Author + information
- Received November 28, 2005
- Accepted December 1, 2005
- Published online April 18, 2006.
- Jay D. Caplan, SB, MBA⁎,a,⁎ (, )
- Sergio Waxman, MD†,‡,c,
- Richard W. Nesto, MD†,§,c and
- James E. Muller, MD⁎,b
- ↵⁎Reprint requests and correspondence:
Jay D. Caplan, InfraReDx Inc., 34 Third Avenue, Burlington, Massachusetts 01803
This review describes efforts to use near-infrared (NIR) spectroscopy to identify chemical components of coronary artery plaques as a means to assess vulnerability. Near-infrared spectroscopy has been well-validated by the physical sciences as a method to characterize chemical composition of various bio-materials and could be ideal to detect vulnerable coronary plaques in patients. Recent studies in aortic and coronary artery autopsy specimens have confirmed the ability of the technique to identify lipid-rich thin-cap fibroatheromas through blood. A catheter-based system has been developed to address the challenges—of access to the coronary artery, blood, motion, and the need to scan—that must be overcome for use in patients. Initial clinical experience in six patients with stable angina demonstrates that high-quality NIR spectra can be safely obtained. Additional studies are planned to validate the ability of the technique to identify lipid-rich coronary artery plaques and ultimately link chemical characterization with subsequent occurrence of an acute coronary syndrome.
Near-infrared (NIR) spectroscopy—a technique routinely used in the physical sciences to determine the chemical composition of substances—is under intensive study as a potential tool to identify vulnerable atherosclerotic plaque. The impetus for this work is the high likelihood that plaques with a distinct chemical composition (i.e., inflamed thin-cap fibroatheromas [TCFAs]) represent vulnerable plaques that are likely to cause acute coronary syndromes (1). The goal of intracoronary NIR spectroscopy is to provide a “chemogram” of the wall of the artery that will serve as an index of vulnerability.
Because of the high level of development of NIR spectroscopy for other applications, there was little doubt that it could be used in an ex vivo setting to determine the chemical composition of tissue. Numerous studies from multiple groups have now confirmed this expectation, as described in this review.
Despite these encouraging ex vivo data, major questions remained as to whether high-quality spectra can be safely obtained from the coronary arteries of living patients in whom problems of access, penetration of blood, cardiac motion, and need for a scanning capability must be overcome. New developments, however, in laser technology, fiber optics, and chemometric analysis of spectra make intra-coronary NIR spectroscopy an approach that might be of considerable utility for the detection of vulnerable plaque.
Principles of NIR spectroscopy
Near-infrared spectroscopy takes advantage of the fact that different substances absorband scatterNIR light (wavelengths from 800 to 2,500 nm) to different degrees at various wavelengths. An NIR spectrometer emits light into a sample and measures the proportion of light that is returned over a wide range of optical wavelengths. The return signal is then plotted as a graph of absorbance (y-axis) at different wavelengths (x-axis) called a spectrum. Figure 1shows the spectrum of cholesterol, a molecule of great interest for vulnerability detection, and other pure substances.
Absorbance is caused by the loss of NIR light when it interacts with certain molecular bonds (CH, OH, NH, and others) whose vibrational frequency is identical to the frequency of the incoming light. The absorbance spectrum of a biologic tissue, which is composed of thousands of different chemical entities, is determined by the net result of the absorbances of its individual chemical components.
Scattering of NIR light, which differs from absorbance, results from deflections of the light by cellular and extracellular structures. A common example of scattering is the redirection of visible light by water droplets when a car headlight is aimed into fog.
Scattering is both a help and a hindrance to tissue spectroscopy. Because scattering varies as a function of wavelength and is different for different tissue components, the degree of scattering might provide valuable information about the sample. Scattering also permits light to travel an indirect path from the source, through the tissue, and back to the detector. These properties aid the spectroscopic determination of chemical composition.
Scattering by tissue, however, also causes the loss of most of the NIR light emitted, thereby creating a need to increase the light delivered and/or the sensitivity of the detectors to obtain adequate signal-to-noise in the return signal.
To identify an unknown tissue type, NIR spectra of known tissue samples are first correlated with characteristic features of the tissue measured by a reference method such as biochemical analysis or histology. Multivariate mathematical techniques, such as principal component regression analysis, are then used to identify spectra associated with the tissue of interest. These algorithms are then applied to NIR spectra of unknown samples, to identify their chemical composition of the sample. These methods are part of a family of analytical techniques known as “chemometrics.”
As previously noted, NIR spectroscopy is in routine use in disciplines such as agriculture, chemistry, food processing, pharmaceuticals, and astronomy (2). Specific uses include the measurement of octane in gasoline, alcohol content in beverages, and composition of pharmaceuticals. An important medical application has been the use of differential NIR absorbance of hemoglobin and oxyhemoglobin to determine oxygen saturation in tissues (3–5).
Near-infrared spectroscopy can provide simultaneous, multicomponent, nondestructive chemical analysis of biologic tissue with an acquisition time of <1 s (6). Furthermore, no sample preparation is required, and physical, biologic, and molecular information can be derived. It has been documented that a large range of biologic constituents can be identified using NIR algorithms (6), and signals can be obtained from structures millimeters- to centimeters-deep relative to the tissue surface.
Characterization of atherosclerotic plaque
The use of NIR spectroscopy for characterization of atherosclerotic plaque was initiated in 1993 by Cassis and Lodder (7). It was first shown that NIR spectra could accurately characterize low-density lipoprotein cholesterol accumulation in hypercholesterolemic rabbit aortas (7). Near-infrared spectroscopy was then used in humans to image lipid content in carotid plaques exposed at the time of surgery (6). After surgery, ex vivo NIR scanning was performed on the excised plaque, and spectra were shown to predict lipoprotein composition as determined by gel electrophoresis.
The early reports that NIR spectroscopy could be useful for characterizing plaque have been supported by similar findings from other groups. In 1999, Jaross et al. (8) compared cholesterol content determined by NIR spectroscopy versus that determined by reversed-phase, high-pressure liquid chromatography in human aorta specimens. A high correlation coefficient (0.96) was observed. The same group later reported that the cholesterol-to-collagen ratio could also be accurately quantified by NIR (9).
In a study of recently excised human carotid plaque samples, Wang et al. (10) found a high correlation between direct ex vivo measurements of lipid/protein ratios and results obtained with a NIR spectrometer fitted with a fiber-optic probe.
Identification of inflamed TCFAs in autopsy specimens
The use of NIR spectroscopy for detection of an inflamed TCFA, a particular form of atherosclerotic plaque suspected to be likely to rupture and cause an acute coronary syndrome, was first reported by Moreno et al. (11) in 2002. Spectra were collected from 199 human aortic autopsy samples with a commercial NIR spectrophotometer (Model 6500, FOSS NIRSystems, Laurel, Maryland). The samples were fixed in formalin and studied without motion and in the absence of blood. The spectra were then compared with histologic findings suspected to be associated with vulnerable plaques. An algorithm was constructed with 50% of the samples as a calibration set. The algorithm was then tested on a validation set to determine its ability to identify high-risk features as determined by histology. Sensitivity and specificity were 90% and 93% for lipid pool, 77% and 93% for thin cap, and 84% and 91% for presence of inflammatory cells, respectively.
These studies were then extended from aortic to coronary tissue, and similar results were obtained. Moreno et al. (12), with the use of the same commercial NIR spectrophotometer (FOSS NIRSystems), performed NIR spectroscopy on fixed human coronary samples collected from 167 sections of 45 arteries. Neither motion nor blood was present. Spectral findings were compared with histologic measurements of lipid area. An algorithm was again constructed with 50% of the samples as a calibration set. The algorithm was then tested on a validation set for its ability to detect lipid areas > or <0.6 mm2. Sensitivity and specificity for the detection of lipid-rich coronary plaques were 83% and 94%, respectively.
In summary, seven studies from three different groups have documented conclusively that NIR spectroscopy, in the absence of difficulties posed by motion and blood, can accurately identify important features of atherosclerotic plaques suspected to represent plaque vulnerability (6–12).
Approaches to overcome the obstacles to performance of clinical NIR spectroscopy
The coronary artery of a beating heart is a difficult environment for the performance of spectroscopy. Four obstacles must be addressed before results similar to those of ex vivo studies can be achieved. First, NIR light must be delivered to the artery and collected with minimal risk to the patient. Second, the undesirable effect of blood on the signal must be overcome. Third, the effect of coronary motion on spectra must be managed. Fourth, a scan of all major coronary arteries is needed for clinical use.
Spectroscopic access to the human coronary artery
A multidisciplinary team (InfraReDx Inc., Burlington, Massachusetts) has constructed a small (3.2-F) coronary artery catheter containing fibers to deliver and collect infrared light within the coronary artery (Fig. 2).The lateral spatial resolution of the device is approximately 1 mm in radius (13). The catheter is similar to an intravascular ultrasound (IVUS) catheter, which has an excellent safety record. As described in the following section, it has been used successfully in patients, indicating that the problem of access has been solved.
Reduction of the effect of blood on acquisition of the spectral signal
Tests were conducted in autopsy specimens to determine if NIR spectroscopy could accurately identify lipid-rich plaques through blood. With the use of a Foss spectrometer, spectra were collected from 63 specimens from 26 unfixed human aorta samples. The system identified large lipid-rich plaques (sensitivity, 88%, and specificity, 79%) through up to 3 mm of blood (14) and could classify them on the basis of cap thickness (15).
Minimizing the effect of coronary artery motion on acquisition of the spectral signal
An ultra-fast system was created to minimize the effect of cardiac motion on NIR signals. This system uses a specialized laser that can obtain spectra within 6 ms by scanning only a limited number of wavelengths. Although the narrower waveband provides fewer spectral features for analysis, it is capable of adequate performance, as judged by studies performed on ex vivo human coronaries in the presence of blood (14). Approximately 3,000 arterial locations from 30 hearts were analyzed by NIR spectroscopy and histology. The area under the curve of the receiver operating characteristics curve for identification of TCFA and disrupted plaques was 0.74.
To evaluate performance of the system during cardiac motion, a human coronary autopsy specimen was attached to the surface of a beating pig’s heart and connected to the porcine circulation (16). The ability of the NIR catheter (placed in the xenograft lumen) to identify the presence or absence of a Teflon target attached to the surface of the graft was determined (17) (Fig. 3).The NIR system correctly identified the presence or absence of the target in all cases, despite cardiac motion.
Development of the ability to scan the coronary artery
A system was developed to rotate and pull the optical probe back so that an entire artery could be scanned. This system was tested for its ability to identify TCFA-sized targets in a blood-filled phantom model of the coronary artery. The system accurately identified all eight simulated TCFAs (Fig. 4)in a reproducible manner (InfraReDx Inc., unpublished data on file, October 2005). The phantom study does not prove that lipid-rich plaques can be identified in patients—for that goal, clinical studies are required.
Clinical experience with the NIR spectroscopy system
In 2001, Dr. Richard Nesto tested a prototype device in patients undergoing coronary stenting at the Lahey Clinic in Burlington, Massachusetts (18). No device-related adverse events occurred, but substantial motion artifact was present. In August 2005, the new ultrafast NIR system was tested in six patients at the Lahey Clinic by Dr. Sergio Waxman and Dr. Nesto. The trial confirmed safety (clinical outcomes did not differ from those expected with stenting and IVUS usage) and demonstrated that signals obtained in the artery differ from those obtained in blood alone (InfraReDx Inc., unpublished data, August 2005) (Fig. 5).
Additional studies are needed to test the ability of NIR spectroscopy to identify lipid-rich plaques in patients (as it has in autopsy specimens). If such studies are positive, in December 2006, InfraReDx Inc. will seek to make the system available to clinicians for the detection of lipid-rich plaques. The cost of the console will be less than an IVUS console, and the costs of the catheters will be within the range of special-use catheters already purchased by catheterization laboratories. Additional optimization strategies include the construction of a non-rotating catheter, the combination of NIR with other modalities such as ultrasound in a single catheter, and the use of molecular imaging agents.
Future studies will be required to determine if, as expected, lipid-rich plaques have a higher likelihood of causing cardiac events. If the NIR system is capable of detecting lipid-rich plaques in vivo that are vulnerable, it might be useful for determining the value of placing a drug-eluting stent at a lesion producing an intermediate (50% to 60%) degree of stenosis and for the identification of patients who could be randomized to a pharmaceutical agent designed to stabilize vulnerable plaques or conventional therapy. Success in either treatment study would establish the value of identification and treatment of vulnerable plaques.
- Abbreviations and Acronyms
- intravascular ultrasound
- thin-cap fibroatheroma
- Received November 28, 2005.
- Accepted December 1, 2005.
- American College of Cardiology Foundation
- Dempsey R.J.,
- Davis D.G.,
- Buice R.G.,
- et al.
- Wang J.,
- Geng Y.J.,
- Guo B.,
- et al.
- Moreno P.,
- Lodder R.,
- Purushothaman K.,
- et al.
- Moreno P.R.,
- Ryan S.E.,
- et al.
- Marshik B.,
- Tan H.,
- et al.
- Marshik B.,
- Tan H.,
- et al.
- Waxman S.,
- Khabbaz K.R.,
- et al.
- Waxman S.,
- Tang J.,
- et al.