A Low Cost Optical Coherence Tomography Machine
Kevin White
EBME 105: Introduction to Biomedical Engineering
Tim Buetler
10/31/2010
Introduction:
Optical Coherence Tomography (OCT) is an emerging non-invasive imaging technique with state of the art designs reaching sub micrometer resolutions (1). While technology is beginning to become widely accepted and used in the developed world, it is out of reach for many third-world countries because it is too expensive for the poorer nations of the world. The goal of this project is to vastly reduce the cost of this technology for glaucoma screening in poorer nations. This project aims to reduce the cost of this OCT machine to $4,000-$6,000 while designing the machine to be relatively small and portable.
OCT is based on the principles of low-coherence interferometry and often employs a Michelson interferometer (2). An interferometer is a device which utilizes lights phase properties to compare the light that returns from two mirrors; one mirror a known distance from the receptor and another mirror an unknown distance. The unknown distance can be calculated by comparing the phases of the light from the unknown length and the light at a known length (2).
In a basic OCT set-up a light source emits a broadband spectrum of light which is carried over fiber optic cable to an evanescent mode coupler. The coupler divides the light evenly, with half of the light directed toward the reference arm and the other half directed to the sample which is to be scanned. The light is reflected back from both the reference arm and the scanning sample and is recorded by a photoreceptor. The data is then translated in a digital signal, processed, displayed by a computer (2).
Since OCT has become a more popular imaging modality, different engineers have tried to eliminate some of the drawback typically associated with OCT such as low scan speed and lack of portability. These different designs have altered some of the basic principle on which the technology was originally based and has led to a few significantly different types of OCT machines. The main difference between the separate designs is the reference arm and how the light that is recorded is processed. The original design for OCT is called Time Domain OCT because the intensity of the light which is recorded is graphed and processed as a function of time. For a functional image, the known length of the reference arm must be changed over the length of one scan to accommodate structures which are at different lengths. In the first OCT machines, this problem was dealt with by simply moving the mirror in the reference arm back and forth. This is a straightforward method, but unfortunately for this project, the piezoelectric motor which would be required proves to be too expensive (3).
All of the other designs for OCT that were considered for this project employ a Fourier transformation which changes the data into a function of time (4). Utilizing this Fourier transformation, engineers have designed other types of OCT. Swept-source OCT was only considered briefly for this project. In swept source OCT, a tunable light source is used so that, instead of changing the location of the reference mirror, the wavelength of the light source is changed (4). It was quickly realized that the price for the tunable light sources were much higher than the budget allowed. Spectral OCT is a popular method of OCT because of its very quick acquisition time. In spectral OCT, a spectrometer is used to read all of the different wavelengths at once instead of both moving the reference mirror and slowly scanning through the separate wavelengths (5). The biggest drawback for the spectral OCT for our project was the cost of the spectrometer. The method that was chosen for this project is called rotational OCT. In this technology, the source light is shown through a diffraction grating which separates the different wavelengths of the light. The light is then reflected off a rotating array of mirrors. The light hits the mirrors at an angle allowing the light of different wavelengths to travel different lengths. When the light is recorded, a Fourier transformation is performed, and the intensity of light can be processed as a function of time (6).
The biggest limiting factor as to which OCT design was used for this project was the price. There are only several components which drive the cost of this technology up, but the prices were significant enough to eliminate several designs from contention. The parts in OCT which drive the price up are the light source, the reference arm, and the photoreceptor (3). Additionally, the price of the processing power that is needed for changing the received light into a usable image is relatively expensive, but competition for the technology has driven the price around as low as it will go.
Time Domain
Time domain OCT was the format first used in early OCT machines. The concepts behind the technology are relatively straightforward and it is an easy technology to fabricate. As described above, the most basic component of time domain OCT is the interferometer. This device splits the light and sends it down both the reference arm and the sample arm. The reference arm is a known length from the photoreceptor while the sample is an unknown length from the photoreceptor. As the image is acquired, the reference arm must be moved either closer to the photoreceptor or away from the photoreceptor. This is required so that the phases of the light returned from the sample arm can be compared with multiple known lengths from the reference arm (7).
There are several advantages to using time domain OCT as the principle design of the low cost OCT system. First, time domain is easy to understand; it is important for all project members to have a complete grasp of the technology involved. The processing involved in time domain OCT is also much simpler than the processing of the other designs. Time domain does not require a Fourier transformation because the data collected is collected as a function of time. The lower processing requirement reduces the cost of the processing equipment involved. The biggest drawback to time domain OCT for our project is the movement of the mirror. Moving the reference mirror quickly is technically challenging. The arm needs to move very quickly and precisely each time it is used. The mirror which is moved cannot vibrate excessively or twist so that the light is reflected away from the detector. The image quality greatly depends on the quality of the reference arm construction. One of the most important factors for time domain OCT is the speed which the in the reference arm moves. The slower the motor moves, the lower the signal to noise ratio and the longer acquisition time that is required (3). The speeds which are required for time domain OCT are too high for the cheap and easily accessible AC and DC motors. Instead, time domain requires the more complex and expensive piezoelectric motor.
A piezoelectric motor uses a material which changes shape when it is exposed to a voltage differential. A voltage is applied to the material and it is forced to change shape. This changing shape forces the motor to spin very rapidly (8). This motor would work very well for our OCT project because it is commercially available and easily attached to the device. The one drawback associated with piezoelectric motors is their cost. The cost of piezoelectric motors makes them out of reach for our low budget model.
Many of the features of time domain make it an attractive option for our low cost project. The design plan is very straightforward and the principles behind the system are easy to comprehend. While the time domain design for OCT theoretically fits the needs of our project very well, the practical limits, namely the cost of the system, make time domain infeasible.
Swept Source
Swept Source OCT works in a very different manner than time domain. The data that is collected from a swept source OCT machine undergoes a Fourier transformation and is therefore considered as an example of Fourier domain OCT (9). In time domain OCT, the reference mirror is moved back and forth, whereas in swept source the reference is stationary. To achieve scans of different depths, swept source OCT uses a tunable laser to vary the frequency which the light source emits. While most OCT machines require a light source with a very broad range of frequencies, swept source demands the opposite (4). A narrower and therefore more specific light source is preferred The varied frequencies of the light source have different penetrating depths and scattering properties which, when processed with a Fourier transformation and other processing techniques, result in an image which has a high resolution and accurately represents the sample which was imaged. The basic setup for swept source OCT is very similar to that of time domain OCT. The light source is connected via fiber optic cables to a coupler which splits the intensity of the light. Half of the light goes down the reference arm while the other half goes to the sample which is to be imaged. The light is reflected back from both of the arms and a photoreceptor detects and records the light which hits it. The procedural difference between swept source OCT and time domain OCT is that in swept source the frequency of the laser is tuned and in time domain the reference mirror is moved (9). The signal processing is significantly more involved for swept source OCT however. When the data is collected in swept source OCT, it cannot be manipulated in its raw form to display a coherent image. Instead, a Fourier transformation must be used. This Fourier transformation changes the data into intensity as a function of time which can then be manipulated to form a recognizable image (4).
As with all of the different methods of OCT, there are several advantages to using swept source OCT. Swept source has the advantage of utilizing a Fourier transformation which is inherently faster than time domain (4). This increased speed decreases the scanning time, which is good for image quality because of decreased motion artifacts. Also, swept source OCT has a better signal to noise ratio than time domain OCT (3). While image quality is not a paramount concern of this project, the images need to be clear enough to be understandable for glaucoma diagnosis. Since cost is the biggest factor for this project, if a higher image quality can be achieved with swept source OCT, the mirrors and lenses can be downgraded to help lower the cost of the machine without bringing the image quality below the acceptable standard.
The main factor which limits the use of swept source OCT is once again the price. While the cost of the set-up is fairly inexpensive, the light source is very expensive. Swept source OCT does not require the use of an expensive CCD camera, however the cost of the tunable laser puts the cost of the system way above our price range. It is very likely that in the future, the prices of lasers in general will decrease, but, until that happens, swept source OCT costs too much to be implemented into the low cost system.
Swept source OCT has many important features which make it an attractive option for designing a low cost OCT system. It has a straightforward system design and is technically simple to build. The image quality is very good and the acquisition time is significantly shorter than that of time domain. However, all of the important positives which are displayed by swept source OCT are overshadowed and rendered useless because of the expensive light source which is required. The cost of the tunable light source makes swept source OCT too expensive for a low cost system. The unrealistic price completely overshadowed all of the benefits offered by a swept source design (9).
Spectral OCT
Another commonly used OCT Fourier domain design is known as spectral domain OCT. Spectral OCT works on the same principle as swept source OCT, but achieves its results in a slightly different manner. In the swept source OCT design, a narrowband light source which could be quickly tuned to different frequencies is desired. In spectral OCT, the opposite is what is required. A very broadband and light source with a low coherence length is what is needed for spectral OCT (5). Instead of the multiple frequencies being scanned at separate times like with swept source, in spectral OCT the light source is shown through fiber optic cables to a coupler and down both the reference and sample arms. When the light returns, instead of passing straight through to a photoreceptor, it must first pass through a spectrometer which separates the light into its spectrum. Because the light is separated into a much broader spectrum, a normal photoreceptor cannot simultaneously record such a broad spectrum. Therefore, a CCD, or Charged-Coupled Device, camera must be used. This camera is able to simultaneously record the entire spectrum of the light that is returned from both the reference and the sample arm. Once the light has returned from both arms, has travelled through the spectrum, and has been recorded by the CCD camera, it is processed by the computer and can eventually be displayed as an image (3). The processing which is required for Spectral domain OCT is similar to that of swept source OCT and requires a Fourier transformation. This Fourier transformation changes the data into intensity as a function of time. Spectral OCT works because the different frequencies of light can all penetrate different lengths into the tissue. When the frequencies returned from the sample arm are separated in the spectrometer, their interference patterns with the light from the reference arm can be used to calculate the distances the light has travelled and can then be constructed into a clear picture (5).
Because both spectral domain OCT and swept source OCT are both example of Fourier domain OCT, they have many of the same advantages and drawbacks. For spectral OCT the acquisition time is much quicker than that of both swept source and time domain (2). Because the spectrometer splits the light into its component frequencies in one step, the machine does not need to wait as the light source is tuned across a spectrum of frequencies like it does in swept source or for the reference mirror to move like it does in time domain. This shorter scanning time is most important for applications which have a very large scanning volume. With glaucoma screening however, the area that needs to be visualized is relatively small. The advantages of the quicker scanning times are still evident; they help with reducing the involuntary eye movements in patients, but it is a much less important factor than the cost. Additionally, spectral OCT benefits from a much improved signal to noise ratio when compared with time domain (5). As explained earlier, this improved sensitivity would leads cheaper lenses and other components which would ultimately bring down the price of the unit.