2.1.

System Description

Figure 1 shows the experimental setup of the optical-resolution photoacoustic microscope for in vivo imaging of cerebral microvasculature. A dye laser (Credo, Sirah Laser und Plasmatechnik, Germany) pumped by a Nd:YLF laser (IS8II-E, EdgeWave GmbH, Germany) was employed as the wavelength-tunable irradiation source. It provided laser pulses with a pulse width of 9 ns and a pulse repetition of 1 KHz. Laser beam from the dye laser was focused by a set of plano-convex lens and then filtered by a 50-μm diameter pinhole. An objective lens was used to focus the laser beam into a single-mode fiber (core diameter: 3.5 μm). Laser beam from the distal end of the fiber was focused into the brain cortex by a long working distance objective lens (NT46-143, Edmund Optics, $4\times$, working distance: 34 mm), after being collimated by another same objective lens. The incident energy density on the brain cortex was estimated to be $18\mathrm{mJ}/{\mathrm{cm}}^2$, which was slightly smaller than the American National Standards Institute safety limit²¹ (${20\mathrm{mJ}/\mathrm{cm}}^2$ in the visible spectral region).

Fig. 1

Schematic of the optical-resolution photoacoustic microscopy system. USD: ultrasonic transducer, PD: photodetector.

A home-made prism which contained a silver gilt with a 45 deg inclined plane was used for acoustic-optical coaxial alignment.²² The inclined plane reflected the incident laser beam, while a contact ultrasound transducer (V2022, Olympus, central frequency: 75 MHz) detected acoustic waves transmitted through the prism. An acoustic lens was ground in the bottom of the prism for acoustic focusing, and a plano-convex lens was inserted before the laser beam incident into the prism for correcting the divergence when laser beam passed through the acoustic lens. The prism we designed provided a more effective way to detect acoustic signals because it avoided the transformation from longitudinal acoustic wave to shear wave when the acoustic wave was reflected.

The imaging head was mounted on a three-axis linear stage, of which the $Z$ axis was used to adjust the focal plane. 3-D imaging was implemented combining time-resolved acoustic detection and two-dimensional raster scan. The data acquisition card (ATS9350, Alazartech, Canada) recorded both photoacoustic signal and photodetector signal, which indicated the deviation of incident laser energy, using a sample rate of 500 MHz.

The lateral and axial resolutions were estimated by imaging carbon nanoparticles with a mean diameter of 100 nm, as illuminated in Fig. 2. The Gaussian fit to the transverse profile of a carbon nanoparticle gave an estimation of the lateral resolution to be 4.07 μm [Fig. 2(a)]. The axial resolution was estimated by applying shift-and-sum to the Hilbert transform of the axial profile of carbon nanoparticle [Fig. 2(b)]. The contrast-to-noise ratio (CNR) curve [inset of Fig. 2(b)] was applied for the estimation of axial resolution.²³ Here, the contrast was defined as the difference between the smaller of the two peaks and the valley. The noise was defined as the standard deviation of the experimental measurement without absorbers. According to Rayleigh criterion, the axial resolution was defined to be the distance where the CNR equaled 6 dB. The axial resolution was estimated to be 19.5 μm from the CNR curve.

Fig. 2

Resolution estimation of the optical-resolution photoacoustic microscopy system. (a) Transverse spread profile of a typical carbon nanoparticle. Red line indicates the Gaussian fit to the experimental data. (b) Simulated photoacoustic shift-and-sum signals. The dot line and dash-dot line indicate two nanoparticles with a distance of 20 μm. The solid line indicates the overlap of photoacoustic signals of the two nanoparticles. The inset shows the contrast-to-noise ratio (CNR) versus the shift distance when summing two nanoparticle signals.

2.2.

Animal Preparation

Experiments were performed on $8\mathrm{babl}/\mathrm{c}$ mice (25 to 30 g, 5 week, Animal Biosafety Level 3 Laboratory, Wuhan, China). Before each experiment, a mouse was anesthetized by intraperitoneally administering a dose of $0.2\mathrm{g}/\mathrm{kg}$ chloral hydrate and $1\mathrm{g}/\mathrm{kg}$ urethane. During the anesthesia, the body temperature of the mouse was kept at 37°C by a temperature stabilized heating pad. The mouse was fixed on a brain stereotactic apparatus to avoid motion disturbance. One dorsal portion of the skull was removed using a dental drill after removing the scalp and temporal muscle over it, thus a cranial opening (${\sim 5\times 5\mathrm{mm}}^2$) was implemented and the dura mater was exposed. A thin layer of artificial cerebrospinal fluid was implemented on the dura mater before the ultrasonic gel was applied to keep the appropriate biochemical environment.

During each experiment, we first used an isosbestic wavelength of 584 nm to image the cerebral microvascular structure through the cranial opening. Then, we randomly chose a cross-sectional scan along the rostral direction. The scan length was about 2 mm which covered both arterioles and venules. The selected cross scan was used for dynamic monitoring of the diameter, HbT, and ${\mathrm{SO}}_2$ during the injection of norepinephrine. Three wavelengths of 576, 580, and 584 nm were used for the monitoring. Each monitoring trial lasted for $\sim 23\min$ (150 B-scans at three wavelengths). During the first 2 to 3 min, the mouse stayed calm, and we could get the base line. Then, 100 μL of norepinephrine bitartrate ($0.2\mathrm{mg}/\mathrm{mL}$, A0937, Sigma-Aldrich, St. Louis, MO) was injected through tail vein in six of the eight mice. The other two mice were injected with 100 μL physiological saline as the control trials. The monitoring went on throughout the injection and lasted for 20 min after injection. After each experiment, the mouse was euthanized with an overdose of pentobarbital. All the procedures in the experiments were carried out in according to the Institutional Animal Care and Use Committee of Hubei Province.

2.3.

Data Analysis

After the experiment, amplitudes of photoacoustic signal were extracted via Hilbert transform and no other data processing was required. In order to extract the vessels from the background, we considered pixels with an amplitude, i.e., 3 dB, above the standard deviation of the noise to be vessels.²⁴ In this way, we could count the number of vessel pixels from the transverse view. Since we scanned the B line with a step size of 2 μm, the vessel diameter could be roughly calculated.

Images acquired under the wavelength of 584 nm mapped the distribution of HbT. Referring to the vascular region obtained above, pixels within the same vessel at the same time were averaged. The relative change was further calculated by normalized to the base value.

Oxygen saturation of the microvessels can be calculated using a multiwavelength method.²⁵ Deoxygenated hemoglobin (HbR) and oxygenated hemoglobin (${\mathrm{HbO}}_2$) were treated to be the dominant absorbing compounds in blood. Using more than two wavelengths, we can get the following by least-squares fitting.

$K$ is the proportionality coefficient that is considered to be constant. $\epsilon$ is the molar extinction coefficient. $\varphi$ is the amplitude of photoacoustic signal. Further data processing was the same as that used in HbT to get the relative change in ${\mathrm{SO}}_2$.

In order for temporal characterization of the response, we quantified the response times from the injection time, until one parameter (i.e., diameter, HbT) monotonically changed from a relatively stable state to 10% of the maximum response. In each trial, the injection time was set to be 0. As each line scan under three wavelengths costs about 9 s, we can quantify the response times $\mathrm{\Delta }t$ (units: minute) to be as follows:

3.1.

Imaging of Mouse Cerebral Microvasculature

Figure 3(a) shows the maximum amplitude projection of region of interest (ROI) Z1 depicted in Fig. 3(c). The imaging wavelength was 584 nm, where HbR and ${\mathrm{HbO}}_2$ have nearly the same molar extinction coefficient. Therefore, the image mainly mapped the concentration of HbT that was independent of blood oxygen level. Figure 3(a) gave a good visualization of the branch structures. The minimum resolvable vasculature in Fig. 3(a) covered 3 to 4 pixels. Considering that the pixel size used here was 2 μm, the minimum resolvable vasculature [marked with CL in Fig. 3(a)] was 6 to 8 μm, which was probably the single capillary vessel. Multiwavelength photoacoustic imaging was conducted in ROI Z2. Three wavelengths were chosen, of which 584 nm is the isosbestic wavelength reflecting the distribution of HbT (${\mathrm{HbT}=\mathrm{HbR}+\mathrm{HbO}}_2$), and 576 and 580 nm are ${\mathrm{HbO}}_2$ absorption-dominant wavelength reflecting mainly the distribution of ${\mathrm{HbO}}_2$. Meanwhile, the relatively high absorption coefficient of whole blood at the three wavelengths guaranteed the high signal-to-noise ratio. The commonly used arithmetic mean filter (window size: $10\times 10\mathrm{pixels}$) was applied to the calculated ${\mathrm{SO}}_2$ value and the ${\mathrm{SO}}_2$ value was visualized with pseudocolors ranging from blue to red, depicted in Fig. 3(b). The red vessels were most probably arterioles, and the blue ones were likely venules. According to the mouse brain anatomy, the region labeled Z2 has the characterization where branch arterioles from the middle cerebral artery and venules from superior sagittal sinus intersect with each other,²⁶ which is consistent with our result.

Fig. 3

Photoacoustic mapping of mouse cerebral microvasculature. (a) shows the morphology imaging in region labeled Z1. (b) shows the oxygen saturation imaging in region labeled Z2. and (c) shows the open-skull and imaging region. CL, single capillaries.

3.2.

Changes of Cerebral Microvascular Diameter

Most of the microvessels under investigation exhibited obvious vasoconstriction immediately after the injection. Vasoconstriction reached the maximum after 3 min. The microvessesls began to recover after severe vasoconstriction which lasted for 2 to 4 min. Normally, the microvessels recovered back to the resting state 7 min after the injection. However, some microvessels sustained a slight contraction throughout the investigation. Figure 4 illustrates typical changing of microvascular morphology during the observation. Figure 4(a) and 4(c) is the transverse view of the morphological changes of an arteriole and a venule, respectively, during the injection of norepinephrine. The horizontal axis shows the time course, and the vertical axis marks the semidiameter along the center of the microvessels. Figure 4(b) and 4(d) is the plot of the diameter of the arteriole and the venule mentioned above, where the blue line shows the raw data, and the red line shows the tendency of the blue line after the application of a mean filter. The arithmetic mean filter set the center point values to be the average of an 8-pixel neighborhood and signals beyond the end points were set to be 0. Both the microvessels constricted severely immediately after injection, reached a constriction stabilization period after about 2.5 min, and recovered thereafter. There are still some differences between the responses of arteriole and venule. The arteriole constricted from 30 μm to a minimum diameter of 15 μm, whereas the venule constricted from 32 μm to a minimum diameter of 22 μm. The constriction ratio of arteriole and venule, which was defined as 1-minimum diameter/resting state diameter, was 50% and 31.25%, respectively. By the end of the observation, the diameter of the arteriole recovered back to the same as the resting state, while the venule preserved a slightly smaller diameter. Ten arterioles and ten venules were conducted in statistical analysis, respectively. The arterioles constricted to a minimum diameter that is $55.93\%\pm 6.12\%$ of the resting state, whereas the venules reached only $75.85\%\pm 9.49\%$, showing a significant difference ($p<0.001$). By the end of the observation, the diameter of arterioles recovered to $97.37\%\pm 4.16\%$ of the base value, the venules only back to $91.66\%\pm 2.37\%$, showing a significant statistical difference ($p<0.05$). In the control trials, no changes of diameter were observed throughout the observation, indicating that the observed phenomenon was not the effect of tail vein injection of exogenous reagents. The observations suggested that cerebral microvasculature of healthy mice constricted severely when intravenously injected with norepinephrine, but recovered in a short time without prolonged drug injection thereafter. The arterioles exhibited more intense constriction than venules.

Fig. 4

Changes of cerebral microvascular diameter during investigation. (a) and (c) shows the transverse view of a typical arteriole and venule during the observation, the vertical axis marks the semidiameter from the center of vessels. (b) and (d) are diameter plots of arteriole in (a) and venule in (c), the blue line indicates the raw data and the red line indicates the tendency of changes in diameter. Horizontal axis of all figures is the same time course, and the injection time is set to be 0 (indicated by arrow flag).

3.3.

Changes of Cerebral Microvascular HbT and ${\mathrm{SO}}_2$

Statistical analysis was conducted on the 10 arterioles and 10 venules, respectively, to get the response of HbT and ${\mathrm{SO}}_2$. The statistical results were illustrated in the form of $\mathrm{mean}\pm \mathrm{standard}$ error of mean (SEM) in Fig. 5. For better visualization, markers of SEM were drawn at only one point every half minute. Figure 5(a) depicted the changes of microvascular HbT during the investigation. Both arterioles and venules showed an immediate decreased HbT along with the vasoconstriction, 4 min after the injection arterioles reached a minimum of $74.51\%\pm 3.36\%$ of the base value, meanwhile venules reached $60.54\%\pm 7.05\%$, showing a significant difference ($p<0.05$). Thereafter, all of the vessels showed a slow increased HbT, and they recovered to the base value about 13 min after the injection. No significant difference between HbT in arterioles and that in venules was observed by the end of the observation. The above observations indicated that HbT in cerebral microvasculature decreased shortly after the vasoconstriction, venules showing a more severe response. But all of the microvessels recovered to the base value without further injection.

Fig. 5

Changes of cerebral microvascular (a) HbT and (b) ${\mathrm{SO}}_2$ during investigation. Red line indicates the changes of arterioles; blue line indicates the changes of venules. The injection time is set to be 0 (indicated by arrow flag).

Changes of cerebral microvascular ${\mathrm{SO}}_2$ were a little bit later than that of diameter and HbT, as illustrated in Fig. 5(b). ${\mathrm{SO}}_2$ in arterioles was $93.6\%\pm 5.4\%$ before the injection and declined about 1.5 min after the injection of norepinephrine, reaching a minimum of $56.9\%\pm 4.5\%$. Finally, it fluctuated at $74.3\%\pm 3.9\%$ after a slight increase. ${\mathrm{SO}}_2$ in venules was $73.8\%\pm 3.7\%$ before the injection. Compared with arterioles, ${\mathrm{SO}}_2$ in venules showed a slight later decline, reaching a minimum of $41.2\%\pm 3.7\%$. It kept fluctuating at this level without the tendency of recovery. In the control trials, no changes of HbT and ${\mathrm{SO}}_2$ were observed throughout the observation. The observations indicated that a one-time injection with high dosage norepinephrine caused a serious decline in ${\mathrm{SO}}_2$ of cerebral microvasculature immediately. Although the ${\mathrm{SO}}_2$ will slightly increase in the arterioles, it sustained a low ${\mathrm{SO}}_2$ level in venules for a long time even without any further drug administration.