|Home | About | Journals | Submit | Contact Us | Français|
Comprehensive volumetric microscopy of epithelial, mucosal and endothelial tissues in living human patients would have a profound impact in medicine by enabling diagnostic imaging at the cellular level over large surface areas. Considering the vast area of these tissues with respect to the desired sampling interval, achieving this goal requires rapid sampling. Although noninvasive diagnostic technologies are preferred, many applications could be served by minimally invasive instruments capable of accessing remote locations within the body. We have developed a fiber-optic imaging technique termed optical frequency-domain imaging (OFDI) that satisfies these requirements by rapidly acquiring high-resolution, cross-sectional images through flexible, narrow-diameter catheters. Using a prototype system, we show comprehensive microscopy of esophageal mucosa and of coronary arteries in vivo. Our pilot study results suggest that this technology may be a useful clinical tool for comprehensive diagnostic imaging for epithelial disease and for evaluating coronary pathology and iatrogenic effects.
Many important diseases arise from and exist within superficial tissue layers. For example, epithelial metaplasia, dysplasia and early cancers may be found in luminal organ mucosa, and high-risk coronary atherosclerotic plaques occupy the intima. Finding these lesions can be difficult, however, as they are characterized by microscopic features not visible to the eye and may be focally and heterogeneously distributed over a large luminal surface area. A catheter or endoscope capable of comprehensively conducting microscopy in patients over large surface areas and throughout the entire mucosa or intima could provide new possibilities for early diagnosis and characterization of these prevalent diseases.
Optical biopsy has recently been championed for nonexcisional histopathologic diagnosis in situ, but three-dimensional microscopy over large mucosal or intimal volumes remains a challenge. Confocal microscopy provides cellular resolution in tissue up to a depth of several hundred microns in vivo1. The limited field of view, similar to that of high-magnification microscopy, and the challenge of engineering small-diameter endoscopes2, however, make comprehensive three-dimensional imaging difficult. Optical coherence tomography (OCT)3, with resolution ranging from 2 μm to 15 μm, is also capable of imaging microscopic structure in individuals. Its inherent vertical cross-sectional imaging capabilities and relative ease of incorporation into fiber-optic catheters and endoscopes4 make it a promising candidate for comprehensive volumetric microscopy. Human studies have shown that OCT is capable of identifying dysplasia in Barrett’s esophagus5 and colonic adenomas6 and can identify all of the histopathologic features of vulnerable coronary plaques7. Until now, however, OCT has been too slow for comprehensive microscopic imaging, and it has therefore been relegated to a point-sampling technique with a field of view comparable to that in conventional biopsy. Recently, it has been shown that the application of frequency-domain ranging techniques8,9, instead of the delay-scanning interferometry of OCT, substantially improves detection sensitivity10–12. This new approach, optical frequency-domain imaging (OFDI)13, leverages high sensitivity to provide imaging speed that is orders of magnitude faster than that in conventional OCT. Here, we present the first catheter-based images obtained with OFDI in vivo and show the potential of this technology for rapidly imaging large tissue volumes at microscopic resolution.
We performed comprehensive volumetric microscopy with OFDI using minimally invasive catheterization and endoscopy to deliver the fiber-optic probe to the target organ or organ system (Fig. 1). The OFDI probe delivers infrared laser light and collects photons that are back-scattered from refractive-index inhomogeneities of microscopic structures within the tissue. We determined the intensity and origin of the reflected light by detecting spectrally resolved interference as the source laser wavelength was tuned9,13. Through rapid rotation and longitudinal pull-back of the internal portion of the probe, the OFDI system records the full three-dimensional microstructure of the tissue. We then used image processing to render volumetric images and to provide a form of virtual histology, in which high-resolution cross-sectional views of arbitrary locations in the tissue can be obtained. The prototype OFDI systems (Supplementary Fig. 1 online) developed for these studies provided an approximately 30-fold increase in A-line acquisition rate (up to 64,000 axial image lines per s) and a 3-fold increase in ranging depth (7.3 mm in air) over conventional OCT systems, corresponding to a 90-fold increase in imaging speed. These technical advances, combined with the use of high-speed rotational catheters, enabled imaging in vivo over large tissue areas (>25 cm2) and depths (up to 2 mm), with a three-dimensional resolution of approximately 15 μm × 15 μm × 10 μm.
To show the potential of OFDI for comprehensive volumetric microscopy, we imaged the esophagus of two female Yorkshire swine in vivo using a cylindrical balloon OFDI catheter. The entire volume of data in a 4.5-cm-long segment of each swine esophagus was acquired in 5.8 min (Fig. 2 and Supplementary Video 1 online). As the complete 14-GB data set cannot be encapsulated in discrete figures, here we present data after compression and downsampling. Volumetric renderings depict the tubular anatomy and allow visualization of the vascular network supplying the esophagus (Fig. 2a). We measured the longitudinal length of the volumetric image directly using a pull-back encoder and confirmed this measurement by visualization of the balloon’s terminal ends, which were separated by 4.5 cm (data not shown). We obtained the radial dimension using interferometry and confirmed it by imaging the diameter of the balloon (1.8 cm). The volumetric data allowed us to render two-dimensional cross-sections in longitudinal (Fig. 2b) and circumferential (Fig. 2c) directions. We acquired a total of ten volumetric images in two swine with no complications or instrument failures. As the physical characteristic of tissues that gives rise to image contrast, optical scattering, is identical in OCT and OFDI, we applied interpretation criteria that have been established for OCT (refs. 14,15) to identify anatomical structures in the OFDI images (Fig. 2d). The architectural assignments were supported by corresponding histological sections that we acquired after the imaging procedures (Fig. 2e). The dimension and optical properties of these anatomical features are similar in swine and humans14.
Intravascular optical imaging is particularly challenging as blood cells cause scatter light and degrade coherence. One solution to this challenge is to temporarily displace blood by injecting saline at the coronary ostium. Previous studies with OCT have shown that this approach can provide several seconds of clear viewing without ischemia or other complications16,17. With sufficiently fast imaging, therefore, long segments of coronaries could be comprehensively imaged during saline injection. Using a prototype OFDI system operating at an A-line acquisition rate of 54 kHz (108 images per s), we acquired volumetric data sets in each of the three main coronary arteries in five living swine. We automated saline delivery using an infusion pump and injection rates of 2–4 ml/s. The internal core of the catheters rotated at a rate of 108 revolutions per s, and a computer-controlled translation stage provided longitudinal (pull-back) actuation. We obtained a total of 56 volumetric data sets: 22 in the right coronary artery, 10 in the left anterior descending artery and 24 in the circumflex artery. In four scans, flushing did not provide clear viewing of the artery wall. In an additional eight volumetric scans, we found that the catheter used contained a fractured optical fiber, which led to excessive optical attenuation. The remaining 44 scans yielded clear visualization of the entire arterial wall over coronary segment lengths ranging from 24.3 mm to 63.3 mm. A three-dimensional rendering of a right coronary artery showed side-branch ostia consistent with the gross appearance of a typical prosected artery (Fig. 3a and Supplementary Video 2 online). The volumetric data could also be viewed in arbitrary circular (Fig. 3b) or longitudinal (Fig. 3c) sections for clear examination of the structure of the vessel wall, including the single-cell-layer intima, media and adventitia as indicated by the image interpretation criteria7 established with OCT. Although the optical properties of these layers are similar in humans and swine16, our previous studies with OCT found substantial pathologic thickening and a richer, more heterogeneous structure in human subjects17 (Supplementary Video 3 online).
In addition to imaging normal coronary arteries, we conducted a feasibility study to assess the potential of OFDI for imaging disrupted arterial structure. We performed angioplasty and stenting in the circumflex artery of one swine. During the angioplasty procedure, we over-inflated a balloon catheter intentionally in order to induce partial dissection of the arterial wall. We then positioned a coronary stent to partially overlap the dissected region. A volumetric OFDI image, acquired after stenting (Fig. 4), delineates the metallic wire mesh of the stent, the intima and media, and the remainder of the vessel and surrounding tissues including denuded adventitia. The luminal diameter is considerably expanded relative to the reference vessel diameters. The wire struts of the stent are opaque to infrared light and appear as bright local reflections at the lumen with occluded sectors radially behind them18. Cross-sectional images revealed portions of the dissected intima and media inside the stent (Fig. 4c,d), and the exposed adventia and flaps of the intima and media (Fig. 4d,e). After an interval of approximately 15 min, we acquired an additional volumetric data set and observed thrombi at locations of disrupted intima and media (Fig. 5 and Supplementary Video 4 online).
Comprehensive volumetric microscopy could bridge the gap between noninvasive radiologic techniques, which provide wide-field imaging but lack sufficient resolution for cellular diagnosis, and point-sampling approaches such as excisional biopsy. Previous research has shown the diagnostic potential of OCT, but, because of the relatively slow imaging speed, wide-field screening has not been feasible. We have shown that by acquiring the interferometric signal of OCT in the frequency domain using a wavelength-swept laser source, detection sensitivity can be markedly improved and the imaging of large luminal areas can be readily performed using minimally invasive, narrow-diameter catheter probes.
The diagnosis of epithelial cancers and their precursors exemplifies the need for comprehensive volumetric microscopy. As these lesions are frequently focal and can be distributed heterogeneously across a wide field, sensitive diagnosis is extremely demanding; diagnosis must be rendered on a microscopic scale over areas larger than 10 cm2. Barrett’s esophagus is a clinically important condition because it is a precursor of esophageal adenocarcinoma, a cancer associated with high morbidity and a dismal prognosis. Previous preliminary clinical studies have shown that conventional OCT imaging can identify high-grade dysplasia and adenocarcinoma when used in a point-sampling mode similar to random biopsy19. As image resolution and contrast in OCT and OFDI are identical, OCT diagnostic criteria5,19 for focal diagnosis of Barrett’s esophagus, dysplasia and intramucosal adenocarcinoma should be directly applicable to OFDI. With the current OFDI prototype, the time required for comprehensively imaging the distal esophagus was less than 6 min. Although this duration may be acceptable for clinical use, shorter procedure times would be attractive. The use of a higher speed computer bus and parallel hard drives may support the use of the OFDI system at an A-line acquisition rate of 68 kHz, which would reduce procedure times to less than 1 min.
A second category of applications for comprehensive volumetric microscopy is the diagnosis and characterization of superficial disease in cases where excisional biopsy is hazardous. The hypothesis of a vulnerable plaque is one of the most important current topics in interventional cardiology20. Autopsy studies have found that most acute myocardial infarctions are secondary to thrombotic occlusion at the site of a disrupted lipid-rich plaque, characterized by large necrotic, lipid-rich cores, thin fibrous caps and abundant macrophages21. It is hypothesized that this type of coronary plaque places patients at a greater risk for heart attack22. To test this hypothesis, develop appropriate therapeutic measures and potentially guide intervention, there is a clear need for diagnostic methods capable of identifying these lesions in living individuals before plaque rupture. The combination of previously validated OCT criteria for the focal characterization of coronary atherosclerotic plaques7,23,24 with comprehensive intracoronary OFDI may meet this need.
The coherent detection used in OCT and OFDI cannot directly detect fluorescence. Several molecular contrast methods, however, have already been suggested for conventional OCT, such as coherent antistokes Raman shift25 and protein-based probes26. Extension of these techniques to OFDI will enhance its utility in biomedical research, as well as in clinical medicine. Furthermore, functional techniques, including Doppler27 and polarization-sensitive28 methods, provide further contrast mechanisms that may be used for the study of disease and for primary diagnosis.
OCT and OFDI use interferometry to measure the same physical entity, the electric field amplitude, AS(z), of light reflected from a depth z within the sample S. Images are then produced by mapping AS(x,y,z), obtained through multiple axial scans, to a grayscale lookup table. The interferometric signal in OFDI, S(k,z0), at optical wavenumber k (= 2π/λ, where λ = wavelength) and reference delay length z0, can be expressed as S(k, z0) = ∫ ARAS(z)cos[k(z − z0)]dz, where AR denotes the amplitude of reference light. This signal is collected while holding z0 constant and while varying k using a monochromatic laser whose wavelength varies over time. AS(z) is computed through Fourier transformation: . Thus, AS is detected at all depth points simultaneously. In OCT, a broadband light source is used and the detector integrates over k-space; the directly measured signal is ∫ S(k, z0)dk = ARAS(z0). The full depth dependence of AS is then captured sequentially by scanning z0. Although the displayed image is a graphical presentation of AS(z) in both OCT and OFDI, the Fourier integration in OFDI processing enhances the detection sensitivity several hundred times. Additionally, as depth scanning is not required in OFDI, this improvement in sensitivity can be leveraged to increase imaging speed.
At the core of the OFDI systems (Supplementary Fig. 1) is a unique laser source that uses a semiconductor optical amplifier (Covega) and custom-made filter based on a polygonal scanner (Lincoln Laser)29. The laser developed for this study provided a sweep range of 111 nm (1,264–1,376 nm), repetition rate of 64 kHz and average power of 30 mW. The fiber-optic core of the systems is also unique in that it provides dual-balanced and polarization-diverse detection along with an acousto-optic frequency shifter (Brimrose) that removes depth degeneracy and doubles the effective ranging depth. The large ranging depth of 7.3 mm of our system accommodates varying distances between the OFDI probe and the tissue surface. The electrical signals from the detectors are digitized with a two-channel analog-to-digital converter (Gage Applied Technologies) at a sampling rate of 100 MHz.
We developed a custom catheter comprising a rotating optical core and balloon sheath for centration within the esophageal lumen. We performed esophageal imaging in two anaesthetized female York-shire swine weighing approximately 35 kg, according to a protocol approved by the Massachusetts General Hospital Subcommittee on Research Animal Care. No effort was made to flush mucus from the esophageal wall before balloon inflation. We operated the imaging system at a reduced A-line acquisition rate of 10 kHz, to facilitate streaming the acquired data continuously to a storage medium (using a standard data bus in a personal computer). The system acquired a total of 3,500,000 axial profiles over 5.8 min while rotating and translating the catheter to provide a longitudinal spacing of 32 μm between adjacent circular cross-sections.
We fabricated OFDI imaging catheters (0.8 mm diameter) based on a design similar to that of commercial intracoronary ultrasound imaging catheters. We performed coronary imaging in five female Yorkshire swine (40–50 kg) according to a protocol approved by the Massachusetts General Hospital Subcommittee on Research Animal Care. After anesthesia and heparin administration (100 U per kg body weight intravenously), we inserted a 7F guide catheter into the right femoral artery and advanced it over a guide wire to the distal coronary artery while monitoring the catheter’s position by angiography. To displace blood from the field of view, we delivered saline through the guide catheter using an injection pump (MedRad) at a constant rate of 2–4 ml/s. The imaging system continuously acquired data at an A-line rate of 54 kHz and temporarily streamed these data to a 2-GB digital memory on the acquisition board itself. The rotary junction rotated the inner core of the catheter at 108 revolutions/s and pulled the core longitudinally within the sheath with a velocity 5.4–16.2 mm/s. Each volumetric image comprised a total of 680 cross-sectional frames (500 radial A-lines per frame) and was acquired over 6.3 s. The longitudinal spacing of the volumetric imaging ranged from 50 μm to 150 μm.
Before the balloon was inflated, the lumen diameter was 3.3 mm. We inflated the balloon to a diameter of 4.0 mm in order to intentionally induce partial dissection of the arterial wall. We then positioned a 13-mm-long coronary stent and expanded it to a diameter of 3.9 mm using 20 atm of pressure. We acquired OFDI images immediately after stent deployment and following an interval of approximately 15 min, starting from the distal circumflex and extending to the normal segment of artery proximal to the stent. After the entire procedure, we prosected the circumflex artery and obtained histology sections 5 mm proximal to the stent and at three distal locations with intervals of 1.0 mm starting at the distal border of the stent.
Supplementary Figure 1. Optical frequency domain imaging system schematic
The interferometer in an OFDI system is assembled using commercially available fiber optic components to ensure environmental stability. Optical circulators are used for efficiently coupling light from the source to the sample and back to the receivers. Dual-balanced and polarization diverse detection avoids birefringence artifacts and provides shot-noise limited signal collection. Inset depicts an OFDI system console.
This research was supported in part by the US National Institutes of Health (contracts R01 HL070039, R33 CA110130, R01 RR0119768, R01 HL076398 and R01 CA103769) and by the Terumo Corporation.
Note: Supplementary information is available on the Nature Medicine website.
COMPETING INTERESTS STATEMENT
The authors declare that they have competing financial interests (see the Nature Medicine website for details).
Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/