1.

Introduction

The arterial network of the cardiovascular system (CVS) is composed of two subsystems, macro- and microcirculation, which interact to enable an optimal adaptation to various physiologic disturbances. With age and/or risk factors such as hypertension and pathologies such as diabetes, modifications appear in both the macro- and microcirculation subsystems (see, e.g., Refs. 1 and 2). The monitoring and analyses of large vessels’ characteristics (such as the arteries) provide a good vital biomarker to assess the status of macrocirculation. Thus, pulse-wave velocity (PWV) is a measure of arterial stiffness. The latter is considered an important predictor of cardiovascular events.^3–5 On the other hand, assessment of microvascular blood flow has been recognized as important for the follow-up of pathologies such as diabetes or Raynaud’s phenomenon but also to analyze the effects of aging.^6,7 Thus, it has been shown that age changes the morphology and quantification of the cutaneous microvasculature.^6,8 Moreover, age is a primary risk factor for cardiovascular disease.⁹

Different optical techniques have emerged to monitor microvascular blood flow.^10–13 Laser speckle contrast imaging (LSCI) has the advantage of being noninvasive, contactless, highly reproducible, leading to high temporal and spatial resolution images of the microvascular blood perfusion, and requiring low-cost devices.^14–16 Moreover, when assumptions are made on the moving scatterers’ velocity profile, LSCI data provide information on red blood cells’ velocity values.¹⁷

Studying the relationship between macro- and microcirculation may lead to an early estimation of many disorders in the CVS. Several authors emphasized that the two subsystems, macrocirculation and microcirculation, must be simultaneously taken into account.^1,18

In this work, we propose to analyze the impact of age on data recorded simultaneously from the macrocirculation (PWV) and the microcirculation (LSCI) in healthy subjects. The analysis in healthy subjects is important as it is the first step before an analysis in pathological subjects. Our goals are, therefore, the following: study the evolution with age of (1) the possible correlation between PWV and LSCI microvascular perfusion; (2) the possible correlation between PWV and microvascular red blood cells’ velocity extracted from LSCI data; and (3) the possible correlation between microvascular perfusion and microvascular red blood cells’ velocity extracted from LSCI data. Furthermore, for the three above mentioned items, three physiological states are analyzed: rest, vascular occlusion, and post-occlusive hyperaemia peak.

In what follows, we first present the LSCI theoretical background and the way to estimate moving blood cells’ velocity from LSCI data. Then the measurement procedure used to acquire data from the macro- and the microcirculation subsystems is described. The processing algorithm used to process PWV and LSCI data is then detailed. Finally, we present our findings and discuss them.

2.

Materials and Methods

2.1.

Laser Speckle Contrast Imaging

2.1.1.

LSCI Principles

LSCI relies on a laser source and a camera. When a laser source diffuses a light over the desired area of tissue, the photons of the laser light are backscattered by both moving (such as red blood cells) and static (such as skin) scatterers. The backscattered light forms an interference pattern of bright and dark pixels, called speckle, on the camera. The movements in the illuminated sample (i.e., movements of blood cells) lead to temporal changes in the speckle pattern. Due to the exposure time $T$ of the camera, a blurring in the speckle pattern is obtained. To quantify the degree of blurring, the spatial contrast $K$ is used and computed as¹⁷

Eq. (1)

$$K=\frac{{\sigma }_{\mathrm{s}}}{\langle I\rangle },$$

where ${\sigma }_{\mathrm{s}}$ refers to the spatial standard deviation in a small region around a pixel of the speckle raw data, whereas $\langle I\rangle$ is the mean intensity around that pixel. A speckle contrast $K$ close to 1 indicates that there is no blurring of the speckle pattern and, therefore, no motion. Alternatively, $K$ values closer to 0 mean that the scatterers are moving fast enough to blur all the speckles.

The value of speckle contrast $K$ can be obtained directly by computing Eq. (1) from the pixels in a surrounding $N\times N$ window. The size of the window at which the speckle contrast is computed is critical:¹⁹ the statistics are compromised with too few pixels. By opposition, the spatial resolution is sacrificed with too many pixels. It has been reported that a square window of size $5\times 5{\text{pixels}}^2$ or $7\times 7{\text{pixels}}^2$ would be convenient.¹⁹ To obtain a two-dimensional contrast image reflecting the local motion due to blood flow, a sliding window is used to move along the raw speckle image. Perfusion is then computed from the inverse of the contrast value (see an example of a perfusion image in Fig. 1).

Fig. 1

Experimental perfusion image ($118\times 234{\text{pixels}}^2$) of forearm in a healthy subject obtained with LSCI technique.

From contrast $K$ values, moving blood cells’ velocity values can be extracted provided assumptions are made on the velocity profile of the moving scatterers. This is due to the fact that the variance ${\sigma }_{\mathrm{s}}^2(T)$ of the spatial intensity distribution in a time-averaged speckle pattern with an integration time $T$ is linked to the autocovariance $C_{\mathrm{t}}(\tau )$ of the temporal fluctuations in the intensity fluctuations of a single speckle:²⁰

Eq. (2)

$${\sigma }_{\mathrm{s}}^2(T)=\frac{2}{T}{\int }_0^T(1-\frac{\tau }{T})C_{\mathrm{t}}(\tau )\mathrm{d}\tau ,$$

where $C_{\mathrm{t}}(\tau )$ depends, among others, on the velocity distribution of the scattering particles.^20,21

The choice of velocity profile has a major effect on the relation between the speckle contrast $K$ and velocity values.^22,23 Thus, it has been shown that the relation between speckle contrast $K$ and the ratio ${\tau }_{\mathrm{c}}/T$ (where ${\tau }_{\mathrm{c}}$ is the correlation time of the intensity fluctuations) is critically related to the velocity distribution.²³ Several authors used the Lorentzian model to link the motion of the scatterers and the speckle contrast $K$.^17,20,24 This is appropriate for a Brownian motion (unordered flow).

2.1.2.

Computation of Moving Blood Cells Velocity from LSCI Data

The autocovariance $C_{\mathrm{t}}(\tau )$ of the temporal fluctuations in the intensity fluctuations of a single speckle is defined as

Eq. (3)

$$C_{\mathrm{t}}(\tau )={\langle [I(t)-{\langle I\rangle }_{\mathrm{t}}][I(t+\tau )-{\langle I\rangle }_{\mathrm{t}}]\rangle }_{\mathrm{t}},$$

where ${\langle \rangle }_{\mathrm{t}}$ indicates a time-averaged quantity. We also have

Eq. (4)

$$g_2(\tau )=1+\frac{C_{\mathrm{t}}(\tau )}{{\langle I\rangle }_{\mathrm{t}}^2},$$

where $g_2(\tau )$ is the intensity temporal autocorrelation function. We can also write (Siegert relation):

Eq. (5)

$$g_2(\tau )=1+\beta {|g_1(\tau )|}^2,$$

where $g_1(\tau )$ is the electric field temporal autocorrelation function and $\beta$ accounts for the loss of correlation related to the ratio of the detector (or pixel) size to the speckle size and to polarization. From Eqs. (2), (4), and (5), and assuming ergodicity²⁰ (we can, in this case, replace the time average by the ensemble average), we have to solve the equation

Eq. (6)

$$K^2=\frac{2\beta }{T}{\int }_0^T(1-\frac{\tau }{T}){|g_1(\tau )|}^2\mathrm{d}\tau ,$$

to obtain an analytical expression of the contrast $K$ for the velocity distributions mentioned above.

The Lorentzian profile is the most commonly used profile to determine the theoretical expression of contrast $K$.^{17,20,22,24,25} If the moving scatterers are assumed to follow a Lorentzian distribution, we have (see, e.g., Ref. 25)

Eq. (7)

$$g_1(\tau )=\exp (-\frac{|\tau |}{{\tau }_{\mathrm{c}}}),$$

where ${\tau }_{\mathrm{c}}$ is the correlation time of the intensity fluctuations. Solving Eq. (6), the contrast $K$ for a Lorentzian distribution can be written as

Eq. (8)

$$K_{\mathrm{Lorentzien}}={\beta }^{1/2}{\{\frac{1}{x}+\frac{1}{x^2}[\exp (-2x)-1]\}}^{1/2},$$

where $x=T/{\tau }_{\mathrm{c}}$.

The relationship between correlation time ${\tau }_{\mathrm{c}}$ and moving blood cells in the microcirculation is assumed to be an inverse one. Therefore, from the experimental values of contrast $K$, the velocity of moving scattering particles can be calculated as¹⁷

Eq. (9)

$$v_{\mathrm{c}}=\frac{\lambda }{2\pi {\tau }_{\mathrm{c}}},$$

where $\lambda$ is the laser wavelength.

2.1.3.

Effect of Static Scattering

The expression of speckle contrast $K$ is used to determine the velocity of the moving blood cells in the microcirculation. Nevertheless, Eq. (8) does not take into account the presence of static scatterers (such as bones, skin, and skull). It has been reported that if the static scatterers effect is not considered, then it results in an underestimation of the spatial and temporal variations in the sample dynamics.^8,26 Moreover, some authors mentioned that the computation of blood flow velocity from LSCI data leads to erroneous values when the presence of static scatterers is not taken into account.²⁷ Only a few authors studied the effect of static scatterers on the estimation of moving blood cells velocity from the expression of speckle contrast $K$ (see, e.g., Ref. 28).

By assuming a Lorentzian velocity profile for the moving scatterers, and when static scatterers are taken into account, the expression of speckle contrast $K$ becomes²⁸

Eq. (10)

$$K_{\mathrm{Lorentzien}}={\beta }^{1/2}{[{\rho }^2\frac{\exp (-2x)-1+2x}{2x^2}+4\rho (1-\rho )\frac{\exp (-x)-1+x}{x^2}+{(1-\rho )}^2]}^{1/2}+C_{\text{noise}},$$

where $x=T/{\tau }_{\mathrm{c}}$, $\rho =I_{\mathrm{f}}/(I_{\mathrm{f}}+I_{\mathrm{s}})$ with $I_{\mathrm{f}}$ the time-averaged intensity of the fluctuating dynamically scattered light, $I_{\mathrm{s}}$ the intensity of the statically scattered light, and $C_{\text{noise}}$ a measurement noise such as shot noise or camera readout noise.²⁸ From this latter expression and using Eq. (9), moving blood cells’ velocity can be computed when static scatterers are considered. Moreover, in the laser speckle contrast imager used for the experimental acquisition (see below), we have (from Perimed documentation)

Eq. (11)

$$\text{Perfusion}\sim \frac{1}{K}-1.$$

Using Eqs. (9) and (10), we thus obtain the relation linking the perfusion and the velocity of moving scattering particles

Eq. (12)

$$\text{Perfusion}\sim {\{{\beta }^{1/2}{[{\rho }^2\frac{\exp (-4\alpha v_{\mathrm{c}})-1+4\alpha v_{\mathrm{c}}}{8{(\alpha v_{\mathrm{c}})}^2}+4\rho (1-\rho )\frac{\exp (-2\alpha v_{\mathrm{c}})-1+2\alpha v_{\mathrm{c}}}{4{(\alpha v_{\mathrm{c}})}^2}+{(1-\rho )}^2]}^{1/2}+C_{\mathrm{noise}}\}}^{-1}-1,$$

where $\alpha =\pi T/\lambda$.

In order to obtain an experimental $\rho$ value, the following expression was used²⁹

Eq. (13)

$$\rho =1-{\beta }^{-1/2}{[\frac{\langle I_1{I}_2\rangle }{\langle I_1\rangle \langle I_2\rangle }-1]}^{1/2},$$

where $I_1$ and $I_2$ are two sequential intensity images, $\langle I\rangle$ denotes the spatial averaging of the intensity over a selected area containing $N$ pixels, and $\langle I_1{I}_2\rangle =(1/N){\sum }_{i=1}^N{I}_1(x_i)I_2(x_i)$.

3.

Results and Discussion

When using Eq. (13), the values of $\rho$ obtained for each population are mentioned in Table 1. We observe that the average value of $\rho$ for the younger group does not differ significantly from the one obtained with the elderly subjects. With these $\rho$ values and during post-occlusive reactive hyperaemia, we observe a strong correlation between perfusion and age and between velocity and age, as well as between PWV and age at rest ($R^2$ equal to 0.67, 0.63, 0.91, respectively). For $\rho =1$ in the computation of the velocity values, we also obtain a strong correlation between velocity and age ($R^2$ equal to 0.72). Our first findings are, therefore, that the studied parameters of microcirculation and macrocirculation are correlated with age.

Table 1

Average values of ρ computed from Eq. (13) in two groups (young and elderly) of eight subjects each (see text for details).

Moreover, Fig. 2 shows the evolution of microvascular blood cells’ velocity with perfusion for the three physiological states (rest, vascular occlusion, and post-occlusive reactive hyperaemia), and for all the subjects when the static scatterers effect is neglected ($\rho =1$). From this figure, we observe that moving blood cell velocity values increase with perfusion values. We also note a strong correlation between them ($\text{Velocity}=0.21\times {\text{Perfusion}}^{1.27}$, $R^2=0.98$). Moreover, we note that the highest moving blood cell velocity values (and, therefore, the highest perfusion values) are obtained for the aged subjects during post-occlusive reactive hyperaemia. When $\rho =1$, strong correlations are also obtained between perfusion values obtained during post-occlusive reactive hyperaemia and PWV values (see Fig. 3), as well as between moving blood cell velocity obtained during post-occlusive reactive hyperaemia and PWV values (see Fig. 4): for perfusion values recorded during post-occlusive reactive hyperaemia and PWV, we have $\text{Perfusion}=-2.4\times {\mathrm{PWV}}^2+46.3\times \mathrm{PWV}-74.9$, $R^2=0.76$; for moving blood cells’ velocity recorded during post-occlusive reactive hyperaemia and PWV, we have $\text{Velocity}=-4.4\times {\mathrm{PWV}}^2+72.3\times \mathrm{PWV}-184.7$, $R^2=0.77$. For the other physiological states (rest and biological zero), the correlation is much lower ($R^2<0.3$).

Fig. 2

Moving blood cells’ velocity values and perfusion values computed from LSCI data for 16 subjects, at rest, during vascular occlusion and post-occlusive reactive hyperaemia (see text for details). Curve shows line of best fit by least squares. The value of $\rho$ has been set to 1.

Fig. 3

LSCI perfusion values, at rest, during vascular occlusion and post-occlusive reactive hyperaemia and PWV values, for 16 subjects (see text for details). Curve shows line of best fit by least squares for data recorded during post-occlusive reactive hyperaemia.

Fig. 4

Moving blood cells’ velocity values computed from LSCI data, at rest, during vascular occlusion and post-occlusive reactive hyperaemia and PWV values, for 16 subjects (see text for details). Curve shows line of best fit by least squares for data recorded during post-occlusive reactive hyperaemia. The value of $\rho$ has been set to 1.

In this study, we also studied the effect of static scatterers on velocity values of moving blood cells computed from LSCI data. The mean velocity values obtained with the $\rho$ values computed from Eq. (13) for each subject are mentioned in Table 2. Our results show that static scatterers have an effect on the velocity values of moving scatterers computed from LSCI data, at rest, during vascular occlusion and during post-occlusive reactive hyperaemia: velocity values of moving scatterers increase when the static scatterers’ effect is taken into account (in this case $\rho <1$). Furthermore, we can observe that at rest and during vascular occlusion, the mean velocities’ values found for young subjects are lower than the ones obtained for elderly subjects. This is true when the static scatterers effect is taken into account and when it is not. Nevertheless, these differences are not statistically significant. From the literature, the value of basal blood flow (product of velocity by concentration of moving blood cells) for young and elderly subjects remains debatable (see, e.g., Refs. 40 and 41). Alternatively, during post-occlusive reactive hyperaemia, there is an obvious difference in the mean velocity values between young and elderly people (see Table 2). Thus, the mean velocities obtained from the elderly group are statistically higher than the ones obtained from the young people. This is true when the static scatterers’ effect is taken into account and when it is not. This may be due to the stiffness of vessels that increases with age. PWV is a measure of arterial stiffness⁴² and it is recognized that it increases with age. Our results regarding PWV are, therefore, in accordance with what was expected. Because stiffness is higher for aged people, the revascularization of vessels after the vascular occlusion (post-occlusive reactive hyperaemia) may lead to higher moving blood cells’ velocity values.

Table 2

Mean velocity results (μm/s) of moving scatterers computed from LSCI data. Three physiological states are analyzed: 1 min at rest, 1 min during biological zero and 3 s during post-occlusive reactive hyperaemia peak. The results are obtained from two groups (young and elderly) of eight subjects each when a Lorentzian velocity profile is assumed for moving scatterers.

The effect of static scatterers on the correlation $R^2$ values between moving scatterers’ velocity and PWV is shown in Table 3. From this table, we observe that the correlation values do not vary much in the presence or absence of static scatterers.

Table 3

Correlation values computed between velocity and PWV when LSCI data are recorded at rest, during biological zero (BZ) and during a post-occlusive reactive hyperaemia (PORH). PWV is recorded at rest. See text for details.

To the best of our knowledge, no other group has worked on the impact of aging on moving blood cells’ velocity computed from LSCI data. Moreover, our work is the first one to simultaneously assess LSCI perfusion, moving blood cells’ velocity and PWV. This analysis has been performed on two populations: a younger and an older one. Skin aging has been thoroughly studied and is still the subject of many works (see, e.g., Ref. 43).

Age-related changes play an important role in the pathogenesis of many diseases.^44–47 Previous studies in elderly subjects suggested impairments of microvascular reactivity upon aging.^48,49 With aging, a number of hypoxia- and metabolism-related changes occur.^40,50,51 These changes are due to alterations in the microcirculation. Moreover, as mentioned recently,⁴⁸ the number of functioning capillaries diminishes with age, and phenomena such as vascular rarefaction, appearance of zones of complete vascular obliteration, irregular caliber of microvessels,^52–54 and inhibition of the processes of angiogenesis⁵⁵ may appear. As pointed out by Gates et al.,⁵⁶ capillary rarefaction may be due to vessel destruction, impaired angiogenesis, and impaired vasculogenesis. Oxidant stress may also contribute to capillary rarefaction by inducing endothelial cell apoptosis and/or reducing the nitric oxide needed for vascular budding and stimulation of vascular endothelial growth factor. Aging is accompanied by suppression of the endothelial function and cellular metabolism, degeneration of the sensor and sympathetic innervation.^57,58 Moreover, it has been reported that aging may involve alterations in nitric oxide, prostanoid, endothelium-derived hyperpolarizing factor(s), and endothelin-1 pathways.⁵⁶ Aging also leads to a degradation of a number of extracellular matrix proteins, including collagen and elastic fibers.^59,60 The study of post-occlusive hyperaemia is, therefore, of importance as it is mediated by two major mediators: sensory nerves and endothelium-derived hyperpolarizing factors. Furthermore, we have to note that skin thickness could vary with age.⁶¹ Thus, the stratum corneum is generally accepted to maintain its thickness during aging. However, dermal, epidermal, and whole skin thickness changes are controversial. Ultrasound reveals the appearance of a subepidermal low echogenic band that thickens with age (due to changes in collagen structure), especially in environmentally exposed areas.⁶¹ Some studies also indicate the presence of an echogenic band in the lower dermis which thins with increased age. However, the whole dermis appears to become more echogenic in elderly people.⁶¹

Studies have reported an attenuated vasodilator response of skin microcirculation to a variety of stimuli with age.^62–65 This attenuation is thought to be the result of endothelial dysfunction.⁵⁶ Thus, Tikhonova et al. recently reported a higher increase of perfusion in young people compared to aged subjects after an occlusion removal.⁴⁸ Hagisawa et al. reported the same conclusion some years before.⁶² These findings are different from ours: during post-occlusive reactive hyperaemia, we found the highest perfusion values for the aged subjects. The discrepancy may be explained by differences in at least three parameters: (1) to monitor microvascular blood flow Tikhonova et al., as well as Hagisawa et al., employed a different technique from the one used in our work: they used laser Doppler flowmetry while we used LSCI. These two techniques do not probe the same volume of tissue (laser Doppler flowmetry probes deeper than LSCI);⁶⁶ (2) the age ranges studied are different: Tikhonova et al. studied a group composed of people between 19 and 30 years old and another group composed of people between 30 and 60 years old.⁴⁸ Hagisawa et al. studied a group composed of people between 22 and 27 years old and another group composed of people between 62 and 68 years old.⁶² In our younger group, people were between 20 and 30 years old, whereas in our older group, people were between 50 and 62 years old; and (3) the effect of fitness. It has been reported that effect of fitness has an influence on the peak of post-occlusive reactive hyperaemia: Tew et al. found a higher post-occlusive reactive hyperaemia peak in aged fit participants than in active young subjects, and the post-occlusive reactive hyperaemia peak in active subjects was higher than for sedentary aged subjects.⁴¹ The latter study was conducted through laser Doppler flowmetry signals. Moreover, microvessel function is worse in older sedentary compared with older active or younger men, and this is attributable to impaired nitric oxide signalling.⁶⁷ Other factors may also explain the differences between our results and the ones of Tikhonova et al., as medical drugs (e.g., at age-specific and hormone replacement therapy) and other vasoactive substances. Furthermore, it has previously been reported that gender has an impact on microcirculation,⁶⁸ which was not studied in our work.

In our analysis, three physiological states have been studied: rest, vascular occlusion, and post-occlusive reactive hyperaemia. The biological zero has been the subject of many studies (see, e.g., Refs. 69 and 70). However, its origin is still not completely known: it is thought to be generated by Brownian motion of the macromolecules within the interstitium⁷¹ or other phenomena related to the function of the autonomic nervous system.⁷² Nevertheless, as recently pointed out,³⁴ this definition may require clarification.

4.

Conclusion

This study demonstrates that, in healthy subjects, perfusion and moving blood cells’ velocity, two correlated and age-related parameters of microcirculation, are correlated with PWV, a marker of arterial stiffness at the macrovascular level. Future works should be performed in diseased patients. It might be of interest to study these relationships to show whether or not these correlations still exist to try to define new criteria for different diseased states. The latter could lead to modifications in patients’ treatment. For example, it has been shown in diabetic patients that PWV is predictive of cardiovascular mortality.⁷³ It could be of interest to study whether microcirculation parameters can also predict this risk and to assess the time-course of the modifications of microcirculation parameters. Do these modifications appear earlier than PWV modifications? Do microcirculation markers (perfusion or moving blood cells’ velocity or both) assessed by optical laser devices predict better the cardiovascular mortality? Do clinicians have an interest in performing both methods to characterize patients’ risk?