Seeing is believing: The development of optical coherence tomography

The Historical Roots of OCT

The origins of OCT can be traced to the birth of modern optics and, more specifically, to Thomas Young’s famous two-slit experiment (1). Young observed that when an opaque barrier with two thin slits cut through it is placed between a light source and a white screen, a pattern of alternating white and black stripes appears on the screen (Fig. 1A). Young interpreted this pattern as arising from constructive and destructive interference between a pair of outwardly propagating waves originating from the two slits. This experiment provided the first clear evidence for the wave nature of light, and it foreshadowed by 60 years James Clerk Maxwell’s insight that light was a propagating electromagnetic wave.

In the 1880s, Albert Michelson used the phenomenon of wave interference to build optical instruments of unprecedented precision (2). As seen in Fig. 1B, in a Michelson interferometer, two light paths are arranged at right angles. Light from the source, located at the end of arm 1 in Fig. 1B, is directed to a half-silvered mirror that is set at a 45-degree angle in the center of the apparatus. Half of the light beam is reflected from the mirror and is thereby redirected at a 90-degree angle from the original light path to a conventional mirror placed at the end of arm 2. The other half of the light beam passes through the half-silvered mirror in a continuation of the light beam’s original direction to a conventional mirror placed at the end of arm 3. After reflecting from their respective mirrors, these two light beams return along each of their arms and are recombined in the center of the apparatus at a second half-silvered mirror set at a 45-degree angle, and the recombined light beam then travels to a detector at the end of arm 4. The resulting interference pattern is determined by the difference in the length and/or the refractive index of the two light paths.

It is fair to say that no physical principle has provided more fertile ground for the invention of high-precision methods and instruments than interferometry. Among its progeny are X-ray crystallography, phase contrast and differential interference contrast microscopy, modern radio astronomy, holography, medical ultrasound, and fiber-optic gyroscopes. The most remarkable example of interferometry is the Laser Interferometer Gravitational-Wave Observatory (LIGO) (3). LIGO can detect a change in the distance between two mirrors that are 4 kilometers apart with a precision less than 1% the diameter of a proton (10⁻² × 10⁻¹⁵ m = 10⁻¹⁷ m). This is equivalent to detecting a change in the distance between the Earth and the Sun (150 million kilometers) with a precision less than the width of a water molecule (~3 Å).

The Invention of OCT

In 1985, Dr. James Fujimoto, currently the Elihu Thomson Professor of Electrical Engineering at the Massachusetts Institute of Technology (MIT), joined the faculty at MIT as an Assistant Professor and set up a research program focused on ultrafast optical phenomena. Inspired by earlier work from Bell Laboratories on femtosecond laser pulses, Fujimoto and his colleagues explored the possibility of using such pulses as a range-finding tool. The idea was to split a light pulse into test and reference pulses and then compare the time for the test pulse to travel to and from a target of interest with the time for the reference pulse to travel a precalibrated distance. From early on, the team focused on imaging the retina within the intact eye, taking advantage of the optical clarity of the cornea, lens, and vitreous. Cow eyes from local slaughterhouses provided a plentiful source of test material.

In the Fujimoto laboratory, research into tissue imaging was also underway with a more conventional and less-expensive low-coherence light source and a Michelson-type interferometer (4). Encouragingly, in 1988, Adolf Fercher at the University of Vienna, who was exploring the use of interferometry with low coherence light for measuring the length of the eye, reported the use of this approach to produce a profile of light scattering surfaces along a single line passing from the cornea to the back of the eye. In 1990, David Huang, currently the Peterson Professor of Ophthalmology in the Casey Eye Institute at the Oregon Health Sciences University and then an M.D.-Ph.D. student with Fujimoto, was joined by Mr. Eric Swanson, a group leader at MIT’s Lincoln Laboratories with expertise in fiber optic communication networks and intersatellite laser communication systems.

By 1991, the team of Huang, Swanson, and Fujimoto had built an interferometer and data-processing pipeline that could i) perform an axial scan (i.e., measure light scattering at different depths within a tissue; referred to as the Z-axis) by measuring the signal strength from the recombined test and reference beams as a moving mirror changed the path length of the reference light beam, ii) perform a series of such measurements along a line in the X–Y plane, and iii) assemble the collection of Z-axis scans into a two-dimensional cross-sectional image of the tissue. This configuration is now referred to as time-domain OCT because, at each X–Y location, the Z-axis information is collected during the time interval required for the moving mirror to sweep out its trajectory (Fig. 1C). In collaboration with clinical colleagues and using ~800-nm light, the MIT team recorded cross-sectional images of the retina and the lumen-proximal wall of a coronary artery from post-mortem human samples (5). A comparison with histologic sections subsequently obtained from the same tissues showed excellent correspondence of the main features. The 500- to 700-μm Z-axis depth of the retina image was more than enough to visualize the full thickness of the retina, the choroid, and part of the adjacent sclera. The more highly scattering composition of the artery wall limited the depth of this image to 150–200 μm, but this was deep enough to clearly observe the structure and location of an atherosclerotic plaque. In these first interferometry images of the retina, the Z-axis resolution was insufficient to resolve the three layers of cells within the neural retina, but the clinical potential of seeing the retina and choroid in cross-section was clear. The resulting publication was simply titled “Optical Coherence Tomography” (5). As of this writing, it has been cited more than 10,000 times.

Technical Advances

A major technical challenge with the initial OCT configuration (Fig. 1C) was the long time required to acquire an image: The instrument used by Huang et al. (5) required several minutes for a single two-dimensional image. Over the ensuing 2 years, a prototype clinical instrument for ophthalmic applications was designed and built with ~100-fold faster data acquisition (4, 6). A quantum leap in data acquisition speed came in the early 2000s with the realization that reflectance data could be collected far more rapidly if the reference mirror remained at a fixed location and either i) the light source included a range of wavelengths that could be interrogated in parallel by adding a spectrometer to the detector or ii) the light source consisted of a narrow bandwidth laser that swept rapidly and repetitively across a range of wavelengths together with a detector that recorded the interference pattern at multiple time points within each spectral sweep. These two approaches are referred to as spectral-domain (SD) OCT and frequency-domain (FD) OCT, respectively (Fig. 1 D and E). [Frequency-domain OCT is also referred to as swept-source (SS) OCT.] Both SD-OCT and FD-OCT use a Fourier transformation to compute the reflectance profile across the Z-axis, and, therefore, they are collectively referred to as Fourier-domain OCT. Using these technologies, current commercial OCT scanners can acquire 100,000 Z-axis scans—referred to in the literature as “A-scans”—per second (Fig. 2 A–C).

One of the most exciting extensions of OCT is OCT angiography, a rapid, safe, and noninvasive method for imaging the microvasculature (8, 10). Defects in vascular structure and integrity play a central role in some of the most common retinal disorders, including diabetic retinopathy, age-related macular degeneration, and venous and arterial occlusions, and this has been a principal motivator for developing OCT angiography. For the past several decades, fluorescein or indocyanine green angiography has been the gold standard for visualizing the retinal vasculature, monitoring blood flow, and detecting and localizing vascular leakage. In this method, the fluorescent dye is delivered intravenously at a peripheral site (the arm) and then the sequential appearance of the dye in arteries, capillaries, and veins is recorded with a time series of images captured through an ophthalmoscope. Fluorescein angiography is used to visualize the retinal, but not the choroidal, vasculature. The choroidal vessels, which lie beyond the retinal pigment epithelium (RPE), cannot be seen with fluorescein angiography because the green emitted light is absorbed by RPE melanin. To visualize the choroidal vasculature, indocyanine green is used as the vascular tracer; its infrared emission is not absorbed by RPE melanin.

In OCT angiography, multiple scans of the target tissue are captured and compared, and differences between the images that are associated with moving blood cells are used to create a map of vessel locations, a protocol that relies on the high data acquisition rates of current OCT scanners. Importantly, OCT angiography can visualize both the retinal and choroidal vasculatures because the infrared light used in OCT is not absorbed by RPE melanin. As seen in Fig. 2D, OCT angiography provides a high-resolution depth-resolved map of the vasculature, including, in this example, a small tuft of pathologic vasculature growing under the retina (8). In Fig. 2D, panel (b), the same region of the retina is visualized with fluorescein angiography. In en face and cross-sectional views obtained by OCT angiography, individual capillaries are resolved within the neovascular tuft, and their Z-axis location between the choroid and retina is clearly defined [Fig. 2D, panels (c) and (d)]. Fluorescein angiography and OCT angiography provide complementary information to guide clinical decision-making (8, 10). Fluorescein angiography surveys a large retinal area, it provides dynamic blood flow information, and it identifies sites of vascular leakage, whereas OCT angiography surveys a smaller retinal area, and it provides high-resolution depth-resolved images of vascular anatomy for both the retina and choroid.

A second exciting advance is the wedding of OCT and endoscope/catheter technologies (both probe types are hereafter referred to as a “catheter”) (11, 12). Accessing internal organs with catheters has revolutionized gastroenterology and cardiology, and equipping catheters with OCT imaging capability was recognized early on as a natural extension of these technologies. Catheter-based OCT presents interesting engineering challenges related to miniaturization. In one design, an intravascular catheter with a 3.2-mm outer diameter houses fiber-optic cables that terminate at a microprism that can be rapidly rotated by a micromotor to acquire images of the artery wall at a frame rate of 400 per second (13). During data acquisition, the catheter is slowly pulled back along the artery (at a speed of ~1 mm per second), generating a series of A-scan images that are then assembled into a single three-dimensional image. Two cross-sections from an OCT series within a coronary artery are shown in Fig. 2E, panels (b) and (c), along with an angiogram [Fig. 2E, panel (a)] that shows an abrupt narrowing of the artery starting just upstream of the panel (c) cross-section. The irregular topography of the narrowed lumen is clearly apparent in panel (c).

Clinical Impact

OCT’s greatest impact thus far has been in ophthalmology, where it has become an indispensable diagnostic tool for virtually all retinal diseases (11). OCT of the retina provides an early and quantitative assessment of nerve fiber layer loss in glaucoma (a correlate of retinal ganglion cell loss), and it visualizes macular edema, macular holes, retinal detachment, and retinal neovascularization. The widespread adoption of OCT within the ophthalmology community in the early 2000s coincided with the development of antiangiogenic therapies for retinal neovascular disease, principally exudative AMD and diabetic macular edema. In these therapies, an anti-VEGF antibody or a VEGF “trap” (a soluble form of the VEGF receptor) is injected into the eye at intervals of one to three months. Reducing the intraocular level of free VEGF leads to regression of neovessels and reduced vascular leakage, the latter leading to reduced retinal edema. The ability of OCT to provide rapid, quantitative, and noninvasive monitoring of disease severity and the response to antiangiogenic therapy has been integral to tailoring treatment schedules to each patient’s needs (14). A 2017 analysis estimated that the use of OCT to guide decision-making in the context of antiangiogenic therapy for AMD saved more than $10 billion in medical expenses between 2008 and 2015 (15). The cumulative savings through 2023 would likely be at least $20 billion.

In cardiology, intravascular imaging is used to identify partial or complete coronary artery occlusions and to assess the integrity of the vessel wall and the structure and stability of the occlusion. For more than 50 years, the mainstay of coronary artery imaging has been angiography, which consists of a time series of X-ray images showing the flow of an intravenous bolus of radio-opaque dye. The dye is released from a catheter tip that has been threaded from its insertion site in a peripheral artery to the base of one or more coronary arteries. Like fluorescein angiography of the retinal vasculature, the resulting movies provide dynamic information about blood flow. As in the retina, one limitation is that the angiographic images are two-dimensional projections of three-dimensional objects. To supplement angiography, catheter-based ultrasound (also referred to as intravenous ultrasound) and, more recently, catheter-based OCT (Fig. 2E) have been developed (11, 12). Catheter-based ultrasound has been found to improve clinical decision-making and clinical outcomes when a coronary artery is targeted for stent placement. Catheter-based OCT produces an image in less than one-tenth the time required for a catheter-based ultrasound image, and the resulting OCT image is approximately 10-fold higher resolution (11). Thus, catheter-based OCT could become at least as useful as catheter-based ultrasound for guiding clinical decision-making in cardiology (12).

Catheter-based OCT is also being evaluated in gastroenterology and pulmonology, especially as related to harvesting cancerous or precancerous tissues for microscopic evaluation (11). The diagnostic yield of catheter-harvested tissue is strongly dependent on sampling the appropriate sites, and the real-time subsurface information provided by OCT can both improve tissue sampling and permit longitudinal monitoring of potentially precancerous lesions. For example, surveillance of the esophagus for precancerous changes can be performed with a rotational OCT probe that generates an image of the full circumference of the esophagus along 6 cm of its length and to a depth of 3 mm. Similarly, OCT coupled to bronchoscopy permits imaging of the airway wall for early detection of lung cancer, wall thickening in obstructive lung disease, airway smooth muscle remodeling in asthma, and fibrosis and dilation of pulmonary tissue in interstitial lung disease.

The Future

Beyond ophthalmology and cardiology, the clinical development of OCT, together with rigorous assessments of its utility, is still at a relatively early stage. As OCT imaging can provide biological information about any near-surface tissue structures, OCT has potential utility in dermatology, in oncologic surgery, and in any medical specialty that uses remote imaging, including gynecology and urology. In ophthalmology, a new generation of small, portable, and inexpensive OCT scanners promises to greatly improve access to OCT in resource-limited settings (16). Finally, it will be interesting to explore synergies between OCT and artificial intelligence, which could enhance clinical decision-making by extracting patterns from large collections of OCT images (17).

With their beautiful synthesis of physics, engineering, and biology, James Fujimoto, David Huang, and Eric Swanson have dramatically changed the practice of ophthalmology, improved the lives of millions of people, mitigated healthcare costs, and set the stage for new applications of OCT in the service of human health.

Data, Materials, and Software Availability

There are no data underlying this work.