1.

Introduction

The brain is one of the most energy-demanding and metabolically active organs of the body. Although it comprises only 2% of body weight, it receives about 15% of cardiac output and uses 20% of total body oxygen and 25% of total body glucose.¹ Under normal physiological conditions, oxidative metabolism of glucose is the primary way to produce adenosine triphosphate, the most important energy source for the brain. Oxygen needs to be delivered to the cerebral tissue at a rate that is biochemically appropriate to neurons’ metabolic needs.¹ The cerebral metabolic rate of oxygen (${\mathrm{CMRO}}_2$) depends on the density of neurons and on their state of functional activation.² Many of the common disorders of the brain, such as Alzheimer’s, Parkinson’s, Huntington’s, and others, have been found to be associated with alterations in the cerebral oxygen metabolism.^3–5 Therefore, the measure of ${\mathrm{CMRO}}_2$ would not only enable a better understanding of the normal physiology during rest, sleep, anesthesia, aging, or functional brain tasks^6,7 but also be important for research and clinical applications related to brain disorders, stroke, and help to improve the management of patients who are at the risk of developing brain hypoxia-ischemia.^8,9

Since the extraction of oxygen from cerebral tissue is closely matched to the brain’s metabolic needs, ${\mathrm{CMRO}}_2$ should be equal to the total amount of oxygen delivered to the cerebral tissue per unit time minus the amount leaving in the venous circulation per unit time.² As shown in Eq. (1), the cerebral blood flow (CBF) and the arteriovenous oxygen difference (${\mathrm{AVDO}}_2$) are the key parameters to estimate ${\mathrm{CMRO}}_2$ based on Fick’s principle.¹⁰ Under steady-state physiologic conditions, ${\mathrm{CMRO}}_2$ remains relatively constant over a range of arterial partial pressure of oxygen (${\mathrm{PaO}}_2$), typically 23 to 100 mm Hg, due to the cerebral autoregulation mechanism.² During mild hypoxia (${\mathrm{PaO}}_2<90\mathrm{mm}$ Hg), the oxygen supply may be inadequate; however, metabolic demand may still be met by elevated CBF or reduced venous blood saturation, which means the higher extraction of oxygen from the blood.¹¹ Severe hypoxia (${\mathrm{PaO}}_2<30\mathrm{mm}$ Hg) may cause significant decrease in ${\mathrm{CMRO}}_2$ due to insufficient blood flow or oxygen extraction compensation.¹² The metabolic autoregulation is a negative feedback system that seeks to balance the oxygen supply to its demand in a certain range¹³

Currently, techniques to obtain ${\mathrm{CMRO}}_2$ are limited. The important criteria for such techniques include noninvasiveness, adequate spatial and temporal resolution, low-radiation exposure, good safety profile, and widespread clinical availability.¹⁴ Unfortunately, no in vivo method can fulfill all of these requirements. In early studies, the jugular vein oximetry involving catheterization has been used for a surrogate measure of ${\mathrm{CMRO}}_2$ but is invasive.¹⁵ Positron emission tomography (PET) with ${}^{15}\mathrm{O}$-labled radiotracers provides the most direct measurement of ${\mathrm{CMRO}}_2$.¹⁶ Yet, the spatial resolution is poor ($\sim 1\mathrm{cm}$ for clinical system), and the high ionizing radiation dose prevents its repeated use on the same patient. Magnetic resonance imaging (MRI) is also applied to image ${\mathrm{CMRO}}_2$ by utilizing the blood oxygenation-level-dependent (BOLD) effect. BOLD MRI is noninvasive and can be performed with high spatial resolution, but it is only sensitive to deoxy-hemoglobin and has trouble distinguishing between changes in oxygen saturation and blood flow.¹⁷ In recent studies, phase-contrast MRI has been proposed to quantify the blood flow in major inflow vessels; however, for the oxygen saturation estimation, only a large-sized vein (superior sagittal sinus) can be targeted.¹⁸ Moreover, the complex setup and high expenses preclude PET and MRI for many bedside clinical applications.

Imaging techniques based on optical contrast, which are sensitive to blood functional parameters, have been widely employed for the assessment of tissue oxygen metabolism. Near-infrared spectroscopy offers a means to estimate ${\mathrm{CMRO}}_2$ noninvasively with the blood flow information provided by diffuse correlation spectroscopy.^19,20 The main drawback is that the spatial resolution is relatively low due to strong optical scattering. Photoacoustic imaging overcomes this limitation by ultrasonically imaging optical contrast through the photoacoustic effect.²¹ Optical-resolution photoacoustic microscopy (OR-PAM) has been used to estimate oxygen consumption in superficial tissues.²² This technique utilizes fine optical focusing to achieve high lateral resolution ($\sim 5\mu \mathrm{m}$); however, the imaging depth is shallow ($<1\mathrm{mm}$) and the detectable flow speed ($<12\mathrm{mm}/\mathrm{s}$) is limited due to low laser repetition rate.²³ While recent work suggests hope for photoacoustic flow estimation in deep tissues ($>1\mathrm{mm}$) using acoustic-resolution photoacoustic imaging,²⁴ it is less established than Doppler ultrasound methods we selected in this paper.

Here, we propose ${\mathrm{CMRO}}_2$ estimation using photoacoustic and Doppler ultrasound methods. The combination of local blood oxygenation estimation and flow estimation in deep vessels is accomplished with a custom photoacoustic and microultrasound scanning system. We previously demonstrated oxygen flux estimation in a phantom study using such a system.²⁵ In this paper, the ${\mathrm{CMRO}}_2$ is estimated in vivo by combining the measurements of arterial and venous oxygen saturation and flow rate of the internal jugular vein.

2.

Materials and Methods

2.1.

Experiment Setup

Our custom photoacoustic and microultrasound system was constructed by a combined light delivery and ultrasound probe, a tunable laser, a computer-controlled scanning, and triggering and data acquisition system. As the irradiation source, a tunable optical parametric oscillator (Surelite OPO Plus, Continuum, California) was pumped by a Q-switch Nd:YAG nanosecond-pulsed laser (Surelite III, Continuum, California) at a repetition rate of 10 Hz. The tuning range is from 410 to 710 nm. The laser output was coupled into a bifurcated fiber-bundle light guide with a core diameter of $185\mu \mathrm{m}$ (900 fibers, CeramOptec Industries, Germany). An acrylic probe was designed to position the bifurcations of the light guide around a 25-MHz single-element-focused ultrasound transducer (V324-SM, Olympus Panametrics-NDT, Massachusetts). The transducer was placed such that the focus of the transducer and the center of the illumination spot ($\sim 1{\mathrm{cm}}^2$) were aligned. The illumination fluence was measured to be about $10{\mathrm{mJ}/\mathrm{cm}}^2$. This probe was mounted on a three-axis motion system for B-scan and C-scan imaging. A peripheral component interconnect (PCI) motion card (NI7350, National Instruments Inc., Texas) was used to control a high-precision three-axis motion stage with three integrated stepper motors (23Y002D-LW8, Anaheim Automation, California). Fast scanning was accomplished with a voice-coil actuator as previously described^25,26 while vertical and elevational translation were accomplished using the stepper motors.

Signals from the transducer were amplified by a variable-gain preamplifier then further amplified by an ultrasound pulser/receiver (5073PR, Panametrics-NDT, Massachusetts). The variable-gain stage (AD603ARZ evaluation board, Analog Devices, Massachusetts) was used to implement time-gain compensation (TGC). The TGC curves in a range of 40 dB for both ultrasound and photoacoustic imaging were controlled via a digital input–output (DIO) card (NI PCI-6542, National Instruments Inc., Texas). Signals were digitized using a multichannel data acquisition card (CS8289 Gage Cobra, Gage Applied Systems Inc., Illinois) with 12-bit dynamic range and sampling rates as high as $125\mathrm{MSamples}/\mathrm{s}$ and postprocessed using MATLAB^® (R2012a, MathWorks, Massachusetts).

Our system provided combined photoacoustic and Doppler ultrasound mode. The details are referred in our previous work.²⁶ In brief, a pulse sequence was generated by the DIO card and sent to interleave laser and ultrasound triggers. Photoacoustic and Doppler ultrasound images were coregistered after signal processing. A schematic of the constructed system is shown in Fig. 1, which also shows a photograph of the probe with bifurcated fiber-bundle light delivery.

Fig. 1

A schematic of light delivery and data acquisition setup and photograph of the fiber bundle delivery surrounding the transducer. DIO, digital input–output; US, ultrasound; P/R, pulser/receiver; and DAC, data acquisition card.

3.

Results

First, the bandwidth broadening Doppler ultrasound approach was validated in phantoms with embedded tubing by comparing estimated flow velocity against preset flow speeds, as shown in Fig. 2. The construction of the cornstarch-gelatin phantom can be referred to our previous work.³¹ Under the assumption of fully developed parabolic flow, we observed excellent consistency and sensitivity to flow velocities of several $\mathrm{mm}/\mathrm{s}$, sufficient for in vivo flow estimation in murine carotid artery and jugular veins. The correlation of determination $R^2$ of linear least-square fitting is 0.99 with a slope of 0.94 and a zero-offset of $4.2\mathrm{mm}/\mathrm{s}$.

Fig. 2

Comparison between the preset and the measured flow velocity of phantom using Doppler bandwidth broadening

Next, we estimated ${\mathrm{CMRO}}_2$ in vivo using Doppler ultrasound for flow estimation and multiwavelength photoacoustic methods for ${\mathrm{sO}}_2$ estimation. The animal was prepared as described in Sec. 2.2 and scanned across the neck at the speed of $5\mathrm{mm}/\mathrm{s}$ under different physiological conditions. The global arterial ${\mathrm{sO}}_2$ was recorded by the pulse oximeter during the modulation of ${\mathrm{FiO}}_2$. Figure 3 shows representative in vivo coregistered photoacoustic (acquired at 580 nm), Doppler bandwidth broadening, and power Doppler images of the carotid artery and jugular vein when global arterial ${\mathrm{sO}}_2$ was measured at 85%. The artery was located $>2\mathrm{mm}$ in depth, and the vein was located $\sim 1\mathrm{mm}$ under the skin. As in Fig. 3(b), the mean flow speed was quantified as $146\mathrm{mm}/\mathrm{s}$. For the task of cross-sectional area estimation, a circular region was selected so that all the color pixels in the power Doppler image [Fig. 3(c)] were fitted into the circle.

Fig. 3

Representative in vivo coregistered images of carotid artery (labeled as “A”) and jugular vein (labeled as “V”): (a) photoacoustic image acquired at 580 nm, (b) Doppler bandwidth broadening image, the color bar shows the absolute value of flow speed and direction, and (c) power Doppler image.

Figure 4 shows the measured venous mean flow velocities as a function of global arterial ${\mathrm{sO}}_2$ measured by the pulse oximeter. The appropriate ultrasound pulse-repetition-rate was chosen to ensure that the flow speeds were within the detectable range.³⁵ The mean flow velocity of the internal jugular vein begins to drop above a global arterial ${\mathrm{sO}}_2$ of $˜85\%$.

Fig. 4

Venous mean flow velocities measured from hyperoxia (global arterial ${\mathrm{sO}}_2$ $\sim 100\%$) to hypoxia (global arterial ${\mathrm{sO}}_2$ $\sim 65\%$) conditions.

Figure 5 shows photoacoustic ${\mathrm{sO}}_2$ estimates in both arteries and veins as a function of measured global arterial ${\mathrm{sO}}_2$ from a pulse oximeter. As described in Sec. 2.4, we used two photoacoustic methods to cross validate our ${\mathrm{sO}}_2$ estimation. Method 1 is the traditional approach, which solves for unknown concentrations of oxy- and deoxy-hemoglobin using measured photoacoustic signals and known molar extinction coefficients, whereas method 2 uses multiwavelength photoacoustic measurements at different ${\mathrm{sO}}_2$ levels to solve a system of equations to estimate the different ${\mathrm{sO}}_2$ levels. Figure 5(a) shows the comparison of arterial ${\mathrm{sO}}_2$ estimations between method 1 and method 2. The estimates are consistent with $p<0.05$. It can also be observed that the ${\mathrm{sO}}_2$ estimates from arteries are close to the pulse oximeter values. The consistency ensures the validation of ${\mathrm{sO}}_2$ estimation in arteries and offers confidence in ${\mathrm{sO}}_2$ estimation in veins.

Fig. 5

Comparison of photoacoustic ${\mathrm{sO}}_2$ estimations between method 1 and method 2 in (a) carotid artery and (b) jugular vein. The arterial ${\mathrm{sO}}_2$ error bars are larger than venous ${\mathrm{sO}}_2$ error bars, likely due to flow pulsatility in arteries.

The representative ${\mathrm{CMRO}}_2$ was quantified according to Eq. (9) in Sec. 2.5. Here, we used a typical $C_{\mathrm{HbT}}$ level of $0.147\mathrm{g}/\mathrm{ml}$ at a hematocrit of 0.42.³⁴ The results are shown in Fig. 6 as a function of global arterial ${\mathrm{sO}}_2$ as measured using pulse oximetry. As anticipated, the ${\mathrm{CMRO}}_2$ remains relatively constant during mild hypoxia and hyperoxia (global arterial ${\mathrm{sO}}_2$ of 65% to 100%), consistent with the hypothesis of brain metabolic autoregulation.

Fig. 6

The estimated representative ${\mathrm{CMRO}}_2$ ($\mathrm{ml}/100\mathrm{g}/\min$) as a function of varying global arterial ${\mathrm{sO}}_2$ measured using pulse oximetry during modulation of ${\mathrm{FiO}}_2$ under anesthesia. Error bars are standard deviations estimated according to the theory of propagation of errors.

4.

Discussion

Given the major significance of oxygen metabolism in health and disease, our work represents an approach to measure this metric using photoacoustic and ultrasound technologies. Current efforts use high-frequency ultrasound and visible light for animal models owing to limited penetration requirements. It should be noted, however, that depths investigated represent unreachable penetrations for previous oxygen metabolism progress using OR-PAM.

In this study, the total CBF is quantified by measuring outflow instead of inflow,³⁶ as internal jugular veins provide the majority ($\sim 80\%$) of the venous drainage from the brain. Since not all the blood vessels are counted for the measurements, the ${\mathrm{CMRO}}_2$ we measure is only a representative measure of oxygen consumption. Given the bifurcation of the common carotid artery into internal and external, our estimate of ${\mathrm{CMRO}}_2$ may be influenced by including facial oxygen flux. Encouragingly, the representative ${\mathrm{CMRO}}_2$ remains relatively constant during the modulation of the ${\mathrm{FiO}}_2$, which agrees with the hypothesis of cerebral metabolic autoregulation in the range of 23 to 100 mm Hg of ${\mathrm{PaO}}_2$. During mild hypoxia, the metabolic demand may be met by elevated CBF or higher extraction of oxygen from the blood.¹¹ We have observed that the mean flow speed of internal jugular vein was dropping at global arterial ${\mathrm{sO}}_2$ of $\sim 85\%$. This may be the consequence of vasodilation, which helps to increase the cerebral blood volume under the hypoxia condition.²

Uncertainty due to measurement variability is quantified by measuring ${\mathrm{sO}}_2$ and flow rates at least three times and estimating standard deviations. Future work should aim to reduce uncertainties by optimization of light fluence, using higher-sensitivity detectors, and by a potential combination of motion compensation and judicious averaging. Array transducers may improve frame rates to real time if sufficient wavelength-switching capabilities are available, whereas current ultrasound B-scan speeds are limited to $\sim 15\mathrm{fps}$ while photoacoustic B-scans are limited by the 10-Hz repetition rate of the laser to acquire a single A-scan.

Future work should investigate conditions where brain autoregulation is impacted, such as stroke, trauma, acute sepsis, brain damage, and potentially neurodegenerative conditions. Preclinical investigation with animal models may offer invaluable insight prior to human translation. Human translation may be possible with infrared light and diagnostic-frequency transducers, especially in infants and children, considering the superficial depths of internal carotid and jugular vessels in these subjects.

5.

Conclusion

We have demonstrated in vivo estimation of ${\mathrm{CMRO}}_2$ metabolism in murine models using a custom combined photoacoustic and microultrasound system. Estimates of ${\mathrm{CMRO}}_2$ are based on arterial and venous oxygen saturation measurements using photoacoustic methods and flow measurements based on Doppler ultrasound. Photoacoustic estimates of oxygen saturation in arteries are validated with pulse oximetry while venous ${\mathrm{sO}}_2$ is cross validated using two independent photoacoustic methods. The estimated ${\mathrm{CMRO}}_2$ remains relatively constant from mild hypoxia to mild hyperoxia conditions, which agrees with the hypothesis of cerebral metabolic autoregulation. The demonstrated method for estimating oxygen consumption may have significant applicability in other tissues and has potential for clinical translation owing to its label free nature.