Andrew M. Rollins
Assistant Professor, Biomedical Engineering
Case Western Reserve University
Doctors and medical researchers have a long list of imaging technologies they use to explore the human body. But none are perfect. Microscopes, for example, are fine for cells and small tissue samples, but aren't well suited for examining them inside the body. Ultrasound, CT, and MRI can peer inside the body but don't have the resolution to capture cellular detail. And although electron microscopy picks up extremely fine detail, it doesn't work with living samples.
Optical coherence tomography (OCT), a relatively new imaging technique, lets physicians and medical researchers capture small features and textures in living tissue and when combined with catheters, is small enough to fit inside the human body. Its development was, in part, spurred by researchers looking for a way to see fine details in the layers of the retina. The optics of the eye tend to limit the resolution to about the thickness of the retina for conventional optical ophthalmic imaging, such as slit-lamp microscopy and scanning-laser ophthalmoscopy. Therefore, it's difficult to image retinal layers without the optics of the eye itself interfering. Ultrasound and MRI also lack the resolution. OCT, on the other hand, bypasses the problem of the eye's optics and generates retinal images with about 25 times the resolution of ultrasound and MRI.
OCT is also finding biomedical imaging applications beyond the retina. For example, we are investigating OCT imaging for the anterior segment of the eye, gastrointestinal endoscopy, and developmental cardiology. Others are looking at OCT imaging for cardiovascular, dermatologic, urogenital, and many other applications. But while it provides unprecedented resolution for exploring living tissues in real time, OCT has its limits. It has a small field of view, roughly measured in tens of millimeters, and it sees only about 2 to 3 mm into nontransparent tissue.
THE ABCS OF OCT
OCT, like ultrasound, bounces waves off of a sample and extracts imaging data from reflected signals. But ultrasound uses acoustic pulses and relatively inexpensive electronics that switch between transmitting and receiving quickly enough to process information. OCT uses light.
Light is too fast, especially over the small distances involved, and electronics would have to work in the terahertz range to be truly analogous with ultrasound, an expensive and impractical proposition. Instead, OCT uses interferometry, splitting off some light into a reference path before the rest of the light is sent into the sample. An interferometer combines the returning sample light with the reference light to produce interference fringes. These fringes are measured electronically and the signal is processed to determine reflectivity values of the sample as a function of depth into the tissue.
To get snapshots at different depths, OCT smoothly varies the distance the reference light travels by scanning an optical delay line, a process called pathlength ranging. It sweeps the interferometer's sweet spot, covering the range of light from specific depths, and building a single line of an image. The light, along with receiving optics, then move to another spot along the sample and the process repeats, until the region of interest is scanned. The lines are then combined into a cross-sectional image.
The light used plays an important role. It has to be spatially coherent (highly directional) like a laser (as opposed to a light bulb) so it can focus in on the tissue sample, then be scanned, and efficiently collected. But it can't be temporally coherent like a laser (having essentially one wavelength), or else light scattered back from deep in the tissue will look just like light scattered from the tissue surface. In other words, the system would have low resolution. OCT uses light sources that are temporally incoherent, like that from a light bulb with lots of different wavelengths, but spatially coherent, like a laser. In this way, only fringes from sample light that travel roughly the same distance as the reference light are used (i.e., those that went in and bounced right back).
The first OCT machines built were extremely slow, taking several minutes to take the 500 or so slices needed to build up an image. This was fine for inanimate samples, but useless with patients who could not hold still for that length of time.
LIGHT AND MIRRORS
Several challenges had to be overcome. First, the machines needed a faster way to sweep the reference path length. The simplest scanning optical delay line, a mirror moving back and forth, just took too much time overcoming its inertia each time it changed direction. Also, the light source had to be brighter. Increasing the scan rate would decrease the signal-to-noise ratio unless more light was used to probe the sample.
One technique used to overcome the mirror problem was borrowed from scientists measuring femtosecond light pulses from ultrafast lasers. The light first goes through a diffraction grating and lens before hitting the mirror. This distributes the spectrum of the light onto the mirror. And instead of moving the mirror, it is only slightly tilted, forcing one side of the light spectrum to travel farther than the other. The mechanism that could tilt the mirror at several kilohertz had already been developed.
A newer approach, Fourier domain OCT, avoids moving the mirror altogether. Instead, the device measures spectral interference using a tunable laser that sweeps wavelengths, or a broadband light source and a highspeed spectrometer. Interference fringes show up in the frequency domain and Fourier-transform processing reconstructs the reflectivity profiles common to OCT. Although this reduces mechanical complexity, it increases the amount of computation needed.
Near-IR light works best for OCT because it allows imaging deeper in nontransparent tissue than other wavelengths. Visible light only images a fraction of a millimeter deep. Near-IR can image penetrates 2 to 3-mm deep. But you can't go too far into IR because water molecules begin absorbing too much of the light. So OCT researchers must balance absorption and scattering.
For imaging the retina, 830 nm is the standard wavelength used. This wavelength has low water absorption, which is necessary because the light has to pass through the entire length of the eye and back. Penetration depth isn't a problem because the retina is thin. For other tissues, 1,300-nm light is becoming the standard. There's a dip in the water-absorption spectrum there, and it happens to coincide with what was once the standard wavelength for optical communications. The communication standard has changed to 1,550 nm, but there are still a lot of fibers, light sources, and other devices built for the communication industry that can be used in OCT.
The light beam plays a major role in resolution. Wider bandwidths yield higher axial resolutions (depth into tissue along the path of the light), while narrower beams improve resolutions transverse (left and right, up and down). So there is still a lot of research into developing better light sources. For example, although LEDs have been employed, they are not bright enough for high-speed OCT. Instead, superluminescent diodes, a device halfway between an LED and a laser diode, are being used. Other sources being explored include semiconducting optical amplifiers, femtosecond lasers, and supercontinuum light sources. The latter take advantage of short-pulse lasers and nonlinear optic fibers. Pumping the laser through these fibers broadens the spectrum, producing extremely broadband light.
These advances in scanning the frequencies and developing better light sources has lead to OCT machines that can take 8 to 16 images/ sec with 500 slices/image. Resolution is typically in the 10 to 20- µm range, but OCT has been demonstrated with resolution on the order of 1 µm.
Researchers are enthusiastic about what OCT could do for ophthalmology: Give them an inexpensive, noninvasive, and painless way to closely monitor changes in the retina. This could help chart the course of macular degeneration, diabetes-related blindness, macular holes, and a host of other progressive eye diseases.
Other early users are looking at endoscopic applications in the digestive (GI) tract and intravascular imaging to detect plaques in the blood vessels. It might find GI cancers earlier in their development than ultrasound probes, which don't have as high resolution. This would let surgeons remove the cancer before the rest of the body even knows it's there. But OCT's small field of view, a few square millimeters, means it would take a long time to image the entire GI tract. One possible solution is to build high-speed, automated OCT to quickly and reliably image large GI segments. Another possible solution is to combine OCT, which can easily probe small areas with high resolution, with a large field-of-view technology, such as MRI or fluorescence, and gain the advantages of both.
In blood vessels, OCT might be able to discern vulnerable from stable plaques. Vulnerable ones can rupture and clog a vessel feeding the heart or brain, causing a heart attack or stroke.
Researchers also want to expand OCT using well-known properties of light to extract more information from scans. One of the most promising ideas is Doppler OCT. It looks for frequency shifts in the interference patterns that indicate moving objects in the sample such as blood cells. This is of particular interests to ophthalmologists because blood flow is affected by many common causes of blindness, including diabetic retinapothy and macular degeneration.
Researchers are also exploring polarization. This technique would use a light source with a known polarization for OCT and measure polarization of returning light and interference fringes. It is believed this will be more sensitive in imaging damage to bifringent tissues (i.e., those that slow light more as it passes through in one direction than another direction). Damage to tissue can change its structure or organization and change its birefringence. Nerve fibers, for example, are bifringent, and so are skin and other connective tissues.
Researchers are exploring polarization-sensitive OCT to assess burn severity. In burns, connective tissue proteins are denatured which reduces the tissues' birefringence. Polarization-sensitive OCT has been shown to detect differences between healthy and damaged tissues better than scatteringintensity imaging alone.
Polarization-sensitive OCT could also prove useful in detecting the early damage to the retinal nerve-fiber layer (RNFL) associated with glaucoma. Glaucoma is a family of diseases characterized by increased pressure inside the eye. This pressure kills the retina and leads to blindness. Doctors believe the RNFL, which is bifringent, first degrades then thins before the retina is permanently damaged. They hope to detect this degradation, or at least the thinning, as a means of early detection of glaucoma.
Another area under investigation is spectroscopic OCT. It examines how much light returns from tissues and how the spectrum of that light changes. This should let doctors characterize the metabolic state or biochemistry of the tissue. The difficulty is that OCT typically uses light with a narrow frequency band compared to light sources used for conventional spectroscopy. This limits the spectral range that can be measured with OCT. Also, the near-infrared light used for OCT is less useful than UV and visible light for probing most common chromophores associated with tissue metabolism. This difficulty is being overcome by researchers developing broadband light sources and new nonlinear optical techniques that use molecular contrast agents for OCT.