CA performance. For this, the absolute value of the maximum and minimum slope variation (Max SV and Min SV, respectively) of DHbT and the corresponding peak-times (Max PT and Min PT, respectively) were extracted as objective parameters as shown in Fig. 2B.
Additionally, in order to compare the feasibility of DHbT with HbT and to provide physiological insight, the decrease time (DT), defined as the time to nadir right after standing, and the increase time (IT), defined as the time to the overshoot peak from the nadir, and the recovery time (RT), defined as sum of DT and IT, were extracted from the transient response of HbT by using the similar approach given in the previous study 19 as shown in Fig. 2A.

Statistical analysis.
We performed statistical analysis using SPSS (IBM Corp. Released 2012, IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY, USA) and GraphPad Prism (GraphPad Prism version 5 for Windows, GraphPad Software, San Diego, CA, USA, www.graphpad.com). Means and standard deviations were extracted from continuous variables, and the Shapiro-Wilk test was used to test normality of the data. Between-groups differences in continuous variables were evaluated using unpaired t-tests (if data were normally distributed) and Mann-Whitney U tests (if data were not normally distributed). Moreover, within-group changes in cardiovascular variables were evaluated using Wilcoxon signed-rank tests. Significance levels were set to P < 0.05.

Results
Subject characteristics. Table 1 shows group characteristics, represented by group-averaged BL values and standard deviations. There were no significant between-group differences other than the symptom score (Mann-Whitney U test, P = 0.00001). Although the symptom group tended to have a higher HR at BL, this difference did not reach significance.
Blood Pressure and Heart Rate Variation. A typical response of the SBP, DBP, MAP, and HR during the maneuver is shown in Fig. 3A. For each subject, BL of SBP, DBP, MAP, and HR values were obtained by averaging the 30-second measurements prior to the squat. Additionally, the block-averaged response was divided into four distinct time intervals to derive a single representative value; one averaged 30-second epoch from the squatting-rest position (sqR) and three averaged 5-second epochs from the standing-rest position (T1, T2, and T3) representing for each of the recovery phases after standing at 60, 120, and 180 seconds. These time intervals were selected to extract the representative values of SBP, DBP, MAP, and HR while the CA progresses into the autoregulatory plateau after the squat or stand. Additionally, from the cardiovascular oscillation right after standing, minimum and maximum peak (MinP and MaxP) of SBP, DBP, MAP, and HR values were obtained for analysis.
The overall pattern of cardiovascular responses during the squat-stand maneuver were similar between the control and symptom groups (Fig. 3B). The control group demonstrated higher BP (SBP, DBP, and MAP) and   Table 1. Subject characteristics. Values are presented as the mean ± SD. Text in bold indicates statistical significance. SD = standard deviation; BMI = body mass index; SBP = systolic blood pressure; DBP = diastolic blood pressure; MAP = mean artery pressure; HR = heart rate. lower HR in general at each interval than the symptom group. However, these differences did not reach significance at BL, sqR, T1, T2, and T3 (n.s., Bonferroni-corrected Mann-Whitney U tests, significance level adjusted to 0.01 [obtained by 0.05/5]). The magnitude of SBP and MAP increased significantly at sqR compared to BL for both the control group (7.15% and 8.17% respectively, P < 0.05) and symptom group (11.45% and 9.99% respectively, P < 0.01). For both groups, when the subjects returned to the standing-rest position, SBP and MAP dropped and recovered to the BL level during T2 and T3.
Similar to SBP and MAP, DBP increased significantly after the squat (during the sqR interval) compared to BL, for both the control group (5.41%, P = 0.046) and symptom group (5.49%, P = 0.002). However, the absolute increment of DBP was lower than that of SBP and MAP, and persisted throughout T1 and then recovered to the BL levels for both groups.
Meanwhile, HR fell substantially during sqR compared to BL, for both the control group (−9.36%, P = 0.028) and the symptom group (−15.23%, P = 0.001) and then increased rapidly subsequent to the squat-stand transition (T1, T2, and T3). BP and HR returned to the BL level within T2 following the squat-stand transition and then showed slight increase in T3 compared to BL for both the control group (1.85%, P = 0.046) and the symptom group (2.35%, P = 0.035). However, the block-averaged variations of SBP, DBP, MAP, and HR in response to the orthostatic challenge were not significantly different between groups.
The differences in cardiovascular measures over time (vs. BL) are shown in Fig. 4A. SBP, DBP, and MAP increased from the BL level for both groups as noted above. However, no between-group differences were observed. SBP, DBP, and MAP gradually returned to BL levels over T1, T2, and T3. BP recovery showed no significant between-group difference.
The symptom group demonstrated significant decrease in HR at sqR (vs. BL) compared to the control group (Mann-Whitney U test; control, −8.2 ± 3.2 bpm; symptom, −14.0 ± 6.4 bpm; P = 0.041). Unlike SBP, DBP, and MAP, HR returned immediately to the BL level at T1 and remained unchanged at T2 and T3. There were no between-group differences in HR recovery to the BL level.
The changes in cardiovascular measures at T1, T2, and T3 vs. sqR are shown in Fig. 4B. The drops in SBP, DBP, and MAP seen at T1, T2, and T3 were not significant. Such correspondingly indicates a negative test result of the classical orthostatic hypotension (OH) test, which is defined as an SBP drop of more than 20 mmHg and/ or a DBP drop of more than 10 mmHg after three minutes of standing 3,20 . By contrast, the HR level change (vs. sqR) was significantly greater in the symptom group than the control group at T1 (P = 0.003), T2 (P = 0.005), and T3 (P = 0.041). Particularly at T1, the variations in HR (vs. sqR) were 9.6 bpm higher in the symptom group (15.0 ± 6.0 bpm) than the control group (5.3 ± 3.9 bpm). The overall HR deviation from sqR for the control group demonstrated gradual increase over time, whereas the deviation for the symptom group showed no significant variations.
The transient variations in SBP, DBP, MAP, and HR are shown in Fig. 4C. The difference between MinP and sqR (MinP -sqR), which implies the initial transient reduction of the cardiovascular response, was calculated. Additionally, the maximum relative change to the MinP was extracted by taking the difference between MaxP and MinP (MaxP -MinP), as shown in Fig. 3A.
MinP -sqR of SBP, DBP, and MAP were not significantly different between the control and symptom groups. Conversely, the initial transient reduction of BP demonstrated a positive result for a few participants with the conventional initial orthostatic hypotension (IOH) diagnosis (Control 83.3% and Symptom 57.1%); a transient decrease in SBP by more than 40 mmHg and/or DBP by more than 20 mmHg within the first 15 second of standing prior to BP recovery under supine-stand test 4 . In addition, MaxP -MinP of the SBP, DBP, and MAP did not show significant differences between the two groups.
HR remained stable when the subjects returned to the standing position with a slight deviation in MinP -sqR for both groups (control, −1.7 ± 2.6 bpm; symptom, −0.6 ± 1.7 bpm). MaxP -MinP of HR increased abruptly over 30 bpm for both groups (control, 33.4 ± 5.16 bpm; symptom, 35.4 ± 7.1 bpm). However, there was no significant difference in MaxP -MinP of HR between the two groups.
Cerebral Microvascular Hemodynamic Response. The objective parameters obtained from the regionally grand-averaged DHbT are summarized in Table 2. The squat-stand maneuver resulted in significantly smaller increase in Min SV in section R4 in the symptom group compared to the control group (P = 0.033). However, the variation of the Max SV did not show significant differences in the symptom group compared to the control group.

Discussion
In this study, we utilized a wireless multi-channel CW-NIRS system and a beat-to-beat finger photoplethysmography device to investigate hemodynamic changes that occur during repeated squat-stand maneuvers. We compared SBP, DBP, MAP, and HR in a control group and an OI symptom group. We also evaluated the individual CA performance through statistical analysis of DHbT variation.
There were no between-group differences in the absolute levels and the variations of SBP, DBP, and MAP after the squat-stand maneuver. This may be because subjects had no history of OH, even though the symptom group had OI symptoms. However, significant variations in HR were observed in the symptom group after the squatting rest, which could indicate a compensatory effort by the heart to maintain cardiac output (CO) constant; this may further indicate impairment of the baroreflex function involved in the pathogenesis of IOH 21,22 . This suggests that the symptom group had a mildly impaired CA compared to the control group. In the same context, from the IOH diagnostic test, a few subjects demonstrated IOH-positive regardless of the presence of OI symptom. However, since the conventional diagnosis method is established for the supine-stand maneuver, the transient BP drop during squat-stand maneuver should employ a different diagnosis method for the categorization of the control and symptom group. It addresses the fact that the measured BP drop was moderate as compared to the previous investigation conducted to young healthy adults (SBP = 56 ± 15.7 mmHg and DBP = 37.0 ± 6.8 mmHg) 23 , which resulted in false-positive outcomes. The cardiovascular responses did not demonstrate clear evidence of degradation in the CA performance, since finger photoplethysmography was an indirect method to assess CA, which is an overlooking method that measures from the finger artery (i.e. a branch of the brachial artery).
Unlike BP and HR, the cerebral hemodynamic response, HbT, is directly related to CA 8,14 . HbT increases when a subject squats; the squatting-rest position results in increase in the venous return due to muscle contraction of the lower body 24 . When a subject stands, gravity pulls the venous blood towards the feet initiates the HbT drop, which might be associated with a decrease in CBF 25 . Accordingly, the baroreceptor activates to stimulate the vasomotor center to recover CBF as a compensatory mechanism that increases systemic vascular resistance or CO through autonomic-reflex activity 26 . It is well known that the relative change of CBV has a linear relationship with CBF 27,28 . Furthermore, the variations in the CBV and HbT are proportionally related 21 . As such, we revealed significant between-group differences in the DHbT response, which is the absolute change of HbT with respect to time, using a CW-NIRS device during the squat-stand maneuver. Our results clearly demonstrated increased PT in the entire sections except L1 and decreased SV in section R4 in the symptom group compared to the control group. The group averages of the DHbT response are shown in Fig. 5. The DHbT response showed a  clear difference between the two groups, which shows that DHbT can be used as an objective parameter for the assessment of an individual's vasoreactivity. Meanwhile, the PT of the DHbT has demonstrated significant increase in the right hemisphere (especially in section R3) among the symptom group as compared from the control group. This result could reflect the hemisphere asymmetry in parasympathetic control of the heart 29 . Namely, previous studies have shown that sympathetic cardiovascular control is lateralized to the right hemisphere 30,31 . In addition, the lack of a right-sided HbT increase in response to pressure is indicative of an impaired vasoreactivity caused by dysfunction of the right hemisphere and the sympathetic nervous system 32 . Thus, the degraded fluctuation of the DHbT response in the right hemisphere is opening up the possibility of the lateralization of the sympathetic cardiovascular control in the brain.
Interestingly, both DT and RT extracted from HbT showed significant increase in the symptom group compared to the control group. This result clearly shows that the reach time to nadir was increased, which evidently shows the dysfunction of vasoreactivity in the symptom group right after squatting rest. However, since the results from DHbT demonstrates higher level of statistical significance than those from HbT, the CA assessment can be made more accurately with the metrics using DHbT than conventional schemes using HbT.
We tested the feasibility of using a multichannel CW-NIRS device as a CA performance assessment tool. The derivative of HbT, DHbT during the squat-stand maneuvers, demonstrated significant between-group difference, such as SV, PT of the DHbT response.
The utilization of CW-NIRS in CA assessment is preferred over a finger photoplethysmography-based BP monitoring scheme, due to its portability and high temporal and spatial resolution. As such, CW-NIRS devices can be used for the CA assessment in patients with neurological and psychiatric disorders. In addition, the proposed method can be used on healthy people on a regular basis to assess CA performance.
For future work, we plan to perform experiments with patients diagnosed with IOH or OH to verify the method by assessing the dysfunction of autonomous nervous system, where the present work was applied limitedly to young adults with OI symptoms. The comparison of the DHbT response between the healthy controls and clinically diagnosed patients will allow us to validate the proposed evaluation method with respect to its sensitivity and specificity.

Data Availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.