Typical swept source OCT setup

a parameter that is freely adjustable and has not been taken into account yet. Consider-ing exclusively receiver noise , which is independent of the light incident on the de-tector, the corresponding signal to noise ratio is proportional to . Regarding solely shot noise (proportional to ), the corresponding signal to noise ratio becomes independent of . If one takes into account only photon excess noise

(proportional to , the according signal to noise ratio is inversely propor-tional to .

2.31

The previous findings clearly indicate that the choice of , determining the heterodyne gain in OCT, is important to achieve the best possible sensitivity for OCT imaging.

Typically, the light in the reference arm has to be considerably attenuated resulting in an optimum reflected power on the order of ~100 µW. If the reflected optical reference arm power is to too high, excess noise usually dominates, which reduces the over-all . On the other hand, if the reflected optical reference arm power is too low, the receiver noise typically dominates, also worsening the overall . Note that the application of high-speed photo receivers (high electronic bandwidth ) with sufficient-ly high amplification, which are required for high speed SS-OCT, often results in a poorer receiver noise performance. Hence, in this case, raising the reflected optical ref-erence arm power (larger ) may become necessary. Generally, the OCT system can only achieve shot noise limited operation if the overall value at the intersection point of and is not smaller than . If the measured sensi-tivity (chapter 2.1.4.1) is constant or almost constant over a large range of reference arm attenuation values , this is a clear indication for shot noise limited operation.

Note that dual balanced operation can considerably reduce photon excess noise [70, 71], simplifying the achievement of shot noise limited operation and allowing for higher values of , if necessary. However, dual balancing is not perfect, particularly due to imperfect power balancing over the whole spectral range. Therefore, residual excess noise remains. Furthermore, another noise contribution, that has been neglected so far but which can make a significant contribution in case of dual balancing, is beat noise

 [71]. Beat noise arises if one takes in account parasitic light that is reflected from the sample arm, due to e.g. spurious reflections from the sample arm optics, with a power reflectivity . Typically, and the photon excess noise is proportional to

, where the beat noise is the part of the result which does not cancel out due to a phase difference of even if one assumes perfect dual balanc-ing.

im-provement of wavelength-swept light sources for OCT application made up a large part of this thesis. Possible operation principles and different implementations of wave-length-swept light sources are discussed in chapter 2.2 and in chapter 3.

2.1.6.1 Experimental setup

Light from the wavelength-swept light source first passes an optical circulator and is then split in the reference and the sample arm using a 50/50 fused fiber coupler. In the reference arm light is collimated, reflected at a mirror with adjustable position, and cou-pled back into the single mode fiber. Moving the mirror shifts the point of zero delay in the sample. Additionally, the reference arm light has to be attenuated by a reasonable factor, using for example neutral density filters, in order to ensure an optimal sensitivity (see chapter 2.1.5.6). In the sample arm light exiting the fiber core of the single mode fiber is Rayleigh imaged on the sample, passing two galvanometer mirror scanners (galvoscanners). The second fast scanner moves the beam transversally on the sample

Figure 2.7: Typical SS-OCT setup as used for OCT imaging for this thesis. The interferometer is an optical fiber based Michelson interferometer using dual balanced detection. The two galva-nometer mirror scanners (galvoscanners) enable transversal scanning of the light beam on the sample. Data acquisition with the analog to digital converter card, generation of control signals for both galvanometer scanners using an analog/digital input/output card and data processing for real-time display of two-dimensional cross sectional images (B-scans) are synchronized to an A-scan trigger signal from the wavelength-swept light source. A personal computer with LabVIEW software is used.

optical circulator

1 3

2

50/50 fused fiber

coupler

attenuator

sample

galvoscanners mirror reference arm

sample arm neutral

density filter

polarization controller

photodiodes differential

amplifier

+

-band-pass filter

PC

high-speed digitizer card wavelength

swept light source

control signals for galvoscanner

drivers A-scan

trigger

x y AD I/O Card

(x-direction). For three-dimensional imaging, another slow scanner is scanned in addi-tion to the fast scanner, resulting in a perpendicular, independent movement of the beam on the sample (y-direction). The slow scanner steps to its next position when the fast scanner has completed a B-scan enabling line-by line scanning of the surface. Note that the galvoscanners can be mounted either prior to the last lens (pre-objective-scanning) or after the last lens (post-objective scanning). In case of retinal imaging (see chap-ter 3.2.2), the arrangement of the lenses has to be adapted. Pre-objective-scanning is used and an additional lens is mounted with a distance to the intermediate focus equal to the focal length (see [74]). In this way, during scanning of the galvoscanner, the OCT light beam ideally pivots around a point which is chosen close to the pupil of the eye and is called the pivot point. Therefore, vignetting of the OCT beam by the pupil can be minimized enabling a wide field of view. Note that fiber lengths in the reference and sample arm are matched.

In order to ensure maximum fringe visibility and therefore a maximum interference sig-nal, a crucial point is that the electric fields returning from both interferometer arms and superposing in the fiber coupler must have the same polarization. Since the polarization state in standard single mode fiber is not preserved and can change considerably, a fiber polarization controller (see chapter 2.2.2.2) has to be used in at least one interferometer arm.

Dual balancing is achieved by detecting light from both outputs of the interferometer with a dual balanced photo detector, which consists of two photo diodes and a differen-tial amplifier. The optical circulator redirects the light that is returning to the light source. A variable attenuator is used in order to guarantee best possible power matching at both photodiodes and therefore enable optimum dual balancing. In case of retinal imaging at 1060 nm (see chapter 3.2.2), two further 50/50 fiber couplers are used. The first is replacing the circulator and the second is inserted prior to the attenuator (see [74]). This solution is preferred to the circulator based approach due to comparably high losses in the circulator at 1060 nm and since it can ensure optimally matched pow-er at both photodiodes ovpow-er the whole sweep spectrum allowing for bettpow-er excess noise suppression. The reason is the fact that light is passing components of the same type with almost identical spectral characteristics. Besides equal power levels at the photodi-odes, another crucial factor for optimum dual balancing is matching the optical path lengths between the coupler, where light from both interferometer arms superpose, and the two photodiodes.

After the balanced receiver, the signal is low-pass filtered with a cut-off frequency that is chosen depending on the desired imaging range but should not exceed the sampling frequency in order to avoid aliasing. The signal is digitized with a high speed digi-tizer card. For OCT imaging performed for this thesis, an analog to digital converter card with a maximum sample rate of 400 Msamples/s, 12-bit resolution and an onboard memory of 4 Gbyte was used (GaGe Applied Technologies - model CS12400) and a personal computer with LabVIEW software was utilized. The typically saw tooth like shaped control signals for both galvoscanner drivers were generated with an analog and

digital input/output card. The data acquisition with the high-speed digitizer card, numer-ical processing of the data (in order to generate real-time two-dimensional preview) and the generation of control signals for the galvoscanners were synchronized using an A-scan trigger signal from the wavelength-swept light source.

2.1.6.2 Signal processing and numerical resampling

In order to remove residual DC signal remaining due to imperfect dual balancing, a background trace is typically recorded with the sample arm blocked. This signal is then numerically subtracted from all A-scan fringe traces. Additionally, an electrical high-pass filter can be used to suppress very small frequency components. However, these means cannot improve excess noise suppression.

As already mentioned in chapter 2.1.2.4, the wavenumber over time characteristic of wavelength-swept light sources typically is not linear, which is a problem, since the PSF is broadened if the samples are not equidistantly spaced in frequency. One solution is to develop a swept light source with sufficiently high sweep linearity, as presented in chapter 3.1.1 [10]. Another approach is to apply sampling with uneven time-spacing realized by externally clocking the analog to digital converter [73]. However, the com-mon way is to numerically resample each A-scan data set prior to discrete Fourier trans-formation. For OCT imaging performed for this thesis, this is accomplished as follows:

Firstly, a fringe signal has to be recorded that is corresponding to a single reflection in the sample arm providing not too large fringe frequencies in order to learn about the relative change of optical frequency within the time interval of each A-scan. The change in optical frequency is then directly linked to the change of the phase of the fringe sig-nal. Practically this can be realized by replacing the sample with a mirror. However, normally this procedure has to be repeated occasionally during imaging. Thus, the placing can become cumbersome. There are alternative solutions which require no re-moval of the sample, as for example the use of a separate Mach-Zehnder interferome-ter [73], the utilization of the unused output of a fiber coupler [74] or the application of a beam splitter and an additional mirror in the reference arm, where one optical path can be blocked. In the latter case an autocorrelation signal can be recorded in non-balanced detection [91, 92].

Secondly, the Hilbert transformation of the acquired fringe signal is calculat-ed, which corresponds to a rotation of in the complex plane. Therefore, the analyti-cal representation of the fringe signal reads:

2.32 Consequently, the fringe envelope , and, by using proper phase unwrapping, also the time dependent phase of the fringe signal can be derived.

Finally, with the knowledge of , each A-scan fringe signal trace acquired during OCT imaging can be numerically resampled with means of interpolation algorithms providing equidistant optical frequency intervals for discrete Fourier transformation.

The time interval after which the determination of has to be repeated, since other-wise axial resolution deteriorates, only depends on the stability and repeatability of the wavelength sweeps from the swept light source. In case of the wavelength-swept light sources used in this thesis, this time interval typically exceeded 30 minutes and was usually limited due to thermal drifts in the source (no active feed-back control).