Optical Coherence Tomography

Use of Optical Coherence Tomography for the Imaging of Cortical Tissue

Abstract In an effort to provide more detailed information about epileptic lesions in neural tissue for the purpose of efficient removal, an Optical Coherence Tomography system was built to provide non-destructive imaging. The device is based upon the principle of a low-coherence interferometer. Such an instrument can detect if a particular point within a sample of tissue reflects light; an integration and interpretation of many scans of many points provides a high resolution image. The device built for this investigation has been successful in scanning semi-reflective objects such as glass and glossy plastic. Objects with low reflectivity or high degrees of scattering did not produce strong enough signals to be detected successfully. One-dimensional scans were accomplished with reasonable speed and under computer control. Limited twodimensional scans were also taken with much lower speed and with constant human intervention.

Introduction Overview Optical Coherence Tomography (OCT) is an imaging technique that is similar in principle to ultrasound, but with superior resolution. It relies on exposing a sample to a burst of light and then measuring the reflective response from different depths and is therefore capable of scanning non-invasively beneath the surface of the sample. In ultrasound imaging, it is relatively easy to measure the time delay of each reflected packet. However, for light pulses, interferometry must be used to measure the displacement with meaningful accuracy. The amount of light reflected from each point within the scanning window in the sample is plotted graphically as an OCT image. The goal of this investigation is to use Optical Coherence Tomography to image epileptic lesions on cortical tissue from rats. Such images would be immensely useful for surgical purposes. They would detail how deep the lesion is, allowing for precise removal that neither removes an insufficient amount of damaged tissue nor extracts too much healthy tissue. Though commerical OCT systems already exist, they typically do not scan very deeply beneath sample surfaces. For the purpose of this study, a system must be constructed that scans up to 2 millimeters into tissue1. Unfortunately, an increase in axial depth necessitates a decrease in transverse (along the surface of the sample) resolution due to focal restrictions of the objective lenses2. However, this loss is acceptable for this investigation, as the main goal is to determine lesion depth and not to achieve perfect image clarity. Physics and Design The ability to detect the positional delay of light reflecting from a tissue sample is at the heart of OCT. Low-coherence interferometry provides just that. A low-coherence light source has the potential to produce interference fringes only when integrated with light from the same source that has traveled nearly exactly the same distance3. This means that if light from such a source is split by a beam splitter into two equal parts, both of them reflect off of different objects, and combine to form one beam again, they will produce an interference fringe pattern only if the distance they traveled while split was exactly the same. Hence, this type of interferometry can

determine if the two reflective objects were the same distance from the splitter or not. The type of device described here is a Michelson interferometer.

INTRODUCTION FIGURE 1: MICHELSON INTERFEROMETER1

Optical Coherence Tomography applies the principle of a Michelson Interferometer by using one of the split beams to be a controlled reference and the other to be the variable length scanning beam. The length of the reference arm can be adjusted by moving a reference mirror back and forth along the axis of the beam. By thus controlling the reference arm and looking for interference patterns, the device can determine if the corresponding point on the objective arm (that is, the point that is exactly the same distance from the splitter as the reference mirror) reflects light. If it does, interference occurs, the system quantifies the strength of the interference, records this, and moves on to a different point by again altering the reference mirror path length. In an OCT system, a free space interferometer such as the one diagramed above is not practical. The components of the system must be able to be configured in a more flexible manner and not be confined to such a rigid design. Therefore, fiber optic cables and connectors are used. The principle of the interferometer remains the same, with the exception of the loss of control over the polarization of the light. When light reflects again and again off of the walls of the fiber optic cables, they do not lose intensity, but are rotated so that it is possible for the polarization if the two split beams to shift out of alignment with each other. When this occurs, interference patterns are seriously weakened, and so it is possible for the two beams to have traveled the same distance and still not be detected. This problem is easily rectified by the addition of a polarization controller. Such a device rotates the polarization of light in a fiber optic cable without serious dampening, allowing the system to be quickly adjusted to maintain polarization alignment. Another modification that is made to the interferometer is an oscillation of the reference mirror. This causes the reference and objective beams to shift in and out of alignment when they are matched. As a result, the detected interference is not a constant phenomenon, but rather it comes in packets that appear and disappear with the vibration of the mirror. A lock-in amplifier looks for these packets and filters out all other noise, allowing for more accurate and more readily quantifiable data. Additionally, it is completely unrealistically to assume that tissue samples will reflect incident light without a significant amount of scattering and absorption4. The lock-in amplifier also serves to enhance these weaker reflections.

INTRODUCTION FIGURE 2: THE OCT FIBER OPTIC INTERFEROMETER1

Materials and Methods Materials In the construction of the Michelson Interferometer that is the OCT device, the following pieces of equipment were utilized: a Super Luminescent Diode broadband light source model BSL-1300A by E-Tek Dynamics, fiber optic cable of type SMF-28-100 by Thorlabs, fiber optic connectors of type FCPSMC-126 by Fiber Control Industries, connectorization kit CK03 by Thorlabs, F220FC-C collimators by Thorlabs, AD11F mounting adapters by Thorlabs, the FCPL-S22131 2x2 50/50 coupler by Newport, photodetector model PDA400 by Thorlabs, an FC adapter model SM1FC by Thorlabs, various mounting bases, clamps, posts, and adapters by Thorlabs, microscope objective K36-131 by Edmunds, polarization controller FPC-1 by Fiber Control Industries, a Tektronix 2865 oscilloscope, a National Instruments Signal Connector Block SCB-68, a Macintosh IIc equipped with a National Instruments Lab-NB DAQ Card and a National Instruments GPIB Card, a Power Macintosh G3 Beige Tower equipped with a National Instruments PCI-MIO-16E-4 DAQ Card, a Compaq Presario 12XL300 notebook computer, a Lock-in Amplifier model 128A by Princeton Applied Research, a generic corner cube reflector, a mechanical vibrator coil model SF-9324 by Pasco Scientific, two Klinger stepper motors, one Klinger programmable stepper motor controller, several coaxial cables and T-splitters, and various adapter mounting plates produced by the Duke Physics Instrument Shop. Methods I: Interferometer Setup (completed before the author began work on this investigation) Prior to the author beginning work on this investigation, the system had already been set up to some extent, following a design described as follows1: The superluminescent diode (SLD) was attached to the mounting grid-board. Using an FC connector and a fiber optic cable, its light output was connected to one orifice of the 2x2 coupler. The amplifier/photodetector assemble was also securely attached to the grid-board. It was connected to its power supply. A fiber optic cable was secured to its input using an FC adapter. The other terminus of the cable was connected to the 2x2 coupler, on the same side as that to which the light source had already been connected.

One collimator was mounted with a steel post to the grid-board such that its collimation axis was parallel to the plane of the table. Using necessary optical cables and adapters, it was connected to one end of the polarization controller, which was in turn connected to one of the two remaining terminals on the 2x2 coupler. The other collimator was also mounted using a post to the grid-board, but with its axis oriented normal to the table and pointed directly into it. Directly below, a microscope mount was attached with a microscope objective. A mounting platform was installed under the objective, near the focal length of the objective lens. The collimator on this assembly was connected directly to the remaining 2x2 coupler terminal.

METHODS FIGURE 1B: ACTUAL INTERFEROMETER SETUP

Methods II: Computer and Electronic Hardware Setup (in progress when author began work on this investigation) The output of the amplifier/photosensor assembly was connected using a coaxial cable to a T-splitter, allowing the signal to be simultaneously distributed to the lock-in amplifier signal input as well as an oscilloscope channel for monitoring.

METHODS FIGURE 2: STEPPER MOTOR CONTROLLER, LOCK-IN AMPLIFIER, AND OSCILLOSCOPE

Two computers with Data Acquisition (DAQ) interface cards and custom-made breakout hardware were used to drive various devices in the apparatus and to take data. A third computer was used to generate the images from the raw data. The Macintosh IIc was set up to step the reference mirror on command from the Power Macintosh. The Power Macintosh was used to generate the reference signal required by the lock-in amplifier, as well as the driving function for the oscillating reference corner cube mirror. Both of these connections are accomplished using coaxial cables. Details of this set up can be seen in the accompanying figures. The output of the lock-in amplifier was connected to an input channel of the Power Macintosh DAQ board via custom-built connectors and the National Instruments Signal Connector Block SCB-68. One output of the Power Macintosh was connected to an analog input channel of the Macintosh IIc; this link serves to command the stepper motor controller to “step” the reference mirror 10 microns.

METHODS FIGURE 3: POWER MACINTOSH DAQ INTERFACE SETUP

Methods III: Operation of the Assembled Apparatus Before meaningful scans can be made, the device must be aligned so that the reference mirror reflects the maximum amount of light back through the interferometer. To accomplish this, the SLD was switched on, the amplifier/photosensor was activated, and the oscilloscope set to Direct Current (DC) monitoring mode. The height of the corner cube as well as its lateral position were adjusted manually so as to maximize the oscilloscope voltage readout. The alignment can be tested by blocking and unblocking the mirror with an opaque object. The oscilloscope reading should drop when the light is blocked and rise again when the obstacle is removed. The lengths of the reference and objective arms of the were then matched exactly. A mirror was placed under the microscope objective to act as a total reflector for the purpose of calibrating the interferometer. The oscilloscope was switched to Alternating Current (AC) monitoring mode. The function generator program for the mirror’s oscillation was started. The distance of the reference mirror to its collimator was then varied using the Klinger stepper motor until interference fringes could clearly be seen.

METHODS FIGURE 4: TYPICAL MIRROR INTERFERENCE PACKET

The object to be scanned could now be placed under the objective lens. The reference mirror could then be moved closer to the collimator as least as much as the estimated thickness of the scanning object. The Power Macintosh could then be programmed to take a designated number of data points as well as the speed at which it should progress between data points. Each data point corresponded to a movement of 10 microns; therefore, the scanning depth corresponded to the number of data points times 10 in microns. The function generator program for the lock-in amplifier must also be started and the amplifier itself switched on. The data collector program itself could then be started. A filename for the raw data file must be input. The program would then proceed automatically, taking data and stepping the mirror when necessary. The resulting file, once completed, could then be transferred to the Compaq laptop computer. A custom written LabVIEW application that looked for peaks in each row of raw data (each of which corresponded to one step of the mirror) would be run, thus isolating the lock-in detected maxima for each depth, which indicated the amount of light reflected from that

particular depth. The data from each of those points would be concatenated into a single text file by the program. The resulting file would then be imported using SCION Image, a program that can convert numeric data into pixel color values, and a graphical representation of the reflectivity at each depth, i.e. the OCT image itself, would be produced.

METHODS FIGURE 5: COMPAQ PRESARIO LAPTOP WITH SCION IMAGE AND LABVIEW

Results

RESULTS FIGURE 1: SCAN OF TEMPORARY REFERENCE MIRROR

RESULTS FIGURE 2: SCAN OF STACK OF GLASS COVER SLIPS ON A MIRROR

RESULTS FIGURE 3: SCAN OF MULTI-LAYERED PLASTIC BAG

RESULTS FIGURE 4: SCAN OF RONCHI RULED GLASS TILE

Discussion The OCT system can now successfully scan semi-reflective media in one or two dimensions. One-dimensional scanning can be accomplished quickly and completely under computer control. Two-dimensional scanning still requires constant human operator supervision and intervention between one-dimensional scan “slices.” Figure 1 above is a one-dimensional scan of only a temporary reference mirror. Despite the nearly complete reflection by this mirror, there is a dark band only at one depth. This is precisely how the system is designed to work; it produces a dark pixel only at those depths corresponding the the location of the reflective surface. It is stretched horizontally for ease of viewing; this is not a true two-dimensional scan. Figure 2 represents a piece of plastic cut from a bag obtained from Duke University Stores. It was of the heavy duty variety, not the ordinary grocery bag type. Its multiple layers, though indiscernible to the naked eye, can clearly be marked in this OCT image. This is also a one-dimensional scan that has been stretched horizontally for viewing purposes. Figure 3 is of a stack of glass cover slips on top of a mirror. The each dark stripe is an interface between two pieces of glass. The darkest stripe is the mirror; each stripe below that results from light being trapped inside the glass and reflected back at a delayed time. As above, this scan is one-dimensional only. Figure 4 is a Ronchi ruling, a special glass tile with parallel raised ridges that are very fine and very close together. This is one of the only two-dimensional images produced by this system so far, since so much manual intervention and time is required. On this image, the regions of high reflectivity seem to be separated by areas of low reflectivity. This probably indicates that the raised ridges are not reflective, and therefore leave the regular gap pattern in the image. The system mechanism, therefore, works. However, images cannot yet be taken of actual cortical tissue, because very little light is reflected from the sample. The device is not yet sensitive enough to detect fringes of this subtlety. There appears to be a great deal of noise generated by the amplifier/photodetector assembly. Even when the light source is not turned on, the amplifier produces noticeable noise on the oscilloscope. The noise is enough to impede detection of the weak interference patterns that would be expected from highly scattering objects. The lock-in amplifier also has limitations; although it can be tuned to detect very weak signals, it also isolates unwanted distortions from the noise when its sensitivity is boosted. It can also

overload and cause further distortion when a strong signal is detected while at heightened sensitivity. This lack of ability to detect very weak reflections is the only remaining major obstacle to the successful imaging of a brain lesion. If the signal could be amplified with enough intensity and with low noise, a scan of even a low-reflectivity medium such as a lesion on neural tissue would likely succeed. The continuation of the project will involve this refinement, as well as a few general improvements to the program code that will make the scanning process smoother and faster. A second stepper motor controller would also help to automate the entire scanning process in two dimensions, reducing the potential for human error. A third motor could also be added to scan in three dimensions; however, each additional dimension radically increases the amount of time needed for scanning.

References 1. Post, Amber; Colsher, Michael. “Optical Coherence Tomography for Imaging of Neural Tissue in the Cortex.” Unpublished work obtained from author. 2. Fujimoto, J. G.; Drexler, W.; Morgner, U.; Kärtner, F.; Ippen, E. “Optical Coherence Tomography: High Resolution Imaging Using Echoes of Light.” Optics & Photonics News, 25 – 31, January 2000. 3. Guenther, Robert D. Modern Optics. New York: Wiley and Sons, 1990. 4. Schmitt, Joseph M. “Optical Coherence Tomography: A Review.” IEEE Journal of Selected Topics in Quantum Electronics5, 1205 – 1215, 1999.

Acknowledgements Thanks to Dr. Robert Guenther for his mentorship and for providing this project. Thanks to Amber Post for her constant assistance and for imparting everything she knows about OCT. Thanks to Dean Mary Nijhout and Rebecca Zufall for organizing the program and assisting the research fellows in writing their papers. Special thanks to the Howard Hughes Medical Institute and the Duke University Trinity College of Arts and Sciences for administering the program and providing the funding for it.