Swept source optical coherence microscopy (OCM) enables cellular resolution en face imaging as well as integration with optical coherence tomography (OCT) cross sectional imaging. A buffered Fourier domain mode-locked (FDML) laser light source provides high speed, three dimensional imaging. Image resolutions of 1.6 μm × 8 μm (transverse × axial) with a 220 μm × 220 μm field of view and sensitivity higher than 98 dB are achieved. Three dimensional cellular imaging is demonstrated in vivo in the Xenopus laevis tadpole and ex vivo in the rat kidney and human colon.
© 2007 Optical Society of America
Confocal microscopy provides an optical sectioning ability in thick samples and is a powerful technique for biological and biomedical studies . The sectioning ability in confocal microscopy depends heavily on tight focusing with high NA objectives and rejection of out-of-focus light with a pinhole in the conjugate focal plane. Therefore, it is vulnerable to aberration and multiple scattering. Optical coherence microscopy (OCM) improves the resistance to aberration and multiple scattering by combining high-sensitivity, coherence-gated detection with standard confocal microscopy, leading to greater image contrast and imaging depth in highly scattered tissues [2–5]. Moreover, the sectioning ability using coherence gating mainly depends on the spectrum of the light source, reducing the requirement for high NA optics and therefore leading to easier miniaturization for endoscopic studies.
Traditionally, OCM is implemented with time domain detection [2–5], which allows high speed en face imaging in order to eliminate motion artifacts. However, time domain systems have a number of features that make the optical system undesirably complex. First, a rapid phase modulation scheme in the reference arm is required for time domain OCM systems. Several methods have been proposed, including rapid scanning optical delay lines , electro-optic modulators [6, 7], and acousto-optic modulators [8, 9]. All of these methods require sophisticated optical designs and specific issues such as scanner synchronization and dispersion compensation need to be carefully addressed. Second, spatial overlap between the coherence gate and confocal gate is critical to ensure optimal image quality in highly scattered tissues. Due to the inhomogeneous nature of tissues, the matching between the two gates generally can not be set a priori without real time measurement. In addition, when a fiber based endoscope system is used, any variation of stress in the fiber can introduce gate mismatch and degrade the image quality. Therefore, a feedback loop and a fast coordination algorithm are required to maintain optimal image quality during the imaging period. Finally, due to the limited field of view, usually a few hundred μm, OCM itself can suffer from sampling error in clinical studies. One solution to this limitation is to use optical coherence tomography (OCT) for large scale survey and to conduct OCM only in the regions where abnormalities are detected using OCT. An imaging modality which incorporates high speed OCT and OCM is thus desirable. However, due to differences between time domain OCT and OCM system designs, it is challenging to achieve both high speed OCT and OCM in one system.
Recent work has shown that spectral / Fourier domain detection enables OCT imaging with dramatically improved speed and sensitivity over conventional time domain detection [10–12]. Spectral / Fourier domain OCM has also been investigated [13–15]. Different from time domain OCM, which acquires only a single en face image, spectral / Fourier domain OCM generates an image by acquiring an entire 3D volume and rendering the en face plane. Importantly, and in contrast to OCT, there is no inherent sensitivity advantage of spectral / Fourier domain OCM compared to time domain OCM. The extremely high data rates necessary to acquire an entire volume with spectral / Fourier domain OCM result in reduction of pixel exposure / dwell time for any given pixel in the en face image, which effectively negates the sensitivity advantage. However, OCM and OCT using spectral / Fourier domain detection has the advantage of sharing the same optics in the reference arm, which makes integration of the two techniques relatively straightforward. In addition, spectral / Fourier domain detection measures optical signals from all depths simultaneously and therefore the need for coherence and confocal gate coordination is significantly relaxed. An en face OCM image at a depth matched to the confocal gate can be digitally extracted from the entire 3D dataset. Similarly, a confocal image with reduced speckle can be generated by summing the dataset in the axial direction. Disadvantages of spectral / Fourier domain detection include the need for a high-speed camera and spectrometer, with typical camera readout rates and spectrometer losses limiting the system speed to ∼29,000 axial scan per second. This corresponds to >2 seconds acquisition time to generate en face images with 256×256 pixels. Thus the image quality will suffer from motion artifacts when in vivo studies are performed.
Swept source / Fourier domain detection has similar advantages to spectral / Fourier domain detection, but does not require a spectrometer and line scan camera and therefore has higher detection efficiency and higher speed. Different swept source designs have been proposed and implemented in OCT [16–20]. The Fourier domain mode-locked (FDML) laser is a new type of frequency swept laser that is especially promising for high speed OCT imaging [19, 20]. By using a cavity with a long optical delay line and scanning a high finesse, band-pass filter synchronously with the cavity round-trip time, the need to build up lasing from spontaneous emission is circumvented and very high sweep rates can be achieved. Axial scan rates up to 370 kHz have been demonstrated recently with a buffered configuration  and could provide OCM frame rates of ∼6 Hz with 256 x 256 pixels, which is comparable to high-speed time domain OCM and is sufficient to eliminate motion artifacts for in vivo imaging. In this paper, we demonstrate swept source / Fourier domain OCM imaging with a buffered FDML laser. We demonstrate initial imaging results at ∼42 kHz axial scan rates, corresponding to image acquisition times of ∼1.5 seconds. Image resolutions of 1.6 μm × 8 μm (transverse × axial) are achieved and cellular level resolution imaging is demonstrated in Xenopus laevis tadpoles, rat kidney, and human colon.
2. Experimental setup
Figure 1 is a schematic diagram of the swept source OCM system. The buffered FDML laser has a full tuning range of ∼110 nm centered at 1290 nm and an output power of ∼12 mW. 3% of the laser power is coupled to an optical spectrum analyzer (OSA) for monitoring the spectrum and a Mach-Zehnder interferometer for recalibration of time to optical frequency. The other 97% of the laser power is then split equally and delivered to a reference arm and a sample arm. A polarization controller and a neutral density filter wheel are used to set the reference arm power to obtain the optimum sensitivity performance. Although the types of glasses in the objective lens are unknown, the dispersion in the sample arm is partially balanced in the reference arm by use of SFL6 and can be compensated numerically as described in reference . The sample arm has a scanning confocal microscope with a pair of closely spaced galvanometer-driven mirrors, a pair of relay lenses, and a 60× water-immersion objective. Due to high loss in the optics at this wavelength, the incident power on the sample is reduced to ∼1.5 mW. It should be noted that the loss in the optics does not only reduce the illumination power but also the back-coupling efficiency. The back aperture of the objective is intentionally underfilled, such that the effective NA is ∼0.35, corresponding to a theoretical confocal gate of ∼16 μm . This provides a good compromise between transverse resolution and depth of field, such that a series of high quality, coherence gated en face images can be extracted from the 3D dataset. The excess laser noise is cancelled by dual-balanced detection.
The confocal gate is measured to be ∼20 μm [solid line in Fig. 2(A)] by blocking the reference arm and recording the DC photodetector signal while translating a mirror in the sample arm around the focus. The slight difference from the theoretical value is a consequence of residual spherical and chromatic aberrations. The coherence gate is measured to be ∼8 μm [dashed line in Fig. 2(A)], which is slightly larger than the theoretical value and is the result of the non-Gaussian spectral shape, nonlinear sweep speed, and the frequency calibration errors as described in reference . Figure 2(B) shows the transverse resolution of the microscope system. The smallest elements measuring 2.2 μm in width on the USAF 1951 resolution target can be clearly visualized. The e-2 radius was estimated to be ∼1.6 μm using an edge-scan measurement method . Figure 2(C) shows the measured axial point spread function on a log scale at the depth of ∼450 μm, which is typically the deepest penetration depth of the OCM system while preserving reasonable image quality in highly scattered tissues. This demonstrates that the sensitivity is higher than 98 dB and the dynamic range is higher than 50 dB within this depth range.
Figure 3(A) shows an in vivo volumetric image of a Xenopus laevis tadpole, a commonly used model organism for developmental biology studies. The 3D dataset is post-processed to generate a series of en face images at different depths near the focal plane. Three representative en face images ∼12 μm apart in depth from each other are shown in Fig. 3(B), 3(C), and 3(D). Degradation of transverse resolution from Fig. 3(C) to Fig. 3(D) is apparent due to the finite depth of field with ∼0.35 NA. The ability to generate images at different depths covering the entire focal range within one acquisition circumvents the need of online coordination between coherence gate and confocal gate. In contrast, for time domain OCM, only one en face image is acquired at a time and therefore precise control of the position of the coherence gate is required to ensure optimal image quality. Figure 4 presents additional in vivo cellular images of a Xenopus laevis tadpole. Figures 4(A) and 4(C) are two en face images acquired when the focus is set at ∼200 μm and ∼400 μm below the surface. Image size is 256 × 256 pixels over the field of view of ∼220 μm × 220 μm. Nuclei and cell membranes can be clearly visualized in both images. Cytoskeletons can also be identified in Fig. 4(C). Figures 4(B) and 4(D) are generated by summing the 3D dataset in the axial direction. Because the coherence gate is shorter than the confocal gate in this study, the images resemble those taken by a confocal microscope, but with reduced speckle. In contrast to standard confocal microscopy, these images were acquired using a relatively low NA and long confocal gate. Obscuration of detailed cellular structures and loss of contrast are evident, indicative of the insufficient rejection of out-of-focus scattered light by confocal gate alone.
Figure 5 shows swept source OCM images and H&E stained histology of a fixed rat kidney, demonstrating the capability to image in highly scattered biological samples. Figures 5(A) and 5(B) are acquired when the focus is at ∼40 μm and ∼120 μm below the surface, respectively. The cell lining along the kidney tubules is readily visible and small regions of bright reflectivity, consistent with nuclei, can also be observed in the images.
The ability to generate high quality cellular images in human tissue with the system is demonstrated in Fig. 6. Figure 6(A) is a swept source OCM image of a normal human colonic mucosa taken at ∼100 μm below the surface while Fig. 6(B) is H&E stained histology. Mucosal specimens can exhibit changes in optical properties with fixation, so the sample was preserved in phosphate-buffered saline (PBS) and imaged within 6 hours of excision, then fixed and processed for histology. Normal colonic mucosa shows the presence of round crypts with goblet cells inside epithelium lining the lumen. In addition, the lamina propria houses many lymphoid cells. All features mentioned above can be clearly resolved in the swept source OCM image. Figures 6(C) and 6(D) are images of a different region at ∼100 μm and ∼150 μm below the surface, respectively. A decrease of lumen size with depth is apparent and detailed structures such as goblet cells and lymphoid cells can still be easily identified deep in the tissue.
4. Discussion and conclusion
Swept source / Fourier domain OCM has several advantages compared to time domain OCM. First, it more easily enables integration with OCT. An imaging modality which can be easily switched between high speed OCT for large scale survey imaging and OCM for a detailed cellular level examination would be a powerful modality for optical biopsy and facilitate clinical studies. Second, coordination of the confocal and coherence gates is not necessary since a stack of coherence gated images which cover the entire focal range is acquired simultaneously. Swept source / Fourier domain OCM also has the advantage of reduced system complexity. Rapid phase modulators in time domain OCM system and spectrometers in spectral / Fourier domain OCM are not required in swept source / Fourier domain OCM.
Although the data rate of spectral / Fourier domain OCM and swept source / Fourier domain OCM is higher than time domain OCM, the frame rate for an en face image at a particular depth is limited by the camera readout rate and the laser sweep rate respectively. Unlike standard swept laser sources, the FDML laser can operate at extremely high speeds and therefore enables rapid, swept source / Fourier domain OCM. FDML lasers overcome limitations of sweep speeds which are present in standard swept lasers and unprecedented high sweep rates up to 370 kHz have been demonstrated with a buffered configuration. At these speeds, ∼6 Hz frame rate can be supported and motion artifacts for in vivo imaging can be minimized. However, the high data rates of 3D volumetric imaging require advanced high-speed data processing solutions in order to perform real-time display. In terms of realtime processing and display, time domain OCM approaches are therefore currently easier to implement.
In this paper, we have demonstrated the feasibility of swept source OCM using a buffered FDML laser. 3D cellular imaging is presented in Xenopus laevis tadpoles, rat kidney, and human colon with resolution of 1.6 μm × 8 μm (transverse × axial), 220 μm × 220 μm field of view, and sensitivity higher than 98 dB. The imaging speed is currently limited by the data acquisition and processing hardware, but has the potential to be significantly increased, enabling in vivo imaging and real-time display. To fully utilize the laser’s high sweep rate and dynamically display an en face image of interest, an analog demodulation scheme should be developed so that the digital processing requirements can be minimized. Unfortunately because of the nonlinear frequency sweep of the FDML laser, a simple band-pass filter is not sufficient to demodulate a particular en face image from the 3D dataset. However, matched filter approaches should be possible and promises to enable very high demodulation speeds. Toward improving resolution, we have recently developed FDML lasers with full tuning ranges of ∼170 nm using specially matched semiconductor optical amplifier (SOA). This suggests that axial resolutions can be improved to ∼5 μm, which is comparable to the thickness of standard histological sections. Furthermore, the superior phase stability of the buffered FDML laser compared with other frequency swept lasers  promises to enable real-time phase microscopy with phase sensitivity compatible to spectral / Fourier domain systems [13, 14].
The authors would like to acknowledge helpful advice from Dr. Yu Chen, and Vivek Srinivasan. We gratefully acknowledge LambdaQuest for providing Fabry Perot tunable filters used in the buffered FDML laser and Dr. James Connolly from the Beth Isreal Deaconess Medical Center for providing the specimens imaged in this study. This research is supported by the Air Force Office of Scientific Research and Medical Free Electron Laser Program FA9550-040-1-0046 and FA9550-040-1-0011, National Institutes of Health R01-EY011289-21, and R01-CA75289-10, and National Science Foundation ECS-0501478 and BES-0522845. A.D. Aguirre acknowledges support from National Institutes of Health EB-005978-02 and the Whitaker Foundation. D.C. Adler acknowledges support from the Natural Sciences and Engineering Research Council of Canada.
References and links
1. T. Wilson, Confocal Microscopy (Academic, London, 1990)
4. A. D. Aguirre, P. Hsiung, T. H. Ko, I. Hartl, and J. G. Fujimoto, “High-resolution optical coherence microscopy for high-speed, in vivo cellular imaging,” Opt. Lett. 28, 2064 (2003) [CrossRef] [PubMed]
5. A. L. Clark, A. Gillenwater, R. Alizadeh-Naderi, A. K. El-Naggar, and R. Richards-Kortum, “Detection and diagnosis of oral neoplasia with an optical coherence microscope,” J. Biomed. Opt. 9, 1271 (2004) [CrossRef] [PubMed]
6. Y. H. Zhao, Z. P. Chen, C. Saxer, S. H. Xiang, J. F. de Boer, and J. S. Nelson, “Phase-resolved optical coherence tomography and optical Doppler tomography for imaging blood flow in human skin with fast scanning speed and high velocity sensitivity,” Opt. Lett. 25, 114 (2000) [CrossRef]
7. V. Westphal, S. Yazdanfar, A. M. Rollins, and J. A. Izatt, “Real-time, high velocity-resolution color Doppler optical coherence tomography,” Opt. Lett. 27, 34 (2002) [CrossRef]
9. M. Pircher, E. Goetzinger, R. Leitgeb, and C. K. Hitzenberger, “Transversal phase resolved polarization sensitive optical coherence tomography,” Phys. Med. Biol. 49, 1257 (2004) [CrossRef] [PubMed]
10. J. F. de Boer, B. Cense, B. H. Park, M. C. Pierce, G. J. Tearney, and B. E. Bouma, “Improved signal-to-noise ratio in spectral-domain compared with time-domain optical coherence tomography,” Opt. Lett. 28, 2067 (2003). [CrossRef] [PubMed]
14. C. Joo, T. Akkin, B. Cense, B. H. Park, and J. E. de Boer, “Spectral-domain optical coherence phase microscopy for quantitative phase-contrast imaging,” Opt. Lett. 30, 2131 (2005). [CrossRef] [PubMed]
16. 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, 1981 (2003). [CrossRef] [PubMed]
18. R. Huber, M. Wojtkowski, J. G. Fujimoto, J. Y. Jiang, and A. E. Cable, “Three-dimensional and C-mode OCT imaging with a compact, frequency swept laser source at 1300 nm,” Opt. Express 13, 10523 (2005). [CrossRef] [PubMed]
19. 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, 3225 (2006). [CrossRef] [PubMed]
20. R. Huber, D. C. Adler, and J. G. Fujimoto, “Buffered Fourier domain mode locking: unidirectional swept laser sources for optical coherence tomography imaging at 370,000 lines/s,” Opt. Lett. 31, 2975 (2006). [CrossRef] [PubMed]
21. 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, 2404 (2004). [CrossRef] [PubMed]
22. M. Gu, C. J. R. Sheppard, and X. Gan, “Image-Formation in a Fiberoptic Confocal Scanning Microscope,” J. Opt. Soc. Am. A 8, 1755 (1991). [CrossRef]
23. M. Rajadhyaksha, R. R. Anderson, and R. H. Webb, “Video-rate confocal scanning laser microscope for imaging human tissues in vivo,” Appl. Opt. 38, 2105 (1999). [CrossRef]
24. D. Adler, R. Huber, and J. G. Fujimoto, “Phase-sensitive optical coherence tomography at up to 370,000 lines per second using buffered Fourier domain mode-locked lasers,” Opt. Lett. 32, 626–628 (2007). [CrossRef] [PubMed]