10121560432338Nat MethodsNat. MethodsNature methods1548-70911548-710525822799442890110.1038/nmeth.3336NIHMS677546ArticleHigh-speed Label-free Functional Photoacoustic Microscopy of Mouse Brain in ActionYaoJunjie1WangLidai1YangJoon-Mo1MaslovKonstantin I.1WongTerence T. W.1LiLei1HuangChih-Hsien2ZouJun2WangLihong V.1Optical Imaging Laboratory, Department of Biomedical Engineering, Washington University in St. Louis, St. Louis, MO, USADepartment of Electrical and Computer Engineering, Texas A&M University, College Station, TX, USACorrespondence should be addressed to L.V.W. (lhwang@wustl.edu)29420153032015520151252015125407410
We present fast functional photoacoustic microscopy (PAM), which is capable of three-dimensional high-resolution high-speed imaging of the mouse brain, complementary to other imaging modalities. A single-wavelength pulse-width-based method was implemented to image blood oxygenation with capillary-level resolution and a one-dimensional imaging rate of 100 kHz. We applied PAM to image the vascular morphology, blood oxygenation, blood flow, and oxygen metabolism in the brain in both resting and stimulated states.
Many biomedical imaging techniques, especially small-animal functional magnetic resonance imaging (fMRI), two-photon microscopy (TPM), and wide-field optical microscopy, have profoundly impacted hemodynamic studies of the mouse brain by providing structural, blood oxygenation, and flow dynamic information at various length scales. However, small-animal fMRI is insufficient to resolve brain hemodynamic activities at microscopic length scales finer than 50 μm 1; phosphorescence-lifetime-based TPM suffers from slow measurement of blood oxygenation 2; and wide-field optical microscopy lacks depth resolution 3.
Given these limitations, photoacoustic (PA) tomography (PAT) can play a complementary role. Previously reported PAT techniques variously lacked capillary-level resolution, wide-field imaging speed, or blood oxygenation imaging capability 4–8. Here, we present fast functional photoacoustic microscopy (PAM), which is capable of high-resolution high-speed imaging of the mouse brain through an intact skull in vivo. PAM has achieved a lateral spatial resolution of ~3 μm, which is 5 times finer than that of our previous fast-scanning system 7, 25 times finer than that of our previous acoustic-resolution system 6, and more than 35 times finer than that of ultrasound-array-based photoacoustic computed tomography 5. By using a new single-wavelength pulse-width-based method, PAM allows three-dimensional (3D) blood oxygenation imaging with capillary-level resolution at a one-dimensional (1D) imaging rate of 100 kHz. PAM’s blood oxygenation imaging speed is 50 times higher than that of our fast-scanning PAM 8, 100 times higher than that of our acoustic-resolution system 6, and more than 500 times higher than that of phosphorescence-lifetime-based TPM 2.
In PAM, both the excitation laser beams and the detection acoustic axis are confocally steered by a customized water-immersible MEMS (i.e., microelectromechanical system) scanning mirror (Fig. 1a, Supplementary Fig. 1, Supplementary Note 1). The lateral resolution—in the direction perpendicular to the acoustic axis—is ~3 μm at the optical focus, and the axial resolution—in the direction along the acoustic axis—is ~15 μm. The high laser pulse repetition rate of 500 kHz, 5 times greater than that in our previous work 7, enables dense sampling for morphological capillary-resolution imaging over a large scanning range. With a 1D time-resolved imaging rate of 500 kHz, PAM has achieved a 2D frame rate of 400 Hz over a ~3 mm scanning range, and a 3D volumetric rate of 1 Hz over a 3×2 mm2 field of view (Fig. 1b).
By using a new single-wavelength pulse-width-based method (PW-sO2), PAM is capable of high-speed imaging of the oxygen saturation of hemoglobin (sO2) (Fig. 1c, Online Methods). The two forms of hemoglobin, oxy- and deoxy-hemoglobin (HbO2 and HbR), have different saturation intensities, defined as the excitation intensity that reduces the absorption coefficient to half its initial value (Supplementary Note 2, Supplementary Fig. 2) 9. When first excited by a picosecond pulse and subsequently by a nanosecond pulse of the same wavelength and pulse energy, HbO2 and HbR display different saturation levels (Supplementary Fig. 3). From PA signals acquired with the two laser pulses, the relative concentrations of HbO2 and HbR are quantified, and thus sO2 can be computed. PW-sO2 does not suffer wavelength-dependent optical attenuation as the traditional wavelength-tuning method does. Nevertheless, the maximum PW-sO2 imaging depth is limited by optical attenuation to the point where saturation becomes insufficient.
A 5×10 mm2 region of the mouse brain was imaged by PAM through an intact skull with the scalp removed (Fig. 1d, Supplementary Video 1) (acquisition time: ~15 seconds). The optical focal plane was fixed at ~250 μm beneath the skull surface. The imaging parameters for all the key experiments are summarized in Supplementary Table 1. By additional depth-scanning of the optical focal zone with a z-step size of 100 μm, PAM provided an imaging depth of ~0.7 mm (Fig. 1e, Supplementary Fig. 4, Supplementary Video 2), giving an effective pixel count of ~47 in focus along the depth direction. The optical scattering of brain tissue degraded the lateral resolution and image contrast of PAM with increasing imaging depth, as in any depth-resolved optical microscopy; thus, deep capillaries cannot be resolved by the current version of PAM. PAM of the brain vasculature was confirmed by TPM (Supplementary Fig. 5). The acoustic sectioning of PAM could not resolve the blood vessels along the z-axis as well as the optical sectioning of TPM. The skull degraded the image quality of PAM by blurring the optical focusing and attenuating the PA signal (Supplementary Note 3, Figs. S6 and S7).
sO2 of the mouse brain was mapped vessel-by-vessel by using the pulse-width-based method(Fig. 1f) (acquisition time: ~40 seconds). A pulse energy of 400 nJ was used for all the sO2 measurements unless otherwise stated. The optical fluence at the optical focus was estimated to be ~0.3 J/cm2. The non-saturated PA signal acquired with nanosecond excitation was used to correct for optical attenuation and the laser spot size. We observed that the averaged sO2 in the skull vessels was lower than that in the cortical vessels, consistent with the low-oxygenation microenvironment in bone marrow 10.
The PW-sO2 method was validated on blood phantoms, with an average measurement error of ~2.7% (Fig. 1g). The PW-sO2 method was also compared with the traditional two-wavelength-based method (TW-sO2) in vivo, with an average difference of <5% for superficial vessels (Supplementary Fig. 8) 4. To quantify the underestimation of PW-sO2 induced by light attenuation with increasing depth, we measured the sO2 in blood phantoms and in a mouse ear in vivo with pulse energies from 50 nJ to 1000 nJ (Fig. 1h, Figs. S9–S10). When the pulse energy was 300 nJ, the measurement error was ~3% for absolute PW-sO2.
We carefully investigated the potential for tissue damage induced by PAM. First, bright-field microscopy of a single layer of mouse RBCs, before and after the PAM imaging, confirmed that the PAM-imaged RBCs were intact, with clear donut shapes (Supplementary Fig. 11). Second, TPM of a mouse brain after PAM imaging with the laser pulse energy intentionally increased to 1000 nJ ruled out its potential to induce bleeding (Supplementary Fig. 12a). A few vessels were imaged by TPM but not by PAM, probably due to the lack of RBC perfusion 11. Last, standard H&E histology on a mouse brain after PAM imaging (Online Methods) showed no burn damage to the brain tissue (Supplementary Fig. 12b). As a positive control, a part of the brain was intentionally burned and was also studied histologically. Representative histological slices from the inside and outside of the burned area, as well as the imaged area, were compared, revealing no burns in the imaged area (Supplementary Fig. 13).
Directly imaging hyperaemia in the brain can help understand neurovascular coupling. Here, we demonstrate the high-speed functional imaging capability of PAM by studying mouse cortical hemodynamic responses to electrical stimulations to the hindlimbs (Supplementary Fig. 1a). Upon stimulations, the PA amplitude in the contralateral somatosensory region started to increase until the end of the stimulations (Fig. 2a, Supplementary Video 3). Meanwhile, the ipsilateral somatosensory region followed a similar trend but responded more weakly (Figs. S14a–b), suggesting vascular interconnection between the two hemispheres 12. We also observed that the sagittal sinus region responded to both left and right hindlimb stimulations, possibly due to the fact that it drains blood from both hemispheres simultaneously 12. The depth-resolved responses revealed that the responding region covered a depth range of 50–150 μm beneath the cortical surface (Fig. 2b). The deep capillary beds showed stronger amplitude responses than the major arteries and veins (Figs. S15a–b) 3.
Meanwhile, the artery dilated substantially in the contralateral hemisphere during the stimulations (Supplementary Fig. 14c, Supplementary Fig. 15a). In the ipsilateral somatosensory region, arterial dilation was also observed but with a much weaker magnitude (Supplementary Fig. 14d). Veins did not show dilations (Supplementary Fig. 15c, Supplementary Fig. 15a) 3. Deep capillary beds are reported to dilate less than 0.5 μm in diameter 13, which is not resolvable by the current version of PAM. Fast line scanning along the vessel axis was repeated to measure the blood flow speed (Supplementary Fig. 16, Online Methods) 8, 14. Stimulations induced a substantial increase in blood flow speed in both arteries and veins (Supplementary Fig. 14e and Supplementary Fig. 15d) 14. However, PAM could not detect the flow speed changes in deep capillaries.
Upon stimulations, sO2 increased substantially in veins and deep capillary beds (Fig. 2c, Supplementary Video 4, Supplementary Video 5). The fractional change in sO2 diminished with increasing distance from the core responding region (Figs. S17a–b), which was ~100 μm below the cortical surface (Supplementary Fig. 17c) 3. The sO2 increase was greater in deep capillary beds than in veins and was insignificant in arteries (Supplementary Fig. 15e). The lack of arterial sO2 response is consistent with the fact that arterial blood has not yet reached capillaries for oxygen consumption and thus maintains a high oxygenation level 3.
In the core responding region, the increase in sO2 in veins also manifested as a decrease in the oxygen extraction fraction (OEF) (Fig. 2d) 15. The fractional change in the cerebral metabolic rate of oxygen (CMRO2) can be estimated from the above hemodynamic measurements (Online Methods). A moderate fractional increase in CMRO2, peaking at ~15%, was observed (Fig. 2d). The ratio between the fractional changes in cerebral blood flow (CBF) and CMRO2 (i.e., the flow-consumption ratio) was ~2.0, consistent with the literature 16.
In summary, using endogenous contrast, PAM has achieved high-speed high-resolution imaging of the vascular morphology, blood oxygenation, blood flow dynamics, and oxygen metabolism of the mouse brain. In particular, PAM has achieved a 1D time-resolved imaging rate of 500 kHz for morphological imaging and 100 kHz for blood oxygenation imaging. In the future, the axial resolution of PAM can be improved by using an ultrasonic transducer with a wider bandwidth or by using nonlinear photoacoustic mechanisms 17. Near-infrared wavelengths can increase the imaging depth of PAM 17. The potential for optical breakdown in RBCs in vivo, which is relevant to excitation pulsewidth, can be further investigated by using photothermal microscopy or more other methods 18.
We would like to end by briefly discussing the advantages and limitations of our PAM for mouse brain hemodynamic imaging. (1) PAM provides better depth resolution and greater absorption-based image contrast than wide-field optical microscopy, but is slower and more expensive. (2) In comparison to TPM, PAM does not need exogenous contrast agents and point-by-point depth scanning, but PAM has worse axial resolution, and at the currently available wavelengths less penetration. (3) In comparison to fMRI, which is sensitive only to HbR, PAM is sensitive to both HbR and HbO2, and has higher spatial and temporal resolutions. However, fMRI can provide deeper penetration. Therefore, PAM is highly complementary to other brain imaging modalities in its contrast mechanism, spatial–temporal resolutions, and functional imaging capability.
ONLINE METHODSPhotoacoustic tomography (PAT)
In PAT, as photons travel in tissue, some are absorbed by biomolecules, and their energy is partially or completely converted into heat. The heat-induced pressure wave propagates in tissue, and is detected outside the tissue by an ultrasonic transducer or transducer array to form an image that maps the original optical energy deposition in the tissue. PAT has a 100% relative sensitivity to optical absorption, which means a given percentage change in the optical absorption coefficient yields the same percentage change in the PA signal amplitude. In PAT operating at 532 nm, the high contrast of cortical microvasculature comes from the absorption of hemoglobin in red blood cells (RBCs), which overpowers the absorption of other absorbers such as water and lipids by three orders of magnitude. As the optical absorption of blood is highly oxygenation dependent, PAT can measure blood oxygenation with high sensitivity. Photoacoustic microscopy (PAM) is a microscopic focused-scanning embodiment of PAT.
Fast functional photoacoustic microscopy
To induce photoacoustic signals, a 3-ns pulsed laser beam at 532 nm (AOT-YVO-100Q, AOT Inc.; pulse repetition rate: 100 kHz) is combined with a 3-ps pulsed laser beam at 532 nm (APL-4000-1064, RPMC Lasers, Inc.; pulse repetition rate: 500 kHz) via a polarizing beam splitter (PBS251, Thorlabs, Inc.). The polarizations of the two laser beams are adjusted by wave plates to maximize the combining efficiency. Laser energy fluctuations are monitored by a fast photodiode that samples a small portion of the laser beams. The laser beams are focused by a plano-convex lens (LA1131, Thorlabs, Inc.), then spatially filtered by a 50-μm-diameter pinhole (P50C, Thorlabs, Inc.). The filtered laser beams are focused to ~3 μm spots by an objective lens (AC127-050-A, Thorlabs, Inc.; NA: 0.1 in air). The optical focal zone (also known as the depth of focus), defined as the full width at half maximum of the peak intensity, is ~83 μm, within which the lateral resolution degrades up to
2; correspondingly, the depth range within which the lateral resolution degrades up to a factor of 2 is ~144 μm. An optical-acoustic beam combiner, composed of an aluminum-coated prism (NT32-331, Edmund, Inc.) and an uncoated prism (NT32-330, Edmund, Inc.), provides optical-acoustic coaxial alignment. Here, the thin aluminum coating reflects light but transmits sound. An optical correction lens attached to the top surface of the combiner corrects the optical aberration due to the prism. The focused laser beams and the generated photoacoustic waves are both directed by a MEMS scanning mirror in a coaxial configuration. The photoacoustic waves are focused by an acoustic lens and detected by an ultrasonic transducer (V214-BB-RM, Olympus-NDT, Inc.; central frequency: 50 MHz; one-way −6 dB bandwidth: 100%).
Driven by sufficiently strong electromagnetic force, the whole MEMS scanning mirror can operate under de-ionized water in a water tank, which is required to provide acoustic coupling from the sample surface to the acoustic lens. The bottom of the water tank is sealed with a piece of membrane that is both optically and acoustically transparent. In PAM, because the acoustic flight time provides depth information along the acoustic axis, each excitation laser pulse generates a 1D depth-resolved image. Volumetric imaging is provided by fast angular scanning of the MEMS mirror along the x-axis and slow linear motor-stage scanning of the sample along the y-axis at a speed of 2–4 mm/s. In PAM, a 3 μm lateral resolution at the optical focus and a 15 μm axial resolution have been achieved in clear media. The axial resolution of PAM is jointly determined by the laser pulse width, the frequency-dependent acoustic attenuation in tissue, and the frequency response of the ultrasonic transducer. The maximum in-focus scanning range is ~3.0 mm along the x-axis, with a cross-sectional frame rate of 400 Hz. When necessary, additional depth-scanning of the optical focal zone extends the focal range of PAM at the expense of imaging speed. By steering both the optical and acoustic axes simultaneously, PAM maintains confocal alignment and high detection sensitivity over the field of view. The two lasers are triggered with a time interval of 500 ns. The 500 ns delay allows the first PA signal to travel 0.75 mm, which is approximately the maximum penetration depth of PAM in the brain.
PAM of oxygen saturation (sO<sub>2</sub>)
In PAM, the two lasers emit the same pulse energy at 532 nm. Because the picosecond pulse has a higher peak intensity than the nanosecond pulse, it results in more saturation (Figs. S3a–b). We define a saturation factor as the ratio of the PA amplitudes under picosecond and nanosecond excitations (Supplementary Fig. 3c).
The relative concentrations of HbR and HbO2 can be estimated by solving the following linear equations:
Pns=ln(10)kF(εHbRCHbR+εHbO2CHbO2),Pps=ln(10)kF[rHbR(F)εHbRCHbR+rHbO2(F)εHbO2CHbO2], where Pns and Pps are the PA amplitudes under nanosecond and picosecond excitations, respectively; k is a proportionality coefficient related to the detection system, the are the molar Grüneisen parameter, and the nonradiative quantum yield; εHbR and εHbO2 extinction coefficients of HbR and HbO2, respectively; CHbR and CHbO2 are the molar concentrations of HbR and HbO2, respectively; and rHbR and rHbO2 are the saturation factors of HbR and HbO2, respectively, which are functions of the local fluence F (i.e., the photon energy imposed over a unit area).
Note that εHbR ≈ εHbO2 at 532 nm and rHbR ≈ 1, Eqs. (1)–(2) are reduced to
Pns≈k1(CHbR+CHbO2),Pps≈k1[CHbR+rHbO2(F)CHbO2], where k1 = ln(10)kεHbR F. Once Eqs. (3)–(4) are solved, the total hemoglobin concentration CHbT is computed by CHbT = CHbR + CHbO2.
Therefore, sO2 can be computed as
sO2=CHbO2CHbT≈1-PpsPns1-rHbO2(F)=k2(1-PpsPns), where k2 = 1/[1 − rHbO2 (F)].
From Eq. (5), we can see that rHbO2 is needed for absolute sO2 measurement. rHbO2 is calibrated for using the local fluence F (Supplementary Fig. 3c), which is proportional to the unsaturated PA signal Pns (Supplementary Fig. 3b). Note that the local fluence change due to the varied laser spot size at different depth is also accounted for by Pns. rHbO2 can also be calibrated for according to the neighboring arteries.
In studies where only fractional changes in sO2 are of interest, we have
ΔsO2sO2≈Δ(PpsPns)1-PpsPns.
Eq. (6) shows that local fluence is not needed for measuring fractional changes in sO2, as long as the saturation is sufficient.
PAM of cerebral metabolic rate of oxygen (CMRO<sub>2</sub>)
If the cortical region of interest has well-defined feeding arteries and draining veins, and the volumetric blood flow rates in the feeding and draining vessels are conserved, CMRO2 can be calculated as
CMRO2=ξ×(sO2-artery-sO2-vein)×CHbT×CBF/W, where ξ is the oxygen binding capacity of hemoglobin (1.36 mL O2/gram hemoglobin or 87.7 L O2/mol hemoglobin); sO2−artery and sO2−vein are the oxygen saturation averaged in the artery and vein, respectively; CBF is the volumetric blood flow rate (L/s); and W is the weight of the region of interest (grams).
Oxygen extraction fraction (OEF) is defined as the fractional difference between the arterial and venous oxygen saturation:
OEF=sO2-artery-sO2-veinsO2-artery.
Under normal conditions, arterial blood is close to fully oxygenated (i.e., sO2−artery ≈ 1). We can rewrite Eq. (7) as
From Eq. (11), we can calculate the fractional change in CMRO2 from the fractional changes in other parameters measured individually.
Experimental animals
Six female ND4 Swiss Webster mice (Harlan Laboratory, Inc.; 16–30 g, 3–10 weeks old) were used for the studies. The laboratory animal protocols were approved by the Animal Studies Committee of Washington University in St. Louis. During the experiment, the mouse’s temperature was kept at 37 °C by a water-circulating heating pad. An intraperitoneal dose of 100 mg/kg α-Chloralose was used for anesthesia, which had a relatively mild effect on the brain functions 14. The mouse was then taped to a lab-made animal holder, which was mounted to the PAM system. The head of the mouse was fixed in a stereotaxic frame. Before imaging, the scalp was surgically removed, while the skull was left intact. Bloodstains on the skull surface—even when invisible to naked eyes—could generate strong photoacoustic signals; thus, the exposed skull surface was carefully cleaned with phosphate buffered saline (PBS) solution. Ultrasound gel was then applied on the skull surface to retain moisture and couple the acoustic signals. A water tank filled with de-ionized water was then placed on top of the mouse head. The membrane at the bottom of the water tank was in gentle contact with the ultrasound gel. The translation of the animal holder by the motor-stage at a speed of 2–4 mm/s did not induce significant disturbance to the animal and the water in the tank.
Electrical stimulations to hindlimbs
Electrical stimulations were introduced by two pairs of needle electrodes inserted under the skin of the right and left hindlimbs, respectively. The electrodes were connected to a function generator (DS345, Stanford Research Systems, Inc.) through a manual switch. The whole procedure consisted of five periods and lasted for five minutes. The first, third and fifth periods were resting states, while the second period, from 60 seconds to 80 seconds, was left hindlimb stimulation, and the fourth period, from 200 seconds to 220 seconds, was right hindlimb stimulation. Each stimulation period consisted of a train of electrical pulses with an amplitude of 2 mA, a pulse width of 0.25 millisecond and a repetition rate of 2 Hz. The stimulation period and intensity were controlled without inducing any paw motions. Five trials were performed on each mouse.
Automatic vessel segmentation
The PA signal amplitude was extracted through the Hilbert transformation of each 1D depth-resolved signal. The data was then processed using a customized vessel-segmentation algorithm. All the data analysis was performed in 3D. Briefly, a cross-sectional image was first converted to a binary image, where the cross-section of each vessel was identified and labeled. By tracking the cross-sections of each vessel throughout all the cross-sectional images, the vessels were individually labeled and thus segmented. All the segmented vessels were visually evaluated and corrected if necessary. The final segmentation information was then stored for future use. The vessel segmentation algorithm can be performed along different orientations.
The vessel segmentation was used to measure blood vessel diameters and correct the sO2 calculation. To measure the vessel diameter, we measured the length of a line across the vessel at different angles relative to the vessel’s axis, and chose the shortest path length as the vessel diameter. To correct the sO2 calculation, any overlapping vessels were separated at the junction through segmentation, and the sO2 of each vessel was calculated individually.
PAM of blood flow speed
Line scanning along the axis of a vessel can be used to measure the blood flow speed. Moving RBCs imaged at a sufficient rate appear as bright–dark streaks in the resulting space–time map. The slope of the bright–dark streaks, measured from the vertical direction, is proportional to the centerline (axial) flow speed. The slope is measured by using a 2D Fourier transformation of the space–time map. The flow direction can be determined from the sign of the slope and the direction of the line scan sweep. To measure high flow speeds more accurately, we imaged the same cells with at least five line scans. Therefore, the maximum measurable flow speed corresponds to a travel distance of 3 mm within 12.5 ms, which translates to a flow speed of ~24 cm/s.
H&E histology
The right hemisphere of a mouse was imaged by PAM with a picosecond pulse energy of 1 μJ and a pulse repetition rate of 500 kHz. Immediately after the imaging, the mouse was transcardially perfused with 0.9% saline followed by 4% paraformaldehyde in PBS. The brain was removed and postfixed in 4% paraformaldehyde for 24 h. Coronal sections (5 μm thick) were cut with paraffin embedding. Standard H&E staining was performed on the sections, which were examined using bright-field microscopy (NanoZoomer, Hamamatsu) with a 20× objective (NA = 0.67). In the positive control experiment, the left hemisphere of a mouse was illuminated by a continuous-wave laser (GM-CF02-100, Information Unlimited, Inc.; wavelength: 532 nm; power: 100 mW; spot size: ~0.25 mm2) for one minute to induce thermal coagulation (i.e., burn). The right hemisphere was imaged by PAM for one minute with a pulse energy of 1 μJ and a pulse repetition rate of 500 kHz. H&E histology was then performed on the brain with the same procedure as above.
Two-photon microscopy imaging
To rule out the potential for causing vessel leakage by PAM, a mouse was imaged by two-photon microscopy (Fluoview 1000, Olympus, Inc.) after the PAM imaging (pulse energy: 1 μJ; pulse repetition rate: 500 kHz). For two-photon microscopy, the skull was thinned to ~30 μm using a dental drill and a microsurgical blade as previously described 14. FITC-dextran solution in PBS (150 μL, 2.5% w/v) was injected via a tail vein before the two-photon imaging. A 4× objective (NA = 0.10) was used to find the same imaging area as that in PAM, and then a 20× objective (NA = 0.70) was used to acquire high-resolution images (excitation wavelength: 800 nm; emission filter wavelength: 495–540 nm). Depth-scanning was performed from the skull surface to a depth of 600 μm into the cortex, with a step size of 5 μm. The same procedure was also used for validating PAM by using two-photon microscopy (Supplementary Fig. 5).
Supplementary Material
Supplementary Table 1. Summary of PAM parameters for mouse brain imaging.
Supplementary Fig. 1. Sequence of electrical stimulations and schematic of MEMS mirror in PAM. (a) Time course of the electrical stimulations to the mouse hindlimbs, consisting of three resting periods and two stimulation periods. Each stimulation period consisted of a train of electrical pulses with an amplitude of 2 mA, a pulse width of 0.25 ms and a repetition rate of 2 Hz. (b) Top view of the water-immersible MEMS scanning mirror, showing the elliptical mirror plate and the supporting hinges. PF, polymer frame. (c) Side view of the mirror, showing the mirror plate driven by the actuation force generated by two permanent magnets and an inductor coil. AF, actuation force; AH, acrylic holder; PM, permanent magnet.
Supplementary Fig. 2. Simulation of PA signal generation and detection in PAM with different excitation pulse widths. (a) Generated PA signal profiles from a 3-μm-diameter sphere excited by a 3-ns pulse (ns PA) and a 3-ps pulse (ps PA) with the same pulse energy. (b) Frequency spectra of the generated PA signals. The picosecond excitation is more efficient in generating high frequency signals (>70 MHz). (c) Frequency spectra of detected PA signals by the ultrasonic transducer. The PA signals are attenuated during propagation. The PA signals generated with the two excitation pulse widths are nearly identical at the detector after propagating ~1 mm in brain tissue (acoustic attenuation: 0.8 dB/cm/MHz) and ~6 mm in water (acoustic attenuation: 0.002 dB/cm/MHz).
Supplementary Fig. 3. Absorption saturation of oxy- and deoxy-hemoglobin (HbO2 and HbR). (a–b) PA amplitudes of HbO2 (a) and HbR (b), as a function of excitation pulse energy with 3-ps and 3-ns pulse widths. To mimic the in vivo studies, the laser beam was focused at ~250 μm beneath the blood surface of a 300-μm-diameter tube filled with whole bovine blood. Note that the light can only penetrate ~40 μm into the blood. The laser spot size on the sample surface was ~50 μm. The results show that HbO2 can be saturated more with picosecond excitation. (c) Saturation factors of HbO2 and HbR. The saturation factor is defined as the ratio of the PA amplitude with picosecond excitation to that with nanosecond excitation.
Supplementary Fig. 4. The average PA amplitude decay with the increasing depth, in natural logarithmic scale. The PA amplitude at a particular depth was represented by the average value of the brightest 0.1% of the pixels from that depth. From the exponential fit to the data, we were able to determine the characteristic attenuation length (CAL), which corresponds to a 1/e fold decay in the PA amplitude. A fast decay component with a CAL of ~200 μm was contributed by the signals from the optical resolution volume, so called Class I signals, and a slow decay component with a CAL of ~500 μm was contributed by the signals from the portion of the acoustic resolution volume that lies outside the optical resolution volume, so called Class II signals. The optical resolution volume in PAM is defined by the cylinder whose base is the optical focal spot and whose height is the acoustic axial resolution. Similarly, the acoustic resolution volume in PAM is defined by the cylinder whose base is the acoustic focal spot and whose height is the acoustic axial resolution.
Supplementary Fig. 5. Comparison of the brain vasculature imaged by two-photon microscopy and PAM. The same mouse was first imaged by PAM, using hemoglobin as the endogenous contrast (left column), and then imaged by two-photon microscopy, using FITC-dextran as the exogenous contrast (right column). Three representative 100-μm-thick layers are shown. Each image was normalized by its own maximum intensity. The results show that PAM and two-photon microscopy have comparable image quality for the pia vessels (<200 μm deep). However, PAM cannot resolve the deep capillary beds as well as two-photon microscopy due to the use of a short wavelength for photoacoustic generation.
Supplementary Fig. 6. Effect of mouse skull on the optical focusing and acoustic transmission in PAM. (a) A CCD image of the optical focus in water, showing a focal spot size of ~3 μm. (b) A CCD image of the optical focus with the laser beam passing through a piece of mouse skull. (c) Light intensity profiles along the dashed lines in (a) and (b), showing that the skull induced marginal degradation in optical focusing. (d) A PAM image of a piece of black tape partially covered by a piece of mouse skull. (e) PA signal profile along the dashed line in (d), showing the signal attenuation by the skull.
Supplementary Fig. 7. PAM of a mouse brain with and without the skull and dura. (a–b) En face PAM images of a 2×2 mm2 region in a mouse brain acquired with the skull intact (a), and the skull and dura removed (b). With the skull and dura removal, large vessels were imaged with greater signal-to-noise ratios, more deep capillaries were imaged, but the spatial resolution was only marginally improved. Some vessels in (a), indicated by the arrows, were absent in (b). (c) An x-z projection of (a), showing that the absent vessels were from the surface of the brain.
Supplementary Fig. 8. Comparative sO2 measurements in a major cortical artery-vein pair, by using the pulse-width-based method (PW-sO2) and two-wavelength-based method (TW-sO2). The TW-sO2 measurement was performed at 570 nm and 578 nm on a slow PA system with a wavelength tunable laser.
Supplementary Fig. 9. Pulse-width-based sO2 measurement on blood phantoms, with different excitation pulse energies. (a) Two 300-μm-diameter tubes were filled with whole bovine blood. The light was focused at ~250 μm beneath the sample surface. The experiment conditions were the same as those in Supplementary Fig. 3. Tube 1 and Tube 2 had an average sO2 of 100% and 45%, respectively, measured by the gas analyzer. The sO2 values in the two tubes were then measured with PAM by using the pulse-width-based method, based on the calibrations in Supplementary Fig. 3. The excitation pulse energy was adjusted from 50 nJ to 1000 nJ. (b) The PAM measured sO2 values averaged in each of the two tubes. The results show that the sO2 measurement error was <3% with pulse energies ≥300 nJ. The sO2 was underestimated by more than 50% when the pulse energy was reduced to ≤200 nJ. (c) The ratio of the sO2 values between the two tubes. As the optical fluence was identical on the two tubes, the relative value was independent of the pulse energy as long as a sufficient saturation effect was maintained. The results show that the relative sO2 measurement error was <2% with pulse energies ≥300 nJ.
Supplementary Fig. 10. Pulse-width-based sO2 measurement in vivo, with different excitation pulse energies. A 1×1 mm2 area of a mouse ear, which contained an artery-vein pair, was imaged by PAM. The sO2 in the artery-vein pair was measured by using the pulse-width-based method with excitation pulse energies from 50 nJ to 1000 nJ, based on the calibrations in Supplementary Fig. 3. The light was focused at ~250 μm beneath the tissue surface. The optical scattering above the artery-vein pair was negligible. The results show that the measured sO2 values in the artery and vein with pulse energies ≥300 nJ were in the normal physiological range, while the measured sO2 values in both the artery and vein with pulse energies ≤200 nJ were underestimated, similar to the results shown in Supplementary Fig. 9. The quantification of the sO2 measurement in vivo is shown in Fig. 1h.
Supplementary Fig. 11. PAM and bright-field microscopy of a single layer of red blood cells (RBCs). (a) A single layer of freshly drawn mouse RBCs fixed on a glass slide was imaged by PAM (pulse energy: 400 nJ). Each picosecond laser pulse was followed by a nanosecond laser pulse with a time interval of 500 ns. The PAM image acquired with the picosecond excitation is shown here. The glass slide, protected by thin plastic film, was immersed in water. To measure the spatial resolution of PAM, a sharp black-ink edge on the glass slide was also imaged. (b) Bright-field microscopy (Zeiss AxioObserver) of the same region of RBCs as (a), conducted at 10× magnification before (left) and immediately after (right) the PAM imaging, showing the densely packed RBCs and their donut shapes after the PAM imaging. (c) Spatial resolution measurement of PAM. The edge spread function (ESF) (solid red line) of PAM was extracted by fitting the ink edge signals (solid black dots) with an error function. The line spread function (LSF) (dashed blue line) of PAM was then computed as the first derivative of the ESF. The lateral spatial resolution of PAM was measured as the full-width-at-half-maximum (FWHM) of the LSF, ~3.4 μm. Note that the 1/e2 width of the LSF is ~7.0 μm. (d) A close-up PAM image of an individual RBC, as indicated by the dashed green box in (a). The donut shape of the RBC could not be resolved by PAM. (e) Bright-field microscopy of the same RBC as (d), at 50× magnification before (left) and immediately after (right) the PAM imaging, showing the cell’s intact donut shape.
Supplementary Fig. 12. Two-photon microscopy and histological examination of a mouse brain imaged by PAM. (a) Two-photon microscopy (TPM) of cortical vasculature conducted after PAM imaging, showing no vessel leakage. FITC-dextran solution in PBS (150 μL, 2.5% w/v) was injected via a tail vein before the two-photon imaging. A 20× objective (NA = 0.70) was used to acquire high-resolution TPM images (excitation wavelength: 800 nm; emission filter wavelength: 495–540 nm). (b) Standard H&E histology of the mouse brain conducted after PAM imaging, showing no tissue damage. H&E stained brain sections were examined using bright-field microscopy (NanoZoomer, Hamamatsu) at 40× magnification.
Supplementary Fig. 13. Histological examination with standard H&E staining of a mouse brain imaged by PAM. (a) A photograph of an extracted mouse brain after the blood was flushed. Before sacrificing the animal, the mouse’s left hemisphere was illuminated by a continuous-wave laser (wavelength: 532 nm; power: 100 mW; spot size: 0.25 mm2) for one minute to induce thermal coagulation (i.e., burn), as a positive control. The right hemisphere was repeatedly imaged by PAM for one minute with an excitation pulse energy of 1000 nJ and a volumetric frame rate of 1 Hz. The PAM focal plane was ~250 μm beneath the skull surface. After the PAM imaging, the mouse brain was extracted, fixed, sectioned and stained with standard H&E procedures. (b) Coronal slices of the mouse brain with H&E staining imaged under a bright-field optical microscope at 0.75× magnification. Five representative slices were selected inside and outside the burned area and the imaged area, as indicated by the dashed lines in (a). (c) Close-up images of the burned area (left column, red boxes in (b)) and the PAM imaged area (right column, blue boxes in (b)) at 40× magnification. Clear burn boundaries and vessel clots were observed inside the burned area in the left hemisphere, while no damage was observed outside the burn area in the left hemisphere or inside the imaged area in the right hemisphere.
Supplementary Fig. 14. Quantification of mouse brain hemodynamic responses to electrical stimulations (n = 6). (a) Time courses of the spatially averaged fractional PA amplitude changes in the right hemisphere (RH) and left hemisphere (LH). The gray boxes outline the stimulation periods. LHS/RHS, left/right hindlimb stimulations. (b) Statistics of the peak fractional PA amplitude changes in the RH and LH. PFC, peak fractional change. (c) Time courses of the fractional arterial diameter changes in the RH and LH. (d) Statistics of the peak fractional changes in arterial diameters. (e) Time courses of the spatially averaged flow speeds in a selected artery-vein pair in the RH in response to the LHS. The data in (a)–(e) are averaged over five trials on each of the six mice, and the error bars denote standard errors. Statistics: paired student’s t-test. P values: *** <0.001; **<0.05.
Supplementary Fig. 15. Quantification of vessel-type-based hemodynamic responses to electrical stimulations to hindlimbs. (a) Representative arteries, veins and capillary beds selected for quantification. (b) Time courses of the changes in total hemoglobin concentration in selected arteries, veins and deep capillary beds. (c) Time courses of the changes in vessel diameters in selected arteries and veins. The diameter changes in capillaries were unresolvable. (d) Time courses of the changes in blood flow speeds in selected arteries and veins. PAM could not resolve the flow speed changes in deep capillaries. (e) Time courses of the changes in sO2 in selected arteries, veins and deep capillary beds.
Supplementary Fig. 16. PAM of blood flow speed by fast line scanning. (a) PAM image of the somatosensory area in the right hemisphere. To measure the blood flow speed, fast line scanning (400 Hz) was performed along a vein and an artery, marked by the dashed lines, before and during the electrical stimulations to the left hindlimb. (b) Left panel: fast line scanning space–time plot acquired along the vein; Right panel: the time traces of the flow speeds in the artery and in the vein, showing the pulsations of arterial blood flow. (c) Fast line scanning space–time plots before (left panel) and during (right panel) electrical stimulations. The blood flow speeds were calculated from the slopes of the bright–dark streaks. (d) The flow speed can be estimated via a 2D Fourier transformation, where the slope between the temporal and spatial frequencies reflects the flow speed and direction. PSD, power spectral density.
Supplementary Fig. 17. Quantification of mouse brain oxygenation responses to electrical stimulations (n = 6). (a) Time courses of the spatially averaged fractional sO2 changes in the three selected subregions shown in Figure 2c. The gray boxes outline the stimulation periods. (b) Statistics of the peak fractional sO2 changes in the three subregions in response to the first stimulation. PFC, peak fractional change. (c) Depth-resolved sO2 responses (shown in color) superimposed on the y-z projected vascular image (shown in gray). The core responding region is marked by the dashed circle. The data in (a)–(b) are averaged over five trials on each of the six mice, and the error bars denote standard errors. Statistics: paired student’s t-test. P values: *** <0.001.
Supplementary Video 1
Three-dimensional rendering of mouse cortical vasculature imaged by PAM with an intact skull.
Supplementary Video 2
Cross-sectional rendering of the mouse brain vasculature. Cross-sectional rendering of the mouse brain vasculature below a 0.6×0.6 mm2 surface region, a composite image from 12 depth-scans of the optical focal zone. The shadows in deeper tissue were due to the blocking of light by superficial blood vessels. MAP, maximum amplitude projection.
Supplementary Video 3
PAM of cerebral responses to electrical stimulations to the hindlimbs. A 3×4 mm2 cortical area covering the somatosensory regions of both hemispheres was imaged at a volumetric rate of 1 Hz. Fractional PA amplitude changes (shown in yellow), in response to the left hindlimb stimulation (LHS) and right hindlimb stimulation (RHS), were superimposed on the vascular image (shown in red). The curve at the bottom of the video shows the average PA signal amplitude in the somatosensory regions.
Supplementary Video 4
Volumetric rendering of oxygen saturation (sO2). sO2 was imaged below a 3×2 mm2 surface region of a mouse brain at resting state. The sO2 values were color-coded from blue (low oxygenation) to red (high oxygenation).
Supplementary Video 5
PAM of cerebral sO2 responses to electrical stimulations to the left hindlimb. Left panel: sO2 map of a 2×3 mm2 cortical region covering the somatosensory region of the right hemisphere, acquired at a volumetric rate of 1 Hz. Right panel: close-up of the core responding region, marked by the dashed box in the left panel. The curve at the right bottom of the video shows the average sO2 in the close-up region. Note that the resting-state frames acquired between 120 seconds and 200 seconds are omitted in the video.
The authors appreciate J. Ballard’s close reading of the manuscript. We thank C. Li, C. Chen, W. Chapman and J. Lee for discussions and technical support. This work was sponsored by US National Institutes of Health (NIH) grants DP1 EB016986 (NIH Director’s Pioneer Award), R01 CA186567 (NIH Director’s Transformative Research Award), 1S10 RR028864 and R01 CA159959 (for L. V. Wang), and US National Science Foundation (NSF) grant CMMI-1131758 (for J. Zou). The bright-field microscopy was performed at the Alafi Neuroimaging Laboratory of the Hope Center for Neurological Disorders, Washington University School of Medicine, which is supported by the NIH Neuroscience Blueprint Center Core Grant P30 NS057105.
AUTHOR CONTRIBUTIONS
J.Y., J.Z. and L.V.W. conceived and designed the study. J.Y., J.M.Y., K.M., L.W. and C.H. constructed the imaging system. J.Y., T.W., and L.L. performed the experiments and analyzed the data. J.Y., L.W. and L.V.W. wrote the manuscript. L.V.W. supervised the whole study. All authors discussed the conceptual and practical implications of the methods.
COMPETING FINANCIAL INTERESTS
L. V. Wang has a financial interest in Endra, Inc., and Microphotoacoustics, Inc., which, however, did not support this work. K. I. Maslov also has a financial interest in Microphotoacoustics, Inc. The other authors declare no competing financial interests.
Fast functional photoacoustic microscopy (PAM) of the mouse brain
(a) Schematic of the PAM system. OAC, optical-acoustic combiner; PBS, polarizing beam splitter; UT, ultrasonic transducer. (b) Scheme of PAM scanning. 3D imaging is achieved by fast MEMS mirror scanning along the x-axis and slow motor-stage scanning along the y-axis. (c) Sequence of PAM excitation and detection. The picosecond pulse incident on oxy-hemoglobin (HbO2) results in more saturation and thus a weaker PA signal than the following nanosecond pulse, whereas the difference for deoxy-hemoglobin (HbR) is negligible. (d) A representative x-y projected brain vasculature image through an intact skull (n = 6). (e) A representative enhanced x-z projected brain vasculature image acquired over a 0.6×0.6 mm2 region with depth scanning, where the signal amplitude was normalized depth-wise (n = 6). (f) PAM of oxygen saturation of hemoglobin (sO2) in the same mouse brain as (d), acquired by using the single-wavelength pulse-width-based method (PW-sO2) with two lasers. The averaged sO2 in the skull vessels was lower than that in the cortical vessels. SV, skull vessel. (g) Comparison of the PW-sO2 measurements in four blood phantoms and the gas analyzer readings. (h) In vivo PW-sO2 measurements in an artery-vein pair in a mouse ear with varied excitation pulse energies. The data in (g) and (h) are averaged within the samples, and the error bars are standard deviations.
PAM of brain responses to electrical stimulations to the hindlimbs of mice (<italic>n</italic> = 6)
(a) Fractional PA amplitude changes (shown in yellow) in response to left hindlimb stimulation (LHS) and right hindlimb stimulation (RHS), superimposed on the vascular image (shown in red). LH/RH, left/right hemisphere. (b) Depth-resolved PA amplitude responses. The responding areas in the LH and RH are shown in red and blue, respectively, and superimposed on the gray-scale y-z projection image. The signal amplitude in the y-z projection image was normalized depth-wise. (c) Fast sO2 imaging before (left panel) and during (right panel) stimulations to the left hindlimb. Three 0.3×0.3 mm2 subregions (i, ii and iii) are further analyzed (Supplementary Fig. 17). (d) Time courses of the fractional changes in the cerebral blood flow (CBF), oxygen extraction fraction (OEF), and cerebral metabolic rate of oxygen (CMRO2) in the core responding region. All the sO2 measurements were acquired with two lasers. The data in (d) are averaged over five trials on each of the six mice, and the error bars are standard errors. Statistics: paired student’s t-test. P values: *** <0.001.