3.1.

Photoacoustic Microscopy Systems

According to the imaging depth and the spatial resolution, PAM can be classified into two groups: optical-resolution PAM (OR-PAM) and acoustic-resolution (AR-PAM) (as shown in Fig. 1). In OR-PAM, the lateral resolution is defined by the optical focus, which ranges from submicrometer to a few micrometers. Because OR-PAM relies on ballistic and/or quasiballistic photons, the imaging depth is less than one transport mean free path (TMFP, typically $<1\mathrm{mm}$ in the brain). This depth is sufficient to cover the mouse cortex in an open scalp experiment. AR-PAM relies on diffusive light to form an image and its lateral resolution is defined by the acoustic focus. The imaging depth of AR-PAM is mainly determined by the penetration depth of the excitation wavelength and may vary from 1 mm (blue to green light) to a few mm (red to NIR light). AR-PAM can image the mouse brain cortex through an intact scalp. For both OR-PAM and AR-PAM, the axial resolution is defined by the bandwidth and central frequency of the ultrasound transducer. OR-PAM systems typically use higher central frequency transducers ($\sim 50\mathrm{MHz}$), which provide better axial resolutions. Image reconstruction in PAM systems is relatively simple because the spherically focused transducer mainly captures signals along the acoustic axis. Therefore, the PAM system forms an A-line of data at each scanning position. Combining all A-lines from two-dimensional (2-D) raster scan forms a three-dimensional (3-D) image.

Fig. 1

Schematics of (a) optical-resolution photoacoustic microscopy and (b) acoustic-resolution photoacoustic microscopy.

While PAM systems play an essential role in high-resolution functional brain imaging,^51–53 they were rarely used in NP studies, probably because (1) the shallow imaging depth does not need highly NIR absorbing NPs, and (2) the raster-scan requires high pulse repetition rate ($>1\mathrm{kHz}$) lasers, which typically lacks the wavelength tunability. Array-based PAM systems have recently been developed and may enable low-repetition rate lasers to be used in PAM.^54–56

3.2.

Photoacoustic Computed Tomography Systems

Both PACT and AR-PAM rely on diffusive photons to generate PA signals. However, transducers used in PACT are unfocused toward the imaging plane and thus require computed tomography algorithms to form an image. For systems with matching numbers of data acquisition channels and array elements, a 2-D or 3-D image can be generated from a single laser shot.⁵⁷ Most PACT systems use high-power Nd:YAG lasers, operating at 10 or 20 Hz. Wavelength tunability can be achieved through an optical parametric oscillator (OPO), a dye laser, or a Ti-sapphire laser. Based on the detection geometry, PACT can be divided into five categories: cylindrical-view PACT,⁵⁸ planar-view PACT,⁵⁹ spherical-view PACT,⁶⁰ linear-view PACT,²³ and circular-view PACT ⁶¹ (Fig. 2). For the first three geometries, unfocused transducer elements are evenly, or semievenly, distributed on the scanning surface, and 3-D images can be formed with near-isotropic spatial resolution. Due to the complete 3-D coverage, exact analytical image reconstruction algorithms, such as the universal back-projection algorithm, can be used in these 3-D geometries.⁶² The latter two geometries have a 2-D detection coverage and exact image reconstruction is still challenging.^63,64 In practice, because linear and circular arrays are cylindrically focused on the 2-D detection plane, most researchers simply use the 3-D-based reconstruction algorithm in their 2-D systems. While the reconstruction may not be quantitatively accurate, the approximation is adequate for recovering dynamic changes in the brain or for extracting strong NP absorption from the hemoglobin background. 3-D PACT systems are commonly used in volumetric neuroimaging,⁶⁰ while 2-D PACT systems are typically used for coronal- (linear-view PACT) or transverse-view (circular-view PACT) neuroimaging. It should be noted that 2-D PACT systems can also generate 3-D images by scanning vertically along the imaging plane. However, the spatial resolution along the third dimension will be poorer than their inplane resolutions.⁶⁵ Advanced reconstruction algorithms or scanning geometries are needed to resolve this issue.^65,66

Fig. 2

Schematics of: (a) cylindrical-view photoacoustic computed tomography (PACT), (b) planar-view PACT, (c) spherical-view PACT, (d) linear-view PACT, and (d) circular-view PACT.

4.1.

Contrast Enhancement

As previously mentioned, hemoglobin is the major intrinsic PA contrast, which enables label-free functional ^67,68 and vascular brain imaging.⁹ However, in the NIR region, the optical absorption of hemoglobin is relatively weak, and contrasts from deep brain vessels are impeded by the surrounding tissue. NPs have been used to address these limitations because their absorption peaks can be systematically varied from visible to the NIR region. Among various types of NPs, gold NPs have been extensively used for PA contrast enhancement due to their strong NIR absorption, good biocompatibility, and durable structure. So far, four types of gold NPs have been used in PAT. They are nanoshells, nanocages, hollow nanospheres, and nanostars. Gold nanoshells are inorganic, optically tunable NPs consisting of a dielectric core (silica) surrounded by a thin-metallic layer (gold). Their optical resonance can vary precisely over a wide spectral range by adjusting the core size and thickness of the gold shell.⁴⁰ Nanocages and hollow nanospheres are both hollow nanostructures, which exhibit improved surface plasmon resonance absorption over core-shell structures.⁶⁹ Gold nanocages can be created through the reaction of Ag nanocubes and ${\mathrm{HAuCl}}_4$, and their name originates from the hollow and porous structures.⁷⁰ Hollow gold nanospheres have a highly uniform structure and exhibit similar optical absorption properties as nanocages.^69,71 Gold nanostars contain multiple sharp tips, which create a “lightning rod” effect to further enhance the local surface plasmon effort.⁷²

For in vivo imaging, gold NPs are commonly coated with polyethylene glycol (PEG) to suppress the immunogenic response and subsequently improve the biostability.⁷³ The feasibility of using PEG-coated nanoshells to enhance PAT was first demonstrated by Wang et al.⁴⁰ in rat brain imaging. The nanoshells used in the experiment had a 10 to 12-nm thick gold shell with peak absorption at $800\pm 5\mathrm{nm}$. The experiment was performed on a circular-view PACT system with a 2-mm-diameter nonfocused ultrasonic transducer scanning around the rat’s head. A tunable dye laser with an 800-nm emission and a 10-Hz pulse repetition rate was used as the excitation source. Due to the slow mechanical scanning, acquiring each slice took 24 min. After three sequential nanoshells injections through the tail vein, the absorption of blood was increased by $\sim 63\%$ [Figs. 3(a) and 3(b)]. Similar setups have also been used to image contrast enhancements from gold nanocages ⁷⁴ and hollow gold nanospheres.⁴¹

Fig. 3

Noninvasive photoacoustic (PA) neuroimaging employing gold nanoparticles (NPs). (a) and (d) PA images acquired through an intact rat scalp before and 20 min after tail vein injection of gold nanoshells, respectively. The experiment was performed on a single-element–based circular-view PACT system. (c) PA image acquired through an intact mouse scalp after nanostar injection. R: rostral rhinal vein. S: sagittal sinus. T: transverse sinus. (d) PA intensity (normalized to the first frame) of sagittal sinus calculated from each frame. Inset, normalized PA intensity monitored for the initial 600 s. The experiment was performed on an array-based circular-view PACT system. Reproduced with permission from Refs. 40 and 72.

Recently, an array-based circular-view PACT system has been used to image contrast enhancement from gold nanostars [Figs. 3(c) and 3(d)].⁷² The system had 512 transducer elements evenly distributed on a 5-cm diameter circle. Each element was arc-shaped with a 1.9-cm radius of curvature along elevation direction. This design ensured rejection of out-of-plane signals.⁷⁵ The array system eliminates mechanical scanning and thus significantly improves the imaging speed—a 2-D image can be captured in 0.625 s. To match the absorption peak of nanostars, a Ti-Sapphire laser with 800-nm output was used as the excitation source. The maximum light intensity on mouse scalp was approximately $15\mathrm{mJ}/{\mathrm{cm}}^2$, which is below the American National Standard Institute (ANSI) safety limit of $30\mathrm{mJ}/\mathrm{cm}$. The high imaging speed allows for monitoring of dynamic changes in PA amplitude. As can be seen in Fig. 3(d), the PA amplitude rose almost instantly after tail vein injection of nanostars and then it increased gradually before reaching the maximum at around half an hour. The PA signal then gradually decreased, indicating the clearance of nanostars the reticuloendothelial system (RES).

4.2.

Tumor Targeting

In addition to vascular contrast enhancement, NPs have also been used in tumor targeting.^76,77 The targeting mechanism can be either passive or active. Passive targeting is based on the enhanced permeability and retention effect (EPR effect),^78,79 which allows NPs to accumulate at a higher concentration in tumors than in healthy tissues.⁸⁰ The EPR effect relies on two mechanisms: first, the porous vascular structures in tumors facilitate the perfusion of NPs; second, once NPs are in a tumor, they tend to stay for a prolonged period of time because of poor lymphatic drainage in the tumor, leading to a higher NP concentration.

Passive brain tumor targeting was recently demonstrated in PA neuroimaging using organic PDI NPs.²³ In this study, PDI NPs had an average size of 50 nm and exhibited strong absorption at 700 nm, which enabled deep-brain tumor imaging. The experiment was performed using a commercial linear-view PACT system (Vevo LARZ, Visualsonics) with a LZ-250 linear array (13–24 MHz central frequency) and a tunable OPO laser system (680–970 nm) with a 20 Hz pulse repetition rate. When all 256 elements were used for imaging, the LARZ system had a frame rate of 5 Hz. The experiment was performed on mice with orthotopic glioblastoma tumors. Before the PA experiment, the authors first scanned the mouse with an MRI to confirm tumor size and location. They then injected $250\mu \mathrm{L}$ of 250 nM PDI NPs through the tail vein and imaged the animal with the LARZ system at 0 hour, 2 hours, 1 day, and 2 days post injection. Figure 4 shows coronal-view images of the mouse brain. The red dashed line indicates the skull, while the blue dashed circle indicates the tumor region. An obvious increase of PA amplitude at the tumor site can be observed starting from day 1. The depth of the tumor was estimated to be 4 mm, which agreed with the MRI image. The tumor contrast continued to increase in day 2, indicating enhanced tumor penetration of NPs with time, while the skull contrast started to decrease, indicating the clearance of NPs in the circulatory system.²³

Fig. 4

Coronal-view PA images demonstrating the effect of passive tumor targeting with PDI NPs. Images were acquired after tail vein injection of $250\mu \mathrm{L}$ of 250 nM PDI NPs in a mouse with an orthotopic glioblastoma tumor. Adapted with permission from Ref. 23.

Despite encouraging results, passive targeting has its inherent limitations. The cells of RES will remove most of the intravascular injected NPs,⁸¹ and the spleen, liver, lung, bone marrow, and kidney would also uptake NPs. Thus, only a small fraction of administered NPs can reach the tumor.^82,83 To increase the circulation time, the size of the NPs needs to be less than 100 nm, and the surface biocompatibility needs to be carefully designed.

Another targeting mechanism is active targeting, which uses the receptors that are highly expressed on the surface of tumor cells, such as the epidermal growth factor receptors (EGFR) and integrin.⁸⁴ When NPs are decorated with ligands that can specifically bind to these receptors, the concentration of NPs in the tumor will increase. Because of the large surface area of NPs, many ligands, such as proteins and peptides, can be coated on NPs.⁸⁵ Active targeting complements passive targeting with the EPR effect to provide prolonged and even targeted accumulation of NPs in tumors.

A preliminary demonstration of activate tumor targeting was carried out using organic Blue NPs, which were made of a PAA matrix with covalently linked CB dye molecules and were conjugated with F3 peptide.²² The F3 peptide binds to tumor endothelial cells while the CB dye enhances optical absorption at 590 nm. The study was performed on rats with 9L glioma tumors in their brains. Animals were euthanized several days after a tail vein injection of NPs to avoid untargeted accumulation. Subsequently, the brains were excised for ex vivo PAT. The experiment was performed on a circular-view PACT system with a single, cylindrically focused 9 MHz transducer. A dye laser (ND6000, Continuum) with a 10 Hz pulse repetition rate and an output at 590 nm (the absorption peak for NPs) was used as the excitation source. The incident light intensity was $3\mathrm{mJ}/{\mathrm{cm}}^2$, which was well below the American National Standard Institute limit of $20\mathrm{mJ}/{\mathrm{cm}}^2$. Because of NPs’ strong optical absorption and targeting properties, the brain tumor could be clearly delineated in the PA image [Fig. 5(a)], even when the dose of NPs ($125\mathrm{mg}/\mathrm{kg}$ of rat mass) was at a level indistinguishable by the naked eye in the photograph [Fig. 5(b)]. This study demonstrates the potential of intraoperative tumor delineation using PAT.

Fig. 5

(a) PA image of an excised rat brain containing a tumor treated with nanoshells. Field of view: $2\mathrm{cm}\times 2\mathrm{cm}$. (b) Photograph of the brain in (a). Adapted with permission from Ref. 22.

4.3.

Multimodal Imaging

Multimodal imaging allows for a more comprehensive characterization of a brain in its healthy and diseased states.⁸⁶ NPs are one of the most attractive multimodal contrast agents because their multicomponent nature offers many possibilities to incorporate various imaging probes. Because the same NPs can be visualized in different imaging modalities, they allow precise coregistration of multimodal images and corresponding tissue information.

A notable multimodal probe used in PA neuroimaging is trimodal MRI-PAT-Raman imaging (MPR) NPs.⁸⁷ MPR NPs consisted of a 60-nm gold core covered with a Raman active layer, which was further coated with silica and ${\mathrm{Gd}}^{3+}$, resulting in a gold-silica–based NP with both Raman and magnetic contrasts. MPR NP performances were measured on glioblastoma-bearing mice with contrasts injected through the tail vein. Imaging experiments were performed several hours after injection to let MPR NPs gradually accumulate in the tumors through the passive EPR effect. PA images were acquired by an AR-PAM system. However, unlike typical AR-PAM systems with kHz repetition rate lasers, their system employed a 10-Hz Nd:YAG pulsed laser (Nd:YAG laser), whose 532-nm output was coupled into a ring fiber optical bundle. A 5-MHz focused transducer was placed at the center of the fiber bundle for signal acquisition. The spatial resolution of the system was 0.6 and 0.38 mm along lateral and axial resolutions, respectively. Figure 6 indicates that MPR NPs can be clearly visualized in all three modalities. The MRI anatomical image allows for the precise localization of tumor position, PAT provides high-resolution 3-D images of the tumor, and Raman enables surface imaging of tumor margins with high sensitivity and high specificity.⁸⁷

Fig. 6

MPR multimodal images of a mouse brain. (a) Magnetic resonance imaging (MRI) image, (b) PA image (green) superimposed on MRI image, and (c) Raman (red) superimposed on MRI image. Reproduced with permission from Ref. 87.

There are many other multimodal NPs that show great potential for PA neuroimaging. For instance, because of strong optical absorption, most NPs also allow PT tumor therapy and PAT.^88,89 PAT provides high-resolution images of NP distribution and tumor vasculature and offers real-time monitoring of temperature rise during PT therapy.^72,90 Single-walled carbon nanotubes have enabled multimodal thermoacoustic tomography and PAT, due to their intrinsic absorption at both optical and radio frequencies.⁹¹ QDs have been used in multimodal PAT and fluorescence imaging, due to their broad excitation spectra (400 to 700 nm) and narrow emission spectra (30 to 80 nm).⁴⁴