We present a novel and cost-effective technique – interleaved optical coherence tomography (iOCT) – to enhance the imaging speed of swept source OCT systems by acquiring data from multiple lateral positions simultaneously during a single wavelength sweep, using a single detector and a virtually imaged phase array (VIPA) as a multi-band demultiplexer. This technique uses spectral encoding to convert coherence length into higher imaging speed; the speed enhancement factor is independent of the source speed or center wavelength, and the effective A-scan rate scales linearly with sweep speed. The optical configuration requires only a change in the sample arm of a traditional OCT system and preserves the axial resolution and fall-off characteristic of a traditional SS-OCT using the same light source. Using 10kHz, 20kHz and 100kHz sources we provide a first demonstration of image speed enhancement factors of up to 12, 6 and 10, respectively, which yield effective A-scan rates of 120kHz, 120kHz and 1MHz for B-scan imaging, with a sensitivity of up to 82.5 dB. We also show that iOCT can image faster dynamics than traditional OCT B-scan imaging and is capable of 3D biological imaging. The iOCT concept suggests a new route to high-speed OCT imaging for laser developers: that is, by focusing on improving the coherence length and linewidth of existing and emerging sources. Hence, iOCT is a nice complement to ongoing research and commercial efforts to enable faster imaging through development of lasers with faster sweep rates, and offers new hope for existing sources with slow sweep rates and potential for enhancement of coherence length to compete with faster sources to achieve high-speed OCT.
© 2013 Optical Society of America
Optical coherence tomography (OCT)  is a non-invasive, high-resolution technique capable of imaging through turbid media like biological tissue. The prevalent adoption of OCT for clinical use – for example, in ophthalmology – has motivated technological advancements on many fronts, including improvements in speed, resolution, and sensitivity. Importantly, next-generation high-speed OCT systems are poised to significantly improve patient diagnosis and treatment. For example, improvements in the imaging speed of OCT systems can be exploited in clinical settings to sample larger fields of view while decreasing the imaging time per patient, which can positively impact both the patient experience (i.e., reduce imaging time) and the quality of the diagnosis (i.e., enable more comprehensive imaging or improve image quality by reducing motion artifacts). Alternatively, the additional speed could be used to sample the same structures multiple times and, thereby, improve the quality of such images through averaging (e.g., to reduce speckle effects) .
The past decade has witnessed marked improvements in the imaging speed of OCT systems, largely through introduction of faster light sources, in the case of swept-source (SS) OCT [3,4], or faster detectors, in the case of spectral domain (SD) OCT [5,6]. One common thread underlying the majority of breakthroughs in system speed across the breadth of possible optical modalities involves novel strategies of parallel acquisition. Indeed, this is the principle behind the revolutionary paradigm shift of Fourier domain (FD) OCT , in which information from all depths of the sample are acquired simultaneously, in contrast with its predecessor, time-domain OCT . Newer schemes like parallel FD-OCT [8–10] and streak-mode OCT  – which are themselves both forms of SD-OCT – have made further gains in acquisition speed for traditional SD-OCT by employing parallel detection at the level of the B-scan: that is, these schemes collect multiple A-scans simultaneously.
While parallel acquisition has been exploited most cleverly in SD configurations, the size and speed of commercially available detectors are limiting factors for SD-OCT systems. In contrast, SS-OCT systems have relatively modest detector requirements by today’s standards and benefit largely from advances driven by the telecommunications industry. Hence, in modern SS systems, the imaging speed is ultimately limited by the sweep rate of the light source and the digitization rate, rather than the detector. This realization has set the tone for much of modern SS development targeting increases in sweep speed: a plethora of new source designs deploying polygonal scanning mirrors [12–15], Fourier-Domain Mode-Locked phenomena [16–18], frequency-to-time mapping [19,20], akinetic sweeps , and MEMS tunable VCSELs [22–24] have all been introduced within the past 10 years.
Though already generally faster than their SD-OCT counterparts, SS systems, too, stand to gain from parallelization. For example, several groups have achieved gains in A-scan speed for SS-OCT by multiplexing multiple sweeps in the time domain with optical delay lines [15,16]; others have used the spatial domain to map multiple images acquired simultaneously (from different sample depths or polarization states) to different imaging depths [25–27]. One group recently employed a similar spatial-domain multiplexing concept to capture images from different lateral points , although the spacing between the points was fairly large and the multi-fiber probe used to illuminate the sample is complicated to design. Still others have demonstrated such multi-probe systems to acquire simultaneous A-scans from different lateral locations of the sample [16,29], though with the significantly added cost associated with using multiple detectors. Hence a problem with most parallelization efforts for SS-OCT is that they are prohibitively complicated and lack the simple elegance of many parallel SD schemes.
In this work, we introduce a fundamentally different and streamlined optical approach to increasing the speed of SS-OCT systems that emphasizes neither changes in the detector nor the source. Instead, we consider a new configuration – interleaved OCT (iOCT) – in which a redesign of the OCT system architecture allows us to parallelize acquisition of sample data – that is, we acquire data from multiple lateral locations during a single sweep of the laser – using only a single probe. Our technique builds on the idea of spectral-encoded endoscopy (SEE) [30,31] to increase imaging speed, but unlike SEE does not compromise the high axial resolution that is the strength of OCT. In short, iOCT offers a new strategy for high-speed, high-resolution OCT imaging that complements – rather than competes with – existing efforts to break barriers in OCT imaging speed through the development of faster sources. At the same time, it offers new hope for “slow” systems to compete with the imaging rates offered by existing high-speed sources. In what follows, we demonstrate that iOCT is not only faster than traditional SS-OCT, but that this general strategy can be applied to enhance the imaging speed of traditional SS-OCT systems regardless of wavelength or sweep speed.
2. General overview of iOCT
The goal of iOCT is to improve the overall imaging speed of SS systems by acquiring high-resolution data from multiple lateral locations simultaneously with a single detector. To sample multiple lateral locations, we use spectral encoding to associate each wavelength within the bandwidth of the source with a unique lateral location. Hence, multiple lateral locations are encoded within a single spectrum. Because axial resolution in OCT is inversely proportional to the source bandwidth, preservation of the axial resolution requires that we also send a substantial portion of the total source bandwidth to each lateral location. Thus, the general strategy of the iOCT concept is to distribute the spectral bandwidth of a broadband source, as is used typically in OCT, so as to encode multiple lateral locations while maintaining the axial resolution at each location by illuminating each with a total bandwidth spanning nearly the entire source spectrum.
Notably, iOCT, SEE and traditional SS-OCT differ only in the design of the sample arm of a typical OCT system employing a wavelength-swept source [Fig. 1]; the remainder of the OCT engine is unchanged. Whereas SS-OCT acquires data from a single lateral location during the sweep time ts, both SEE and iOCT acquire a full B-scan image (i.e., multiple A-scans from different lateral locations) in this same time; however, as shown in the inset of Fig. 1, the latter two methods differ in the way that they distribute wavelengths to the sample. SEE employs a single grating to distribute the spectrum laterally across the sample in a contiguous fashion. iOCT instead requires that the scanning apparatus (usually a fast galvanometer) that creates the B-scan in SS-OCT be replaced by a multi-band demultiplexer (MBDX). Note that multiple schemes for conceiving an MBDX are possible. The role of the MBDX is to divide the spectrum into small, non-overlapping wavelength bands; the number of bands generated will ultimately set the number of wavelengths sampled at each lateral location. In iOCT, each lateral location is illuminated by a unique set of wavelengths, and all locations are illuminated by the same total number of bands.
The inset of Fig. 1 illustrates schematically the spectral distribution of light for a sample comprising three lateral locations, designated as A, B, and C, when interrogated by SS-OCT, SEE or iOCT. In SS-OCT, the different lateral locations are illuminated sequentially: one lateral location is sampled during each wavelength sweep, lasting time ts. In SEE, the grating causes locations A, B and C to be illuminated sequentially during the course of a single sweep: that is, each lateral location is illuminated by a different set of contiguous wavelengths, and the bandwidth that illuminates each location is a portion of the total bandwidth of the light source. In iOCT, the MBDX causes the lateral locations to be repeatedly illuminated in an interleaved fashion such that all locations are illuminated multiple times during a single wavelength sweep. Note that the wavelengths illuminating a given location are not contiguous, but rather are separated by a fixed interval. Hence, the spectrum of the illumination pattern for a single lateral location resembles a frequency comb , and each lateral location is essentially illuminated by a comb with a different starting wavelength or wavenumber. One difference between this scheme and other schemes employing frequency combs for OCT imaging is that the combs used here are all generated simultaneously so that only a single wavelength sweep is needed to sample all locations. In other work, combs were generated sequentially.
While the time to acquire an A-scan with iOCT is the same as in SS-OCT (that is, one wavelength sweep), the time to generate a B-scan (of length N) with iOCT, as with SEE, now equals the time to generate an A-scan. Thus, the iOCT configuration can image a B-scan or, by extension, a 3D volume, N-times faster than traditional OCT. Moreover, since the total bandwidth illuminating each lateral location is nearly that of the entire source bandwidth, all positions maintain the same axial resolution as in a traditional SS-OCT system. Also, since all positions are sampled with combs having the same frequency spacing, their fall-off characteristics are identical.
3. Experimental design
3.1 Optical design
We modified the sample arm of a traditional SS-OCT system to implement iOCT, as shown in Fig. 2. The sample arm of the modified system utilizes as the MBDX a custom-designed, solid glass virtually imaged phased array (VIPA) (thickness = 0.953mm, Light Machinery). A VIPA is essentially a tilted Fabry-Perot etalon with an anti-reflective coating over its entrance window; the operation and design equations for a VIPA have been described elsewhere [33,34]. Collimated light entering the sample arm is first line-focused onto the VIPA using a cylindrical lens. The output at Image Plane 1 (IP1) is a line-focused beam orthogonal to the line-focusing axis of the cylindrical lens and possessing the desired spectral distribution (interleaved wavelengths) due to the intentional overlap of multiple diffraction orders induced by the VIPA. In general, the VIPA can overlap hundreds of diffraction orders due to its high angular dispersion. In our experimental system, light from IP1 was relayed to a second image plane (IP2), thereby providing additional working distance to incorporate folding mirrors for standard OCT imaging. This optical relay system further provided flexibility to adjust the lateral resolution of the system and the lateral range of the line-focused beam for iOCT by choice of the focal lengths of lenses L2 and L3. The sample was placed at IP2. The forward propagating optical loss from the interferometer to the sample, including through the VIPA, is ~4dB. The dominant system loss occurs during back-coupling through the VIPA: the total back-coupling loss through the system is ~15.6dB.
The same sample arm was coupled to one of two OCT swept source engines operating in balanced detection mode. The first comprises a source (HSL-2100-ST, Santec) with a center wavelength near 1310nm and an optical bandwidth of 150nm. The sweep rate of this system could be adjusted to 10kHz or 20kHz with the turn of a knob. The second source (SSOCT-1060, Axsun) has a center wavelength of 1064nm, an optical bandwidth of 100nm and sweeps at 100kHz. The coatings for the VIPA were designed to operate over both wavelength ranges, and the same VIPA was used with both engines. The power on the sample was 1.2mW and 0.7mW, and the sensitivity was measured to be 78.5dB and 82.5dB, as calculated from the ratio of the signal peak intensity versus the noise RMS (root mean square), for the Santec and Axsun systems, respectively. The decreased sensitivity for the Santec system was mostly due to the additional diffraction orders to the non-overlapping spatial regions in IP1, which was not the case with 1060nm system (see Discussion).
The free spectral range (FSR) of the VIPA is a function of its geometry (thickness), and it ultimately determines the spacing between adjacent wavelength samples for a given lateral location (that is, the spectral length of a wavelength band). This spacing determines the sampling frequency and consequently the ranging depth of the system (we use the term ranging depth to refer to the maximum depth that can be imaged without aliasing, although it is clear that other factors, such as tissue scattering, may impose additional limits on the maximum depth from which photons can be collected). The total number of possible points in the A-scan for a single lateral location is thus given by the source bandwidth divided by the FSR. Because of the limited number of resolvable points (~1000-2000) in the light sources we owned due to their relatively short coherence lengths, we designed the VIPA to have an FSR of 107.5GHz (0.61nm-wavelength spacing) or 105.6GHz (0.40-nm wavelength spacing) in order to provide 244 or 251 points in the A-scans for the Santec or Axsun sources, respectively. The number of points differs for the two sources because the FSR at a fixed geometry is a function of the center wavelength of the source. Also, the linewidths of the sources are different, leading to a different number of resolvable points for each source. Though the number of points used here is small, it is sufficient to prove the principle of the iOCT method, and would be easily improved using sources with longer coherence lengths. The parameters of the VIPA can be customized to work also with light sources with a higher number of resolvable points.
The smaller of the linewidth of the VIPA (FSR / finesse) or the linewidth of the source determines the number of lateral locations that are multiplexed onto the spectrum. Essentially, this number equals the number of resolvable spectral points that fall within a single FSR, and for which there is no cross-talk. We designed the VIPA to have a finesse of nearly 103.3 at 1310nm (even higher values of > 200 are possible); hence, our systems were limited by the linewidth of the source, rather than the VIPA. The sources at 10/20kHz and 100kHz had coherence lengths of 5mm and 10mm, and corresponding linewidths of 26.2GHz (0.15nm) and 13.3GHz (0.05nm), respectively. Based on the FSR of the VIPA at the center wavelengths for each of our sources, the numbers of resolvable lateral points in our iOCT B-scans (calculated by FSR of VIPA / linewidth of laser) are 4 and 8 for the Santec and Axsun sources, respectively (Table 1). This number may differ from the number of A-scans collected per B-scan depending on the digital sampling rate.
Some aspects of the system construction beg notice. Firstly, because of the intrinsic dispersive characteristics of the VIPA, it is possible for a single wavelength to appear at more than one lateral location. This situation must be avoided, however, to preserve the integrity of the spectral encoding scheme. Judicious adjustment of the output angle range by inclusion of slit apertures at the entrance pupil of the VIPA, or proper selection of the focal length of the cylindrical lens preceding the VIPA, are two sufficient solutions to address this, though they contribute to reducing the system efficiency. Secondly, the accuracy of the spectral encoding is tied to the stability of the wavelength mapping, which depends on the repeatability of the sweep profile of the laser source (e.g., timing jitter). Instabilities of the source will manifest as the blurring of lateral structures (i.e., reduced lateral resolution) in the same way as might cause blurring of wavelength channels or data at different depths (i.e., degraded fall-off performance) in a standard OCT system. One way to ensure stability is to use a k-clock to trigger the acquisition . Note that the k-clock output should be chosen to enable sufficient sampling of the A-scan; otherwise, it can become a limiting factor in setting the number of resolvable points. For example, the number of resolvable points for the Axsun system decreases from eight to six when the built-in k-clock is used. Alternatively, jitter can be fixed using additional hardware for optical clocking  or real-time phase calibration. In our source, jitter was observed to be rare (1:10 sweeps) and was ~one pixel with the k-clock.
3.2 Data processing
The photodetectors we used had an electronic bandwidth of 80 MHz (Model 1817, New Focus) or 800 MHz (Model 1617, New Focus) and signals were digitized (Alazartec) at sampling rates of 50 MSa/s or 500 MSa/s for the Santec or Axsun sources, respectively. All iOCT and SS-OCT data were processed using customized Matlab software. Figure 3 describes the processing steps for iOCT. In brief, the captured, raw interferogram, which multiplexes interferograms from all lateral locations, first underwent DC removal. Because the internal clock of the digitizer was used to sample the signal for the Santec source, cubic interpolation was then used to convert wavelength data into linearly spaced wavenumbers based on pre-saved calibration data. For the biological images in Section 4.2, the Axsun source data were externally clocked using the k-clock output of the source; thus, no resampling was required. A Hanning window was then applied to the multiplexed interferogram to reduce the sidelobes in the A-scan profile, and an interpolation step was performed to oversample the data in some cases (i.e., for in vivo data) by a factor 2. The oversampling step reduces distortions that could manifest after demultiplexing due to differences in the amplitudes of adjacent lateral positions. Next, separate interferograms for each lateral locations are generated by first applying a simple sequence of circular shifts on the multiplexed interferogram and then down-sampling using an array of indices corresponding to the positions of the wavelengths associated with a given lateral location. For example, a raw iOCT interferogram that multiplexes two lateral locations onto a spectrum comprising 1000 samples would map the odd-numbered indexes to one position and the even-numbered indexes to the other position. Finally, numerical dispersion compensation  was implemented on the individual A-scans and an inverse-Gaussian weighting function (“Envelope correction”) was applied to account for differences in amplitude caused by the spectral response function of the VIPA. This spectral response is a consequence of the Gaussian-weighted intensity of the input beam to the VIPA, which is a result of the Gaussian beam-shaped output of the cylindrical lens. All the same processing steps were done for traditional OCT data, with the exception of the demultiplexing step.
4.1 Validation of operation of the iOCT system
The design of the iOCT system requires that each lateral location be associated with a discontinuous set of discrete wavelengths (i.e., a frequency comb); furthermore, the set of wavelengths associated with each location must be unique and span nearly the entire bandwidth of the source in order to maintain the same resolution as in traditional OCT. Figure 4 illustrates the output of the iOCT system for a reflective mirror covered by a slit, positioned at IP2, whose width was set to be smaller than the lateral resolution of the system. The reference arm was blocked to remove complications due to interference; hence, the peaks indicate which wavelengths are associated with a single lateral location. The same sample arm was used to acquire data at three different speeds, validating the speed scalability of iOCT.
Figures 4(a), 4(c) and 4(e) show that the wavelength data from a single lateral location spans the entire bandwidth of the source, as desired. The shape of the spectrum is a combination of the spectral output of the source and the spectral responses of various system components, including the VIPA itself. The red dotted region is enlarged in Figs. 4(b), 4(d) and 4(f). The solid black and dashed blue line represents data reflected from different lateral locations, which verifies that the wavelengths associated with each lateral location are unique.
4.2 Imaging with iOCT
iOCT acquires B-scan images faster than SS-OCT
Figure 5 shows B-scan images acquired from a test sample comprising a moveable mirror positioned behind a stationary glass coverslip. The position of the coverslip is fixed in all images while the position of the mirror (arrows) is varied laterally and in depth. Each frame of the images was acquired during a single sweep of the laser and verifies that iOCT can acquire B-scans with variable image content. The acquisition time for each image was 100μs, 50μs and 10μs, respectively. The same-sized images would have required 1200μs, 300μs, and 100μs, respectively, with the standard OCT system, since data were digitally sampled by factors of 12, 6 and 10 in this case, though the number of optically resolvable lateral points is only 4, 4, or 8 in these cases. The speed enhancement factor due to iOCT depends on the number of lateral locations sampled in the B-scan. Thus, in this case, the speed enhancements are factors of 12, 6 and 10, respectively.
iOCT preserves the axial resolution and fall-off characteristic of SS-OCT
To demonstrate the ability of iOCT to preserve the axial resolution of traditional OCT, Fig. 6 compares A-scans taken of a single mirror using the same system parameters for both iOCT [Figs. 6(a) and 6(b)] and standard OCT [Fig. 6(c)] imaging modes at 100kHz. After applying dispersion compensation, the axial resolutions are 7.5μm, 7.2μm, and 7.2μm for systems at 10kHz, 20kHz and 100kHz, respectively. These values compare favorably with those obtained for standard OCT (7.9 μm, 7.5μm, and 7.3μm, respectively). Theoretical axial resolutions of the 10/20kHz and 100kHz sources are 6.9μm and 6.2μm with 110nm and 75nm FWHM bandwidth, respectively (the bandwidths were reduced compared to the source optical bandwidth because of changes in the spectral envelope from the optics and the effect of the Hanning window). As expected, the axial resolution of iOCT is nearly the same as in traditional OCT; the variation between these numbers is due to the imperfection of dispersion compensation.
Because iOCT maps the position of all A-scans to the beginning of the ranging depth but does not change the spectral resolution (linewidth) of the source, the sensitivity fall-off over the ranging depth is also preserved. Hence, the amount of fall-off is minimal throughout the iOCT imaging depth, as shown in Figs. 6(a)-6(b), whereas the fall-off increases as expected for larger depths in standard OCT as shown in Fig. 6(d), which shows the full imaging range for the traditional OCT system. Note that Fig. 6(c) is essentially a magnified image of the data to the left of the vertical red line in Fig. 6(c).
iOCT captures faster dynamic events than SS-OCT
The faster speed of iOCT enables it to capture dynamics that cannot be observed with an equivalent SS-OCT system. When running in B-scan mode, traditional OCT systems can measure dynamic events up to half the B-scan frame rate (due to the requirements for sampling at the Nyquist rate). Because iOCT forces the B-scan rate to equal the A-scan rate, iOCT can measure B-scan dynamics at up to half the A-scan rate. This represents rates of 5kHz, 10kHz and 50kHz for the systems we constructed. Figure 7 shows the frequency response of three iOCT systems to the motion of a mirror mounted on a piezoelectric transducer driven with a sinusoidal voltage signal at frequencies ranging from 600Hz to 64.1kHz. In each case, the amplitude of the motion of the mirror was constant and less than 1μm; only the oscillation frequency was varied. The displacement of the mirror was extracted from the phase of the iOCT signal. Figure 7(a) shows representative phase traces as a function of time for the 10kHz system; the traces have been displaced for ease of viewing. In Figs. 7(b)-7(d), it is clear that frequencies above the Nyquist limit for a given A-scan rate (red) are aliased to lower frequencies.
These data were impossible to collect with the traditional OCT system because systems employing galvanometer scanners are additionally limited by the small-signal step response of the galvos. Hence another advantage of iOCT is that it provides an alternative to high-speed scanning that overcomes the limitations of mechanical scanners, which become increasingly significant as the base sweep speed of the light source increases. Costly high-speed scanning alternatives to galvanometers include resonant scanners and acousto-optic modulators .
iOCT can perform biological imaging
Figure 8 demonstrates the ability of iOCT to perform 3D biological imaging. Figures 8(a)-8(f) show iOCT B-scan slices of a pepper seed; the arrow denotes a location with clear differences in the appearance of the images from different lateral positions. Mirror image-like artifacts in the images are possibly due to multiple scattering in the VIPA. Figures 8(g)-8(i) show three representative B-scans (from a set of 6) of the finger of the author acquired simultaneously with iOCT using the 100kHz source at an effective A-scan rate of 600kHz (6 lateral points). The lateral dimension of the VIPA dispersion spanned 140μm and was controlled by the choice of lenses L2 and L3 in Fig. 2. The arrow highlights the differential appearance of a sweat duct in each due to the slightly different lateral positions captured. Figures 8(j) and 8(k) show orthographic slices and a 3D rendering, respectively, of the volumetric data set of a pepper seed (400 × 480 A-scans) that was collected in 0.32 sec.
5. Discussion and conclusions
We have introduced a new paradigm for high-speed imaging with SS-OCT – interleaved OCT – that enables faster imaging and nearly equivalent axial resolution to the traditional OCT system. Moreover, as we have shown, the iOCT concept is wavelength-independent and the “effective A-scan rate,” which is the product of the laser sweep rate and the number of lateral locations sampled with iOCT, scales linearly with the sweep rate. Because of the parallel nature of this scheme, iOCT can be implemented to make state-of-the art, high-speed swept source systems effectively even faster. Given the proper choice of dielectric coatings, the same MBDX can even be used for multiple wavelength regimes, as was done for the VIPA we demonstrated. As a bonus, the system design for iOCT is relatively simple, requiring only a slight change in the design of the traditional OCT sample arm. Hence it integrates nicely with existing OCT systems and also complements other technological advances made for OCT that implement changes in other parts of the system (e.g., designing faster wavelength-swept lasers).
What iOCT gains in speed, however, it loses in ranging depth. That is, deeper structures are aliased more quickly with iOCT than with traditional OCT. For standard light sources with small ranging depths, such as the commercial sources we used here, this trade-off may or may not be desirable. The ranging depth achieved with iOCT is ultimately determined by the number of resolvable points in a given source. This number is critically linked to the instantaneous linewidth of the source: the number of resolvable points RSRC is equal to the product of the number of points one desires per A-scan (i.e., frequency samples) and the number of lateral locations that can be multiplexed with iOCT. In effect, the multiplexing power of a given source for use with iOCT is nearly the coherence length of the source divided by the desired ranging depth.
Actually, there are several factors that can limit the number of resolvable points in an iOCT system. Assuming RSRC is sufficiently high, one potential bottleneck can be the sampling rate of the digitizer. Note that the sampling rate of the digitizer can be artificially reduced if a k-clock is used (this was the case for our Axsun source; iOCT can also be done without the k-clock but requires more processing to improve phase stability [38,39]). The required sampling rate for iOCT depends on the number of resolvable points and the speed at which they will be sampled. The required digitizer speed is therefore equivalent to what one would need to properly sample a traditional system whose A-scan rate was equivalent to the effective A-scan rate achieved with iOCT. For example, if one would choose to use a 100MHz detector to perform SS-OCT with a 100kHz source, one would instead need a 1GHz detector to perform iOCT with the same source when sampling 10 lateral points simultaneously. Such detectors are commercially available, and are similar to what one would use to achieve long ranging depth imaging with sources with long coherence lengths.
The VIPA also has a characteristic RVIPA that is based on its design parameters (e.g., the reflectivity of the surfaces and its thickness). The value of multiplexing with iOCT may be limited by the smaller of RVIPA or RSRC. In our case, we were limited by RSRC. Due to the limited linewidth of the sources we had, which resulted in small total numbers of resolvable points in those sources, here we demonstrated the iOCT concept by multiplexing only a few lateral locations onto the spectrum to yield A-scans comprising nearly 250 axial points each. The resulting, achievable axial resolution is reasonable, 7.2-7.5μm and 0.7mm for the Santec source (10 or 20kHz), but the imaging depths are shallow: 7.2μm and 0.68mm for the Axsun source (100kHz). Fortunately, new light sources are emerging with ranging depths that are very long for standard OCT imaging. For example, new VCSEL light sources have been demonstrated with ranging depths exceeding 155mm . This additional ranging depth, when used purely for imaging deeply, is not always practically useful for two reasons: 1) in scattering media, light penetration is ultimately limited by scattering, not ranging depth; 2) long ranging depths force a trade-off between depth of focus and lateral resolution that may not be appropriate for some samples, even those where a large portion of the ranging depth is expected to be occupied by free space (or a non-scattering medium). Because the iOCT concept allows one to convert excess ranging depth into a new mechanism for increased imaging speed, sources with “unusable” ranging depth are ideal candidates to use with iOCT. Hence, with iOCT, a VCSEL with a 100-mm ranging depth and 200-kHz A-scan rate could be used to perform imaging of the eye with a 10-mm ranging depth and effective A-scan rate of nearly 2MHz (i.e., a speed enhancement of 10, which is achievable with a VIPA with a 50-pm FSR and a linewidth of 5pm).
As with all high-speed OCT techniques, the optical power delivered to the sample via iOCT should increase with speed to maintain image quality. Keep in mind, however, that the power is spread over a larger area than in the case of SS-OCT, which means that higher power sources could be used. Also, when used with systems with low sweep speeds, the longer dwell time at any given lateral position can lead to higher SNR with less demand for sample power. For biological samples, the maximum power may be limited by ANSI standards. The calculation for the maximum allowable power in iOCT should be similar to that implemented for systems with multiple sample probes.
We make one final note about this first demonstration and the potential for improvement of the system. Though modestly sufficient for biological imaging, the sensitivity we demonstrated here can be drastically improved in future work with several modifications to the system. For example, ~6dB of sensitivity can be gained for 1300-nm imaging by redesigning the system to use a Mach Zender interferometer, as used in ; another ~3dB can come by switching to a source with higher power (e.g., VCSEL source 36mW), and a final ~2dB is expected from using galvanometers with larger scanning mirrors, as our current galvos caused some clipping of the dispersed beam which resulted in loss of power.
Moreover, the design of the VIPA has clear implications for the sensitivity. As with any Fabry-Perot etalon, the VIPA has both forward and backward loss due to the high surface reflectivities used. The forward loss we experienced in our system was mainly caused by diffraction at the edge of the anti-reflection-coated VIPA window: such losses do not appear in theoretical calculations and are likely a fabrication artifact. The loss, which resulted in power leakage to undesirable diffraction orders, was less severe for 1060nm than 1310um, and necessitated the inclusion of the slit in Fig. 2. We have observed that air-spaced VIPAs do not exhibit such diffraction losses. The backward loss is due to the rejection of the back-coupled photons at the second surface of the VIPA. The reflectivity of the second surface of our VIPA was designed to be 97%, leading to an unavoidable loss of nearly 13dB. However for practical implementation of iOCT one does not need a high finesse. Since the number of desired resolvable lateral points is equal to the required VIPA finesse, a reasonable finesse of 10 – which would yield a 10-fold improvement in imaging speed – can be achieved using only 75% reflectivity for this second surface, leading to back-coupling losses of only ~6 dB. Such a VIPA would exhibit a 7-dB improvement over our current implementation.
In conclusion, the iOCT concept suggests and motivates a new route to high-speed imaging with OCT for laser developers: that is, by improving the coherence length and linewidth of the source. This new imaging scheme is another step towards real-time volumetric imaging with OCT and dovetails nicely with ongoing work to create extremely high-speed swept-wavelength sources. Moreover, the completely complementary nature of this strategy means that it can be used in combination with other demonstrated strategies to increase speed through time- or space-division multiplexing in one or two dimensions, and as such could potentially enable full 3D volumetric imaging without any mechanical scanners.
This project was funded by Samsung Advanced Institute of Technology (SAIT).
References and links
1. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, “Optical coherence tomography,” Science 254(5035), 1178–1181 (1991). [CrossRef]
5. A. F. Fercher, C. K. Hitzenberger, G. Kamp, and S. Y. El-Zaiat, “Measurement of intraocular distances by backscattering spectral interferometry,” Opt. Commun. 117(1-2), 43–48 (1995). [CrossRef]
6. M. Wojtkowski, R. Leitgeb, A. Kowalczyk, T. Bajraszewski, and A. F. Fercher, “In vivo human retinal imaging by Fourier domain optical coherence tomography,” J. Biomed. Opt. 7(3), 457–463 (2002). [CrossRef]
7. E. A. Swanson, J. A. Izatt, M. R. Hee, D. Huang, C. P. Lin, J. S. Schuman, C. A. Puliafito, and J. G. Fujimoto, “In vivo retinal imaging by optical coherence tomography,” Opt. Lett. 18(21), 1864–1866 (1993). [CrossRef]
8. D.-H. Choi, H. Hiro-Oka, K. Shimizu, and K. Ohbayashi, “Spectral domain optical coherence tomography of multi-MHz A-scan rates at 1310 nm range and real-time 4D-display up to 41 volumes/second,” Biomed. Opt. Express 3(12), 3067–3086 (2012). [CrossRef]
9. T. Bonin, G. Franke, M. Hagen-Eggert, P. Koch, and G. Hüttmann, “In vivo Fourier-domain full-field OCT of the human retina with 1.5 million A-lines/s,” Opt. Lett. 35(20), 3432–3434 (2010). [CrossRef]
10. R. N. Graf, W. J. Brown, and A. Wax, “Parallel frequency-domain optical coherence tomography scatter-mode imaging of the hamster cheek pouch using a thermal light source,” Opt. Lett. 33(12), 1285–1287 (2008). [CrossRef]
11. R. Wang, J. X. Yun, X. Yuan, R. Goodwin, R. R. Markwald, and B. Z. Gao, “Megahertz streak-mode Fourier domain optical coherence tomography,” J. Biomed. Opt. 16(6), 066016 (2011). [CrossRef]
13. S. M. R. Motaghian Nezam, “High-speed polygon-scanner-based wavelength-swept laser source in the telescope-less configurations with application in optical coherence tomography,” Opt. Lett. 33(15), 1741–1743 (2008). [CrossRef]
14. S. H. Yun, C. Boudoux, G. J. Tearney, and B. E. Bouma, “High-speed wavelength-swept semiconductor laser with a polygon-scanner-based wavelength filter,” Opt. Lett. 28(20), 1981–1983 (2003). [CrossRef]
15. W.-Y. Oh, B. J. Vakoc, M. Shishkov, G. J. Tearney, and B. E. Bouma, “>400 kHz repetition rate wavelength-swept laser and application to high-speed optical frequency domain imaging,” Opt. Lett. 35(17), 2919–2921 (2010). [CrossRef]
16. W. Wieser, B. R. Biedermann, T. Klein, C. M. Eigenwillig, and R. Huber, “Multi-megahertz OCT: High quality 3D imaging at 20 million A-scans and 4.5 GVoxels per second,” Opt. Express 18(14), 14685–14704 (2010). [CrossRef]
17. R. Huber, M. Wojtkowski, and J. G. Fujimoto, “Fourier domain mode locking (FDML): A new laser operating regime and applications for optical coherence tomography,” Opt. Express 14(8), 3225–3237 (2006). [CrossRef]
18. M. Gora, K. Karnowski, M. Szkulmowski, B. J. Kaluzny, R. Huber, A. Kowalczyk, and M. Wojtkowski, “Ultra high-speed swept source OCT imaging of the anterior segment of human eye at 200 kHz with adjustable imaging range,” Opt. Express 17(17), 14880–14894 (2009). [CrossRef]
21. M. P. Minneman, J. Ensher, M. Crawford, and D. Derickson, “All-semiconductor high-speed akinetic swept-source for OCT,” Proc. SPIE 8311, 831116, 831116-10 (2011). [CrossRef]
22. V. Jayaraman, J. Jiang, B. Potsaid, G. Cole, J. Fujimoto, and A. Cable, “Design and performance of broadly tunable, narrow line-width, high repetition rate 1310nm VCSELs for swept source optical coherence tomography,” Proc. SPIE 8276, 82760D, 82760D-11 (2012). [CrossRef]
23. V. Jayaraman, J. Jiang, H. Li, P. J. S. Heim, G. D. Cole, B. Potsaid, J. G. Fujimoto, and A. Cable, “OCT imaging up to 760 kHz axial scan rate using single-mode 1310nm MEMS-tunable VCSELs with > 100nm tuning range,” CLEO:2011-Laser Applications to Photonic Applications 1–2 (2011).
24. B. Potsaid, V. Jayaraman, J. G. Fujimoto, J. Jiang, P. J. S. Heim, and A. E. Cable, “MEMS tunable VCSEL light source for ultrahigh speed 60kHz - 1MHz axial scan rate and long range centimeter class OCT imaging,” Proc. SPIE 8213, 82130M, 82130M-8 (2012). [CrossRef]
25. A.-H. Dhalla, D. Nankivil, T. Bustamante, A. Kuo, and J. A. Izatt, “Simultaneous swept source optical coherence tomography of the anterior segment and retina using coherence revival,” Opt. Lett. 37(11), 1883–1885 (2012). [CrossRef]
26. V. X. D. Yang, N. Munce, J. Pekar, M. L. Gordon, S. Lo, N. E. Marcon, B. C. Wilson, and I. A. Vitkin, “Micromachined array tip for multifocus fiber-based optical coherence tomography,” Opt. Lett. 29(15), 1754–1756 (2004). [CrossRef]
27. K. König, M. Speicher, R. Bückle, J. Reckfort, G. McKenzie, J. Welzel, M. J. Koehler, P. Elsner, and M. Kaatz, “Clinical optical coherence tomography combined with multiphoton tomography of patients with skin diseases,” J Biophotonics 2(6-7), 389–397 (2009). [CrossRef]
29. M. K. K. Leung, A. Mariampillai, B. A. Standish, K. K. C. Lee, N. R. Munce, I. A. Vitkin, and V. X. D. Yang, “High-power wavelength-swept laser in Littman telescope-less polygon filter and dual-amplifier configuration for multichannel optical coherence tomography,” Opt. Lett. 34(18), 2814–2816 (2009). [CrossRef]
32. T. Bajraszewski, M. Wojtkowski, M. Szkulmowski, A. Szkulmowska, R. Huber, and A. Kowalczyk, “Improved spectral optical coherence tomography using optical frequency comb,” Opt. Express 16(6), 4163–4176 (2008). [CrossRef]
33. S. Xiao, A. M. Weiner, and C. Lin, “A dispersion law for virtually imaged phased-array spectral dispersers based on paraxial wave theory,” IEEE J. Quantum Electron. 40(4), 420–426 (2004). [CrossRef]
35. H. C. Hendargo, R. P. McNabb, A.-H. Dhalla, N. Shepherd, and J. A. Izatt, “Doppler velocity detection limitations in spectrometer-based versus swept-source optical coherence tomography,” Biomed. Opt. Express 2(8), 2175–2188 (2011). [CrossRef]
36. W. Choi, B. Potsaid, V. Jayaraman, B. Baumann, I. Grulkowski, J. J. Liu, C. D. Lu, A. E. Cable, D. Huang, J. S. Duker, and J. G. Fujimoto, “Phase-sensitive swept-source optical coherence tomography imaging of the human retina with a vertical cavity surface-emitting laser light source,” Opt. Lett. 38(3), 338–340 (2013). [CrossRef]
37. M. Wojtkowski, V. J. Srinivasan, T. H. Ko, J. G. Fujimoto, A. Kowalczyk, and J. S. Duker, “Ultrahigh-resolution, high-speed, Fourier domain optical coherence tomography and methods for dispersion compensation,” Opt. Express 12(11), 2404–2422 (2004). [CrossRef]
38. B. Braaf, K. A. Vermeer, V. A. Sicam, E. van Zeeburg, J. C. van Meurs, and J. F. de Boer, “Phase-stabilized optical frequency domain imaging at 1-µm for the measurement of blood flow in the human choroid,” Opt. Express 19(21), 20886–20903 (2011). [CrossRef]
40. I. Grulkowski, J. J. Liu, B. Potsaid, V. Jayaraman, J. Jiang, J. G. Fujimoto, and A. E. Cable, “High-precision, high-accuracy ultralong-range swept-source optical coherence tomography using vertical cavity surface emitting laser light source,” Opt. Lett. 38(5), 673–675 (2013). [CrossRef]