1.

Introduction

Over the past 10 years, advances in technologies have opened opportunities to utilize spectroscopic modalities for biological and medical applications. The spectral information obtained by observing the interaction of a sample with electromagnetic radiation is very versatile and can offer qualitative and quantitative characterization of its biochemistry. Among the very popular spectroscopic techniques that can be easily coupled to optical elements are methods based on electronic or vibrational excitations in the visible or infrared (IR) region of the electromagnetic spectrum. Vibrational excitations can be generated by either absorption or inelastic scattering of light, which are the bases of IR/NIR and Raman spectroscopy, respectively. Several spectroscopic techniques have been applied to characterize atherosclerotic plaques in order to obtain information about the biochemical composition of the affected tissue. From post mortem studies, it is well known that the severity of a plaque and its stability are strongly correlated with its biochemical composition.¹ The identification of vulnerable plaques especially remains one of the most important and challenging aspects in cardiology. Thus, specific information about the composition of a plaque would greatly improve the risk assessment and management. Furthermore, knowledge about the composition can offer new therapeutic and medication strategies. The application of spectroscopic techniques to biological specimens is meanwhile relatively straightforward. The greatest challenge is the direct application in vivo. Currently, there are several indispensable noninvasive and invasive imaging modalities to diagnose atherosclerosis and localize affected areas. Techniques based on computer tomography (CT), such as electron beam CT (EBCT) and multidetector CT (MDCT), can identify calcification levels, lipid cores, and vessel remodeling, but are either too specific (EBCT only detects calcification) or not sensitive enough (MDCT exhibits high negative but low positive predictive values).² Positron emission tomography can identify inflammation, which is an indicator of vulnerability, but not the only one.² The most common invasive techniques are intravascular ultrasound (IVUS) and optical coherence tomography (OCT). Both catheter-based modalities deliver very precise images of plaque morphologies that can be indicative for certain plaque types. IVUS detects the amplitude of soundwaves and has a penetration depth of several millimeters that allows visualizing the lamina of the vessel wall. The obtained grayscale image can be further interpreted by radiofrequency analysis, which can be depicted in a false color image termed as virtual histology (VH-IVUS).³ OCT on the other hand is based on interferometry and has a better spatial resolution, but cannot penetrate much more than 1 to 2 mm. IVUS and OCT are complementary to each other and hybrid modalities are in development.^4,5 In any case, both modalities deliver grayscale images that deliver rather unprecise information about the biochemical composition of the plaque and require a very high amount of long time empirical expertise from the point of inspection. Spectroscopic techniques would, therefore, be an ideal tool to complement the imaging information. Among the first methods, near-infrared spectroscopy (NIRS) has been introduced that is especially sensitive to the identification of lipid pools.^6,7 NIRS probes for overtones and combinations of molecular vibrations and can contrast lipid molecules such as cholesterol or triglycerides against the collagen of the vessel wall. It is not the most sensitive technique with little penetration, but offers molecular information of especially bulk material without interference from water. In combination with sophisticated data analysis, pullback chemometric images can be generated. To support the plaque visualization, NIRS is commercialized in combination with IVUS and has very recently been successfully applied in clinical studies.^8,9 Another promising spectroscopic approach is fluorescence lifetime imaging (FLIM), which is based on the different, molecule specific, decay time of laser-induced autofluorescence. The technique can distinguish different plaque components and has been applied ex vivo on open vessel carotenoid plaques by an en face illumination and in vivo in combination with ultrasound.^10–13

Along with the development of NIRS for the characterization of plaque compositions, the potential of Raman spectroscopy has been investigated. As a vibrational spectroscopic technique, it is closely related to the IR techniques. The first Raman studies of atherosclerotic tissue were reported at about the same time as the NIR data.¹⁴ Meanwhile, there have been a few attempts in applying the technique under in vivo conditions.^15,16 Several probe designs have been tested on different animal models. Because molecular vibrations and not overtones are probed, the obtained spectra contain much more information and are more sensitive with respect to the overall composition of the plaques. Main components such as lipids and calcifications can be readily identified. Raman-probe spectroscopy has been combined with other spectroscopic modalities as for instance intrinsic fluorescence and diffuse reflectance spectroscopy to evaluate the potential of multimodal spectroscopy for the detection plaques with a particular emphasis on the identification of vulnerable plaque sites.¹⁷ Recently, the potential of intravascular Raman spectroscopy has been demonstrated in vivo on a rabbit model.¹⁸ Main aspects in the probe design can now be addressed and measuring conditions, such as the incoming laser power and acquisition time, are feasible for in vivo conditions. The greatest challenge at the moment is the development of Raman probes that have geometrical parameters that are compatible for applications in humans. In order to overlay the obtained spectral information with plaque morphology, it will be ideal to combine such a probe with an intravascular imaging modality such as IVUS or OCT. Since the signal generation in both OCT and Raman spectroscopy is realized by laser radiation, the combination of the latter is technically much more suitable. In order to test the feasibility of OCT and Raman spectroscopy under similar conditions, we applied both technologies to a rabbit model and correlated the spectral information with the OCT images.

2.

Methods

3.

Results

After the application of the catheter through the carotid artery of the rabbit, the Raman probe was inserted and positioned at the lower aorta just before the bifurcation of the iliac arteries. Raman measurements were started within the abdominal aorta at illumination times between 1 and 3 s. Spectra were taken approximately every 0.5 cm. During the measurements, the body of the rabbit was constantly monitored by x-ray to record the exact position of the Raman probe. The appearance of the spectra was inspected for Raman bands and band patterns characteristic for lipids. Principally, spectral patterns of lipids, such as cholesterol, cholesterol esters or triglycerides can be easily distinguished from spectra of the vessel wall, which mainly consists of collagen and, therefore, exhibits spectral features associated with molecular vibrations of the protein backbones and typical functional groups. During the experiment, the Raman probe was carefully withdrawn through the abdominal and thoracic aorta into the aortic arch. The first noticeable spectral features of lipids were found within the upper part of the aorta above the heart of the rabbit. Figure 1 shows the spectrum of a normal part of the aorta from the abdominal region in comparison with a spectrum from the region close to the aortic arch. The Raman spectrum of the normal vessel shows the typical characteristics of vibrational bands of proteins, whereas the spectrum of the vessel of the upper part of the aorta exhibits Raman bands typical for lipids. Easy to notice are enhanced band intensities in the region between 2800 and $3100{\mathrm{cm}}^{-1}$, as well as within the lower spectral region between 900 and $1800{\mathrm{cm}}^{-1}$. All Raman bands can be clearly assigned to either proteins or lipids. The spectral features of arterial walls and lipids have been previously described in detail.¹⁹

Fig. 1

Raman spectra of an aorta position without plaque depositions (a), plotted in blue, and with lipid depositions (b), plotted in red. Spectrum (a) shows the typical spectral features of proteins, reflecting the protein/collagen composition of the arterial wall. Spectrum (b) exhibits all spectral characteristics typical for lipids.

After the recording of the Raman spectra of normal and abnormal appearance, the Raman probe was exchanged with an OCT probe. Again assisted by the guidance of the catheter, the probe was moved to the position where the Raman spectra with lipid features were measured. The associated OCT images are shown in Fig. 2. Figures 2(A)–2(C) depict horizontal OCT images of different positions within the thoracic aorta that show rough deformations at the walls of the vessel. The diameter of the aorta is about 3.5 mm and the penetration of the OCT is about 0.75 cm, which can be estimated by the obtained OCT images shown in Fig. 2. The white arrows indicate clearly noticeable irregularities of the inner vessel wall, which are on the order of several $100\mu \mathrm{m}$. Figure 2(D) shows a vertical profile of the aorta. The positons of the horizontal images are indicated between 15 and 20 mm. The region that was imaged by OCT is 3 cm long, before the probe enters the end of the catheter. The aorta can be generally characterized as normal with a few positions of early plaque formations.

Fig. 2

OCT images of a rabbit aorta with early plaque formations indicated by the white arrows in (A–C) (a–c zoomed in). (D) is a vertical image of the aorta. The positions of the horizontal images are indicated by the white lines.

After the acquisition of the Raman and OCT data, the animal was euthanized, the aorta perfused with a physiological buffer solution and subsequently fixed in formalin for routine histology. The thoracic fraction of the aorta was dissected for histopathological inspection. Figure 3(a) shows an enlarged area of the OCT Fig. 2(A) as indicated by the inset in comparison with an elastic van Gieson (EvG) stained section from the same region, shown in Fig. 2(B). Although precisely overlapping OCT and microscopic images are difficult to obtain, the similarity in morphology is obvious. Clearly noticeable is a swollen intima, which is about 100 to $200\mu \mathrm{m}$ thick. The magnification shows optically empty vacuoles, indicating former lipid depositions. Figure 3(c) shows a Raman spectrum measured in that region with the typical spectroscopic features associated with lipids. The CH stretching vibrations between 2800 and $3100{\mathrm{cm}}^{-1}$ are dominated by the symmetric methyl stretches at $2850{\mathrm{cm}}^{-1}$. The bands below $1800{\mathrm{cm}}^{-1}$ can be assigned to the $\mathrm{C}═\mathrm{C}$ stretching vibrations of unsaturated lipids at $1640{\mathrm{cm}}^{-1}$, ${\mathrm{CH}}_2$ scissoring vibrations at $1440{\mathrm{cm}}^{-1}$, and CH deformations at saturated and unsaturated positions between 1150 and $1300{\mathrm{cm}}^{-1}$. The lower wavenumber region is dominated by background scattering generated by the quartz fibers of the probe. The obtained $S/N$ ratio reveals most of the Raman bands typical for lipids. The overall spectral appearances indicate plaque depositions mainly consisting of triglycerides.

Fig. 3

Enlarged OCT image (a) from the section shown in Fig. 2(A) in comparison with an EvG stain (b) of a section from the same region; L, lumen of artery; M, muscularis media; and A, adventitia. Graph (c) shows a Raman spectrum with characteristic lipid bands collected within the region where the OCT image was obtained. The position of the Raman probe is shown in (d).

4.

Discussion

It has been demonstrated that the acquisition of OCT images and Raman spectroscopic data is generally possible. The most important outcome of the application in vivo was that both methodologies can be applied consecutively under the same conditions. The current technology level is sufficiently suitable to detect lipids and potentially to characterize plaque composition by Raman spectroscopy on a New Zealand white rabbit model. As it is a common animal model to study atherosclerosis with similarities to human plaques,²⁰ the outcome of the experiments describe the possible conditions for in vivo investigations of plaque developments and treatments, such as balloon angioplasty or stent implantation. Second-generation balloons for angioplasty and stents, which are coated with drugs to prevent early restenosis, are currently under development.^21–23 The efficiency of such drug-coated balloons and drug eluting stents can be investigated by directly monitoring the concentration and long-term release spectroscopically, as the active compounds also exhibit distinct spectroscopic properties.^24,25

The applied conditions of several seconds of nonfocused laser illumination at 100 mW within the artery did not lead to any noticeable alterations of the walls of the blood vessels, as confirmed by subsequent histopathological staining. Raman spectroscopy proved to be sensitive enough to detect early plaque formations. Since the scattering cross section of lipids is generally high, lipid depositions can be easily contrasted against normal tissue. To address concentration thresholds for lipids within a tissue matrix is relatively difficult, as the lipids are not dissolved and will always form crystalline aggregates or droplets, varying in size. Schie et al. have investigated ratios of different lipids within lipid droplet organelles and were able to differentiate lipids of different saturation levels down to a few mass percentages.²⁶

The obtained data quality was sufficient to assign the major depositions as triglycerides, which was surprising as the animals were treated with a fat- and cholesterol-enriched diet.²⁷ Previous results showed depositions dominated by cholesterol and cholesterol esters, but also revealed accumulations of triglycerides.^18,28 Generally, the animals did not show the expected amounts of plaque formations. The presented results were obtained by inserting the Raman probe first, followed by the OCT imaging pullbacks. In principle, the application order does not influence the results. Here, practical aspects, such as the animals’ limited wellbeing, affected by the injected contrast agents for the OCT, were taken into consideration. For studies in human or other in vivo experiments, a reverse order may be more appropriate. Since both methodologies utilize laser light in the NIR region, a probe that combines OCT and Raman spectroscopy is technologically feasible. Because Raman spectroscopy is inherently slower than OCT, at the current state-of-the-art technical development, it cannot be utilized for intravascular imaging as a standalone modality. One of the main challenges will be a coregistration of the two signal types. A combined OCT-Raman fiber would resolve this issue as the two incident laser spots can be overlapping. A combination with IVUS is certainly less feasible as the associated signal generation and detection technologies are completely different. However, signal transduction for NIRS and IVUS also needs to be realized by two different types of fibers. The main reason to combine NIRS with IVUS is that both techniques are of similar speed, so the signal reception can be coregistered. Apart from intravascular vibrational spectroscopy, several other spectroscopic approaches have been suggested and are currently under investigation. Fluorescence-based techniques such as two-photon excitation fluorescence (TPEF) and FLIM are certainly fast enough to record images and have been utilized to characterize atherosclerotic plaques.^11,12,29,30 TPEF is often combined with other multiphoton processes, such as second harmonic generation and coherent anti-Stokes Raman scattering imaging, a much faster, but also more complicated variant of Raman spectroscopy.^28,31–33 However, for these multimodal techniques, reliable optical fibers that are applicable for imaging still need to be developed.