1.

Introduction

Percutaneous coronary intervention is performed in almost 0.5 million cases each year in the United States and involves placement of a stent with a balloon that is inflated in the diseased coronary artery.¹ This procedure has improved in recent years with the advent of the drug-eluting stent (DES). There is still concern about failure rates from vessel renarrowing, restenosis, with cases occurring in up to 20% of patients.^1–5 We hypothesize that these failures would be minimized by evaluating in situ coronary artery drug delivery.^6–14

Restenosis was minimized by improving drug retention and transport in the coronary artery. Drug delivery to the tunica media has reduced or prevented migration and proliferation of vascular smooth muscle cells that cause restenosis.¹⁵ Drug diffusion and retention has also increased by modifying the drug’s physiochemical properties. For example, the 42’ hydroxyl group on rapamycin (a common drug on DES) has been replaced with a tetrazole ring (zotarolimus) that improves hydrophobicity, promoting retention and prolonged release, while minimizing washout.^14,16

The design and performance of a DES may be enhanced by monitoring or visualizing drug delivery to the coronary artery. This is supported by our findings that demonstrate drug transport and mechanical properties of the porcine coronary arteries vary as a function of vessel and location.^6–8 However, currently, there is no ideal imaging modality to monitor and quantify drug delivery to the coronary artery with sub-mm precision.

Although noninvasive imaging modalities, such as magnetic resonance imaging, computed tomography, and ultrasound (US), are capable of anatomical and functional imaging of the heart and coronary artery, they typically lack sufficient contrast to identify and track pharmacologic compounds in the vasculature and myocardium. Optical imaging techniques have drawbacks as well. Fluorescence microscopy and histology require toxic dyes and tissue destruction. Multiphoton confocal microscopy has limited depth penetration of less than 1 mm.^6,17,18 As a result, there are drawbacks to popular techniques used for tracking drug delivery from a DES in an intact heart. On the other hand, photoacoustic (PA) imaging and spectroscopy combined with pulse echo (PE) US potentially overcomes these limitations by offering real-time volumetric imaging with sub-mm resolution deep into tissue. With the benefit of spectroscopy, photoacoustic imaging and spectroscopy (PAIS) also provides high contrast and specificity for detecting and tracking optically absorbing compounds in the vascular tissue and myocardium. These attributes are not only favorable for testing and improving the performance of a DES, but the technology is potentially translatable for in vivo imaging in large animals and humans.

PA imaging is based on the PA effect, the generation of acoustic waves caused by the absorption of a short pulse of light, transient heating, and thermoelastic expansion. These acoustic waves are detected to form an US image with contrast proportional to the absorption of light, which depends on the optical wavelength. The distribution and relative concentration of optically absorbing drugs and contrast agents can also be estimated with high specificity by sweeping the optical wavelengths. In this study, we propose PAIS as a method to track transport of a drug surrogate in the porcine myocardium and coronary arteries. We also demonstrate imaging of a laser-sintered DES coated with a drug surrogate (also a PA contrast agent) and track its position in the intact porcine brachiocephalic trunk.

2.

Materials and Methods

Optical and US setup: a wavelength-tunable laser (Surelite-II OPO, Continuum, 5-ns pulses, 20-Hz pulse repetition rate, and $20\mathrm{mJ}/\text{pulse}$) operating between 410 and 670 nm was coupled into a 1.6-mm-diameter multimode optical fiber (CeramOptec) that connected to a “photoacoustic enabling device” (PED). Specifically, the PED held both the optical components and linear US array (Zonare L10-5, 7.1-MHz center frequency, and 5-MHz bandwidth) for coregistered PE and PA imaging, which has been previously described.^19,20 Furthermore, the PED allowed for coalignment between the optical illumination and plane of US detection. Briefly, light from the fiber passed through a cylindrical lens to provide line illumination at the surface of the tissue. The backscattered PA waves were reflected by an inline acoustic reflector before being detected by the linear array and recorded in real-time by a clinical US scanner (zOneUltra, Zonare Medical Systems). The PED was filled with water for acoustic coupling. Raw frame data (amplitude and phase) were transferred from the scanner to a personal computer running MATLAB™ for sum-delay beamforming and postsignal processing. The axial and lateral resolutions of the PA imaging system were 300 and $400\mu \mathrm{m}$, respectively.¹⁹ The slice thickness was $\sim 2\mathrm{mm}$. The fluence on the surface of the heart was $\sim 10\mathrm{mJ}/{\mathrm{cm}}^2$ and below the American National Standards Institute safety limit of $20\mathrm{mJ}/{\mathrm{cm}}^2$ per pulse light at visible wavelengths.²¹

Drug surrogate: 1,1’-dioctadecyl-3,3,3’,3’-tertramethyllindocarbocyanine perchlorate (DiI) was selected for PA imaging due to its weight match to rapamycin, hydrophobic nature, cost, and favorable absorbance spectrum at visible wavelengths [Fig. 1(b)].⁶ DiI was acquired from Invitrogen (Carlsbad, California).

Fig. 1

PA imaging setup and optical properties of DiI. (a) Nd: YAG laser was tripled to 355 nm to pump an optical parametric oscillator and produce tunable 5 ns pulses between 410 and 670 nm. Laser light was coupled into a 1.6-mm-diameter multimode glass fiber. The fiber connected to a PA flashlight adapter, including an optical assembly with a cylindrical lens to produce a line pattern on the heart.²⁰ The adapter also contained a water-filled cavity for acoustic coupling and an acoustic reflector, which ensures coalignment of the light with the plane of US detection. A clinical US scanner (zOneUltra, Zonare Medical Systems) and 7-MHz linear array (L10-5) recorded the raw acoustic signals (phase and amplitude) for US and PA imaging. The frame data were transferred to a PC for beamforming, postprocessing, and analysis in MATLAB™. (b) Optical absorption spectrum of DiI as measured with a photospectrometer (Ocean Optics, USB4000).

3.

Results

3.1.

Validation and Calibration in Excised Myocardium

Images of four of the seven myocardial samples are displayed in Fig. 2. Wavelengths of 550 and 660 nm represented the local maximum and minimum for DiI absorption in the visible range. Whereas at 550 nm, the PA image was dominated by DiI absorption, at 660 nm the image was dominated by the intrinsic absorption of the myocardium. In general, the PA signal intensity increased linearly with concentration. The final column represents the DiI concentration after unmixing the PA signal at all the wavelengths.

Fig. 2

US and PA imaging of porcine myocardial samples stained with DiI. Each row represents a sample stained with a different concentration of DiI (0, 38.4, 76.8, and $120\mu \mathrm{g}/\mathrm{ml}$). (a) Cross-sectional PE US images of samples embedded in Agarose™. Samples are outlined in green. (b) Corresponding PA signal at 550 nm (near peak of DiI absorption). (c) PA signal at 660 nm (low DiI absorption). (d) Relative concentration of DiI after spectral decomposition. Green scale bar represents 2 mm.

These data were used to generate a calibration curve for DiI in the excised myocardium. The average PA signal (over the ROI) was plotted against the prepared DiI concentration (Fig. 3). The axes are displayed in octaves to match the distribution of prepared DiI concentrations.

Fig. 3

PA calibration curve for DiI. The pixel values over the ROI were averaged and plotted against the prepared concentration of DiI for each sample. The equation for the best fit line (dashed red) and $R^2$ value are also indicated. All concentration values are in $\mu \mathrm{g}/\mathrm{ml}$. Based on the PA signal without DiI (blue-dashed line), the sensitivity of PA for detecting the dye in the samples was $\sim 2\mu \mathrm{g}/\mathrm{ml}$.

Figure 3 indicates a strong linear relationship spanning almost eight octaves between the DiI sample concentration and the amplitude of the PA signal. In addition, minute concentrations of DiI ($<2.4\mu \mathrm{g}/\mathrm{ml}$) could be detected at high spatial resolution ($<300\mu \mathrm{m}$) deep in tissue ($>5\mathrm{mm}$). These favorable attributes of PA imaging were extended to study transport in the porcine coronary artery and myocardium.

3.2.

Imaging of the Porcine Heart (In Situ)

DiI at $2.4\mu \mathrm{g}/\mathrm{ml}$ was perfused into the LADCA for 15 s (see Fig. 4). The PA signal at 550 nm near branching vessels increased linearly during the perfusion, demonstrating real-time capabilities of PA imaging for tracking drug transport and diffusion.

Fig. 4

Increase in PA signal near LADCA during DiI perfusion. (a) Cross sectional PE image of an ex vivo porcine heart. (b) Coregistered PA image at 550 nm during infusion of $2.4\mu \mathrm{g}/\mathrm{ml}$ DiI. Vessel pressure was maintained at 70 mm Hg. (c) Average PA signal over time inside dashed green box during infusion of DiI. The PA signal increase is observed in myocardial tissue adjacent to the coronary (CA) vessel. The scale bar denotes 2 mm.

Estimates of DiI diffusion in the porcine heart are presented in Fig. 5. The PA spectrum was dominated by absorption of the myocardium except at regions with high DiI. Areas of DiI diffusion for 2.4, 12, and $60\mu \mathrm{g}/\mathrm{ml}$ were calculated at 1.30, 1.75, and $3.30{\mathrm{mm}}^2$, respectively. Therefore, an increase in DiI concentration led to an increase in diffusion area. These results demonstrated that PA imaging can track transport of DiI, a drug surrogate.

Fig. 5

Estimation of DiI diffusion in the heart near the LADCA. (a) Cross sectional PE US image over LADCA. (b) Coregistered PA image at 550 nm with a dynamic range of 40 dB. The PA spectra over regions of interest marked with “X” are displayed in (c). Brown and blue regions are similar to the known absorption spectra of the myocardium and DiI, respectively. (d) Coregistered PE (gray) and unmixed PA signal for $60\mu \mathrm{g}/\mathrm{ml}$ DiI concentration (following flush). Green box identifies the ROI for cumulative diffusion analysis. (e) PA images over ROI (green box) for DiI concentration of 0 (100% water), 2.4, 12, and $60\mu \mathrm{g}/\mathrm{ml}$. (f) Diffusion area for each concentration based on PA signal over background. Scale bar is 2 mm.

In addition to diffusion area, we also calculated the diffusion distance from the center of LADCA, which depended on the concentration (see Fig. 6). DiI diffused an average of $0.95\pm 0.37\mathrm{mm}$, ranging from 0.6 to 1.5 mm [Fig. 6(c)]. These results suggest that PA imaging can track transport several millimeters from the coronary into the myocardium.

Fig. 6

Diffusion distance from LADCA. (a) Cross sectional US image of the LADCA in the porcine heart. (b) Overlay of PE US (gray) and PA (hot) images. $\text{Scale bar}=2\mathrm{mm}$. (c) Calculation of diffusion distance from LADCA based on signal above threshold noise inside dashed box in (b). Diffusion distance for this example ranged from 0.6 to 1.5 mm. $\text{Scale bar}=1\mathrm{mm}$.

3.3.

Ultrasound and Photoacoustic Imaging of a DiI-Coated Stent

A cardiovascular stent was fabricated to test if PA imaging could detect DiI coated on a stent placed inside a blood vessel. Images of the DiI-coated cardiovascular stent are presented in Fig. 7. PA imaging was able to differentiate the coated versus uncoated region in water and in the brachiocephalic trunk [Figs. 7(d) and 7(f), respectively]. These results demonstrated that PA imaging was able to detect DiI on a stent placed inside a large blood vessel.

Fig. 7

PE and PA imaging of DiI-coated cardiovascular stent. (a) Zoomed-in photograph of sinusoidal pattern of stent; (b) top view PE US image of stent at 7.1 MHz. (c) Top view photograph of stent with $\mathrm{DiI}/\mathrm{PCL}+\mathrm{PDMS}$ coating on top half. (d and e) Coregistered PE US (gray) and PA (hot) at 550 nm image ($\mathrm{d}=\text{long view}$, $\mathrm{e}=\text{cross view}$) of stent in water with coating on top half. (f) Coregistered US and unmixed PA signal of stent implanted in the brachiocephalic trunk. $\text{Scale bar}=2\mathrm{mm}$. Green box highlights the stent in the blood vessel.

4.

Discussion

Our results demonstrated that PAIS is a nondestructive method that tracks drug delivery in the porcine vasculature while providing excellent contrast, depth, and resolution. PAIS was able to quantify the distribution of a drug surrogate at concentrations from 2.4 to $120\mu \mathrm{g}/\mathrm{ml}$. Perfusion of DiI into the coronary artery increased the PA signal in the surrounding tissue greater than 1.5 mm away from the targeted vessel. PAIS was also able to differentiate the brachiocephalic trunk from the coated and uncoated regions of a stent. Our results suggest that PAIS may be a valuable tool for tracking drug delivery to a specific vessel and provide feedback on drugs loaded onto a DES. In this regard, PAIS may help improve the design of a DES, increase drug retention and delivery to the coronaries arteries, and reduce in-stent restenosis.

Our approach has not only demonstrated that PAIS may be a useful tool for assessing drug delivery in vitro but could also be used in preclinical animal models of vascular disease. For example, a porcine thoracic window to the coronary artery could provide time course information about in vivo drug delivery.²³ Specifically, DES polymer degradation and therapeutic drug transport could be monitored over time in the same animal, thus minimizing the number of animals used in preclinical studies. Our results further suggest that the PA signal from the stent and surrounding tissues could be used to estimate drug concentrations thus providing real-time feedback for improving DES designs.

To visualize drug delivery, this study implemented a drug surrogate with a peak absorption near 550 nm. At this wavelength, penetration depth in cardiac tissue is about $\sim 5\mathrm{mm}$ due to overlap in intrinsic background absorption (e.g., hemoglobin). However, if the drug surrogate is exchanged for a near-infrared (NIR) absorber, PAIS should be able to track drug delivery even deeper ($>10\mathrm{mm}$) into tissue with less interference from background absorption.²³ In this case, it may be possible to use fewer wavelengths for spectral decomposition, thereby simplifying the imaging procedure and scan time. In general, most restenosis drugs absorb in the ultraviolet range, which limits penetration depth.^24,25 Therefore, it could be desirable to conjugate a small-molecule dye onto the drug to track delivery using PAIS. For example, the 42’ hydroxyl group on rapamycin has shown great promise for conjugation of an optically absorbing dye. One group has successfully attached 7-bromo-4-nitrobenzofurazan, a visible spectrum dye, onto rapamycin.²⁶ Conjugating rapamycin with a visible or NIR absorber would facilitate PAIS for quantifying drug delivery to the coronary artery.

In order to track drug delivery, there are many confounding factors that affect the PA signal in the heart: dye concentration, laser fluence, sample geometry, dye distribution, background absorption, and photobleaching. To reduce effects of pulse-to-pulse variations of the laser, we averaged the signal at each location over several pulses. Photobleaching of the dye was also kept to a minimum by only exposing the samples to light while recording data (typically less than 10 pulses at each location). To reduce effects of the geometry on the PA signal during the calibration experiment, each sample was cut to a similar size and shape and placed in the agarose gel at the same depth. The ROIs were chosen inside the sample to reduce boundary effects of the PA signal. Finally, dye variations inside each sample were reduced by soaking the samples for 24 h in the dye. A similar protocol was employed by Keyes et al. prior to PA imaging in porcine LADCA samples.⁶ In that study, it was assumed that 24 h was sufficient for saturation of the dye in the tissue.

There were two other primary limitations in this study. First, the spatial resolution of $\sim 300\mu \mathrm{m}$ was limited by the 5-MHz bandwidth of the US transducer. A higher frequency US probe (e.g., $>50\mathrm{MHz}$) could be incorporated into our system to improve the resolution to less than $50\mu \mathrm{m}$, which would be sufficient to track diffusion through layers of the coronary wall.²⁷ Another limitation was the inability to directly apply our setup to image the heart in vivo. Although transthoracic PA imaging is possible, penetration would be limited to about 2 cm. In this case, intravascular US/PA (PA-IVUS) would be the most practical solution for in vivo cardiac imaging of a DES. Currently, PA-IVUS has been used to track the position of a stent and detect vulnerable plaque in animal models.^28–31 Such probes are being developed and tested by other groups with the ultimate goal of clinical translation.^29,32–35

In summary, this study demonstrated that PA imaging offers excellent spatial resolution, penetration, and contrast for tracking transport of a drug surrogate in the porcine vasculature. PAIS also improves the specificity for estimating drug concentrations by separating the signal from intrinsic background absorbers in cardiac tissue. In vivo PA imaging of drug retention and delivery in the coronaries could help improve the design and performance of a DES and reduce the probability of restenosis during interventional treatment of CHD.