1. Introduction

Photoacoustic imaging (PAI) is a promising molecular imaging technique. It uniquely combines optical excitation and ultrasonic detection, thus effectively utilizing the low ultrasonic scattering and rich optical molecular contrast in biological tissues. This enables high-resolution visualization of structural, molecular, and functional features in biological tissue with a centimeter-scale penetration depth [1,2]. Photoacoustic endoscopy (PAE) further overcomes the limitations of light penetration, making it particularly promising for assessing various internal organs within the human body, including the esophagus, trachea, gastrointestinal tract, and prostate [3–5].

PAI primarily utilizes hemoglobin in the blood vessels as the main contrast for imaging. Through multi-spectral detection, PAI can simultaneously capture the distribution of various endogenous chromophores, such as oxy-hemoglobin (HbO₂), deoxy-hemoglobin (Hb), water, and lipids, based on their distinctive absorption spectral signatures. With advancements in exogenous contrast agents, such as small organic dyes, nanoparticles, quantum dots, and genetically encoded proteins, molecular PAI has experienced remarkable progress [2,6–8]. These innovations have brought several advantages to PAI, including heightened sensitivity and specificity, improved molecular stability, and enhanced biological compatibility. The introduction of the pump-probe detection scheme in PAI has also enabled the suppression of background signals [9–13], facilitating highly specific detection of target molecules even at low concentrations. All these technical advancements can be readily translated to PAE, which holds exciting prospects for life science and preclinical studies.

So far, significant progress has been made in PAE in terms of scanning speed, spatial resolution, and multimodal imaging capabilities [3]. However, the development of molecular PAE has been relatively slow, limiting its biomedical applications [14]. Most PAE systems are optically-resolution-based and employ high-repetition lasers with fixed wavelengths, providing good spatiotemporal resolutions but not necessarily suitable for molecular imaging [15–18]. Although non-linear optical techniques like fiber-based stimulated Raman scattering (SRS) can be used for multispectral imaging, the selection of wavelengths is still restricted [19]. Besides, they have limited penetration depth (typically less than 1 mm), similar to pure optical microscopy. In contrast, acoustic-resolution-based photoacoustic endoscopy (AR-PAE) utilizes a tunable optical parametric oscillator (OPO) laser with high pulse energy, offering significant advantages in terms of penetration depth and molecular imaging capability [20–23].

A flexible and slender AR-PAE probe is crucial for endoscopic examinations of organs such as the colon and esophagus. However, building such probes is much more challenging compared to endoscopic ultrasound, as it requires consideration of both optical illumination and detection. In this study, we have developed a low-cost, flexible AR-PAE probe with an 8 mm diameter. The probe is equipped with a 6 mm-diameter micro-step motor at the distal end, enabling precise angular scanning control to synchronize with the OPO lasers. One notable feature of this probe is its ability to produce high-specificity photoacoustic molecular images using a novel pump-probe detection method. We evaluated the performance of the probe through phantom, ex-vivo, and in-vivo experiments. With its thin and flexible design, coupled with its remarkable molecular imaging capabilities, the proposed AR-PAE method holds great potential as a highly valuable diagnostic tool across various biomedical research domains.

2. Materials and methods

2.1 System setup

Figure 1(a) illustrates the schematic diagram of the implemented system. Two tunable OPO lasers (SpitLight OPO 600, Innolas, München, Germany) operating at 20 Hz were utilized for pump-probe photoacoustic imaging. One OPO (pumped with 355 nm) emitted light at 665 nm for methylene blue (MB) excitation. The other OPO (pumped with 532 nm) was tuned to 820 nm for probing. In conventional spectroscopic photoacoustic imaging, only the 532 nm pumped OPO was used. The light from both OPO lasers was combined using a dichroic mirror and coupled into the guiding fiber bundle (3 mm diameter) of the probe. The optical influence on the sample was around 5 mJ/cm². The ultrasound signal collected by the probe was amplified with a JSR pulser/receiver (DPR500, Imaginant, NY), which was also utilized for pulse generation in ultrasound imaging. The data acquisition was performed using a PCI-5124 data acquisition (DAQ) card (100 MS/s, 12-Bit, National Instruments, TX).

 

Fig. 1. Illustration of the multi-modal AR-PAE system. (a) Schematic of the system. (b) The structure of multi-modal AR-PAE probe. (c) The synchronization control of the micro-motor and the data acquisition. (d) The timing diagram of the pump-probe-based photoacoustic data acquisition. MB, methylene blue; DAQ, data acquisition.

Download Full Size | PDF

The design of the probe was based on a coaxial optical/ultrasound configuration, as depicted in Fig. 1(b). The probe's polyethylene (PE) housing had an outer diameter of 8 mm and a wall thickness of 0.5 mm. At the distal end of the probe, a 6 mm-sized micro-step motor (with a 1:47 gearbox and a total length of 12 mm) was mounted, driving a parabolic off-axis mirror (6.25 mm in size, 12.6 mm of focusing, #37-282, Edmund Optics, NJ) for focused ultrasound detection. The micro-step motor was driven by a regular digital stepping driver (DM430, Yingpengfei, Shenzhen, China). The ultrasound collected by the parabolic off-axis mirror was redirected by a 45° quartz rod towards a 20 MHz piezoelectric transducer (3 mm square in size). The rigid portion of the probe measured 4 cm long (please see Visualization 1), and the viewing angle was limited to 300° due to the obstruction by the phase wires. Additionally, this probe was equipped with a micro-camera module (OV6946, OmniVision, CA) mounted at the tip, providing a high-resolution front view. It has a diameter of 2 mm and a dead zone of only 3 mm. The small inset shows an image of leaf veins captured by this micro-camera.

The synchronization of the entire system was controlled by a multi-channel digital output module (NI USB-6218, National Instruments, TX), as shown in Fig. 1(c). To enable the pump-probe photoacoustic imaging, one of its channels sent 20 Hz trigger signal to a digital delay/pulse generator (DG645, Stanford Research Systems, CA, USA) to control the delay adjustments of the OPO lasers for the three interleaved data acquisition sequences [17], as shown in Fig. 1(d). The second channel triggered the JSR pulser/receiver for concurrent ultrasound imaging. There were three extra channels to provide the Enable/Disable, Direction, and Pulse signals to the digital stepping driver. The Enable/Disable control was implemented to avoid the strong electrical noise generated by the micro-step motor during data collection.

The small size of the micro-step motor limited the number of its poles, resulting in a large intrinsic angular step size of 18° (compared with 1.8° for a regular 42-size step motor). The digital stepping driver's subdivision capability helped eliminate the vibration of the step motor and allowed for a finer step size. Due to the existing of the gearbox, the maximum rotating speed of the micro-step motor was approximately 1.2 frames per second. Additionally, because the minimal current of the digital stepping driver (0.14 A peak) was approximately double the rate current of the micro-step motor, we paralleled two micro-step motors for current protection. In this way, we adapted the micro-step motor to a regular digital stepping driver. Compared to directly driving the micro-step motor with complex multiplexing analog signals, our method greatly facilitated its control.

2.2 Experimental design

The probe's capability was explored through various experiments, including system performance test, spectroscopic photoacoustic imaging, and pump-probe photoacoustic imaging. The system performance test involved both a field test and a phantom experiment. In the field test, a 50-µm-thick tungsten wire was scanned at different imaging depths to determine the lateral resolution and depth of focus (DOF). The wire was constantly illuminated by a multimode fiber (740 nm), and simultaneous recording of photoacoustic and ultrasound signals was performed. At each depth, 240 A-lines were recorded, covering an angular view of 60°. The 2D improved back-projection method was used to restore the elongated lateral resolution in the out-of-focus regions. The phantom experiment aimed to further evaluate the probe's performance in structural imaging. Two layers of metal grids were buried in a phantom made of 2% agar and water. A 3D cylindrical scan was performed, with a total of 100 layers (20 mm in the z-direction), and each layer consisted of 300 A-lines. In the phantom experiment, the signal was averaged only once, and the total scanning time was approximately 1 hour. The reconstructed 2D images were stacked to obtain the 3D photoacoustic and ultrasound images.

To evaluate the probe's capability for high-penetration spectroscopic imaging, a murine tumor on the hind leg was imaged. Indocyanine green (ICG), an FDA-approved dye commonly used as a photoacoustic contrast agent, was injected (0.05 mL @ 1 mM). Photoacoustic images were acquired before and after the injection at three different wavelengths: 760 nm, 840 nm, and 805 nm. The probe was positioned approximately 1 cm away from the tumor, resulting in a pulse influence of about 5 mJ/cm² on the tumor surface. The wavelengths 760 nm and 840 nm are sensitive to deoxyhemoglobin (Hb) and oxyhemoglobin (HbO₂) respectively, while 805 nm corresponds to the peak absorption of ICG in tissue [24]. Assuming that the photoacoustic signals were solely contributed by Hb and HbO₂ and neglecting light attenuation, the concentration distribution of Hb and HbO₂ can be obtained by solving the spectral matrix equation [25]. However, for simplicity, we assumed that the photoacoustic intensity at 760 nm represents the concentration of Hb, and the intensity at 840 nm represents the concentration of HbO₂. Subsequently, the oxygen saturation was calculated as HbO₂/(Hb + HbO₂). Additionally, the 3D photoacoustic images under each wavelength were also obtained and compared. The dose was 3 times lower than existing studies [26] and the mouse survived after the experiment. This in-vivo experiment protocol was approved by the Department of Laboratory Animals of the Central South University (No. 2020KT-39).

To demonstrate the pump-probe photoacoustic imaging with the probe, we conducted a tube phantom experiment and an ex-vivo imaging experiment. In these experiments, we used MB as the contrast agent. MB is a small porphyrin molecule commonly used in clinical diagnostic and therapeutic applications. In the pump-probe technique, MB is excited by a pump laser at a wavelength of 665 nm [27], which puts it into a triplet state (T1). The T1 state of MB has a relatively long phosphorescence lifetime (up to 79 µs) and strong absorption of the probe laser at 820 nm. A long T1 lifetime is crucial for pump-probe photoacoustic imaging, as it allows for a high population of excited molecules when the probe laser arrives, ensuring high detection sensitivity. By using the transient triplet differential (TTD) detection method, we can remove the background photoacoustic signal and selectively detect only the MB in the T1 state [13,28]. The TTD signal can be calculated as

Here ${S_{pump + probe,\tau }}$ is the photoacoustic signal obtained with both the pump and probe lasers on, ${S_{pump,\tau }}$ and ${S_{probe}}$ are the photoacoustic signals with the pump and probe laser. $\mathrm{\tau }$ is the delay time between the pump laser and the probe laser. Here we set this value to be 0.5 µs, which is much smaller than MB’s lifetime to achieve a high sensitivity.

In the tube phantom experiment, we inserted five tubes vertically into the phantom. Three of the tubes were filled with a solution of 100 mM MB, while the other two tubes were filled with diluted ink at a ratio of 1:40. The inner diameter of the tubes was 2.8 mm, and the tube thickness was negligible. We collected 105 A-lines, covering a 210° angle, with 100 times averaging to reconstruct the 2D images. The total scanning time for this experiment was approximately 30 minutes. In the ex-vivo experiment, we imaged a piece of fresh pig ileum. Before burying the biological sample, we injected a dose of ink (approximately 0.2 mL, with a 1:20 dilution) and a dose of MB (approximately 0.2 mL, 0.2 mM) into two separate positions. The MB doses was about 10 times lower than our previous in-vivo experiments [13]. We then waited for about 30 minutes to allow for sufficient diffusion of the dyes in the ileum wall. The two injection positions were at about the same height, enabling their localization in the same 2D image. For the 2D pump-probe photoacoustic imaging, we collected 90 A-lines, covering a 180° angle, with 200 times averaging. Additionally, we performed a 3D ultrasound scan of the pig ileum, consisting of 100 layers, which was completed in 8 minutes. It’s noted that the ink has a high absorption at both 665 and 820 nm [29], but it doesn’t have a long lifetime excited state, so its signal is expected to be reduced down to the background level in the TTD images.

In this study, extensive data analysis has been conducted to effectively demonstrate the performance of the proposed method. This analysis includes the calculation of signal-to-noise ratio (SNR), structural similarity index (SSIM), and the Q-test. For detailed instructions on these methods, please refer to Supplement 1.

3. Results

3.1 System performance test

Figure 2(a) displays the reconstructed photoacoustic and ultrasound images obtained from the field test. After signal filtering, the central frequency was 19.4 MHz, with a bandwidth of 53%, and the resulting axial resolution was about 0.12 mm. To assess the lateral resolution, we extracted the lateral profile of each target and calculated the full width at half maximum (FWHM). Utilizing the improved back-projection (IBP) method [30], Fig. 2(b) reveals a consistent increase in both photoacoustic and ultrasound lateral resolutions with increasing imaging depth. At the depth of 19.2 mm, the lateral resolutions for the photoacoustic and ultrasound images were measured as 0.87 mm and 0.43 mm, respectively, as shown in Fig. 2(c). These results are significantly improved compared with our previous work owing to the higher transducer frequency [21].

 

Fig. 2. System performance test results of the AR-PAE probe. (a) Reconstructed photoacoustic (hot colormap) and ultrasound (hot colormap) images with a point target at different imaging depth. The axial positions are relative to the center of the probe. (c) The changes in FWHMs with the axial positions. (b) Photoacoustic (red) and ultrasound (blue) lateral profiles of the point target obtained at 19.2 mm; (d)-(e) Photoacoustic images for the inner and outer layers of the grid meshes, respectively. (f)-(g) Two different views of the reconstructed photoacoustic images of two-layer phantom. (h)-(k) The corresponding ultrasound images to (d)-(g). (o) Photo image of the metal grids. (m)-(n) MAP images of extracted from the 3D photoacoustic and ultrasound images of the inner mesh, respectively. (o) The lateral profiles of an edge extracted from (l)-(n), and the location of the edge is indicated in (l). The mesh size of the mesh grid was 3 × 6 mm, and the thickness of the grid was about 0.5 mm, as shown in the small inset in (h). PT, photo; PA, photoacosutic; US, ultrasound.

Download Full Size | PDF

Figures 2(d)-(k) exhibit the reconstructed 3D photoacoustic and ultrasound images of the two-layer phantom. The inner layer is clearly distinguishable in both the photoacoustic image (Fig. 2(d)) and ultrasound image (Fig. 2(h)). The outer layer, located approximately 6 mm deeper than the inner layer, exhibits well-resolved details of the grids despite the obstruction of much of its signal, as depicted in Fig. 2(e) and (i). The peak SNR reached 51.2 dB for the photoacoustic images even with only a single-time average, and the peak SNR was 50.9 dB for the ultrasound images. To better assess the structural imaging capability of our probe, we outlined a region within the inner mesh (as indicated by the dotted white rectangle in Fig. 2(h)) and obtained maximum amplitude projection (MAP) images along the axial direction from both the 3D photoacoustic and ultrasound images, as shown in Figs. 2(m) and (n). With the photo image in Fig. 2(l) as a reference, it was observed that the mesh grids were clearly distinguishable in both MAP images. We selected an edge within the grid, indicated by the arrow in Fig. 2(l), and extracted and compared its lateral profiles from the three 2D images (Figs. 2(l)-(n)), as seen in Fig. 2(o). Quantitative studies showed that the structural similarity index (SSIM) between the lateral profiles from the photoacoustic and photo images was 0.818, while the SSIM between lateral profiles from the ultrasound and photo images was 0.892. All these results effectively demonstrate the high spatial resolution and imaging depth capabilities of the probe.

3.2 Spectroscopic photoacoustic imaging

Figure 3(a) presents the experimental diagram, with the position of the 2D cross-section image indicated by a dotted arrow in the photo of the tumor. Figure 3(b) displays the oxygen saturation map before the ICG injection, obtained using wavelengths of 760 nm and 840 nm. The corresponding 3D images for Hb, HbO₂, and oxygen saturation are shown in Figs. 3(d)-(f), respectively. The approximate region of the tumor was outline with dotted red lines for better illustration. These images are based on the assumption that the photoacoustic signals are solely contributed by Hb and HbO₂. It is evident that the photoacoustic signal mainly originates from the tumor surface.

 

Fig. 3. Spectroscopic photoacoustic imaging of a murine tumor with ICG injection. (a) The experimental diagram. (b)-(c) Oxygen saturation map before and after the ICG injection, respectively. (d)-(f) 3D photoacoustic images for Hb, HbO₂ and oxygen saturation, before the ICG injection. (g)-(i) 3D photoacoustic images of different views at 805 nm, before the ICG injection. (j)-(l) Corresponding results to (g)-(i), after the ICG injection. The photoacoustic images are overlayed onto the ultrasound images, and the approximate tumor region is delineated with red dotted lines to facilitate better interpretation of the results.

Download Full Size | PDF

Following the ICG injection, a significant signal rise in the tumor is observed (at least 5 mm deep from the surface) due to the strong absorption of ICG, as indicated by white arrows in Fig. 3(c). It should be noted that in this case, measurements from all three wavelengths are required to accurately map the distribution of ICG and oxygen saturation by solving the spectral matrix equation. However, such a calculation may not be appropriate for the case in Fig. 3(b) since accounting for ICG's absorption in the matrix equation can introduce fake ICG distribution. In summary, current spectroscopic photoacoustic imaging faces substantial technical challenges due to model inadequacy and measurement errors. Therefore, single-wavelength PAI is still commonly employed, especially for mapping deep targets.

Figures 3(g)-(l) depict the 3D photoacoustic images at 805 nm (the absorption peak of ICG) before and after the ICG injection. By comparing Figs. 3(g)-(i) with Figs. 3(j)-(l), the injected ICG in the tumor can be roughly localized. The peak SNR in Figs. 3(g)-(i) was 35.5 dB, and it was 50.6 dB after the ICG injection in Figs. 3(j)-(l). For the 3D images with other wavelengths, please refer to Visualization 2. These results collectively demonstrate the probe's capability for high-penetration spectroscopic imaging.

3.3 High-specific photoacoustic molecular imaging

Compared to spectroscopic photoacoustic imaging, pump-probe photoacoustic imaging provides background-free, high-specificity images of the target molecule, which is crucial for accurate diagnosis. Figure 4 illustrates the imaging results of the tube experiment. In the ultrasound image (Fig. 4(a)), the upper and lower boundaries of each tube are clearly visible, indicating the locations of the five tubes. The three tubes containing MB are indicated with blue dotted circles, while the other two tubes containing ink are indicated in white. In the photoacoustic images reconstructed with ${S_{pump + probe}}$ and ${S_{pump}}$, all five tubes are clearly detected due to the strong absorption of 665 nm (Figs. 4(b) and (c)). In contrast, only the ink tubes are well detected in Fig. 4(d) since the absorption of MB at 820 nm is much weaker. After the differential of the photoacoustic signals, only the tubes containing MB are left in the TTD images (Figs. 4(e) and (f)). Only the boundaries of the tubes being imaged in both the photoacoustic and ultrasound images are due to the limited bandwidth of the transducer. Because the most left MB tube was not well vertically positioned, it showed lower signal intensity than the other two MB tubes.

 

Fig. 4. Experimental results for pump-probe photoacoustic imaging of a tube-containing phantom. (a) Ultrasound imaging of the tubes. (b)-(e) Photoacoustic images reconstructed with ${S_{pump + probe}}$, ${S_{pump}}$, ${S_{probe}}$, and ${S_{TTD}}$. (f) Overlaid TTD image on the ultrasound image, and lines marked with MB and ink indicating the positions of the Alines in (g) and (h), respectively. The MB, ink, and the five background regions are about 2.5${\times} $1 mm in size, with their locations indicated in (c). (g) Extracted Aline of ${S_{pump + probe}}$, ${S_{pump}}$, ${S_{probe}}$, and ${S_{TTD}}$ for an MB tube. (h) The corresponding Alines to (g) for an ink tube.

Download Full Size | PDF

For better illustration, Fig. 4(g) shows the A-line signals of ${S_{pump + probe}}$, ${S_{pump}}$, ${S_{probe}}$, and ${S_{TTD}}$ for an MB tube (position indicated in Fig. 4(b)). It is evident that the photoacoustic signal from the pump laser is almost completely canceled out after differentiation, and the remaining TTD signal is attributed to the increase in the probe laser's signal in ${S_{pump + probe}}$. This increase is a result of the pump-induced excitation of the MB molecule. Figure 4(h) shows the case for the ink tube. Here, it can be observed that both the pump and probe photoacoustic signals are canceled out in the TTD signal, resulting in a clean background. This is because the lifetime of ink is much shorter (picoseconds or nanoseconds level), so when the probe laser arrives, no ink remains in the excited state during the acquisition of ${S_{pump + probe}}$.

To quantitatively assess the sensitivity and specificity of the proposed pump-probe PAE method, we selected one MB tube and one ink tube, and delineated the regions enclosing their upper boundaries. Additionally, we arbitrarily chose five background regions for reference, as shown in Fig. 4(c). We then extracted and compared the maximum value, mean, and standard deviation (Std) in these seven regions in the ${S_{pump}}$, ${S_{probe}}$, and ${S_{TTD}}$ images (Figs. 4(c)-(e)), which are summarized in Supplement 1, Table S1. With these data, we calculated the SNRs of the MB and ink tubes, and also performed a Q-test to evaluate the specificity for MB imaging, as seen in Table 1. Results showed that the SNR of the MB tube in the ${S_{TTD}}$ image is lower than the ${S_{pump}}$ and ${S_{probe}}$ images. However, by comparing the Q values for the MB and ink tubes with the critical value Q (n = 6, P = 0.99) = 0.74, it’s clear that both the MB and ink are statistically higher in signal amplitude than the background in the ${S_{pump}}$ and ${S_{probe}}$ images. In contrast, only the MB signal is statistically higher than the background in the ${S_{TTD}}$ image, while the ink signal is almost down to the background level. This can also be seen from Table S1, where the max, mean, and Std values for the ink are close to the background regions. In summary, although the proposed pump-probe PAE method is lower in sensitivity, it’s capable for high-specificity imaging of MB.

Table 1. The SNR and Q-test^a results for the tube imaging experiment

View Table | View all tables in this article

Figure 5 presents the results for the ex-vivo pig ileum imaging experiment. The 3D ultrasound images in Figs. 5(a) and (d) depict the high-resolution structure of the sample, with both the inner and outer surfaces clearly distinguishable. The approximate positions for the ink and MB injections are indicated in Fig. 5(a), and the cross-section for pump-probe photoacoustic imaging is marked in the x-z view image of Fig. 5(d). It is observed that MB produces a stronger signal than ink in the ${S_{pump}}$ image (Fig. 5(b)), while the signal intensities of MB and ink were closer in the ${S_{probe}}$ image (Fig. 5(c)). After the differential operation, only the MB signal remains in the TTD image (Fig. 5(e)). For better visualization, the TTD signal of MB is rendered in green and overlaid on the ${S_{pump}}$ image (Fig. 5(f)). As indicated in Fig. 5(c), we selected the MB, ink, and five background regions and calculated the SNRs and the Q values, as seen in Table 2. Results show that although the SNR of MB was decreased after the signal differentiation in the ${S_{TTD}}$ image, the ink signal was almost completely canceled out, thus enabling the high-specific detection of MB. For detailed maximum, mean, and Std values of these regions, please see Supplement 1, Table S2. These results demonstrate that with the pump-probe photoacoustic imaging method, the probe can effectively suppress background signals and perform high-specificity endoscopic molecular imaging in biological samples.

 

Fig. 5. Pump-probe photoacoustic endoscopic imaging of an ex-vivo pig ileum. (a) 3D ultrasound images of the pig ileum, which was injected with MB and ink before imaging. The injection positions were indicated with arrows. (b) The x-z view of the 3D ultrasound images, with the position of the cross-section imaged in (c)-(f) marked. (c)-(e) Reconstructed images of ${S_{pump}}$, ${S_{probe}}$, and ${S_{TTD}}$. (f) Overlaid TTD image on the ${S_{pump}}$ image. The photoacoustic images in (b), (c), (e), and (f) are overlaid onto the ultrasound image, and the TTD signals in (e) and (f) are rendered in green. The MB, ink, and background regions selected for the SNR calculation and statistical study are indicated in (c), with a size of 5${\times} $2 mm.

Download Full Size | PDF

Table 2. The SNR and Q-test^a results for the ex-vivo pig ileum imaging experiment

View Table | View all tables in this article

4. Discussion

Molecular imaging is widely recognized as a crucial tool in biomedical research, clinical diagnosis, and drug development, offering immense value in advancing our understanding and management of complex diseases in both research and clinical settings [31,32]. While conventional medical imaging modalities like fMRI and PET have the ability for molecular imaging, they are hindered by limitations such as bulky size, high expenses, limited spatial and temporal resolution, and potential radiation risks. Consequently, they may not be optimally suited for small-scale laboratory research or as bedside diagnostic instruments.

In this study, we have developed an 8-mm flexible distal-driven AR-PAE probe for pump-probe-based photoacoustic molecular imaging, which is the most prominent feature of the probe. As discussed in Sections 3.2 and 3.3, unmixing different chromophores in deep tissues is challenging with spectroscopic photoacoustic imaging due to measurement errors and model inadequacies. In contrast, the pump-probe photoacoustic method provides more reliable molecular imaging results, which are crucial for accurate diagnosis. In addition to the pump-probe-based photoacoustic imaging mode, our probe is also capable for conventional single-wavelength or spectroscopic photoacoustic imaging. Equipped with a 20 MHz small transducer, our probe achieved a photoacoustic/ultrasound lateral resolution of about 0.6/0.3 mm at a depth of 1 cm from the probe wall. The 3D photoacoustic and ultrasound images of the mesh grids clearly demonstrate the high-resolution and high-penetration imaging capabilities of our probe (Fig. 2). Furthermore, we have integrated a 2mm-sized front-view micro-camera on the probe tip, enabling high-resolution images with a 120° field of view. This small camera, combined with the high-frame-rate ultrasound, can serve as image guidance to locate the region of interest and approach the target organ safely before performing the photoacoustic scan. This practical approach enhances scanning efficiency and provides complementary information for accurate diagnosis.

The unique design of the distal-driven micro-step motor plays a crucial role in pump-probe-based photoacoustic imaging. It provides precise angular step control to synchronize with the OPO lasers, enabling start-and-stop data acquisition while maintaining probe flexibility. In contrast, most flexible PAE probes rely on torque coil-based proximal scanning [5,16,33] or utilize a distal-driven DC motor [34], so they cannot synchronize with the OPO lasers. Some proximal-driven AR-PAE probes use a step motor, but they often have rigid shafts [21,22], significantly limiting their biomedical applications. Additionally, while small-array-based probes (including ring-array-based and linear-array-based) might be capable of pump-probe-based photoacoustic imaging, they typically have bulky sizes (ranging from 1.2 to 3 cm) and suboptimal light illumination [35–38], restricting their applications to specific scenarios such as transrectal prostate imaging. However, driving a micro-step motor is notably more challenging technically than using a DC motor in PAE. In this study, we successfully devised a method to adapt the micro-step motor to a regular digital stepping driver, greatly simplifying the control of the micro-step motor. To the best of our knowledge, this marks the first development of a distal-driven-based flexible AR-PAE probe.

It is important to note that this study is a preliminary investigation, and there are several areas in which the proposed multimodal probe can be improved. One limitation is the current size of the probe (8 mm), which is determined by the 6.25 mm reflection mirror. By customizing a smaller reflection mirror, the probe diameter can be reduced to 5 mm. Additionally, in this study, we periodically enabled/disabled the digital stepping driver to mitigate electrical noise. In future work, optical ultrasound detection techniques such as micro-ring resonators and fiber Bragg gratings (FBGs) can be employed [39]. These techniques can save scanning time and eliminate the need for a gearbox, allowing for a smaller micro-step motor and further reducing the probe size and improving scanning speed. To enhance the sensitivity and durability of the probe for animal experiments, measures need to be taken in future work. This may involve optimizing the probe design and improving the materials used for fabrication. Furthermore, strategies need to be developed and implemented to ensure good acoustic coupling for in-vivo endoscopic imaging. These are important considerations for future works that aim to translate this technology to clinical settings.

Indeed, the current limitation of the probe is its relatively slow photoacoustic imaging speed, which is constrained by the repetition rate of the OPO lasers used in this study. However, with advancements in laser technologies, it is possible to develop a probe that is thin, flexible, multimodal, low-cost, and disposable, while also achieving faster photoacoustic scanning times. Future improvements in laser technology could potentially enable a probe that can complete a pump-probe photoacoustic scan in 2 to 3 minutes. As discussed above, incorporating other high-speed imaging modalities can greatly enhance the utility of pump-probe photoacoustic scanning. Compared to existing technologies such as fMRI, PET, or biopsies, which are expensive, labor-intensive, and may take several hours to days to obtain molecular diagnostic results, a multimodal endoscopic probe with improved scanning speed would offer unparalleled convenience and capabilities for both basic biomedical research and clinical applications. It has the potential to significantly enhance the efficiency and accuracy of molecular imaging, enabling faster and more precise diagnosis and treatment planning.

5. Conclusions

In this study, we have introduced a novel micro-step-motor-based distal-driven flexible AR-PAE probe with a diameter of 8 mm. This probe not only possesses the high spatial resolution, high penetration depth, and spectroscopic imaging ability, but also has the capability to perform background-free photoacoustic molecular imaging using a pump-probe detection technique. The experimental results clearly demonstrate the exceptional molecular imaging capability of the probe, representing a significant technical breakthrough in the field of PAE. With the rapid advancements in laser technology, the development of a thin, flexible, and disposable AR-PAE-based multimodal endoscopic probe for molecular imaging at expected frame rates is becoming feasible. This innovation has the potential to revolutionize various biomedical applications.