2.1.

Subject Recruitment and PA Imaging Protocol

The human subject study was approved by the Research Ethic Board of Ryerson University (REB 2017-040). Healthy subjects were recruited based on the inclusion criteria listed in Table 1.

Table 1

Subject criteria and information.

The subject sat down on a chair and immersed their left arm in a degassed warm (36°C) water bath (Fig. 1). PA imaging was performed with a linear-array probe equipped PA imaging system (Vevo LAZR; LZ250-21 MHz of central frequency, 13 to 24 MHz of bandwidth and 256 elements, FUJIFILM Visualsonics, Toronto, Ontario, Canada)⁴⁴ at the near-infrared wavelength (700, 750, 800, 850, and 900 nm). The pulsed-wave Doppler velocity ($V$) was measured to locate the radial artery, and the measurement system was then switched to the PA imaging mode. A PA B-mode image was acquired to measure the pulsatile blood flow in the radial artery for 10 s at each optical wavelength ($\lambda$). The total time that subjects’ arms were immersed in water was $<5\min$. The time trace was acquired for each single wavelength then time-shifted to create a combined multispectral dataset in postprocessing. The detailed methods were described in our previous study.^42,45

Fig. 1

Photograph of PA imaging system and a PA probe positioned over the region of the interest close to the wrist of a subject.

2.2.

Data Acquisition and Postprocessing: V, Pa, and sO₂

A representative coregistered US (gray scale) and PA (color scale) image of the radial artery was shown in Fig. 2(a). The second quarter of the full field of view was chosen as a region of interest (ROI) to maximize the PA image acquisition rate to 20 Hz. To avoid boundary-buildup or edge detection artifact from the upper and lower vessel walls, the upper and lower boundaries of the ROI were chosen at 10% margin within the vessel lumen. In the coregistered US image, the upper and lower vessel walls were tracked at each frame. For each wavelength and each subject, 200 frames of US images ($20\text{frame}/\mathrm{s}$) were analyzed to track the vessel wall. From the upper and lower walls tracked, the upper and lower boundaries for the ROI were computed by 10% margin within the vessel lumen. For example, if the locations of two walls were at 1 and 2 mm in depth (corresponding to $t_{\mathrm{U}}$ and $t_{\mathrm{L}}$, respectively), the locations of two boundaries of ROI were 1.1 and 1.9 mm in depth (corresponding to $t_1$ and $t_2$, respectively) as shown in Fig. 2(a). The horizontal width of the ROI was fixed as a second quarter of the full field of view, whereas the vertical length (related to the vessel diameter) of the ROI was dependent on each frame. The VevoLAZR PA imaging system provides access to the prebeamformed RF data for all 256 transducer elements. These data are then beamformed postacquisition, and the amplitude of each signal is used to reconstruct the PA images shown in Fig. 2(a). The ROI is then selected from this reconstructed PA image. From beamformed radiofrequency (RF) PA signals (64 out of 256 elements) in the ROI, the PA power ($Pa$) was computed by taking an average of the root-mean-square of each RF signal for each $\lambda$, and addressed by

Fig. 2

(a) Representative US (gray scale) and PA, PA (color scale) image of the radial artery. The PA RF power ($Pa$) from the ROI was computed by taking an average of the root-mean-square of each RF signal for each optical wavelength ($\lambda$). (b) Representatives of $Pa$ (900-nm-thick blue and 700-nm-thin blue), oxygen saturation (${\mathrm{sO}}_2$, red), and the Doppler velocity ($V$, black). The subscripts max and min represent the maximum (denoted as a circle) and minimum (denoted as a diamond) value for each parameter.

The minimum ($P{a}_{\min }$ and ${\mathrm{sO}}_{2\text{-}\min }$) and maximum ($P{a}_{\max }$ and ${\mathrm{sO}}_{2\text{-}\max }$) values of $Pa$ and ${\mathrm{sO}}_2$ were, respectively, averaged for 10 s to compare the variation in $Pa$ and ${\mathrm{sO}}_2$ during the pulsatile blood flow as a function of $\lambda$ and age. Two sample $t$-test was conducted using built-in MALTAB function “ttest2.m,” in terms of the ${\mathrm{sO}}_2$ and $Pa$ changes versus age group.

3.1.

PA Power versus Optical Wavelength

The PA magnitude was dependent on the optical $\lambda$ as well as the study subjects. To compare this parameter, the $P{a}_{\max }$ and $P{a}_{\min }$ for all $\lambda$ were normalized to the $P{a}_{\min }$ at an isosbestic point 800 nm (the point at which HbO and HbD have the same optical absorption) for each subject. The $Pa$ dependence on the $\lambda$ for each subject age group is shown in Fig. 3(a). The $P{a}_{\min }$ at 800 nm was set to 0 dB as a reference [green circles in Fig. 3(a)]. Both $P{a}_{\max }$ and $P{a}_{\min }$ increased with $\lambda$ for all groups, as shown in Fig. 3(a). In addition, the difference between $P{a}_{\min }$ and $P{a}_{\max }$ ($\mathrm{\Delta }Pa$) increased with the $\lambda$. The average values of $P{a}_{\max }$, $P{a}_{\min }$, and $\mathrm{\Delta }Pa$ at 700, 800, and 900 nm for all groups were tabulated (Table 2). The $Pa$ difference between $P{a}_{\min }$ and $P{a}_{\max }$ at 800 nm ($\mathrm{\Delta }P{a}_{800}$) for all subjects is shown in Fig. 3(b)-left. The error bars represent the standard deviation from the number of cycles of the pulsatile blood flow for each subject. The box and whisker plot based on the average values of $\mathrm{\Delta }P{a}_{800}$ of each subject for age groups is shown in Fig. 3(b)-right. The $\mathrm{\Delta }P{a}_{800}$ increased with age, i.e., 1.8, 2.2, and 2.7 dB for the age group of subjects in their 20s, 30s, and 40s, respectively.

Fig. 3

(a) PA power ($Pa$) as a function of optical wavelength for each group of age 20s, 30s, and 40s. The thick and thin blue lines represent $P{a}_{\max }$ and $P{a}_{\min }$, respectively. The thick red line represents the variation from $P{a}_{\min }$ to $P{a}_{\max }$ ($\mathrm{\Delta }Pa$). The green circle indicates the isosbestic point (HbO and HbD have the same optical absorption). The error bars represent the standard deviation from the number of cycles of the pulsatile flow for all subjects, i.e., 75, 39, and 38 cycles for each age group, respectively. (b) The $Pa$ difference between $P{a}_{\min }$ and $P{a}_{\max }$ at an isosbestic point 800 nm ($\mathrm{\Delta }P{a}_{800}$) for all subjects and the corresponding box and whisker plot (based on the average value of $\mathrm{\Delta }P{a}_{800}$ for each subject) for each age group. The error bars represent the standard deviation from the number of cycles of the pulsatile flow for each subject. $ns$: not significant, *** : $p\le 0.001$, **** : $p\le 0.0001$.

Table 2

The values of minimum (Pamin) and maximum (Pamax) of PA power and the difference between Pamin and Pamin (ΔPa) at 700, 800, and 900 nm for all groups. The error bars represent the standard deviation from the number of cycles of the pulsatile flow for all subjects.

3.2.

sO₂ versus Subject Age

The ${\mathrm{sO}}_2$ estimated from $P{a}_{\min }$ (${\mathrm{sO}}_{2\text{-}\min }$) and $P{a}_{\max }$ (${\mathrm{sO}}_{2\text{-}\max }$) for each group of age 20s, 30s, and 40s were shown in Fig. 4, respectively. The average ${\mathrm{sO}}_{2\text{-}\max }$ for subjects in their 20s, 30s, and 40s were 98.7%, 97.2%, and 96.7%, respectively [Fig. 5(a)]. This indicates that the difference in ${\mathrm{sO}}_{2\text{-}\max }$ between the youngest and oldest subjects was 2.0% ($p\le 0.01$). On the other hand, the average ${\mathrm{sO}}_{2\text{-}\min }$ for subjects in their 20s, 30s, and 40s were 97.1%, 94.7%, and 93.0%, respectively [Fig. 5(a)]. The ${\mathrm{sO}}_{2\text{-}\min }$ difference between the 20s and 40s groups was 4.1% ($p\le 0.0001$). The ${\mathrm{sO}}_2$ difference between ${\mathrm{sO}}_{2\text{-}\min }$ and ${\mathrm{sO}}_{2\text{-}\max }$ ($\mathrm{\Delta }{\mathrm{sO}}_2$) during a pulsatile cycle increased with age, i.e., 1.6%, 2.5%, and 3.8% for the age group of subjects in their 20s, 30s, and 40s, respectively [Fig. 5(b)].

Fig. 4

Oxygen saturation (${\mathrm{sO}}_2$) for all subjects—subjects in their 20s (blue dots), subjects in their 30s (red diamonds), and subjects in their 40s (green square). The ${\mathrm{sO}}_2$ was computed for $P{a}_{\max }$ (${\mathrm{sO}}_{2\text{-}\max }$) and $P{a}_{\min }$ (${\mathrm{sO}}_{2\text{-}\min }$), respectively. The error bars represent the standard deviation from the number of cycles of the pulsatile flow.

Fig. 5

(a) Oxygen saturation (${\mathrm{sO}}_2$) for each group based on the average value of ${\mathrm{sO}}_2$ of each subject. The thick (${\mathrm{sO}}_{2\text{-}\max }$) and thin (${\mathrm{sO}}_{2\text{-}\min }$) symbols represent the ${\mathrm{sO}}_2$ computed using the two wavelength (700 and 900 nm) methods at $P{a}_{\max }$ and $P{a}_{\min }$, respectively. (b) The difference between ${\mathrm{sO}}_{2\text{-}\max }$ and ${\mathrm{sO}}_{2\text{-}\min }$ ($\mathrm{\Delta }{\mathrm{sO}}_2$) for each group. $ns$: not signigicant, *: $p\le 0.05$, **: $p\le 0.01$, ****: $p\le 0.0001$.

4.1.

PA Power versus Age – Aggregability of RBCs

Several studies reported on the correlation between RBC aggregation and age.^51–53 Woodward et al.⁵¹ measured hemorheological variables (blood viscosity, RBC aggregation, and fibrinogen) from subjects (25 to 74 years old) in association with cardiovascular risk factors. They found that RBC aggregation and fibrinogen increased with age. Christy et al.⁵² reported a significant increase in RBC aggregation with age (20 to 59 years old), establishing a significant correlation between phagocytic activity and RBC aggregability. According to Simmonds et al.⁵³ the mechanism of the age-related increase in RBC aggregation is that aging decreased the electrostatic repulsive forces between cell surfaces, thus promoting RBC aggregation alongside with increased plasma fibrinogen concentration.

During pulsatile blood flow, the dominant hemorheological characteristic is the cyclical aggregation and disaggregation phases of RBCs.^5,32 During systole, the blood flow velocity is maximum, generating maximal shear rate within vessels, and as a result, the RBCs in the radial artery flow as single cells. During diastole, on the other hand, the blood flow velocity is at a minimum, resulting in a minimal shear rate, leading to the formation of RBCs rouleaux.^5,32 In PA imaging, the RBC aggregates form a larger effective absorber compared to single cells, increasing the $Pa$ signal.^41,42,54,55 In Fig. 2(b), the $Pa$ and Doppler velocity $V$ were out of phase during a systolic-diastolic cycle. A higher $Pa$ is expected in the presence of aggregates and a lower $Pa$ from single RBCs. This occurred at all wavelengths of illumination, and as we have shown in earlier studies, it can approximate the aggregate size.⁴¹

The PA amplitude is a function of ${\mu }_a$ (including the absorption cross-section) and an absorber size. According to Eq. (1), ${\mu }_a$ can be represented as a linear combination of ${\epsilon }_{\mathrm{HbO}}$, ${\epsilon }_{\mathrm{HbD}}$, [HbO] and [HbD], depending on $\lambda$. The ${\mathrm{sO}}_2$ can be derived by combining ${\mu }_a$ at two wavelengths as addressed in Eq. (2). Given the ${\mathrm{sO}}_2$ also varies with the absorber size affecting the ${\mu }_a$, the $Pa$ must have a nonlinear relation with the absorber size. However, the $Pa$ at an isosbestic point 800 nm is not dependent on the ${\mathrm{sO}}_2$, resulting in a linear relation between the $Pa$ and the absorber size, as previously demonstrated by our group.⁴¹ The $\mathrm{\Delta }P{a}_{800}$ increased with age [Fig. 3(b)], suggesting that the RBC aggregability also increased with age (since the PA signal is a surrogate metric of the aggregate size).

4.2.

Effect of RBC Aggregation on PA Power as a Function of Absorber Size and sO₂

None of the abovementioned studies have examined the impact of RBC aggregation on the blood ${\mathrm{sO}}_2$ measurements in humans. It is well known that RBC aggregation and plasma viscosity play an important role in determining the overall blood viscosity.⁵⁶ Realizing the fundamental function of RBCs in oxygen transport, Tateishi et al.^57,58 were the first to postulate on the correlation between RBC aggregation and ${\mathrm{O}}_2$ release from erythrocytes and its impact on the ${\mathrm{sO}}_2$.⁵⁹ In addition, the relation between RBC aggregation and the ${\mathrm{sO}}_2$ was theoretically and experimentally investigated through in vitro PA imaging by our group.⁴¹ Similar to the Tateishi’s findings and our in vitro experiments, the in vivo results in this work demonstrate the impact of RBC aggregation on the ${\mathrm{sO}}_2$ measurement.

The $Pa$ increases due to two factors: an increase in the size of the absorber (single cells versus RBC aggregates) and the change in ${\mathrm{sO}}_2$. The average values of $Pa$ as a function of $\lambda$ for the group of subjects in their 40s are shown in Fig. 6(a). The $P{a}_{\min }$ represents measurements when nonaggregated RBCs flow in the radial artery at maximal velocity (when the shear rates inside the radial artery are highest). RBC aggregation increases the $Pa$ due to an increase in optical absorber size. Since the optical absorber size increases as RBCs aggregate, the $P{a}_{\min }$ increases by $\mathrm{\Delta }P{a}_{800}$ (2.7 dB for subjects in their 40s) for all $\lambda$ ($P{a}_{\min }+\mathrm{\Delta }P{a}_{800}$, depicted by the blue-black arrow). This can be schematically represented by $\mathrm{\Delta }P{a}_{800}(\mathrm{\Delta }a)$ in Fig. 6(b). The second reason for the changes in the $Pa$ is the change in the ${\mathrm{sO}}_2$ due to RBC aggregation. The difference between $P{a}_{\max }$ and “$P{a}_{\min }+\mathrm{\Delta }P{a}_{800}$” can be identified as the contribution to the change in ${\mathrm{sO}}_2$ caused by RBC aggregation to the overall PA signal, ($P{a}_{\min }+\mathrm{\Delta }P{a}_{800}+{\mathrm{sO}}_2=P{a}_{\max }$). Combining Eqs. (1) and (2), the ${\mu }_a$ can be expressed as a function of ${\mathrm{sO}}_2$,

Fig. 6

(a) The effect of RBC aggregation on PA power ($Pa$) in relation to an optical absorber size and oxygen saturation (${\mathrm{sO}}_2$). (b) Schematic diagram of the effect of RBC aggregation on $Pa$ in relation to the absorber size and the ${\mathrm{sO}}_2$. $a$ and $\mathrm{\Delta }a$ represent the size of single RBC and the increased size due to aggregation, respectively.

Since ${\epsilon }_{\mathrm{HbO}}$ is larger than ${\epsilon }_{\mathrm{HbD}}$ for $\lambda >800\mathrm{nm}$,⁶⁰ the ${\mathrm{sO}}_2$-induced increase in ${\mu }_a$ results in an increased $Pa$ with ${\mathrm{sO}}_2$, as shown in Fig. 6. On the other hand, since ${\epsilon }_{\mathrm{HbO}}$ is smaller than ${\epsilon }_{\mathrm{HbD}}$ for $\lambda <800\mathrm{nm}$, the ${\mathrm{sO}}_2$-induced decrease in ${\mu }_a$ results in a decreased $Pa$. This can be schematically denoted by $+Pa(\mathrm{\Delta }{\mu }_a)$ and $-Pa(\mathrm{\Delta }{\mu }_a)$ for $\lambda >800\mathrm{nm}$ and $\lambda <800\mathrm{nm}$, respectively, in Fig. 6(b). The in vivo experimental results demonstrating the effect of RBC aggregation on the $Pa$ and how this depends on the RBC ${\mathrm{sO}}_2$ are also supported by our group’s previous in vitro work mimicking the radial artery flow conditions.⁴¹ This interpretation applies to all age groups.

4.3.

Relationship between sO₂ versus Age

$P{a}_{\max }$ could be derived from $P{a}_{\min }$ by combining the effect of the increase in both absorber size and ${\mathrm{sO}}_2$ due to RBC aggregation. Hence, the ${\mathrm{sO}}_2$ computed from the $P{a}_{\max }$ (${\mathrm{sO}}_{2\text{-}\max }$) were higher than that from the $P{a}_{\min }$ (${\mathrm{sO}}_{2\text{-}\min }$) for all groups [as shown in Fig. 5(a)]. In addition, both ${\mathrm{sO}}_{2\text{-}\max }$ and ${\mathrm{sO}}_{2\text{-}\min }$ decreased with increasing age. It has been reported that older persons have a lower ${\mathrm{sO}}_2$ than their younger counterparts.^61–65 Even though the direct relation between the ${\mathrm{sO}}_2$ and age has not been fully investigated, it was reported that dysphagia (swallowing difficulties) could mediate the relation between ${\mathrm{sO}}_2$ and age.⁶⁵ Older people are more likely to experience dysphagia which impairs pulmonary function and lowers ${\mathrm{sO}}_2$. The present experimental results indicated that the ${\mathrm{sO}}_2$ decreases with age, despite the subject cohort comprising healthy individuals.

The difference between ${\mathrm{sO}}_{2\text{-}\max }$ and ${\mathrm{sO}}_{2\text{-}\min }$ ($\mathrm{\Delta }{\mathrm{sO}}_2$) increased with age, as shown in Fig. 5(b). This phenomenon was consistent with the relation between $\mathrm{\Delta }P{a}_{800}$ and age, as shown in Fig. 3(b). The $P{a}_{800}$ represents the effects of RBC aggregability, independent of ${\mathrm{sO}}_2$, which increased with age. Since oxygen release is inhibited by RBC aggregation,^57,58 higher aggregability results in more hemoglobin molecules to be bound oxygen molecules. Despite a negative correlation between the ${\mathrm{sO}}_{2\text{-}\min }$ and age, the RBC aggregation-induced increase in the ${\mathrm{sO}}_2$ ($\mathrm{\Delta }{\mathrm{sO}}_2$) resulted in a positive correlation between $\mathrm{\Delta }{\mathrm{sO}}_2$ and age. This is why the correlation between the ${\mathrm{sO}}_{2\text{-}\min }$ and age is steeper than the correlaiton between the ${\mathrm{sO}}_{2\text{-}\max }$ and age, as shown in Fig. 5(a).

4.4.

Limitations of the Study

The findings of this study suggest the feasibility of the PA assessment of both ${\mathrm{sO}}_2$ and its correlation to age, in vivo. However, there are limitations to this study that form the basis for future extension of this work. The study, whose recruitment is suspended during the global COVID-19 pandemic, enrolled a smaller than desired subject group and was mainly composed of male volunteers. The sex dependence on $Pa$ related to the correlation between ${\mathrm{sO}}_2$ and age must be further investigated. The laser fluctuation which sometimes occur from pulse to pulse or from wavelength to wavelength should be measured and corrected. The absolute values for the ${\mathrm{sO}}_2$ presented in this paper have not been corrected for the effects of laser fluence.⁶⁶ The radial artery depth from the skin surface is $\sim 5\mathrm{mm}$. As such, the effects of fluence (spectral coloring)⁶⁷ might not be as significant as it is in other PA applications. Specifically, in Eq. (3), the ${\mathrm{sO}}_2$ is calculated by a ratio of ${\mu }_a$. The PA amplitude ($P$) is a function of ${\mu }_a$ and fluence ($\phi$), so that the ${\mathrm{sO}}_2$ can be represented by the ratio of “$P/\phi$.” Moreover, since the position and geometry were the same, variations in fluence are minimized. In addition, the normalization to the laser energy at each wavelength could also contribute to a reduction of the fluence effects. As such, the reported absolute values might not represent the true ${\mathrm{sO}}_2$ for each subject. In addition, this quantity is dependent on the subject’s skin color (leading to a larger optical path length to the radial artery, requiring fluence correction). An additional limitation is that the upper limit of age group was in 40s. Further investigation for the older age groups (50s, 60, and 70s) should be conducted to study the age dependence further.

Despite these limitations, this study demonstrates the feasibility of the measurements and confirms previous in vitro findings. Future studies could be done using a portable probe, opening the potential for doing this investigation more easily. In addition, quantitative measurement of RBC aggregation by US such as the structure-factor-size-estimation⁶⁸ should be further applied to this study to correlate both PA and US modalities.