2.1.

Experimental Setup

All procedures were approved by the Johns Hopkins University Institutional Animal Care and Use Committee and all experiments were conducted at the Johns Hopkins Minimally Invasive Surgical Training and Innovation Center. Two Sus domesticus crossbred domestic pigs (5 and 6 months, 65 and 90 lbs., both female) were obtained for this study. Animals were treated humanely. Pigs were anesthetized and subsequently intubated to provide gas anesthesia throughout the surgery. Respiration was controlled with a ventilator that supplied medical air. A midline incision exposed the abdominal organs. Blood pressure and heart rate were monitored throughout the surgery and imaging procedures.

The photoacoustic imaging system, which was placed next to the operating table, as shown in Fig. 1, consisted of an Alpinion ECUBE 12R US system (Seoul, Korea) synchronized with a Phocus Mobile laser (OPOTEK, Carlsbad, California). The US system was connected to a 128-element Alpinion L3-8 linear array transducer, which has a center frequency of 5 MHz, emitted an US transmit focus of 3 cm, and contained an active receive aperture limited to 64 elements. The laser was connected to a 5-mm-diameter fiber bundle. Although we also connected the laser to a 1-mm-diameter optical fiber and attempted to image blood vessels in the pancreas with this fiber, we were unsuccessful. Thus, all images in this paper show our success with the 5-mm-diameter fiber bundle.

Fig. 1

Imaging equipment and surgical environment for in vivo porcine laparotomy.

A fixed laser wavelength of 750 nm was chosen to visualize blood vessels throughout the experiments. The wavelength was fixed, given our primary goal of visualizing (rather than characterizing) major blood vessels during surgery and our reasoning that one wavelength is sufficient to inform the surgeon of the presence or absence of a blood vessel during surgery. The US transducer was either held by a Sawyer robot (Rethink Robotics, Boston, Massachusetts) or by hand. In addition to acquiring photoacoustic images, the US system was also used to acquire coregistered US images and color Doppler images to confirm vessel locations.

2.2.

Surgical Procedure

A laparotomy was performed to gain access to the liver and pancreas, and photoacoustic imaging was used to identify abdominal blood vessels in and around the pancreas and liver. To mimic a surgical scenario where the light source would surround a surgical tool and the US probe would be relatively stationary, while the surgical tool is operated,^{19,21,22,25,51} the fiber bundle was held at a variable angle relative to the transducer. Given this motivation and the existing literature describing that the maximum photoacoustic signal is achieved when the light axis intersects both the photoacoustic imaging target and the imaging plane,^30,52 we prioritized flexibility of the fiber bundle axis relative to the orientation of the imaging plane. Building on knowledge obtained from controlled measurements of light delivery placement relative to the imaging plane and a target of interest,^30,52–55 photoacoustic images were optimized by manipulating the light source to maximize the amplitude and appearance of photoacoustic signals, similar to the free-hand optimization procedure that a surgeon would perform with a specialized light delivery system attached to the surgical tool.

When imaging the pancreas, the superior mesenteric vein (SMV), right gastroepiploic vein (RGEV), and right gastroepiploic artery (RGEA) were visualized. These vessels are annotated in the schematic diagram of Fig. 2(a). The section of pancreatic tissue exposed to the laser was resected. In addition, pancreatic tissue that was not exposed to laser light was resected as a control specimen. The energies used to image the pancreas with the 1- and 5-mm-diameter light sources are shown in Fig. 3(a) (top).

Fig. 2

Vessels imaged in the (a) pancreas and (b) liver. The hepatic veins are inside of the liver but are displayed overlaid on the liver to allow for visualization of the vessel locations.⁵⁶

Fig. 3

Energy versus time graphs and fluence ($\mathrm{\Phi }$) versus time graphs for imaging the (a) pancreas and (b) liver. The 1- and 5-mm-diameter fiber bundles were used to image the pancreas during the time points indicated. The 5-mm-diameter fiber bundle was used to image the liver. The methods implemented to measure energy and calculate fluence are detailed in Sec. 2.3.

The hepatic veins [see Fig. 2(b)] were visualized when imaging the liver. The range of energies used to image the pancreas with the 5-mm-diameter fiber bundle is shown in Fig. 3(b) (top). Energies of 20 to 40 mJ in 5-mJ increments were used to excite the blood vessels to determine the minimum required energy needed to visualize the liver vasculature. A three-dimensional (3-D) reconstruction of the hepatic vein was created with the robot-held US probe translating in 1-mm increments across the liver surface, pausing at each increment to acquire end-inspiration photoacoustic or color Doppler images. The liver tissue exposed to the laser was resected as well as a control from a portion of the same liver lobe that was not in direct contact with the 5-mm-diameter fiber bundle. A control from the liver was additionally resected prior to any imaging. The resected pancreas and liver tissues underwent histopathological analysis with hematoxylin and eosin staining by coauthor S.B., a board-certified veterinary pathologist.

2.3.

Energy Measurements and Fluence Calculations

Energy was manually adjusted throughout the procedure. The input energy into the optical fiber or fiber bundle was recorded with a power meter internal to the laser system. These input energies were calibrated to output energies (recorded with a second external power meter) using the average of both pre- and post-experiment input-to-output energy measurements. The recorded energy as a function of time to image the pancreas and the liver prior to tissue resection is shown at the top of Figs. 3(a) and 3(b), respectively. Fluence, $\mathrm{\Phi }$, was calculated from these recorded energies, $E$, and the area of the incident light, which is defined as

2.4.

Ex Vivo Experiment

An ex vivo experiment was performed to investigate the in vivo observations of potential out-of-plane vessels. The ex vivo phantom consisted of bovine liver tissue that surrounded tubes filled with human blood. The three parallel blood vessels were set up to mimic the RGEV, SMV, and RGEA. A perpendicular blood vessel was used to mimic the out-of-plane vessel, as illustrated in Fig. 4. Energies of 5.5 to 70.1 mJ were used to visualize blood in this ex vivo experiment.

Fig. 4

Photographs of the ex vivo experiment setup. (a) Plastic tubes were filled with human blood to mimic the RGEV, RGEA, SMV, and the out-of-plane blood vessel. (b) Bovine liver tissue surrounded the phantom blood vessels while the L3-8 linear transducer and 5-mm-diameter fiber bundle were used to visualize the blood vessels.

2.5.

Image Display Parameters and Data Analysis

Photoacoustic images were created using delay-and-sum (DAS) beamforming and short-lag spatial coherence (SLSC) beamforming.^34,57 SLSC images were generated with the short-lag value equal to one (or 2% of the receive aperture) and the axial correlation kernel equal to one wavelength. These two SLSC image parameters are defined as $M$ and $k$, respectively, in previous publications.^34,57 Images were normalized to the brightest pixel, log-compressed, and a minimum threshold value was selected to optimize signal display.

Contrast was calculated using beamformed data prior to normalization and log compression, according to the following equation:

3.1.

In Vivo Pancreas Imaging

Figure 5 shows one result from the in vivo pancreas study. The US image with photoacoustic overlay is shown in Fig. 5(a). These images were created with DAS beamforming. The same photoacoustic channel data were beamformed using the SLSC approach, as shown in Fig. 5(b). This SLSC image is overlaid on the US image in Fig. 5(c). A Doppler image confirming the location of blood vessels in the pancreas is shown in Fig. 5(d).

Fig. 5

(a) DAS beamformed photoacoustic image of signals in in vivo pancreas overlaid on coregistered US image. (b) Corresponding SLSC beamformed photoacoustic image created with the same channel data. (c) SLSC beamformed photoacoustic image overlaid on US image. (d) Color Doppler confirmation of blood flow in vessels. (e) Contrast of the three signals in DAS photoacoustic images. The photoacoustic images shown here and the photoacoustic images used for the contrast measurements were acquired with an energy of 36 mJ.

Throughout the respiratory cycle, the pancreas moved in a relatively predictable motion due to the mechanical ventilator that was used to maintain respiration. However, the transducer and fiber source were fixed (i.e., they were not moving with the respiratory cycle). Thus, any observed signal motion over time indicates movement due to respiration. Contrast was measured to assess this type of motion. Figure 5(e) shows the measured contrast of the three signals in Fig. 5(b) as a function of time for image frames acquired during multiple respiratory cycles.

We observed three distinct coherent signals in the in vivo photoacoustic images of the pancreas [labeled as 1, 2, and 3 in Fig. 5(b)]. Each signal consistently appeared and disappeared with respiration, as shown in Fig. 5(e). The source of these signals could be blood vessels (either in the imaging plane or outside of the imaging plane), reflection artifacts, or signals from the location of the light source due to the higher fluence where the light source is incident on the tissue, as seen in previous experiments with interstitial light sources.³⁴ The color Doppler image in Fig. 5(d) shows that there is blood flow in the RGEV, RGEA, and SMV, which indicates that signal #2 could be associated with blood, although photoacoustic signals are not present in the entire RGEV.

The reason for the partial display of the signal is unknown, but one possible explanation could be locally higher absorption due to the location of a valve in the vessel. Due to this uncertainty, it could be argued that signal #2 did not originate from the blood in the RGEV and instead could be an artifact, particularly if signals #1 and #2 in the in vivo pancreas image of Fig. 5(a) were not distinct from each other. However, because signals #1 and #2 are distinct in the DAS image and have distinctly different contrast measurements with changes in respiration, we hypothesize that signal #2 originates from the blood.

The use of coherence-based SLSC beamforming supports our hypothesis that these signals originate from two independent coherent sources, because they appear as two distinct signals in the SLSC images. If there was a single source causing these signals, these signals would instead be merged together as one coherent source in the SLSC image. The SLSC image also confirms that coherent signals from the blood inside of the entire RGEV were not obtained.

While signal #1 could be associated with the location of the light source, this is unlikely because the appearance of this signal changed with the respiratory cycle, while the US probe and fiber bundle were held stationary, as demonstrated in Fig. 5(e). If this signal were solely caused by the stationary light source primarily interacting with tissue, then its contrast would be more constant with changes in respiration. Another possible explanation is that this signal could be a reflection artifact. However, reflection artifacts typically appear deeper than true signals. Thus, we hypothesize that signal #1 is most likely an out-of-plane blood vessel. Similarly, due to signal #3 changing with respiration, as shown in Fig. 5(e), we hypothesize that this signal also originates from an out-of-plane blood vessel. These out-of-plane vessels can be resolved with 3-D imaging, as demonstrated in Sec. 3.4.

3.2.

Out-of-Plane Vessel Confirmation

The ex vivo experiment described in Sec. 2.4 was performed to test our hypotheses about the source of the signals in the in vivo pancreas data. Although the fiber bundle was stationary while the tissue moved with respiration during the in vivo pancreas experiment, by inverting this process and keeping the ex vivo tissue stationary while moving the fiber bundle, we recreated our in vivo observations of photoacoustic signals appearing and disappearing from the image plane in this more controlled ex vivo experimental environment.

Figures 6(a)–6(c) show that the photoacoustic signal from the ex vivo out-of-plane vessel mimicked in vivo photoacoustic signal #3 in Fig. 5(b). Figure 6(d) shows contrast as a function of time for this ex vivo experiment, and this result supports the hypothesis that signals #1 and #3 are associated with out-of-plane blood vessels. In particular, the cyclic temporal appearance of the measured photoacoustic signal contrast in Fig. 6(d) is similar to that of the in vivo signals #1 and #3 in Fig. 5(e).

Fig. 6

(a) DAS beamformed photoacoustic image overlaid on US image of ex vivo blood vessels. (b) Corresponding SLSC beamformed photoacoustic image created with the same channel data. (c) SLSC beamformed photoacoustic image overlaid on US image. These photoacoustic images were acquired with an energy of 30.5 mJ. (d) Contrast of the out-of-plane vessel signal, measured in the DAS photoacoustic image as the light source location was moved to mimic one respiratory cycle.

3.3.

In Vivo Liver Imaging

Figure 7(a) shows the US image of an in vivo hepatic vein and a coregistered photoacoustic image created with DAS beamforming overlaid on this US image. The photoacoustic SLSC image created with the same channel data is shown in Fig. 7(b). This SLSC image is overlaid on the US image in Fig. 7(c). The color Doppler image confirming the location of blood vessels in the liver is shown in Fig. 7(d). Similar to the pancreas image, the blood signals in this photoacoustic image of the liver do not encompass the entire vessel location.

Fig. 7

(a) DAS beamformed photoacoustic image of signals in in vivo liver overlaid on coregistered US image. (b) Corresponding SLSC beamformed photoacoustic image created with the same channel data. (c) SLSC beamformed photoacoustic image overlaid on US image. The photoacoustic images were acquired with 40.5 mJ energy. (d) Color Doppler confirmation of blood flow in the hepatic vein.

Figure 8 shows an US image of the hepatic vein, which was confirmed with Doppler imaging. The diameter of the vein in Fig. 8 is $\sim 3\mathrm{mm}$, which is smaller than the diameter of the vessels shown in Figs. 5 and 7. When imaging these smaller-diameter vessels, we observed a change in photoacoustic signal appearance based on the location of the light source. In particular, when light from the fiber bundle was directed at this vein, we saw a significantly more focused photoacoustic signal. In contrast, when the light source was purposely directed away from this vein, the signal was more diffuse. This change in signal appearance is also evident in Video 1, and it is likely due to laser excitation of the blood inside of the hepatic vein (producing more focused signals) or laser excitation of the surrounding tissue (producing more diffuse signals).

Fig. 8

US image of the hepatic vein and corresponding DAS photoacoustic images (overlaid on US image), demonstrating focused and diffuse signals acquired with a laser energy of 30 mJ. The photoacoustic images are displayed with dynamic ranges of 15 dB (top) and 10 dB (bottom), as noted by the maximum and minimum values on the color bars. Reducing the dynamic range emphasizes the appearance of the focused signal and limits the appearance of the diffuse signal. More examples of focused and diffuse signals from this image acquisition sequence are shown in Video 1 (Video 1, MPEG, 2.8 MB [URL: https://doi.org/10.1117/1.JBO.24.12.121905.1]).

Although the appearance of focused versus diffuse signals were highly repeatable in real time, it is understandable that the more diffuse signals associated with the tissue can be mistaken for photoacoustic signal artifacts and can also confuse the reading of the still images. Therefore, one option to mitigate this confusion is to reduce the dynamic range of the photoacoustic images, as shown in Fig. 8 (bottom). This reduction is sufficient because the diffuse signals have consistently lower amplitude than the focused signals. However, the distinction between the two types of signals is more obvious with the higher dynamic range shown in Fig. 8 (top).

Figure 9(a) shows additional examples of the focused photoacoustic signal from the hepatic vein acquired at multiple energy levels. Each image is displayed with a dynamic range that was chosen to optimize signal visibility. The focused signal was not present at energies of 20 and 25 mJ, even with the higher dynamic range values. Beyond 25 mJ, the focused nature of the signals in the vessels was more apparent, particularly with reduced dynamic range values.

Fig. 9

(a) Focused photoacoustic signals in the hepatic vein, acquired with varying laser energy. Each image is annotated with the dynamic range chosen to optimize the signal display. (b) Measured contrast of focused and diffuse signals at the various laser energies.

Figure 9(b) summarizes contrast measurements of the focused signals shown in Fig. 9(a) and diffuse signals acquired with the same energy levels. Although the diffuse signals at each energy level are not shown, they look similar to the examples shown in Fig. 8 and in Video 1. The contrast measurements are independent of the displayed dynamic range, so they can be relied upon as objective measurements of signal visibility. At each energy level, the focused signal has consistently higher contrast when compared to the diffuse signals, which further supports the dynamic range reduction in Figs. 8 and 9(a), in order to emphasize the appearance of focused signals and limit the appearance of diffuse signals.

The focused and diffuse appearance of the photoacoustic signals observed in the liver is likely caused by the blood in the liver, particularly when considering that the prominent diffuse appearance was not observed in the pancreas, which generally has lower optical absorption and scattering than the liver (see Table 1).

3.4.

Three-Dimensional Vessel Imaging

Figure 10 shows multiple display formats for the 3-D vessel image data that we acquired with robotic assistance (as described in Sec. 2.2). The left column of Fig. 10 displays results associated with the photoacoustic signal, while the right column displays results associated with the color Doppler signal, which was used for confirmation of blood flow. Figure 10(a) shows the 3-D stack of images used for forming the volumetric data. Biplanar views of intersecting elevational and lateral slices are shown in Fig. 10(b), with the chosen slices outlined in green in Fig. 10(a).

Fig. 10

The 3-D representations of the hepatic vein created from DAS photoacoustic and color Doppler images. (a) Volume stack of photoacoustic and Doppler images acquired with a robot translating in 1-mm increments. (b) Biplanar cross sections through the volume stack. (c) 3-D reconstructions of the hepatic vein segmented from photoacoustic and Doppler images. (d) Segmented vessels from photoacoustic and Doppler images combined to show similarities between the two segmented vessel structures.

An isosurface of the segmented vessels based on the photoacoustic and Doppler volumetric image data is displayed in Fig. 10(c). There are size and shape differences between these two segmented vessels that occur because the photoacoustic signal is only present near the location of the 5-mm-diameter light source, which is unable to simultaneously illuminate all regions containing the blood signals. This capability to confidently limit the field of view (FOV) to a specific ROI at the surgical site is one of the main advantages of using photoacoustic imaging to guide surgeries. The presence of a major blood vessel near the surgical ROI can be highlighted, rather than displaying multiple regions with blood flow through the use of Doppler imaging. Although limiting the FOV may be possible with Doppler imaging, doing so would require coordination of identifying tool tips and possible adjustment of knobs on the US scanner to manually limit the FOV.

Figure 10(d) shows the two segmented vessels of Fig. 10(c) overlaid in the same spatial coordinate system. In addition to the smaller FOV with photoacoustic images (and hence the smaller vessel volume), differences between the two segmented vessels also exist because of the differences in the image acquisition procedure. Photoacoustic images were acquired as the robot stepped forward across the liver surface, while Doppler images were acquired as the robot retraced its steps in the opposite direction. Considering these two different motion directions, it is highly possible that the tissue deformed differently during the acquisition of each volume, which would cause the vessels to be shifted relative to each other.

3.5.

Histopathology

Figure 11 shows the histology results for the liver and pancreatic tissues after applying the multiple energies over the time duration shown in Fig. 1. No pathology was observed in the pre-experiment control in either the pancreas or the liver. There was no necrosis observed in the pancreas after being exposed to the laser. However, Fig. 11(a) shows an example of the mild degeneration in the post-experiment control and lasered pancreatic tissue. This mild cell degeneration observed in the pancreas after imaging was likely caused by the pressure of the US probe and other tissue handling. In the liver, necrosis was observed in both the lasered tissue and the post-experiment control tissue, as demonstrated in Fig. 11(b). There was also hemorrhaging and inflammation present in the lasered section of liver tissue. We interpret these findings within the scope of our overall study design in Sec. 4.

Fig. 11

Histopathology results of (left) pre-experiment control, (middle) post-experiment control, and (right) post-experiment lasered tissue for the (a) pancreas and (b) liver at $200\times$ magnification.