2.1.

Photoacoustic Imaging

A reflection-mode photoacoustic (PA) imaging system (Fig. 1 ) was used for identifying a SLN following methylene blue injection. This imaging system employed a dark-field ring-shaped illumination.¹³ By reducing the surface PA wave generation, this illumination allows more efficient detection of PA waves from the deep regions, compared with a bright field one. The surface PA waves can interfere with those from deeper regions and obstruct their detection.^{13, 14} The dark-field illumination was formed by a concave lens, a spherical conical lens, and an optical condenser arranged on an optical rail in tandem. To deliver high optical energy to a deeply located SLN, we used two prisms mounted on an $xy$ linear translation stage instead of an optical fiber, which has lower damage threshold.

Fig. 1

Schematic of a reflection-mode photoacoustic imaging system. PR, prism; CC, concave lens; CL, conical lens; OC, optical condenser; and UT, ultrasonic transducer.

For PA excitation, a $635\text{-}\mathrm{nm}$ wavelength from a pulsed laser was chosen. This wavelength is close to that of methylene blue’s peak optical absorption ( $677\text{-}\mathrm{nm}$ wavelength), which is ten times greater than hemoglobin on a per molar basis (95% oxygenated hemoglobin).¹⁵ The light source was a tunable dye laser (ND6000, Continuum, Santa Barbara, California) pumped by Q-switched Nd:YAG [LS-2137, LOTIS (Minsk, Belarus) II] laser, which provides $<15\text{-}\mathrm{ns}$ pulse duration and a $10\text{-}\mathrm{Hz}$ pulse repetition rate. Laser light delivered to the skin was sufficiently broadened by a concave lens to conform to the maximum permissible exposure (MPE) limitation (actual fluence: $3.4\mathrm{mJ}∕{\mathrm{cm}}^2$ , MPE: $20\mathrm{mJ}∕{\mathrm{cm}}^2$ ).¹⁶ We chose a $635\text{-}\mathrm{nm}$ wavelength, since the dye laser produces maximum energy at this wavelength with a DCM dye.

A 3.5- or $5\text{-}\mathrm{MHz}$ central frequency ultrasonic transducer (V380, V308, Panametrics-NDT, Watham, Maryland) was used for the PA signal detection, since ultrasonic waves at these frequencies suffer less acoustic attenuation in tissue than higher frequency waves. Both transducers were spherically focused, and had 4.95- and $2.54\text{-}\mathrm{cm}$ focal lengths; 2.54- and $1.91\text{-}\mathrm{cm}$ -diam active elements; and 70 and 72% bandwidths based on the full width at half maximum (FWHM) amplitudes, respectively. Transverse resolution of the system was estimated by $1.22\lambda f∕d$ ( $\lambda$ is the measured central wavelength, $f$ is the focal length, and $d$ is the diameter of an active element). Thus, the $5\text{-}\mathrm{MHz}$ transducer was more suited for higher transverse resolution imaging than the $3.5\text{-}\mathrm{MHz}$ one. The $3.5\text{-}\mathrm{MHz}$ transducer was appropriate for deeper imaging, since it has a longer focal length. The spatial resolutions of the system with the $5\text{-}\mathrm{MHz}$ ultrasonic transducer at $19\mathrm{mm}$ deep were $\sim 144\mu \mathrm{m}$ in the axial direction and $\sim 560\mu \mathrm{m}$ in the transverse direction.¹³

To obtain 3-D PA data, raster scanning was employed using an $xy$ linear translation stage (XY-6060, Danaher Motion, Washington, D. C.). This stage was controlled by a computer and synchronized with the data acquisition system. A continuous scan was used to shorten data-acquisition time without signal averaging. On the other hand, a stop-and-go scan was used to average PA signals to improve their SNR for deeper SLN imaging. The PA signal obtained along the depth direction ( $z$ direction) at a single point is called an A-scan. A 2-D B-scan was acquired by a 1-D scan, and was composed of multiple A-scans ( $x$ direction). With a 2-D scan ( $xy$ directions), a 3-D image was acquired. An A-scan, resolved in time along the depth direction, was converted into a 1-D depth-resolved image by multiplying it by the speed of sound in soft tissues $(\approx 1500\mathrm{m}∕\mathrm{s})$ .

Scanning step sizes were $\sim 0.2$ and $\sim 0.4\mathrm{mm}$ for 1-D and 2-D scans, respectively. With 0.1- and $0.2\text{-}\mathrm{mm}$ step sizes and a $10\text{-}\mathrm{Hz}$ pulse-repetition-frequency laser system, the acquisition time was $\sim 25\sec$ for a B-scan and $\sim 42\min$ for a 3-D image, covering a $20\times 20\text{-}\mathrm{mm}$ field of view. The transducer scanned an object through a $5\times 5\text{-}\mathrm{cm}$ opening sealed with a thin, clear membrane in a water container. The opening was located on the bottom of the water container. Ultrasonic gel was used between the clear membrane and the object to couple the ultrasound.

Following amplification by an amplifier (5072PR, Panametrics-NDT), PA signals were digitized by an oscilloscope (TDS 5054, Tektronix, Natham, Massachusetts) with a 50 mega-sampling rate. Since the laser system has pulse-to-pulse fluctuations in energy, PA signals were compensated by the signals from the photodiodes (DET110, Thorlabs, newton, New Jersey), which sample the energy of each laser pulse.

2.2.

Samples

Two types of phantom experiments were performed to assess the device’s sensitivity and estimate in vivo and ex vivo the concentration of methylene blue accumulated in a SLN. The PA signal from a SLN was compared with those from phantoms that had known concentrations of methylene blue to measure the concentration in the SLN. First, laboratory tubing (Dow Corning Corporation, Corning, New York) with $1.47\text{-}\mathrm{mm}$ inner diameter (ID) was placed between two layers of chicken breast tissue, as depicted in Fig. 2b . The tubing was filled with methylene blue of concentrations ranging from 0.1 to 1%, in 0.1% increments. The top tissue layer was $\sim 12$ and $20\mathrm{mm}$ thick in two consecutive sets of measurements. The effective attenuation coefficient of chicken breast tissue is $\sim 0.9{\mathrm{cm}}^{-1}$ at the $635\text{-}\mathrm{nm}$ wavelength, similar to that of human breast tissue.^{17, 18} PA signals were measured from methylene blue within the tubing. Before measurement of PA signals at each concentration, the tubing was cleaned with 0.9% saline to remove residual dye. To estimate the in vivo concentration of methylene blue, the PA signal from an in vivo SLN was compared with that from this phantom experiment. Second, similar to the first case, pseudolymph nodes of comparable dimensions were made of porcine gelatin mixed with methylene blue of various concentrations (0.009 to 0.7%). The pseudolymph node experiments were performed to estimate the concentration of the dye from an excised SLN ex vivo. These pseudolymph nodes were cylinders $3\mathrm{mm}$ in diameter and $4\mathrm{mm}$ in length, similar to the dimensions of a rat SLN. PA signals were directly measured from these pseudolymph nodes [Fig. 3b ]. The PA signal from the excised SLN was compared with those from the pseudolymph nodes to estimate the dye concentration. By using comparable dimensions, the effect of the size of the optical absorber was ruled out.

Fig. 2

(a) Amplitude of photoacoustic signal versus concentration of methylene blue dye at 12- and $20\text{-}\mathrm{mm}$ depth in chicken breast tissue, respectively. (b) Experimental setup for sensitivity measurement and in vivo estimation of concentration. (c) Estimation of concentration at $0\text{-}\text{to}20\text{-}\mathrm{mm}$ depth. UT, ultrasonic transducer; and WC, water container.

Fig. 3

(A) Amplitude of photoacoustic signal versus concentration of methylene blue dye at a pseudolymph node. (B) Setup for ex-vivo estimation of concentration. (C) Photoacoustic B-scans of the phantom, the dyed SLN, and the normal SLN. The concentration in the phantom is $\sim 0.07\%$ , and the dyed SLN was estimated to be the same concentration. (D) Photographs of SLNs with (left) and without (right) dye deposit. ultrasonic transducer, UG: ultrasonic gel, and Wc: Water container.

2.3.

Animal and Drug Information

All animal experiments were carried out in compliance with the guidelines on the care and the use of laboratory animals at Washington University in Saint Louis. Adult male Sprague Dawley rats weighing $250\text{to}350\mathrm{g}$ were used. The rat was initially anesthetized with a mixture of ketamine $(85\mathrm{mg}∕\mathrm{kg})$ and xylazine $(15\mathrm{mg}∕\mathrm{kg})$ . Before imaging, the hair on the region of interest was gently removed using a commercial hair-removal lotion. Intradermal injection of $0.07\mathrm{ml}$ of 1% methylene blue ( $10\mathrm{mg}∕\mathrm{mL}$ , American Regent, Incorporated) was performed on a left forepaw pad. After administration of methylene blue, photoacoustic images were acquired. During all image acquisitions, anesthesia was maintained using vaporized isoflurane ( $1\text{-}\mathrm{L}∕\min$ oxygen and 0.75% isoflurane, Euthanex Corporation, Shirley, New York), and vitals were monitored using a pulse oximeter (8600V Nonin Medical Incorporated, Plymouth, Minnesota). During image acquisition, $8\mathrm{ml}$ of 0.9% saline was administered for hydration. After image acquisition, the animal was euthanized by pentobarbital overdose.

3.1.

Sensitivity of the System

An imaging system with high sensitivity is required to detect low concentrations of methylene blue accumulated in a SLN located deep below the skin surface. The sensitivity of the photoacoustic (PA) imaging system was obtained from the chicken tissue phantom experiments. Figure 2a shows the peak PA signal amplitude versus methylene blue concentration at 12- and $20\text{-}\mathrm{mm}$ depths, respectively. The detection limit of the system was calculated to be $\sim 0.07\mathrm{mV}$ (three times the standard deviation of the photoacoustic signal without the dye) at $20\mathrm{mm}$ . This voltage corresponds to a 0.009% concentration based on the extrapolation of the $20\text{-}\mathrm{mm}$ curve in (Fig. 2a). Therefore, the imaging system is expected to be able to detect a SLN containing greater than 0.009% $\left(0.28\text{-}\mathrm{mM}\right)$ methylene blue located $20\mathrm{mm}$ from the skin surface. However, the depth capability of the system is not limited to $\sim 20\mathrm{mm}$ deep.

3.2.

Concentration Estimation in Vivo and ex Vivo

To identify a SLN from a group of lymph nodes, it is necessary to measure the concentration of methylene blue accumulated within the SLN. in vivo concentration within a SLN can be estimated by measuring a PA signal from the SLN and reading the concentration corresponding to the measured PA signal in Fig. 2a at the 12- and $20\text{-}\mathrm{mm}$ depth. Since the amplitudes of PA signals are proportional to optical fluence in a medium, and the fluence between two depths can be estimated based on the diffusion equation,¹⁹ the amplitudes of PA signals for other depths were interpolated using the diffusion equation for all concentrations of methylene blue. Interpolated values were plotted in mesh [Fig. 2c]. The imaging depth $(\sim 18\mathrm{mm})$ was demonstrated by placing two layers of chicken breast tissue on top of the axillary surface of the rat in vivo. At this depth, the peak amplitude of the PA signal from the SLN was $\sim 0.3\pm 0.008\mathrm{mV}$ (standard error), corresponding to $\sim 0.07\%\pm 0.05\%$ ( $2.2\mathrm{mM}$ , standard error) methylene blue [Fig. 2c]. For concentrations below 0.1% in (Fig. 2b), the amplitudes of the PA signals were extrapolated based on the diffusion equation.

After all in vivo PA images of the rat SLN were acquired, the SLN was excised to measure the PA signal more accurately. The peak amplitude of the PA signal, which was acquired using the method depicted in Fig. 3b, was $\sim 17\pm 0.09\mathrm{mV}$ (standard error). This voltage corresponds to $\sim 0.07\%\pm 0.0006\%$ (standard error) methylene blue [Fig. 3a], which is equivalent to the in vivo estimation at $18\text{-}\mathrm{mm}$ depth. These results show that the imaging system has high sensitivity and accuracy in identifying a SLN and in estimating its dye content both in vivo and ex vivo. [Figure 3c] shows the PA B-scans of the pseudolymph node, the dyed SLN, and the normal SLN. The pseudolymph node in Fig. 3c “Phantom,” had a concentration of $\sim 0.07\%$ , equivalent to that of the dyed SLN in Fig. 3c “Dyed SLN.” Therefore, both the pseudolymph node and the dyed SLN have PA signals of the same amplitude. By contrast, the normal SLN did not generate a PA signal, showing that this did not exhibit optical contrast. Figure 3d shows photographs of a dyed SLN and a normal SLN, respectively. Should this technology be accepted clinically, the average concentration in a SLN can be measured and used for SLN identification.

3.3.

Sentinel Lymph Node Imaging in a Rat in Vivo

A rat SLN was clearly imaged by noninvasive in vivo PA imaging. Figure 4a is a photograph of a rat taken prior to image acquisition and after hair removal of the axillary surface; Fig. 4b is a photograph of the same rat with the skin removed after PA imaging. The SLN is unclear in the photograph because it is surrounded by fatty tissue. Before methylene blue injection, a PA control image was obtained, and is shown in the form of a maximum amplitude projection (MAP) [Fig. 4c].²⁰ The vasculature near an axial node was clearly imaged with high contrast ( $51\pm 7$ , standard deviation with respect to the background) and good resolution $(\sim 500\mu \mathrm{m})$ . Since methylene blue molecules are small ( $0.7\mathrm{nm}$ in diameter and $1.6\mathrm{nm}$ in length)²¹ and migrate into a SLN quickly, scanning started immediately after administration of methylene blue. The SLN appeared at the left top corner with a much higher contrast ( $146\pm 41$ , standard deviation with respect to the background) than the surrounding blood vessels [Fig. 4d], since methylene blue has higher optical absorption than blood at the $635\text{-}\mathrm{nm}$ wavelength. The contrast of the SLN with respect to the blood vessel was $\sim 2$ . The signal amplitude of the surrounding blood vessels was also increased by $\sim 19\%\pm 3\%$ compared to that in the control image, since some of methylene blue flowed into the blood stream directly and/or through the SLN. Accordingly, optical absorption in blood was increased. In all PA images, the brighter the object, the higher the absorption. Another PA image was acquired at $\sim 50\min$ after dye injection with the scan head repositioned [Fig. 4e]. The SLN was still observed. In humans, the mean depth of SLNs is $12\pm 5\mathrm{mm}$ (standard deviation, from the SLN’s top surface to the skin surface (unpublished data), To check the feasibility of this imaging depth, chicken breast tissue was placed on top of the axillary surface of the rat, and then a PA image was obtained [Fig. 4f]. The demonstrated imaging depth was $\sim 18\mathrm{mm}$ . Since methylene blue accumulated in the SLN still had higher absorption than the nearby blood vessels, the SLN is dominant in the PA image, while the blood vessels have faded away. Figure 4g shows an image of B-scans along the dotted line in Fig. 4e, showing the depth of the SLN. A 3-D volumetric image was also constructed [Fig. 4h]. In some clinical cases, the depth of a SLN is greater than $18\mathrm{mm}$ ( $18\pm 5\mathrm{mm}$ , standard deviation, from the SLN’s bottom surface to the skin surface). To increase imaging depth as well as SNR, higher incident energy of a laser is necessary for the PA excitation, while still staying within the MPE. We note that in this study, the incident fluence of the laser was five times less than the MPE, so increasing the laser fluence can enhance the PA signal by five times.

Fig. 4

Noninvasive in vivo photoacoustic MAP images of the SLN in a rat. (A) Photograph with hair removed before photoacoustic imaging. (B) Photograph with skin removed after photoacoustic imaging. (C) Control photoacoustic image without methylene blue injection. Bright parts represent optical absorption, here, blood vessels. $x$ and $y$ denote B-scan and 3-D imaging. (D) After-injection photoacoustic image. (E) After-injection photoacoustic MAP image $\sim 52\min$ after injection with the scan head repositioned. (F) Photoacoustic MAP image of $\sim 18\text{-}\mathrm{mm}$ -deep SLN. Chicken breast tissue was placed on top of the axillary surface of the rat. (G) Photoacoustic B-scan image corresponding to the dotted line in (E), showing the depth of the SLN. The vertical axis in this image corresponds to depth. (H) Photoacoustic 3-D volumetric visualization. The color bars represent optical absorption. (C) and (D) have the same color bar range as (E). (Color online only.)

Accordingly, in vivo PA SLN imaging of another rat was conducted with twice the incident optical fluence. The $3.5\text{-}\mathrm{MHz}$ central frequency ultrasonic transducer (V380, Panametrics-NDT) was employed instead of the $5\text{-}\mathrm{MHz}$ one, since the former has a longer focal length and is more suitable for deeper SLN identification. Figure 5a is the PA control image before injection of methylene blue, showing vasculature in the skin. [Figure 5b] reveals the SLN after methylene blue injection. For demonstration at a greater depth, chicken breast tissue about $18\mathrm{mm}$ thick was placed on the axillary region. Figure 5c is an $\sim 20\text{-}\mathrm{mm}$ -deep PA SLN image. In Fig. 5c, the contrasts of the SLN and the blood vessel with respect to the background were $24.5\pm 3.3$ and $13\pm 2.3$ (standard deviation), respectively. The contrast of the SLN with respect to the blood vessel was $\sim 0.9$ . Figure 5d is an image of B-scans acquired along the far-right dotted line in Fig. 5c, showing the depth of the SLN. Here, the skin surface is not shown due to the limited record length of the acquisition system. We further tried deeper SLN imaging by adding a second layer of chicken breast tissue. Three B-scan images [Fig. 5e] were obtained at three locations of the SLN area with 50 times averaging. The total imaging depth demonstrated was $\sim 31\mathrm{mm}$ , greater than the mean depth of the SLN in humans (from the SLN’s bottom surface to the skin surface).

Fig. 5

Noninvasive in vivo photoacoustic MAP images of the SLN in another rat. (A) Control photoacoustic image without methylene blue injection. (B) After-injection photoacoustic image. (C) Photoacoustic MAP image of $20\text{-}\mathrm{mm}$ -deep SLN. (D) Photoacoustic B-scan image corresponding to the dotted line (d) in (C). (E) Photoacoustic B-scan images acquired at three different locations [(a), (b), (c)] in (C). Total imaging depth was $\sim 31\mathrm{mm}$ . For this depth, the $3.5\text{-}\mathrm{MHz}$ transducer and 50 times averaging were employed.