The brain is an important organ that governs the physiological activities of humans. Neuroscience is based on the whole brain and focuses on revealing the brain functions that are associated with all kinds of phenomena and the pathogenesis of biological brain diseases^[1–4]. Magnetic resonance imaging (MRI)^[5], functional ultrasound imaging (FUS)^[6], and two-photon microscopy (TPM)^[7] are some of the common imaging modalities used in mouse model brain studies. Conventional MRI is primarily used for anatomical and functional imaging of centimeter-scale specimens, with a spatial resolution of approximately 1 mm^[8]. Although high-Tesla MRI has higher spatial resolution, the system becomes bulky and more expensive^[9]. FUS is able to dynamically image deep brain activity by directly measuring small changes in cerebral blood volume caused by neurovascular coupling^[6]. In particular, FUS is insensitive to blood flow parallel to the probe surface due to the angular dependence of the Doppler effect and requires the use of additional contrast media^[9]. TPM allows functional neuroimaging with single-cell resolution to visualize neurovascular coupling activities in vivo^[7]. However, the disadvantage of TPM lies in its poor penetration depth and limited field of view (FOV)^[10]. In sum, a brain imaging technique with a high spatiotemporal resolution, a wide FOV, and without labeling mechanisms is needed to better understand brain structure and function.

Photoacoustic microscopy (PAM)^[11,12], a branch of photoacoustic imaging (PAI) that combines the advantages of optical resolution and acoustic penetration depth^[13–17], has been used to research whole brain metabolism^[18], resting-state connectivity^[19], and brain-related disorders (stroke, Alzheimer’s disease, brain tumors, and epilepsy)^[20–22]. Most conventional PAM systems utilize a mechanical scanning scheme with a two-axis linear platform capable of scanning the entire mouse cerebral cortex with fine lateral resolution. Although this method has an unrestricted FOV, the low imaging speed makes it difficult to acquire dynamic information in the brain^[23]. In order to speed up the imaging cycle, a combined acousto-optical scanning method has been proposed, which uses a 1-axis/2-axis mirror scanner [e.g.,  a galvanometer scanner or a microelectromechanical systems (MEMS) scanner] to achieve rapid scanning of both the acoustic and optical beams^[24–26]. This method captures the rapid oxygen saturation of cerebral hemoglobin at high imaging speeds (∼hundreds of kHz)^[27,28]. Inevitably, coupling media such as water can reduce the stability of the mirror scanner at high speeds, which can lead to reduced image quality at high imaging speeds^[23]. Purely optical scanning uses static photoacoustic (PA) signal detection, where the sensor remains fixed relative to the sample to receive the signal and the scanner deflects the beam above the sensor^[29]. This method avoids the effects of water on the scanner and provides more stable and faster imaging than combined acousto-optical scanning^[30]. It is important to note that this method has an FOV of only a few millimeters^[31]. In brief, it remains a challenge for PAM to achieve both high-speed and large FOV imaging.

In addition, since ultrasonic transducers are usually opaque to light, an acoustic-optical combiner is usually required to colocalize the excitation and acoustic beams, which can complicate the optical path and make it difficult to combine with other optical imaging modalities. These problems can be overcome using transparent sensors, such as optical microring resonators and transparent ultrasonic transducers (TUTs). Li et al.^[32] proposed the use of an optical microring resonator for PA scanning of the whole mouse brain, which effectively simplified the structure of the system because the laser beams could travel directly to the brain through the resonator. Although optical detectors have good sensitivity and bandwidth, additional equipment for the system and the high cost of the sensors limit their applications. Chen et al.^[33] imaged the cerebral cortex of non-anesthetized mice by implanting a TUT into the thinning skull. However, due to the limited size of the TUT, only a small area (1.5mm×1.5mm) could be scanned, which is unfavorable for whole-brain imaging.

In this Letter, a fast and large FOV TUT-based PAM system is presented and used for in vivo whole-brain imaging in mice. The system combines a TUT with a two-dimensional scanning galvanometer, avoids the use of an acoustic-optical combiner, and has an FOV of 20mm×20mm and an imaging speed in the purely optical scanning mode. The utility of TUT-based PAM for brain imaging in small animals is demonstrated by imaging the vascular structure of a mouse brain injury model, a nude mouse glioma model, and by detecting hemodynamic changes in the brain after cerebral hemorrhage in mice.

Figure 1(a) depicts the schematic view of the PAM system. The laser beams were excited by an Nd:YAG solid laser (PR-532-8-A, 532 nm, 10 ns, Power-Laser Inc.) with a repetition frequency (RPF) of 50 kHz. The laser beam with a diameter of 1.5 mm was then collimated and magnified four times by means of a biconvex lens (f=25mm, L1, Daheng Optics Inc.) and a plane lens (f=100mm, L2, Daheng Optics Inc.). Later, an electrically tunable lens (L3, EL-10-30-Ci-VIS-LD, Optotune Inc.) focused the beams into a 2-axis galvanometer scanner (GVS002, Thorlabs Inc.), which can be used for optically scanning the sample. The focused laser passes through the TUT, is delivered to the sample surface, and generates a PA signal. The photoacoustic signal is coupled between the TUT and the sample via a water bladder. When the signals were amplified by amplifiers (54 dB gain, YS001, Yunsheng Inc., respectively), a data acquisition system (DAQ, NI-PXIe-5160, National Instruments Inc.) digitized the PA signals with a sampling rate of 100 MS/s. Meanwhile, the triggering of the laser, the scanning of the galvanometer, and the acquisition of the PA signal are all controlled by the field programmable gate array (FPGA, AX 301, Altera Inc.). The mouse brain was imaged with a scanning step of 10 µm and a FOV of 10mm×10mm. During the testing of the system’s imaging speed and monitoring of cerebral hemorrhage in mice, the PA signals were averaged four times to obtain a faster imaging speed. For other experiments, the PA signal was averaged ten times to obtain better image quality. The laser energy in the tissue surface was ∼440nJ/pulse. By adjusting the optical focus at ∼0.1mm below the skin surface, the surface laser fluence was about 18.52mJ/cm2, which was close to the American National Standards Institute safety limit (20mJ/cm2 for visible wavelengths).

The performance characterization of TUT and the main parameters of the PAM system can be found in our previous work^[34]. In all, the TUT has a transmittance of 72.65% in the 532 nm band, a center frequency of 10 MHz, a bandwidth of 16%, an electromechanical coupling coefficient of 0.62, and a size of 33mm2. The lateral and axial resolutions of the system are 17 and 270 µm, respectively. As shown in Fig. 1(b), the leaf vein skeleton was imaged to characterize the imaging FOV of the system, and the maximum imaging FOV of the system was measured to be 20mm×20mm. The tungsten filament was inserted obliquely into an agar phantom, as shown in Fig. 1(c), to evaluate the depth imaging capability of the system. The maximum imaging depth of the system was 1.64 mm. Figure 1(d) is an optical photo of the TUT. The RPF of the laser was increased to 50 kHz, and the imaging speed of the TUT-based PAM system was further increased. The imaging speed of the system was measured by imaging the whole cortical vasculature of the mouse brain in the range of 10mm×10mm. The imaging results are shown in Fig. 1(e), with a scan time of 80 s for the case of averaging the PA signal four times.

All experimental animal procedures were conducted in accordance with the approved experimental animal protocol of the Guangdong University of Technology. BALB/c mice (aged 5–6 weeks, male) were purchased from Guangdong GemPharmtech Co., Ltd. (Foshan, China). The nude mouse (aged 5–6 weeks, male) was purchased from Guangzhou Ruige Biotechnology, Ltd. (Guangzhou, China).

Before the experiment, BALB/c mice were injected with urethane (6 mL/kg) to maintain anesthesia. Then, after placing the mice on a thermal pad, the hair on the mice’s heads was removed with a razor and depilatory cream. Next, the scalp of the mice was cut with ophthalmic scissors to expose the intact skull. To prepare the mouse brain injury model, a needle with a diameter of approximately 0.5 mm was inserted into the right cerebral cortex to a depth of approximately 1 mm. To prepare the mouse cerebral hemorrhage model, a laser with an energy of 10 µJ was focused and then irradiated on the cerebral cortical vessels of the mice for 5 s. Immediately after the model was prepared, the mice were imaged. At the end of the experiment, all mice were returned to their cages to wait for natural recovery.

To make the glioma model, 4 µL of GL261 glioma cells suspended in a phosphate-buffered saline (PBS) were injected into the right hemisphere of the nude mice at a rate of 2 µL/min after carefully drilling through the skull with a 1 mm diameter dental drill. After 5 min of injection, the needle was slowly extracted, and the bone hole was closed with bone wax. Two weeks later, our PAM system and a small animal 3D MRI system (Aspect M3, Israel) were used for imaging the nude mouse. Finally, the hematoxylin and eosin (HE)-stained section was prepared at Guangzhou Yuebin Medical Research Company.

The in vivo imaging capability of the TUT-based PAM system was demonstrated by imaging cortical blood vessels in brain-injured mice. As shown in Figs. 2(a)–2(d), the optical photos and PA images of mice before and after brain injury are inverted. The location of the mouse brain injury is circled using a thick white dotted line. The major blood vessels were imaged with sufficient lateral resolution [Fig. 2(b)]. The TUT-based PAM system makes it difficult to image microvessels with weak PA signals compared to conventional PAM, which possesses higher detection sensitivity. Following surgery for lesions on the right hemisphere of the brain, vascular morphology is disrupted, and bruising occurs at the site of injury [Fig. 2(d)]. The profiles of the blood vessels in the cross section where the thin white dashed lines in Figs. 2(b) and 2(d) are drawn, as shown in Fig. 2(e). Figure 2(e) shows that after brain injury in mice, there is a decrease in the amplitude of the vascular signals in the uninjured region [the part outside the black box in Fig. 2(e)], which may be related to the hemorrhage in the injured region. In contrast, the blood vessels in the injured region [the part of the black box in Fig. 2(e)] are damaged, and the bruising produces a strong PA signal.

The brains of nude mice implanted with glioma cells were imaged to characterize the tumor imaging capabilities of the TUT-based PAM system. As shown in Figs. 3(a)–3(d), the optical photo, the PA imaging result, the MRI imaging result, and the HE staining result of the brains of nude mice with glioma disease are shown inversely. Among them, glioma locations are circled by white dotted lines. By comparing Figs. 3(a) and 3(b), it can be seen that the blood vessels around the glioma clearly tend to grow towards the glioma and are more densely packed compared to the normal blood vessels in the left hemisphere of the mouse brain, forming the boundary of the tumor. In addition, the shape of the blank area in Fig. 3(b) is nearly triangular, which corresponds to the MRI imaging results in Fig. 3(c). In the tumor staining results, as shown in Fig. 3(d), a large number of GL261 glioma cells were stained with dark colors, while the surrounding normal cells were kept in light colors, which confirmed the authenticity of the presence of the tumor. The altered morphology of the tumor in Fig. 3(d) was caused by the anatomical sectioning of the mouse brain. The imaging experiments on glioma in nude mice indicated that the TUT-based PAM system can be used as an effective tool for imaging tumor diseases, complementary to other imaging modalities.

Monitoring of hemodynamic changes during cerebral hemorrhage in mice demonstrated the rapid imaging capability of the system. Figure 4(a) shows a series of PA MAP images 6 min 40 s after cerebral hemorrhage. The area of cerebral hemorrhage has been circled with white dotted lines in the figure. Figure 4(b) is a statistical plot of the mean normalized amplitude and the number of pixels above half the peak of the amplitude curve over time within the white elliptical region in Fig. 4(a). In the absence of the hemorrhage, the major blood vessels throughout the cerebral range are demonstrated at high resolution. After causing damage to the blood vessels by laser, a trace amount of blood flowed out of the damaged area at 1 min 20 s. Then, at 2 min 40 s, the bleeding rate reaches the fastest, and the blood spreads to the tissue around the blood vessel. Then, at 5 min 20 s, the blood continued to flow out, but the bleeding rate slowed down. At 6 min 40 s, the signal produced by the blood decreased, which may be related to blood coagulation. The imaging experiments on cerebral hemorrhage in mice demonstrated the ability of our system to dynamically monitor the changes in hemodynamics in the brain and also showed the potential of our system in monitoring the early development of some brain-related diseases.

In this study, a TUT-based wide-field PAM system was developed to study the in vivo vascular structure and hemodynamics of the whole brain range in mice. This PAM system has an imaging range of up to 20mm×20mm and an imaging speed of 50 kA-line/s. In addition, a TUT-based PAM system eliminates the need for an acoustic-optical combiner, making it simpler to configure and more suitable for integration with other imaging modalities than traditional PAM systems. The structural changes in the vascular system of the mouse cerebral cortex induced by injury have been analyzed through the study of a mouse brain injury model. Imaging of a mouse glioma model demonstrated that the TUT-based PAM system for tumor imaging was feasible, and the system could be used as a tumor imaging tool complementary to other imaging modalities. The mouse cerebral hemorrhage experiments demonstrated the ability of the system to image cerebral hemodynamics in mice.

Although promising, the performance of the TUT-based PAM system needs to be improved in several aspects. First of all, the imaging speed of the TUT-based PAM system is limited only by the laser RPF, which can be further increased by increasing the laser repetition frequency to a few hundred kHz, which would allow real-time monitoring of fast hemodynamic responses in the brain, such as neurovascular coupling activity induced by fast electrical stimulation. Second, the lateral resolution of the system is 17 µm, which may result in a system that is insufficient for fine resolution of capillaries. Increasing the numerical aperture of the objective lens is the best way to solve this problem. Furthermore, the system can be further extended, for example, to image blood oxygenation using a dual-wavelength laser system. Alternatively, imaging modalities such as ultrasound imaging and optical imaging can also be combined with the system to form a multimodal system^[3]. In addition to imaging, lasers can also be used to produce and treat various disease models of the mouse brain via TUT, such as cerebral ischemia^[35] and brain tumors^[36]. In future work, the system can be used to carry out whole-brain imaging experiments on the cranium of large animals with more cerebral volume, such as whole-brain vascular characterization, functional activity in multiple brain regions, and ultrasound modulation monitoring of brain function at multiple targets.

In conclusion, this work presents a novel TUT-based PAM for whole-brain imaging and cerebral hemodynamic studies in mice. The system has superior spatial and temporal resolution as well as a centimeter-level imaging FOV, which is difficult to achieve with other medical imaging modalities. The potential of whole-brain imaging was demonstrated by three experiments on brain injury, brain tumor, and brain hemorrhage. These experiments demonstrate that the system is able to detect related diseases in the mouse cerebral cortex and also has certain dynamic imaging capability, which is suitable for studying the neurovascular coupling activities associated with certain brain diseases.