Optical Coherence Tomography for Optical Biopsy : Properties and Demonstration of Vascular Pathology

OCT measures cross-sectional tomographic images in tissue and is similar to ultrasound B-mode imaging except that it uses light rather than sound. OCT also differs from ultrasound in that the detection in OCT is based on interferometry.^{19 20 21 22 23} In ultrasound, the time the ultrasound pulse takes to travel to a surface and be reflected back can be measured by an electronic clock. However, this is not possible with optical techniques because of the high speeds associated with the propagation of light. This limitation is overcome with the use of a reference light path and interferometry.

A detailed presentation of the principles of operation of OCT and factors that govern its performance have been published elsewhere.^{14 15} In addition to modifications in optics and electronics, the source has been changed to image in nontransparent tissue, which is described below. The OCT system uses fiber optics and a compact diode light source similar to those used in compact disc players. A schematic of the system is shown in Fig 1. Precision distance measurements are performed by Michelson-type interferometry. Light from the source is split evenly by an optical fiber splitter, which functions as an interferometer. One of the fibers directs light to the tissue and the other to a moving reference mirror. The distal end of the optical fiber can be interfaced to a catheter. The position of the reference mirror is precisely controlled by the system electronics. The light signal reflected from the tissue is recombined with the signal reflected from the mirror. Interference between the light reflected from the tissue and the light reflected from the reference mirror occurs only when the two path lengths are matched to within the coherence length of the light source. This allows precise (micron scale) determination of the distance within the sample from which the light was reflected.

OCT therefore measures the intensity of backscattering (reflection) light from within the tissue plotted as a function of depth. A cross-sectional image is produced in a manner similar to radar by recording axial reflectance profiles while the transverse position of the optical beam on the tissue specimen is scanned. The image is displayed either in gray scale or false color in order to differentiate tissue microstructure.

A 1300-nm wavelength, superluminescent diode source with a 50-nm bandwidth (wavelength distribution) was used for all studies presented here except where indicated. System parameters including the optical beam confocal parameter (focusing properties), signal-to-noise ratio, incident power at the sample, and reference arm power were carefully controlled to maintain consistency. Intralipid (Kabi Pharmacia Inc) and barium sulfate, which have well-established optical properties, were used as controls to confirm that system performance was reproducible throughout the imaging runs.²⁴ The relationship between optical bandwidth (Δλ) and ranging resolution (ΔL) is well established by previous results and from electrodynamic theory and is given mathematically by the formula¹⁷

The 50-nm bandwidth (wavelength spectrum of incident light) for the low-coherence light source used in these measurements yielded an axial spatial resolution of less than 20 μm as experimentally confirmed by measuring the point spread function with a mirror.^{20 25 26 27} The confocal parameter (depth of focus) for the optical beam was 520 μm. The lateral resolution was 30 μm and determined by the optical spot size.

The measured signal-to-noise ratio was 109 dB, using a power of 160 μW at the sample. The signal-to-noise ratio was measured by measuring the maximum signal when the optical beam was reflected from a high reflecting mirror divided by the background noise level of the instrument.²⁷ This signal-to-noise ratio determines the dynamic range with which it is possible to image.

In general, the signal-to-noise ratio for this type of optical measurement can be predicted by using results from optical communications theory.^{25 27} The signal-to-noise ratio (SNR) determines the minimum detectable reflectivity in the tissue and can be mathematically described by

where η is the detector quantum efficiency, ℏω is the photon energy, NEB is the noise-equivalent bandwidth of the demodulation filter, and P_s is the power received by the detector from the sample arm. The sensitivity to weakly reflected light depends only on the detection filter bandwidth and the available optical power.

Cross-sectional images of backscattering intensity versus longitudinal depth and transverse position were displayed as gray scale or false color images. The image size was 250 (transverse) by 500 pixels (longitudinal) unless stated. The image acquisition times ranged from 3 to 45 seconds.

Aorta was obtained within 5 hours of the initiation of the autopsy. The specimens were placed in normal saline with 0.05% sodium azide and stored at 0°C. The tissue was cut into segments smaller than 5×5 cm with the luminal surface exposed. Over 50 plaques of different morphologies from 9 consecutive patients were sampled.

Imaging was performed at room temperature through air. Similar procedures were used with nonvascular tissue (myocardium, peritoneal adipose, epiglottis, and clavicle). The position of the beam on the sample was monitored with a monocular infrared scope or visible light-guiding beam. The peripheral areas of imaged arterial sections were marked with microinjections of dye. After imaging, the sample underwent routine histological processing, being fixed in 10% buffered formalin for 24 hours and subsequently decalcified in a standard Cal Ex solution (Fisher Inc) for 5 hours. The arteries were processed for routine paraffin embedding. Five-micron-thick sections of the arteries were cut at marked imaging sites. Staining was performed with hematoxylin and eosin to identify different components of the vascular wall. Stained histological sections were compared with the OCT images, allowing a better qualitative understanding of tissue properties that alter backscattering (reflection) intensity and produce contrast. The maintenance of spatial relationships among tissue microstructures was confirmed by comparison with histology. Histology from nonvascular tissue is included only to confirm tissue identification. During imaging, the tissue samples were wet with 5 mL of normal saline to prevent drying.

OCT achieves high image contrast and differentiation between lipid- and water-based tissue, a limitation of conventional imaging technologies. Fig 2A shows an in vitro OCT image of cardiac muscle abutting adipose tissue. Because the optical backscattered reflectance of fat cells is significantly lower than muscle, the image dramatically differentiates these two tissues. Fig 2B demonstrates the corresponding adipose histology. Fig 2C shows a quantitative plot of the backscattered reflectance as a function of depth for both tissue morphologies. The high reflection peaks observed in the measurement of adipose tissue may be attributed to supportive tissue structures, while lipid structures have very low reflectivity compared with muscle. Thus, in contrast to ultrasound imaging, lipid- and water-based tissues have distinct optical reflectance properties that enable OCT imaging to sharply delineate the microstructure within lipid-laden atherosclerotic arteries.^{28 29}

Fig 3 demonstrates the dependence of imaging penetration depth on the wavelength of light used. A human epiglottis was imaged in vitro with the use of low-coherence light sources at wavelengths of 850 and 1300 nm with equal incident intensities (100 μW). The image obtained at 850 nm only shows structures predominately near the tissue surface. In contrast, the image at 1300 nm permits identification of underlying cartilage because light at this wavelength suffers less scattering and absorption, thus penetrating deeper into the tissue. In general, both imaging depth as well as tissue contrast are determined by tissue absorption and scattering properties, which vary with wavelength.³⁰ Thus, use of different wavelengths provides an approach for optimizing imaging depth as well as contrast and differentiation between different tissue morphologies. The change to 1300-nm diodes is an important modification from the ophthalmolgic studies, which allows imaging in nontransparent tissue.

A major limitation of ultrasound for intravascular imaging is that it cannot penetrate heavily calcified tissue.³¹ Fig 4 shows an OCT image with the corresponding histology of an in vitro human clavicle, demonstrating imaging at depths of over 1.5 mm into the cortical bone. Unlike sound waves, infrared light is less strongly reflected from calcified tissues and thus OCT imaging is possible even within these structures. A different false color map was used that gives better detail of deeper structures. Similar results were achieved in calcified aortic lesions.

Rupture of lipid-filled atherosclerotic plaques in coronary arteries is now believed to be the most common mechanism initiating acute myocardial infarction.^{2 3 4} However, current imaging techniques only have a limited ability to identify these lesions. Furthermore, these technologies do not have sufficient resolution to precisely guide the microsurgical removal of plaque by interventional catheter-based procedures. For these reasons, other methods of imaging have been aggressively pursued recently to overcome these limitations, including MRI, angioscopy, and intravascular ultrasound (IVUS). MRI (and closely related magnetic resonance angiography) is a promising technology to screen for the presence of coronary artery disease.³² Its advantages are that it is noninvasive, it can give information on flow in addition to anatomic obstruction, and it has a limited ability to provide spectroscopic information on tissue composition. Its limitations are the low resolution (greater than 100 μm even in animal studies), the very poor visualization of the circumflex and only proximal images of the other major arteries, the expensive instrumentation, and the fact that mechanical interventions cannot be deployed simultaneously.³³ Furthermore, though it may eventually be able to screen for areas suspicious for unstable plaques, it is unlikely that its low resolution will be able to identify these small areas definitively. Angioscopy is the direct visualization of the surface of the blood vessel with a fiber optic bundle. It has the advantage over other imaging methods in diagnosing surface fissuring and the presence of thrombus.^{11 34 35} Its disadvantages are that it depends on a clear field of vision, it cannot accurately measure luminal dimensions for guiding interventions, and it cannot image through the vessel wall.³⁶ High-frequency ultrasound (20 to 30 MHz) is superior to angiography in its ability to diagnose dissections, assess the adequacy of coronary interventions, guide stent deployment, and determine the extent of arterial stenosis, most notably at ostial sites.^{37 38} However, the maximum attainable resolution of 100 μm, which decreases linearly with distance, limits its ability to delineate structure.³⁹ Furthermore, it is limited in its ability to assess lipid content (lower limit, 250 μm) and has a poor interobserver agreement as to whether lipid is present in vivo (in some instances not significantly different from chance).^{28 29} Also, clot is very poorly differentiated from plaque.⁴⁰

The intravascular use of OCT has the potential of overcoming these limitations. This study demonstrates the ability of OCT to delineate in vitro plaque morphology. The structural details such as the thickness of intimal caps, extent of lipid collections, and presence of fissures were assessed at a level of resolution (20 μm) not achievable by other imaging modalities. Further refinements in instrumentation, including the use of different light sources based on broader bandwidth diodes or femtosecond pulse lasers could increase this resolution to 4 μm or less.^{41 42} In addition to high resolution, the high dynamic range of OCT (109 dB) provides high contrast between tissue constituents, and the ability to penetrate heavily calcified tissue makes OCT well suited for intravascular diagnostics. The use of light rather than acoustic waves has the additional advantage that since absorption of light, unlike sound, is strongly dependent on molecular composition, OCT has the potential for performing in vivo biochemical analysis based on spectroscopic properties.

Imaging in this study was performed in air but can be performed in saline, which is transparent to both visible and infrared light without appreciable loss of image quality. Preliminary results suggest that it will also be possible to image through varying depths of blood. Although blood strongly absorbs visible light, it has very low absorption in infrared wavelengths (1300 nm). However, optical scattering effects can reduce the amount of optical signal that is reflected. This in turn will limit the penetration depth of the imaging. In addition, it should be noted that there are other ways to improve performance in this scenario. The effect that blood has on imaging is a clinically important question that requires further investigation. If the presence of blood produces excessive degradation of imaging, simultaneous injection of saline may be required for intravascular use.

Future investigations will focus on adaptating OCT for in vivo imaging, including the development of an intravascular imaging catheter, optimization of incident light source wavelength, and reduction of image aquisition times. Since OCT is based on fiber optic technology, it can readily be engineered and adapted to perform imaging through a disposable, fiber optic–based delivery catheter. The diameter of a standard optical fiber is 125 μm (smaller fibers also can be used); thus it is small enough to be integratable with a catheter. The catheter can be implemented in a manner similar to transluminal ultrasound. In this case there must be a reflecting and focusing element on the distal end of the catheter. These can be miniaturized with the use of micro-optical fabrication techniques. The ultimate catheter technology should resemble transluminal ultrasound catheters in size. The remainder of the system is compact and portable.

Since imaging depth and contrast are functions of optical wavelength, the choice of optimal imaging wavelengths should further improve imaging performance. Finally, although the current acquisition times of 3 to 45 seconds used in this study are inadequate for in vivo imaging, imaging speed may be reduced significantly. Instrument developments including increasing source power, scanning speed, and redesign of electronics are straightforward and should result in a one to two orders of magnitude reduction in image acquisition time.

This work was supported in part by the National Institutes of Health (grants NIH-9-RO1-EY11289-10 and NIH-9-RO1-GM35459-09) and the Office of Naval Research, Medical Free Electron Laser Program (grant N00014-94-1-0717). The authors wish to thank Arthur E. Weyman, MD, Geoffrey Rose, MD, Stephen A. Boppart, and Neal Weissman, MD, for their comments. The technical support of Cathryn Blackwell and Joseph Gamba is also appreciated. Thanks to Cindy Kopf for her aid in the preparation of the manuscript.