Wearable speckle plethysmography (SPG) for characterizing microvascular flow and resistance.

In this work we introduce a modified form of laser speckle imaging (LSI) referred to as affixed transmission speckle analysis (ATSA) that uses a single coherent light source to probe two physiological signals: one related to pulsatile vascular expansion (classically known as the photoplethysmographic (PPG) waveform) and one related to pulsatile vascular blood flow (named here the speckle plethysmographic (SPG) waveform). The PPG signal is determined by recording intensity fluctuations, and the SPG signal is determined via the LSI dynamic light scattering technique. These two co-registered signals are obtained by transilluminating a single digit (e.g. finger) which produces quasi-periodic waveforms derived from the cardiac cycle. Because PPG and SPG waveforms probe vascular expansion and flow, respectively, in cm-thick tissue, these complementary phenomena are offset in time and have rich dynamic features. We characterize the timing offset and harmonic content of the waveforms in 16 human subjects and demonstrate physiologic relevance for assessing microvascular flow and resistance.


Introduction
Photoplethysmography is a well-established technique in which light is transmitted through tissue in order to interrogate vascular fluctuations caused by the cardiac cycle [1,2]. These fluctuations are due to volumetric expansion of blood caused by variations in pressure which ultimately modulate the attenuation of transmitted light [3,4]. Analyzing the Photoplethysmographic (PPG) cardiac waveforms has been used extensively as a means for hemodynamic characterization [5][6][7][8]. One successful example is pulse oximetry in which the AC components of the PPG waveform at multiple optical wavelengths are used to calculate arterial hemoglobin oxygen saturation [5]. However, PPG has failed to achieve wide-spread clinical adoption in other important applications such as continuous noninvasive arterial pressure, arterial stiffness characterization, and noninvasive cardiac output monitoring [7,8]. One reason for this is the tendency of the PPG waveform to deteriorate in situations of low peripheral blood flow such as hypovolemia or thermoregulatory vasoconstriction [1]. In addition, PPG may lack sufficient information content to properly characterize the cardiovascular system.
Affixed Transmission Speckle Analysis (ATSA) addresses some of the shortcomings of LSI by utilizing transmission geometry. One format that ATSA can be performed in is a miniaturized form factor attached to a digit, similar to a conventional pulse-oximeter. In this setup, the highly diffuse photons probe the full thickness of tissue and provide a high-SNR signal that can be used to determine average flow. The clip-on design further improves signal quality by reducing motion artifacts. Importantly, ATSA is also capable of detecting fluctuations in laser coherence due to the cardiac cycle. We refer to the methodology of obtaining this signal as speckle plethysmography (SPG). Additionally, this instrument is sensitive to the PPG waveform; volumetric vessel changes associated with cardiac pulsatility modulate light intensity. Thus, ATSA ultimately provides simultaneous, co-registered SPG-PPG output.
In this work we introduce ATSA and validate its sensitivity to flow in finger-like tissue phantoms. In a series of human volunteers, we demonstrate the process by which the SPG and PPG waveforms can be extracted from the same raw data and provide a physiological mechanism for the origin of both signals. We show that SPG and PPG signals are offset in time, and evaluate the harmonic content of the dynamic, temporally varying SPG waveform. Time-delay and frequency content signal decomposition strategies are applied to in vivo measurements in 16 volunteers and used to evaluate correlations between SPG-PPG parameters and age. We assess sensitivity to vascular tone (the degree of constriction or resistance in blood vessels [49]) by measuring four subjects undergoing two controlled physiologic perturbations: exercise and cold pressor challenges. To evaluate repeatability in this same context, a single patient was followed for seven measurements for each physiological perturbation. Our results clearly highlight the sensitivity of the time-delay and waveform features in response to these challenges, and underscore how this new approach can be used to assess important vascular parameters: microvascular perfusion, vascular tone and vessel resistance. In a broader context, these fundamental parameters are critical to understanding derived metrics such as blood pressure, cardiac output and fluid status.

Instrumental assembly
The ATSA instrument lends itself particularly well to being incorporated in a compact fingerclip. We obtained a commercially available ATSA device (Flowmet, LAS, Inc., Irvine, CA). This instrument takes advantage of highly diffuse speckle signals that have penetrated the skin at least several mm. The light source consists of a 785 nm laser diode while the detector consists of a 752-pixel x 480-pixel CMOS array. An aperture is placed in front of the detector to selectively filter the coherent signal from ambient light. The source and detector are placed opposite one another in a form factor practically identical to a commercially available pulseoximeter.
To clarify nomenclature, ATSA refers to the specific instrument setup used in this manuscript, while SPG refers the process of obtaining pulsatile speckle fluctuations due to the cardiac cycle. The signal directly recovered using SPG will be referred to as the SPG waveform or SPG signal, keeping with the same convention as PPG. Figure 1 shows how the SPG and PPG signals are extracted from raw data. Within the ATSA device there is a 785 nm laser diode placed opposite to a 752-pixel x 480-pixel CMOS camera. The camera acquires images of the diffuse speckle interference pattern at approximately 200 frames per second. On a frame-by-frame basis, a 7-pixel x 7-pixel sliding window filter is used to extract the mean intensity <I> and the standard deviation <σ> of the raw images. The single black and red squares underneath the <I> and <σ> signify that the 7pixel by 7-pixel sliding window array is condensed into single values. This data is used to construct the average intensity and the speckle contrast (K) matrices. K is the ratio of the standard deviation to the mean intensity. At this point, the average intensity array is used produce the PPG signal array using the following PPG equation: 50]. The average of the entire PPG array yields to the single PPG signal timepoint. At the same time the speckle contrast is used to produce the SPG signal array using the equation shown in Fig. 1. This equation has been validated experimentally to produce a metric that is linear with respect to volumetric flow [51]. Finally, this array is averaged to produce a single SPG signal timepoint. This process is repeated for each frame to construct the SPG and PPG time-series -all from the same raw data.   Figure 2(a) depicts how the sample, typically a digit on the subject's hands or feet, is placed within the finger clip such that it falls between the source and detector. As demonstrated in Fig. 2(b), light from the coherent source enters the tissue and is dispersed randomly by a combination of static and dynamic scatterers [52,53]. Static scatterers are comprised of matrix materials (e.g. collagen) and membranes belonging to stationary cells and organelles [53]. Dynamic scatterers consist of the membranes belonging to moving red blood cells [52]. Light passing through the tissue is also attenuated by absorbers contained in the tissue; the most important of these is hemoglobin, the oxygen-transporting molecule confined in the cytoplasm of red blood cells [53]. Dynamic scattering modulates the speckle pattern incident on the CMOS detector, producing the SPG signal ( Fig. 2(c)). Likewise, attenuation due to absorbers modulates light intensity, producing the PPG signal ( Fig. 2(d)). The pulsatile nature of blood pressure within the arteries and arterioles is well-established, but this pulsatility wanes downstream in the capillary and venous plexuses [49]. Therefore, it is likely that pulsatility in both attenuation and coherence originates from the arterial side of the vasculature. Pulsatility in the SPG waveform is due to variation in dynamic scattering related to volumetric flow, whereas pulsatility in the PPG waveform is due to variation in the concentration of absorbers caused by vessel expansion [1]. Thus, the SPG and PPG waveforms embody two pulsatile signals originating from different electromagnetic properties and are related to distinct physiological processes.

Theory
The SPG waveform in the black dotted box on the right side of Fig. 2(c) contains arrows indicating components of the cardiac cycle referenced later. The thick red arrow points to the systolic upstroke in which ejection of blood through the aorta rapidly increases. The black dotted arrow shows the systolic peak, the apex of blood flow. Lastly, the solid black arrow designates a secondary reflection, a reverberation caused by refractive index mismatches at vessel branch-points.

Data pro
In this work w how time-dela a beat-by-bea global peakcomputationa from the PPG least 120 hear  of discrete pea monics of the w roduce harmon the next heart or each ratio. I wn in Fig. 4 The post-occlusive reactive hyperemia (PORH) challenge consisted of occluding a subject's upper arm for three minutes using a pneumatic cuff pressurized to 215 mmHg [55,56]. The occlusion leads to a buildup of metabolic byproducts which vasodilate the arm causing a significantly increased blood flow when the cuff is depressurized. The state of relatively increased blood flow in PORH amplifies the PPG waveform. It should be noted that PORH was not used for its usual intended purpose of comparing the hyperemic state to baseline measurements for measuring endothelial function [57]. In this work, 16 subjects aged 23 to 81 (11 females and five males) were each measured with ATSA using a protocol consisting of a 5-minute baseline, 3-minute occlusion, and 5-minute post-release period. Intervals within the baseline and post-release periods were analyzed to obtain the physiological parameters discussed above.

Cold pressor challenge
The cold pressor challenge consisted of a subject submerging one hand in ice water for 30 to 90 seconds inducing peripheral vasoconstriction along with a slight increase in heart rate and blood pressure [58]. In this work, the ATSA device was attached to the subject's left index finger while a baseline measurement was acquired for two to three minutes. At this point the subject inserted his contralateral hand (right hand in this case) into ice water for as long as tolerable -usually 30 to 90 seconds -while data continued to be acquired on the nonsubmerged hand. This data was processed to calculate TD. In order to assess repeatability, this process was repeated on one subject seven times. In order to assess the effects in general, this test was performed on four subjects aged 25 -35; there were three males and one female. The study was carried out under a UC Irvine IRB approved protocol and informed consent was obtained (HS #2004-3626).

Exercise challenge
The exercise challenge consisted of 10 minutes of cardiovascular exercise on a stationary bicycle. The left index finger was measured with the ATSA for two minutes before and two minutes following the challenge. Cardiovascular exercise is known to decrease peripheral vasoconstriction especially in the superficial vascular beds [59]. It also increases blood pressure and cardiac output by increasing the heartrate and stroke volume. The decreased peripheral vasoconstriction is required to vent excess heat and the increased cardiac output is necessary to supply working muscle with oxygen and nutrients needed for heightened metabolic activity. Like the cold pressor challenge, repeatability was assessed with seven repeat measurements on one subject, and general effect was assessed with one measurement on four subjects.  Fig. 5(b), where the ATSA instrument shows significant linearity (R 2 = 0.98). For each 10-second measurement, the standard deviation of the flow index was less 0.5%. The specified flow range encompasses values observed in prior literature and in healthy subjects [60][61][62][63]. This suggests the ATSA output is linearly proportional to the volumetric flow rate in the physiological range of interest.    Figure 7 sho directly follow methods secti including a hi is a 23-year-o chosen as an dual-axis plot Fig. 7(b) sho maxima of th waveform. No much more pr resembles the TD of Subjec distribution w that Subject A ws the TD re wing the arteri ion). Subject A istory of smok old male with n archetypal exa t of a 4-second ws the same d he SPG wavefo otably, the PPG ronounced sign e SPG wavefor ct A. Figure 7(c was derived fr A has both a gr 7. Two Subject Co ed for Subject A. Fig. 6(a) Fig. 8(b) and

Change
consists of the shows the THR n THR and a b ct A) had les petitive structu he 300-second active hyperem diovascular ris pe-2 diabetes. S These two subj ts. Figure 7( Figure 10 presents SPG-PPG data obtained from a 29-year-old healthy male subject undergoing cold pressor and exercise challenges. Figure 10(a) contains a dual-axis plot of the SPG (blue) and PPG (red) waveforms during 5 seconds of a 2-minute baseline measurement. Figure 10(b) contains a 5-second interval of the cold pressor challenge. Compared with the baseline condition in Fig. 10(a), the PPG waveform in this example is more rounded and contains a steeper systolic upshot. This observation potentially explains why TD is increased in the cold condition. It is also noteworthy that systolic upshot the SPG-signal (blue) in Fig.  10(b) retains its steep slope.  Figure 10(c) shows a five-second interval acquired after 10 minutes of exercise. The postexercise PPG waveform has a more pronounced and narrow systolic peak and more closely resembles the SPG waveform than the baseline PPG waveform. In contrast, the baseline PPG waveform has a rounded-off plateau region near the apex. Thus, the location of the PPG peak at baseline is subtler and relatively delayed compared to post-exercise. Figure 11(a) summarizes repeatability testing of TD on a single subject performing exercise and cold pressor challenges. Each box represents seven repetitions/measurements, and each measurement is an average over approximately 60 heartbeats. For the box marked cold, each measurement consists of average TD at baseline subtracted from average TD during cold-shock (TD cold -TD baseline). Similarly, the box marked exercise contains measurements of baseline subtracted from post-exercise (TD post-exercise -TD baseline). Finally, the box marked baseline is derived from two sequential baseline measurements -one subtracted from the other (TD baseline 1 -TD baseline 2). This was done to emphasize sensitivity to dynamic changes associated with cold and exercise, and to minimize variability in TD among different measurements. The average increase in TD for the cold challenge was found to be 0.044 seconds, with a range of 0.031 to 0.068 seconds. The average change in TD post-exercise was −0.0578 seconds, with a range of -0.0110 to -0.1334 seconds. Notably, the differences between each pair of the three distributions (i.e., cold, exercise and baseline) are statistically significant according to Student's t-test. The p-value for the difference between the cold and exercise distributions is 0.0013, the p-value for the difference between cold and baseline distributions is 0.0031, and the p-value for the difference between exercise and baseline distributions was 0.0286.  Figure 11(b) summarizes the results from four subjects performing the exercise and coldpressor challenges. As in Fig. 11(a), each box shows how TD changes from baseline for each of the challenge among the individuals. As before, the box labeled "baseline" is the difference between paired 2-minute baseline measurements. The average increase in TD for the cold challenge was found to be 0.0294 seconds, with a range of 0.0051 to 0.0671 second. The average decrease in TD for the exercise challenge was found to be 0.0430 seconds, with a range of 0.0238 to 0.0654 seconds.

Discussion
This work introduces methods for extracting physiological parameters from the SPG-PPG signal obtained with ATSA. We demonstrate how the dual SPG-PPG signal is derived from the same raw data obtained with an instrument consisting of a CMOS camera and coherent light source with the SPG originating from flow pulsatility and the PPG originating from volumetric pulsatility. We demonstrate that both waveforms are offset in time and introduce an algorithm characterizing time delay between the systolic peaks. Additionally, we introduce a framework for analyzing SPG structure by calculating harmonic content on a pulse-by-pulse basis. Two signal decomposition methods (time delay and harmonic content) are applied to in vivo measurements to demonstrate physiological significance, with our analysis revealing age correlations for both. We proceed to demonstrate changes in time delay associated with exercise (decreased time delay) and cold pressor challenges (increased time delay).
The correlation between age and extracted SPG-PPG parameters suggest that these methods are sensitive to arterial stiffness and vascular aging. Aging arteries undergo mechanical changes resulting in decreased compliance, referred to as arterial stiffening. Thus, the correlations between SPG-PPG signal parameters and age indicate potential sensitivity to these mechanical changes. This implies the methods introduced in this work may be used for noninvasive vascular assessment at early ages as a predictive measure of arterial health outcomes. We observe that the age-related time delay increases are due to pulse-shape changes in the PPG signal caused by decreased arterial compliance. In fact, these changes in PPG signal shape have been previously observed in the field and reported as an increase in crest time [1,64]. Our work presents this phenomenon in a new light using the SPG waveform as the reference signal, rather than relying on the PPG waveform which is often subject to artifacts and low signal to noise ratio (Fig. 6) [65,66]. The mechanism for changes in SPG signal and harmonic content comes from a distinct aspect of vascular physiology. Refractive index mismatch at vessel bifurcations increases with atherosclerosis [67]. This results in the reflection of high-frequency harmonic components as the pulse propagates down the arterial network explaining why there is less harmonic content seen in the older individuals. Our analysis methods could potentiate a noninvasive means of characterizing vascular age for cardiovascular risk stratification. Additionally, it could be applied to early detection of atherosclerosis for early medical and lifestyle interventions.
In this work, harmonic content was characterized using THR rather than other harmonic ratios for two main reasons. First, preliminary analysis demonstrated THR magnitude corresponds with the amount of pulsatile structure; if there are numerous or prominent secondary reflections then THR tends to be larger. Second, the regression of THR onto subject age showed a much stronger correlation than other harmonic ratios. Because of these characteristics, THR was used exclusively in this work to interrogate harmonic content.
The changes in TD caused by cold pressor and exercise challenges indicate a completely different functionality related to vascular tone. Decreased temperature enhances vascular tone to preserve thermal homeostasis by retaining heat in the body's core, whereas increased temperature decreases vascular tone to shed excess heat in peripheral circulation. Overall, the changes in time delay observed in the post-exercise and cold pressor conditions demonstrate that this parameter is sensitive to vascular tone. This is particularly important because vascular resistance is a confounding factor in calculating blood pressure, cardiac output and fluid status [68] meaning that this technique could provide solutions for these medical applications.
This study only partially elucidates the implications of information embedded in the dual SPG-PPG signal. The cardiac pulse is progressively modified by mechanical properties of the underlying vasculature as it propagates down the arterial network. As light transmitted from a single coherent source interacts with the tissue, the pulsatile characteristics of the underlying cardiovascular system are ultimately contained in the SPG-PPG signal. We only present two ways of processing these signals, but there are many other methods that have not yet been explored. Examples include analyzing systolic upstroke time, dicrotic notch prominence, and other harmonic content ratios. Further, almost every processing technique applied to PPG alone could be applied to this new framework. The SPG-PPG signal therefore constitutes a promising technology for continued research.
Overall, this study provides initial data to support the potential clinical significance of the SPG-PPG waveform. The SPG is distinct and possibly more robust than the PPG alone, as shown by the fact that the PPG waveform becomes weak and distorted with vascular constriction, while under these conditions the SPG signal retains its strength and pulsatility (Fig. 6). This may be important because medical applications such as non-invasive cardiac output monitoring have failed due to issues with the PPG signal quality. We have also shown that the SPG and PPG signals provide different and complementary information, specifically flow speed pulsatility from the SPG and vascular expansion pulsatility from the PPG signal. The TD and pulse-shape characteristics of the SPG and PPG waveforms are evidence that they provide unique information, which is further supported by their differential response to contrasting physiological challenges. We conclude that wearable, affixed transmission LSI sensors can generate highly complementary SPG and PPG signals. They potentially provide a single, non-invasive platform for assessing vascular dynamics. Because of their relative simplicity and ease-of-use, ATSA devices may create new possibilities for non-invasive diagnostics and therapeutic guidance in circulatory diseases.