Interferometry gives eye camera a new lease of life

28 Apr 2003

A retina camera that combines adaptive optics with optical coherence tomography is generating images of retinal cells with unprecedented resolution. Rob van den Berg discovers that it could lead to early detection of glaucoma.

From Opto & Laser Europe May 2003

Ever since German physiologist Hermann von Helmholtz first observed the inside of the human eye with his ophthalmoscope, physicians have striven to obtain better images of the retina and unlock its secrets.

Now a US expert in retinal imaging is pushing ophthalmoscope technology to its limits to try to achieve that goal. Donald Miller - who carried out much of his initial research at the University of Rochester, US, but now heads his own group at the University of Indiana - has developed a highly sensitive camera that can generate image slices of the retina to a lateral and depth resolution of just a few micrometres (2 and 4 µm respectively). At the recent Photonics West conference in San Jose, US, Miller presented the first results of his work.

Overcoming distortion "In contrast to what you might expect, the human eye has significant optical defects - aberrations that distort a passing wavefront, blur the retinal image and degrade our visual experience," explained Miller. "Another, and often larger, source of retinal image blur is diffraction, which is caused by the finite size of the eye's pupil. Both effects not only limit what the eye sees looking out, but also determine the smallest internal structures that can be observed when looking into the eye with a microscope."

This is where adaptive optics, a technique developed for astronomy that compensates for wavefront distortion induced by atmospheric turbulence, can help. By analysing the amount of distortion and correcting for it with a deformable mirror, it is possible to obtain sharper images of stars and other stellar objects. In a similar way, it is possible to measure the aberrations of the eye and compensate for them to obtain higher-resolution images of the retina. This is important as photodetector cells in the retina can be as small as 4 µm in diameter.

In 1997 Miller's team constructed the first retina camera based on adaptive optics. "With this camera we were able to see individual photoreceptor cells in the fovea [the part of the retina that gives the sharpest vision]. This was a major breakthrough, but these cells turned out to be the easiest to detect due to their relatively high contrast and brightness," said Miller. "Studying other parts of the retina at the cellular level has proved much more difficult. The retina is organized into well-defined cell layers that are tightly stacked one on top of the other. Reflections from the layers create a host of superimposed images at the detector with the brightest reflections masking the fainter ones."

One way to image the retina's different layers is to use a confocal scanning laser ophthalmoscope. A focused laser beam is scanned across the retina at the desired depth, and the reflected light is detected. This makes "optical sectioning" possible.

By combining this approach with adaptive optics, scientists have already achieved good images of the retina. However, Miller wanted to see if performance could be further improved. "The first and only confocal scanning laser ophthalmoscope (cSLO) using adaptive optics was developed in 2002 at the University of Houston, and achieved a 2-3 times increase in axial [depth] resolution over commercial cSLOs," he said. "Its 110-150 µm axial resolution was a big improvement, but is still much larger than cells in the retina and so provides only coarse sectioning of the retina, whose total thickness is a few hundred micrometres."

Miller decided to try to solve the sectioning problem using an entirely different approach: optical coherence tomography (OCT). In OCT the "time of flight" and intensity of reflected optical waves is measured using interferometry. Typically, a Michelson interferometer is used to split the light beam into a reference beam and a signal beam. While the former is reflected from a mirror, the latter is backscattered from the sample under investigation. An interference signal (image) is then generated for reflections with a path length that matches that of the reference beam. By precisely controlling the length of the reference arm, it is possible to perform imaging at different depths within the sample.

Miller and his Indiana colleagues Junle Qu, Ravi Jonnal and Karen Thorn have now built a sophisticated retina camera that combines the enhanced lateral resolution of adaptive optics with the high axial resolution of OCT (see figure 1).

The camera can image the living human retina with a tiny voxel (3D pixel) volume of 44 µm3 - more than 340 times smaller than that of the commercial cSLO.

Record resolution "We managed to record high-resolution images at different depths in the human retina. Although we are not certain yet how to interpret these, the variation in pattern that we see at the cellular level suggests that we are looking at real retinal structure," said Miller. "We have achieved a depth resolution of 14 µm - an order of magnitude better than the adaptive optics cSLO - using a superluminescent diode (SLD) with a very short coherence length."

The Indiana camera consists of three independent optical circuits. There's an adaptive optics system for compensating any optical distortion; a one-dimensional (1D) OCT imaging system for depth profiling and reference purposes; and a two-dimensional (2D) OCT for capturing 2D image slices of the retina at any depth.

The light source for each system is an SLD operating at a distinct wavelength. SLDs are favoured over other laser sources because their short coherence length diminishes the problems associated with laser speckle, an optical interference effect that degrades image quality in OCT systems.

"To reduce the effects of speckle we are looking at light sources with better spatial properties," said Miller. "We are also evaluating faster read-out detectors and real-time image-processing electronics to reduce the acquisition and processing time, making the instrument more clinically viable."

The adaptive optics system consists of a Hartman-Schack wavefront sensor, which measures wavefront error, and a deformable mirror that uses 37 actuators to adjust its shape and make wavefront corrections. This wavefront correction system uses a low-power 788 nm SLD and closed-loop control to perform up to 22 wavefront measurements and corrections per second. The exposure level at the eye is less than 7 µW - 80 times less than the maximum exposure recommended by the American National Standards Institute.

The 2D OCT imaging system comprises a scientific-grade CCD camera, an interferometer and a 10 mW SLD operating at 679 nm. The length of the interferometer's reference arm is adjusted by a voice-coil connected to a piezoelectric mirror.

The 1D OCT system is needed because even minimal motion can degrade or mis-register the image. Involuntary head and eye movements and pulsations of the eye can induce random changes in the distance between the retina and the CCD camera in the 2D OCT system.

In order to compensate for these movements, Miller has come up with a tracking scheme that uses an 856 nm SLD to perform a 1D OCT scan 20 times per second, traversing the retina's full depth at just one location. This serves as a reference to keep the retina in exactly the right position.

Potential applications include the detection of local pathological changes and the investigation of the function of retinal layers. Its sensitivity could mean more accurate observations at the head of the optic nerve, or population counts of ganglion cells.

The camera may even be able to detect the onset of glaucoma, a leading cause of blindness in the West that is caused by the gradual loss of optic nerve fibres. Currently, glaucoma can only be detected after significant damage has occurred. The detection of diabetes could also be improved - the disease leads to microaneurysms in the retinal blood vessels, also causing blindness.

"The retina camera might also contribute to a deeper scientific understanding of the light-collecting properties of the retina," said Miller, "in particular, the spatial arrangement and relative numbers of cone and rod photoreceptors, which transform photons into neural signals."