Optical coherence tomography (OCT) (1- 7) is a high-resolution cross-sectional imaging modality that offers the unique possibility of visualizing and quantifying clinically important coronary plaque microstructures, including macrophages (8- 10), thin fibrous caps (2,7,11- 12), collagen (13- 15), thrombus (7,16- 17), stent strut apposition (18- 21), and stent coverage (4- 6,22- 30). Although OCT has shown promise for intracoronary imaging, its clinical applicability has been severely limited by the attenuation of its optical beam by intraluminal blood, which prevents clear visualization of the vessel wall. The 2 most common approaches for displacing blood include: 1) proximal balloon occlusion of the coronary artery followed by saline perfusion; and 2) brief, nonocclusive saline purge through a guiding catheter. Proximal balloon occlusion is undesirable because of its potential to cause myocardial ischemia and vessel injury. Intermittent, nonocclusive saline purges are safe, but the slow image acquisition speed of OCT relative to the duration of clear viewing provided by the saline flush, limits OCT to “snap-shot” imaging at a few discrete locations.

A second-generation OCT technology, termed optical frequency domain imaging (OFDI) (31- 32), recently has been developed that solves this problem by imaging at much higher frame rates. Optical frequency domain imaging is the imaging form of optical frequency domain reflectometry, a telecommunications technique that is used to determine fiber fault locations for short- and long-haul optical networks (33). In combination with a short, nonocclusive saline flush and rapid spiral pullback, the higher frame rates generated by OFDI enable imaging of the 3-dimensional (3D) microstructure of long segments of coronary arteries in vivo in swine (31). We report the first clinical experience with intracoronary OFDI.

Patients undergoing percutaneous transluminal coronary intervention for coronary artery disease were enrolled at the Lahey Clinic (Burlington, Massachusetts) from May to September 2007. Partners and Lahey Clinic institutional review boards approved the study. All patients gave informed consent prior to their participation.

The OFDI system used in this study (31) provided images at 100 frames/s with a total of 512 radial scans per circular cross-sectional image. Axial resolution was 7 μm in tissue (refractive index n = 1.4), and the signal to noise ratio was 110 dB. All unprocessed data were continuously archived to hard disk drive at a rate of 320 MB/s. During the procedure, every fifth image was reconstructed and displayed in real time at a rate of 20 frames/s. Immediately after the imaging session, all images were reconstructed and stored as a digital movie, which required approximately 15 to 30 s, depending on the number of frames acquired.

The OFDI catheter comprised an optical imaging core within a transparent outer sheath (2.6-F) (3) that incorporated a rapid-exchange guide wire provision. The transverse resolution of the catheter was 30 μm, and the focal distance was 2.5 mm. For rapid spiral imaging, an optical rotary junction rotated the optical imaging core at 100 revolutions per second and pulled the imaging core back at speeds ranging from 5 to 20 mm/s (longitudinal resolutions 50 to 200 μm, respectively) in a manner analogous to that used in rotational intravascular ultrasound (IVUS). The OFDI system, rotary junction, and catheters were fabricated in the Wellman Center for Photomedicine at the Massachusetts General Hospital. The OFDI catheter was assembled under design control guidelines and approved under a U.S. Food and Drug Administration investigational device exemption.

All patients received intracoronary stents (Cypher, Cordis Corp., Miami Lakes, Florida) as per the standard of care. After intervention, the OFDI catheter was advanced over the 0.014-inch guidewire and through a 7-F guide catheter. The OFDI catheter was then positioned so that its imaging core was distal to the stent. Radiopaque markers on the OFDI catheter sheath and imaging core were used for confirming catheter position during angiography. After catheter placement, a cardiovascular injection system (Avanta, Medrad Warrendale, Pennsylvania) was used to provide a nonocclusive saline (lactated Ringer's solution) flush through the guide catheter at a rate of 2 to 4 ml/s for approximately 3 to 6 s. The maximum saline delivered for any given imaging session was 20 ml. When a blood-free image was observed, the OFDI system operator initiated rapid pullback of the OFDI imaging core by using standalone electronic control of the pullback motor. The pullback and catheter rotation was terminated with the same mechanism when significant intraluminal blood was observed in the OFDI image or when the catheter imaging core entered the guide catheter. After reconstruction of the digital movie, the catheter imaging core was repositioned, and data acquisition was repeated in each patient several (i.e., 1 to 3) times. Different saline purge parameters (2 to 4 ml/s) and pullback rates (5 to 20 mm/s) were attempted.

After imaging and reconstruction of the OFDI images, 1024 × 1024 × N (N = number of frames) pixel data sets were processed offline with the use of ImageJ (34). Image contrast was normalized for each frame of the OFDI movie. Cross-sectional and longitudinal sections were displayed with an inverse gray scale lookup table. Clear visualization lengths were determined by measuring the pullback length over which blood did not significantly alter the signal from the coronary wall. For some patients (OFDI-02 and -03), to remove rotational artifacts caused by cardiac motion, we performed frame-to-frame registration by image registration and rigid body transformation (35). For 3D visualization, the following components of each of the images were manually segmented with previously established OCT plaque and stent characterization criteria (8- 9,19,36): 1) artery wall; 2) lipid pool; 3) calcific nodule; and 4) stent. We have previously shown that macrophages can be identified in OCT images by their highly backscattering, punctate appearance, a finding that was validated by comparing OCT data to corresponding histopathology slides stained with CD68 (8). Macrophages were automatically segmented with normalized image intensity metrics on a per-image basis (8) with manual removal of outliers. Three-dimensional median filtering with a 3 × 3 × 3 kernel was conducted for each segmented volume.

After segmentation, the artery data were rendered in color using the following scheme: 1) red = artery wall; 2) yellow = lipid pool; 3) white = calcific nodule; 4) blue = stent; and 5) gray = guide wire. The volumes corresponding to each component were added to generate the final segmented and colored volume data set. Segmented data were imported into a 3D, volume-rendering program (OsiriX 2.75, The OsiriX Foundation, Geneva, Switzerland). Maximum intensity projection and volume rendering techniques were used for visualization and were implemented with several different opacity tables. Cut surfaces and fly-throughs of the rendered data sets also were constructed.

A total of 3 patients (OFDI-01 through OFDI-03) were imaged between May and September 2007. Baseline characteristics for the patients are presented in (Table 1). Imaging was performed successfully in all patients. None of the patients experienced chest pain or electrocardiogram changes during intracoronary OFDI. There were no complications related to the OFDI procedure. A range of saline flow rates (2 to 4 ml/s) and pullback speeds (0.5 to 2.0 cm/s) were used, providing imaging of arterial segments from 3 to 7 cm in length.

Patient OFDI-01 had a drug-eluting stent (DES) placed distal to a proximal left anterior descending artery (LAD) aneurysm (Figure 1A). The OFDI imaging of the proximal LAD originated distal to the stent and terminated within the stent (Figure 1A). Three-dimensional visualization ((Figure 1)B and Figure 1C) of the segmented data set, obtained at a saline purge rate of 3 ml/s and a pullback velocity of 2.0 cm/s, shows the proximal LAD stent (blue) as well as coronary atherosclerosis, including calcific nodules (white), lipid-rich plaque (yellow), and punctate macrophages (green). A longitudinal OFDI section shows a large lipid pool at the distal aspect of the coronary segment ((Figure 2)A and Figure 2B). An expanded portion of a cross-sectional image of this lesion (obtained from location denoted by arrow in Figure 2B) demonstrates that this lipid pool has OFDI features consistent with a thin-capped fibroatheroma (TCFA), including a thin fibrous cap (Figure 2C, black arrow), and macrophages at the cap–lipid pool junction (Figure 2C, green arrow). A flap of tissue can be seen over the cap (arrowhead in Figure 2C), which could represent disrupted intima or fibrin. In this image, the minimum cap thickness near the shoulder is approximately 60 μm (Figure 2B arrow; corrected for a refractive index n = 1.4).

For Patient OFDI-02, imaging was conducted from the distal vessel, across the posterior descending artery, to the proximal segment of the right coronary artery (RCA) (Figure 3A). Cutaway longitudinal views of volume renderings ((Figure 3)B to Figure 3D), acquired at a saline perfusion rate of 3.0 ml/s and a pullback velocity of 2.0 cm/s, show a bare-metal stent (BMS) (Figure 3C), placed 9 years previously, and a DES, placed immediately before OFDI imaging (Figure 3D). The entire DES can be visualized, crossing the ostia of the posterior descending artery. Only portions of the BMS can be seen in the cutaway view (Figure 3C) because of neointimal hyperplasia covering the stent and the transparency chosen for rendering. However, some BMS struts viewed from the luminal aspect appear to have little tissue coverage.

Fly-through views show the BMS and DES in greater detail ((Figure 4)A and Figure 4B). Although a portion of the BMS struts (Figure 4A) have a similar appearance to the DES struts (Figure 4B), further inspection of cross-sectional OFDI images reveals microscopic distinctions; as opposed to the DES struts (Figure 4D), the BMS struts (Figure 4C) appear thicker, have a surface reflectivity that is comparable with that of arterial tissue, and have edges that continuously blend into the luminal surface. These observations are consistent with the presence of a covering over the BMS struts that is thinner than the resolution of the OFDI system.

The angiogram shows minimal luminal stenosis proximal to the DES (yellow arrowhead in Figure 3A). Volume renderings depict a large lipid pool at this location (see dotted horizontal line in Figure 3D), which is partially covered by the DES and present at the junction between the DES and artery wall. A fly-through view at this location (Figure 5A) shows that the lesion is circumferential. A cross-sectional OFDI image (Figure 5B) at the corresponding site demonstrates that this lesion is a TCFA, with cap thicknesses <65 μm at multiple locations (black arrowheads in Figure 5B). As in Patient OFDI-01, this TCFA has a cap that contains many macrophages (green arrowheads in Figure 5B), some of which appear to extend to the luminal surface. Cholesterol crystals can also be identified within the lipid pool (red arrows in Figure 5B).

Imaging of Patient OFDI-03 began distal to a mid-RCA DES and terminated within the stent. (Figures 6, 7) depict images obtained with the slowest pullback rate of 5 mm/s and a saline purge rate of 3.0 ml/s. The clear visualization length was 3.0 cm. The stent was deployed over a large calcific nodule, resulting in distortion of the stent struts ((Figure 6)B and Figure 6C). Because of the finer longitudinal resolution (50 μm vs. 200 μm), the stent strut microstructure is better appreciated in 3 dimensions and contains fewer discontinuities (Figure 6B) in comparison with the other 2 cases (Figures 1, 2, 3, 4, 5). Improved visualization of the stent struts is also clearly evident in the longitudinal cutaway view (Figure 7A). A longitudinal reconstruction from this session is shown in (Figure 7)B, demonstrating a calcific nodule beneath the DES. The longitudinal image's resolution is superior to that obtained with 200-μm longitudinal resolution ((Figure 2)A and Figure 2B) and is comparable with that of OFDI image cross sections.

Our results demonstrate that all of the coronary microanatomy and pathology previously visualized by OCT, including lipid pools, calcium, macrophages, thin fibrous caps, cholesterol crystals, and thrombus, can also be detected by OFDI. Importantly, the ability of OFDI to comprehensively image long coronary segments removes the diagnostic uncertainty that hindered the interpretation of data obtained from first-generation OCT systems. In addition to native coronary microscopic pathology, stents and stent neointimal hyperplasia were clearly identified in OFDI images from patients in this study. Whether or not OCT or OFDI can be used to visualize single-cell re-endothelialization remains an important subject of future research that will impact its utility for evaluating risk of late stent thrombosis (37).

This was a proof-of-principle study with a limited number of patients, and as such, should be considered preliminary. Nevertheless, the results presented here are promising. The saline perfusion and rapid pullback-imaging scheme worked for all patients, even in cases when it should have been difficult. For example, in Patient OFDI-01, excellent images were obtained even in the presence of a proximal LAD aneurysm and incidental perfusion of the left circumflex coronary artery (LCX) through the left main. To date, we have not imaged the LCX. In other OCT imaging studies, we did not find differences between clear imaging durations for the 3 main coronary arteries (7,12). We therefore expect that imaging the LCX will be similar to that of the LAD and RCA in terms of clear visualization time. Like the RCA, the LCX sits in the atrioventricular groove, so we additionally anticipate that intracoronary motion will be comparable to that of the RCA.

We have not attempted to optimize the size and positioning of the guide catheter, flow rates, or synchronization of the flow rates with actual coronary flow. Additional gains in imaging time and/or decreases in the required saline flow rates may be identified upon further optimization of these parameters. Furthermore, a recent report has demonstrated that nonocclusive purging with isosmolar contrast media can allow high-quality OCT imaging for longer pullback durations (∼10 s) (38). The results of this study suggest that the OFDI pullback imaging time may be safely increased and that viscous contrast media may be more efficacious than crystalloids.

Although OFDI is an important advance over OCT, it still has some of the same limitations as the first-generation technology. The most important such limitation is the penetration depth of light, which prohibits visualization of the external elastic lamina in significant plaques. Like OCT, this limitation prevents the use of OFDI for evaluating certain coronary features, such as remodeling. Because this is a fundamental limitation of OFDI and all OCT-related technologies, the best solution to this problem may be a hybrid device that combines IVUS and OCT into one instrument. In the interim, before this multimodality device is developed, it will be necessary to use both IVUS and OFDI if visualization of both superficial and deep coronary pathology is required.

There is substantial interest in obtaining longitudinal microscopic information because of recent findings that coronary pathology is highly dependent on shear forces in the direction of blood flow (39). At the maximum pullback velocity of 20 mm/s, our imaging speed of 100 frames/s resulted in an undersampled longitudinal resolution (200 μm). In comparison, for the slowest pullback velocity of 5 mm/s, longitudinal reconstructions were adequately sampled and image quality was comparable with that of the cross-sectional images (Figure 7). Because both longitudinal information and long imaging lengths are highly desirable, future development of the OFDI method should focus on increasing longitudinal image sampling without sacrificing length. In addition to improving clearance of blood by optimizing saline delivery, these imaging characteristics may be achieved by increasing the frame rate of the OFDI system (40- 41).

Although 3D visualization provides a powerful manner to represent the OFDI data, it is currently very time consuming. The volume representations presented here took several hours to create, as many of the segmentation steps were semi-automatic or manual. If 3D visualization becomes important for the use of OFDI in the cardiac catheterization laboratory, additional research must be conducted to develop automatic volume data processing methods that can be completed in real-time during the interventional procedure. The 3D reconstructions also are prone to artifacts caused by linear, rotational, and deformational motion of the artery during the pullback. Electrocardiogram gating (42) or post-acquisition image processing techniques (43) that have been used for IVUS reconstructions can potentially correct these artifacts. Finally, we note that the reproducibility of 3D OFDI volume reconstructions has not yet been established. Studies to evaluate the reproducibility of intracoronary OFDI are currently under way.

We have presented the first human experience with intracoronary OFDI, a second-generation OCT technology. We found that, compared with conventional OCT, the greater speeds of OFDI used in this study enabled comprehensive volumetric microscopy of coronary arteries in living patients after a short and nonocclusive saline flush. The ease of use of this new method along with the wealth of information contained in the data bring it closer to becoming a useful imaging technique in cardiac catheterization laboratories.

The authors thank Judith Pendleton and Patricia Baum for their invaluable assistance in this study.