1058

OPTICS LETTERS / Vol. 35, No. 7 / April 1, 2010

Dual detection full range frequency domain optical coherence tomography

Kye-Sung Lee,1,* Panomsak Meemon,2 William Dallas,3 Kevin Hsu,4 and Jannick P. Rolland1,2
2 1 The Institute of Optics, University of Rochester, Rochester, New York 14627, USA CREOL, The College of Optics and Photonics, University of Central Florida, Orlando, Florida 32816, USA 3 College of Optical Sciences, University of Arizona, Tucson, Arizona 85721, USA 4 Micron Optics, Inc., Atlanta, Georgia 30345, USA *Corresponding author: kyelee4@gmail.com

Received January 4, 2010; revised February 8, 2010; accepted February 8, 2010; posted March 1, 2010 (Doc. ID 122111); published March 31, 2010 It has been shown that frequency domain optical coherence tomography (FD-OCT) systems achieve higher sensitivities compared to time domain optical coherence tomography (OCT) systems. However, the obscure object structure due to the mirror image generated by the Fourier transform is one of the remaining issues in the FD-OCT. We designed and developed what we believe to be a novel full range FD-OCT system that we refer to as the dual detection full range frequency domain optical coherence tomography (DD-FDOCT) that enables the instantaneous retrieval of quadrature components of the complex interferometric signal. The DD-FDOCT system enables full range imaging without loss of speed, and it may be less sensitive to phase error generated by involuntary movements of the subject compared to the other established full range OCT systems, because it uses two signals with a phase difference of / 2 obtained simultaneously from two detection arms to remove mirror images at all depths. 2010 Optical Society of America OCIS codes: 110.3010, 110.4280, 110.4500.

Optical coherence tomography (OCT) is an imaging modality that provides in vivo, non-invasive, and high-resolution cross-sectional images of biological tissues. Recently it was shown that with the frequency domain optical coherence tomography (FD-OCT) one may achieve higher sensitivities as compared to time domain OCT [1]. However, the obscure object structure caused by the mirror image generated by the Fourier transform is one of main problems in the FD-OCT [2]. It is similar to the twin image in holography, as the mirror image is created by the sign ambiguity in the recorded phase used in the process of image reconstruction. The removal of twin images in holography has been investigated using hardware methods as well as numerical methods [3,4]. The removal of the mirror image in the FD-OCT is a key step in securing a sufcient imaging depth especially in high axial resolution FD-OCT employing a broadband source where it is hard to achieve higher spectrally resolved spectrum corresponding to the deeper point in space. The removal yields a long depth of imaging range required by the long depth-of-focus optics incorporated in the sample arm of a FD-OCT system [57]. FD-OCT systems using phase or frequency shifting methods were developed to remove the mirror image. The phase shifting itself has different implementations; it is achieved either by displacing the reference mirror using a piezo translator [8] or by exploiting the inherent phase shifts of 3 3 ber-optic couplers [9]. Both implementations directly or indirectly derive the real and imaginary components of the complex signal, which are always / 2 out of phase with each other. The phase shifting implementation using a piezo translator in the reference arm, however, requires two sequential measurements for a single full range
0146-9592/10/071058-3/$15.00

FD-OCT image, which decreases the imaging speed and is sensitive to any interferometer drifts between the / 2 phase shifted acquisitions. The implementation using the inherent phase shifts of 3 3 beroptic couplers enabled the instantaneous retrieval of the complex interferometric signal. The two signals were simultaneously obtained, respectively, in the two detectors. However, the uneven wavelength dependent splitting ratios in the 3 3 ber coupler lead to an imperfect performance. Also it can be employed only in ber based OCT. The frequency shifting method requires an acousto-optic or an electro-optic phase modulator in the swept source based FD-OCT [10,11] or a sequential modulation of the phase offset of the reference beam (M-scan) during lateral scanning of the probing beam (B-scan), which is referred to as the B-M mode scanning method [12,13]. Although the two frequency shifting implementations require only one measurement to make a full range FD-OCT image, any phase error in the sequential phase modulation in the B-M method, which can be generated by movements of the subject, may limit the performance of the full range OCT. The implementation using an acousto- or electro-optic modulator in the reference arm can be applied only to the swept source based FD-OCT. Dual detection schemes using a beam splitter cube employed in the detection arm have been commonly used in polarization sensitive OCT to obtain two orthogonal polarization signals simultaneously [14]. In this Letter, we report on the design and development of a full range FD-OCT system, referred to as the dual detection full range frequency domain optical coherence tomography (DD-FDOCT), which enables the instantaneous retrieval of the two quadrature components of the complex interferometric signal by adjusting a beam split 2010 Optical Society of America

April 1, 2010 / Vol. 35, No. 7 / OPTICS LETTERS

1059

ter cube in the dual detection scheme as will be further detailed. Also, a newly developed widebandwidth Fourier-domain mode-locked (FDML) laser employing a single semiconductor optical amplier (SOA) is described that builds on the pioneering work of Huber et al. who rst demonstrated a FDML laser to increase the sweeping speed [15]. However, the swept source in [15] had a limited full-width 105 nm bandwidth (BW) using a single gain medium. Jeon et al. reported a high-speed wide-bandwidth FDML laser using two gain media that together achieved a full width of 160 nm [16]. A schematic diagram of the DD-FDOCT system is shown in Fig. 1(a). The DD-FDOCT was developed by modifying the detection and the reference arms of a Michelson interferometer based swept source FD-OCT system as shown in Fig. 1(a). For this system, we developed a custom designed FDML laser to achieve a high-speed and wide-bandwidth swept source employing a custom developed single SOA centered at 1300 nm. The ber FabryPerot tunable lter (FFP-TF) in the FDML has a free spectral range of 158 nm, a BW of 0.2 nm, and a loss L of

Fig. 1. (Color online) (a) Layout of the dual detection full range swept source based FD-OCT using a broadband FDML laser: NPBS, nonpolarizing beam splitter; COL, collimator; M, mirror; Gal, galvanometer; MZI, Mach Zehnder interferometer; OBJ, objective; DET, detector. (b) Spectrum of the FDML laser measured with an optical spectrum analyzer. (c) Adjustment of NPBS2 to obtain / 2 phase shift between the two interference signals.

2 dB. The FFP-TF has a rst-order electromechanical resonance near 45 kHz. Therefore the total ber cavity length including the ber delay loop (Corning SMA28-e) was designed for a length of 4500 m. The SOA has a peak gain of 32.9 dB at 1304 nm, a BW of 46 nm, and a 3 dB saturation power of 12.4 dBm. The output coupling ratio of the fused coupler is 20%, which provides an average output power of 5.6 mW. Figure 1(b) shows the power spectrum of the FDML laser measured with an optical spectrum analyzer in 50 spectra average mode. The FWHM was measured to be 158 nm (from 1252.01410.0 nm) corresponding to 7 m axial resolution in air. The light from the reference arm was guided by two mirrors to the non-polarizing beam splitter (NPBS) cube [i.e., NPBS2 shown in Fig. 1(a)] and then split into two detectors. The sample beam was also split by NPBS2 into the same two detectors. Each reference beam interfered with its associated sample beam. The bulk cube in the reference arm compensates the dispersion mismatch between the two arms. The two interferometric signals were acquired on each channel of a two-channel, highspeed, and 12-bit-resolution analog-to-digital (A/D) converter operating at 200 Msamples/s (National Instrument, Inc., model NI PCI 5124). The Mach Zehnder Interferometer signal used for the calibration in the system was recorded on another 8-bitresolution A/D converter operating at 250 Msamples/s (National Instrument, Inc., model NI PCI 5114). The three channels were synchronized with the external periodic triggering signal generated by the swept source. The / 2 phase shifted interference signal (compared to another interference signal) was obtained by precisely rotating and shifting the NPBS2 as shown in Fig. 1(c). There is no difference between the sample Lsam and reference Lref beam path lengths generated by the NPBS2 when the beam splitter is parallel to the two optical axes. However, the path lengths are mismatched by the amounts of Lsam1 Lref1 for detector 1 and Lsam2 Lref2 for detector 2 if the NPBS2 is shifted and rotated as shown in Fig. 1(c). The rotation and shift were adjusted to make the difference [i.e., Lsam2 Lref2 Lsam1 Lref1] between the two interference signals equal to / 4 at 1300 nm, which corresponds to a / 2 phase shift for the central wavelength. This method of creating a / 2 phase shift reveals to be robust. The maximum phase shift deviation across the spectrum was calculated to be 0.1 rad, which corresponds to 6% deviation from / 2. To investigate the removal of the complex conjugate artifacts from the image with the developed DD-FDOCT, we used a mirror (single interface) as a sample. The two interference signals each recorded on a separate detector are shown in Fig. 2(a). One of the two signals shown in the solid curve (red online) has a / 2 phase difference compared to the interference signal shown in the dotted curve (blue online) in Fig. 2(b). A complex spectral signal I was then generated using the two acquired signals I and I , = / 2 as

1060

OPTICS LETTERS / Vol. 35, No. 7 / April 1, 2010

Fig. 2. (Color online) (a) Two interference signals generated from the dual detection FD-OCT. (b) The zoom-in signals in the narrow time interval showing a / 2 phase difference between the two signals. A-scan depth proles of the single reector in the sample arm (c) with single detection and (d) with suppression of the complex conjugate peak of 35 dB in the DD-FDOCT.

cident power on the sample was 2.3 mW. The average of all the collected spectra in the B-scan was subtracted from each unprocessed spectrum to remove coherent noise in the image. In Fig. 3(a) the image using only a single detector shows the folded artifact image, while the artifact-free image generated in the DD-FDOCT system is shown in Fig. 3(b). Both images are 2 mm 2 mm. In conclusion, we have experimentally demonstrated that the DD-FDOCT system enables full range imaging by the instantaneous retrieval of quadrature components of the complex interferometric signal. Using the setup, we achieved a complex conjugate artifact suppression of 35 dB. The DD-FDOCT may be less sensitive to the phase error generated by involuntary movements of the subject compared to the other established full range OCT systems, because it uses two signals with a phase difference of / 2 obtained simultaneously from two detection arms to remove mirror images at all depths. Importantly, it may assist one in the search for an approach that uses the Doppler FD-OCT together with full range imaging without the need for any other processing to remove phase ambiguities. This is currently under investigation in our laboratory. The NYSTAR Foundation as well as the Photonics Technology Access Program (PTAP) sponsored by the Defense Advanced Research Projects Agency (DARPA) and the National Science Foundation (NSF) supported this research. References
1. R. Leitgeb, C. Hitzenberer, and A. Fercher, Opt. Express 11, 889 (2003). 2. M. Wojtkowski, R. Leitgeb, A. Kowalczyk, and T. Bajraszewski, J. Biomed. Opt. 7, 457 (2002). 3. E. Leith and J. Upatnieks, J. Opt. Soc. Am. 54, 1295 (1964). 4. J. Arocena, T. Rothwell, and M. Shegelski, Micron 36, 23 (2005). 5. K. Lee and J. P. Rolland, Opt. Lett. 33, 1696 (2008). 6. S. Murali, K. S. Lee, and J. P. Rolland, Opt. Express 15, 15854 (2007). 7. S. Murali, K. P. Thompson, and J. P. Rolland, Opt. Lett. 34, 145 (2009). 8. R. Leitgeb, C. Hitzenberger, A. Fercher, and T. Bajraszewski, Opt. Lett. 28, 2201 (2003). 9. M. Sarunic, M. Choma, C. Yang, and J. Izatt, Opt. Express 13, 957 (2005). 10. S. Yun, G. Tearney, J. de Boer, and B. Bouma, Opt. Express 12, 4822 (2004). 11. J. Zhang, J. Nelson, and Z. Chen, Opt. Lett. 30, 147 (2005). 12. Y. Yasuno, S. Makita, T. Endo, G. Aoki, M. Itoh, and T. Yatagai, Appl. Opt. 45, 1861 (2006). 13. R. Wang, Appl. Phys. Lett. 90, 054103 (2007). 14. M. R. Hee, D. Huang, E. A. Swanson, and J. G. Fujimoto, J. Opt. Soc. Am. B 9, 903 (1992). 15. R. Huber, M. Wojtkowski, and J. G. Fujimoto, Opt. Express 14, 3225 (2006). 16. M. Jeon, J. Zhang, Q. Wang, and Z. Chen, Opt. Express 16, 2547 (2008).

I = I + iI, = /2 .

I was then inverse Fourier transformed and was plotted in Fig. 2(d) together with other different delays. The suppression of the complex conjugate peak of 35 dB was obtained in the DD-FDOCT system compared to the complex conjugate peak obtained from the FD-OCT using one of the dual detections. Finally, the DD-FDOCT was applied to in vivo human nger, and we achieved a full range OCT image shown in Fig. 3 that consisted of 500 A-scans and 500 points per A-scan. The lateral resolution of 22 m was achieved using a 0.02 NA objective lens. The in-

Fig. 3. Biological imaging (in vivo human nger): (a) real and mirror images overlapped OCT image using a conventional FD-OCT system; (b) mirror image removed OCT image using the DD-FDOCT system.