Hank Hogan, Contributing Editor
In explaining why he’s interested in developing an ultrafast three-dimensional tomographic endoscope, Kohji Ohbayashi, a physics professor at Kitasato University in Kanagawa, Japan, pointed to a few basic facts. Optical coherence tomography, or OCT, has the best resolution of any tomographic technique that images biological samples and tissue. Hence, it offers the ability to spot cancer at the smallest possible size.
So a high-speed 3-D OCT endoscope would provide a better cancer diagnostic tool, according to Ohbayashi. “With the system, we would be able to detect the earliest stage of cancer without a biopsy.”
A 3-D optical coherence tomography image of a finger shows the structure of the outermost layer (epidermis) and the underlying dermis. A new system can capture more than 60 such images per second. Courtesy of Kohji Ohbayashi, Kitasato University.
As with all tomographic methods, OCT images samples as a series of slices. A light beam is divided into two arms, one probing the sample and the other acting as a reference. The two are then combined via interferometry, giving rise to an image. OCT can probe deeper into tissue than other optical techniques and offers a resolution in the microns.
The problem confronting researchers like Ohbayashi is how to acquire 3-D OCT data as quickly as possible. A faster acquisition rate would reduce the total time required for a scan, making the approach more useful in a clinical setting. A faster rate also would cut down on the amount of imaging artifacts resulting from movement – again, an important attribute in a diagnostic tool.
Ohbayashi was lead author of a Photonics West paper outlining a new approach to attaining this goal. The group demonstrated a system that can image 60 million axial scans per second – a data rate that’s comparable to the fastest of any OCT method.
The key to achieving this rapid-fire data acquisition was the use of optical demultiplexers. This was a first in OCT, Ohbayashi said. The technique builds upon earlier work from the group, which showed that Fourier domain OCT could be performed with a discrete source.
Optical demultiplexers disperse incoming light into outgoing discrete separable wavelengths, with an equal frequency interval in each segment. This approach enables detection at all the discrete wavelengths simultaneously, and that is why the data acquisition rate is so high.
In their demonstration setup, the researchers used a superluminescent diode from NTT Electronics Corp. of Tokyo as a light source, with an output centered at about 1560 nm. They divided this light equally into two arms. By adjusting the length of the reference arm, they could change the depth of the image within the tissue. Using a scanner from Electro-Optical Products Corp. of Glendale, N.Y., and a galvo mirror from Cambridge Technology Inc. of Lexington, Mass., they moved the sample beam across the tissue, thereby providing a scan in X and Y.
After collecting the light from the sample, they used interferometry to combine it with that from the reference arm. They sent the result into two banks of optical demultiplexers, which were custom-made arrayed waveguides from NTT Electronics.
The demultiplexers separated the OCT signal into 256 narrow bands, which the researchers detected with an equal-number photoreceiver array from New Focus Inc. of San Jose, Calif. They captured the output from the array using a fast multichannel data acquisition system built with boards from National Instruments Corp. of Austin, Texas. From there, the information flowed into a computer, which displayed the image.
Results from the system when doing a lateral scan at a rate of 16,000 frames per second were a 3-mm-depth range with 23-μm resolution. These tests, although encouraging, showed also that the dynamic range of the system is only marginal for measurement of biological tissues. Improvements in this area are planned.
Another area that needs work is the scanning speed, which is determined by the data acquisition subsystem. The digitizing hardware currently allows about a million 1000-point fast Fourier transforms to be performed each second. While faster than what was possible before, this is still only about two-thirds of what is needed for a real-time 3-D raster scan of a sample. Therefore, live rendering of images isn’t yet possible. It is being worked on, with engineers implementing changes to the data acquisition system to boost its speed. In the meantime, Ohbayashi noted that the system can store the data and render the image later.
The system is currently being used in new areas, he added. “We are now applying the system to biological organs which show time-dependent deformation. An example is the iris, which deforms following the intensity of illumination.”
He also noted that an ultimate goal is a commercial unit. It’s unknown when – or if – that will happen. Various companies have expressed interest in such a system; however, the total cost of the product could be a problem.
Contact: Kohji Ohbayashi, Kitasato University, Kanagawa, Japan; e-mail: firstname.lastname@example.org.