Endoscopic optical coherence tomography with a ... - OSA Publishing

4040

Letter

Vol. 42, No. 20 / October 15 2017 / Optics Letters

Endoscopic optical coherence tomography with a focus-adjustable probe WENCHAO LIAO, TIANYUAN CHEN, CHENGMING WANG, WENXIN ZHANG, ZHANGKAI PENG, XIAO ZHANG, SHENGNAN AI, DEYONG FU, TIEYING ZHOU, AND PING XUE* State Key Laboratory of Low-Dimensional Quantum Physics and Department of Physics, Tsinghua University and Collaborative Innovation Center Quantum Matter, Beijing 100084, China *Corresponding author: [email protected] Received 1 August 2017; revised 8 September 2017; accepted 11 September 2017; posted 12 September 2017 (Doc. ID 303780); published 3 October 2017

We present a focus-adjustable endoscopic probe for optical coherence tomography (OCT), which is able to acquire images with different focal planes and overcome depthof-focus limitations by image fusing. The use of a twoway shape-memory-alloy spring enables the probe to adjust working distance over 1.5 mm, providing a large scanning range with high resolution and no sensitivity loss. Equipped with a homemade hollow-core ultrasonic motor, the probe is capable of performing an unobstructed 360 deg field-ofview distal scanning. Both the axial resolution and the best lateral resolution are ∼4 μm, with a sensitivity of 100.3 dB. Spectral-domain OCT imaging of phantom and biological tissues with the probe is also demonstrated. © 2017 Optical Society of America OCIS

codes:

(170.4500)

Optical

coherence

tomography;

(170.2150) Endoscopic imaging; (170.3880) Medical and biological imaging; (220.0220) Optical design and fabrication; (220.4000) Microstructure fabrication; (230.3990) Micro-optical devices. https://doi.org/10.1364/OL.42.004040

Optical coherence tomography (OCT) is a powerful imaging technology that provides non-invasive cross-sectional images with high-resolution. As tiny endoscopic probes work as its sample arms [1–8], the applied range of OCT is expanded to various internal organs, such as arteries and the esophagus. High-resolution OCT is highly desired in many applications. However, the increase of lateral resolution leads to the decrease of depth of focus (DOF). To overcome this limitation, various approaches have been proposed. It is still challenging in the endoscopic field, because the solutions should be miniaturized into a small space. Annular apodization [1] could generate a Bessel-like annular beam in an endoscope with high transverse resolution and extended DOF. However, the gain of the DOF is very limited, and light efficiency is low, resulting in the loss of sensitivity. The methods of using a binary phase filter [2], a phase mask, [3] and an axicon lens [4,9] suffer from the same problem as the 0146-9592/17/204040-04 Journal © 2017 Optical Society of America

apodizing method [1]. The wavefront division method [5] could produce multiple foci, providing a 20-fold DOF extension, but the long DOF is achieved at the expense of 10–20 dB sensitivity loss. The membrane-based microlens probe [7] provides tunable focus, but the lateral resolution is low (13 μm) due to its high spherical aberration. The GRIN lens rod-based probe [8] has dynamically adjustable focus, but with low lateral resolution of 11 μm, and the lateral scanning range is restricted by the diameter of the lens rod. Besides, most of these endoscopes have a driving mode of proximal scanning, which has problems of unstable polarization distortion and nonuniform scanning speed, especially in curve lumens. The eletrowetting lens [10], liquid crystal lens [11], and electrothermal bimorph [12] could be useful in an optical tunable system, but they still have difficulties in being miniaturized, stable, or achieving high resolution for use in an endoscopic probe. Other methods, such as multiple-aperture synthesis [13], adaptive optics [14], and depth-encoded synthetic aperture [15], require a complex optic system, which is hard to implement in a small endoscopic probe. In this Letter, we present a focus-adjustable endoscopic OCT probe, which is able to perform C-mode scanning (acquiring images with several focal planes and fusing them together, similar to the concept in ultrasound imaging) to overcome depth of field limitations. The best lateral resolution is ∼4 μm, corresponding to a DOF of only 160 μm, while the focus-adjustable probe enables a depth-scanning range of 1.5 mm, with high resolution and no sensitivity loss, equivalent to a large DOF extension. A homemade hollow-core ultrasonic motor enables an unobstructed 360 deg field-of-view distal scanning. To the best of our knowledge, this is the first demonstration of C-mode scanning in endoscopic OCT. Figure 1(a) shows the schematic diagram of the prototype focus-adjustable probe, which is based on our previous work [16] and is naturally developed for the new function to enlarge the depth of field. A single-mode fiber (SMF) fixed in a protecting needle goes through a two-way shape-memory-alloy (SMA) spring and a homemade hollow-core ultrasonic motor. One end of the spring attaches to the motor with glue, and the other end is fixed with an SMF through a metal connector. As we adjust the electric current through the spring, it stretches

Letter

Fig. 1. (a) Schematic of the focus-adjustable probe. The focal position is adjusted as the SMA spring stretches out or draws back. (b) Schematic of the endoscopic SD-OCT system. (c) Photograph of the fabricated focus-adjustable probe. FC, fiber coupler; PC, polarization controller; C, collimator; M, mirror; G, grating; L, lens.

out or draws back, driving the SMF moving back or forth. In this way, the distance between the fiber tip and GRIN lens changes, resulting in the changing of the focus position. The outer diameter of the spring is 2 mm, with an original length of 3.3 mm. The two-way SMA spring wire has electrical resistivity of 0.05 Ω/mm. As the electric current passes through the spring, it heats up and stretches. The stretch stops when the balance between heat production and heat loss is established. Similarly, as we reduce the current, the spring cools down and draws back. The working temperature of the spring is 20°C– 55°C. The adjustment range of the electric current is from 150 to 220 mA, which gives the spring a maximum stretch length of 1 mm. As a result, the working distance of the probe is able to be adjusted from 0 to over 1.5 mm. The period for one full cycle of expansion/contraction of the spring is ∼12 s, which is slower than that of the electrowetting lens, but acceptable in practical use. Further improvement is in progress. The light emitted from the fiber is focused by an 8°-angled 0.31 pitch gradient-index (GRIN) lens with 1.8 mm diameter, and reflected by a right-angled prism. The prism and GRIN lens are glued to the hollow-core rotor, and are driven by the stator of the motor to perform 360 deg scanning without any shadow resulting from light blocking by power wires. The speed of rotation is tunable by adjusting the driving voltage of the motor. The maximum rotation speed is ∼30 r∕s. The hollow-core motor used in this probe has a diameter of 1.9 mm, slightly larger than the one we reported previously [16], but the way they work is the same. The outer diameter of the probe is 2.5 mm, and the rigid length is 32 mm. As show in Fig. 1(b), the probe works with a spectral domain (SD)-OCT system. The light source is a broadband super luminescent diode (Inphenix, Inc., Livermore, California, USA) with


4041

6.5 mW output power and a bandwidth of 90 nm centered at 850 nm, corresponding to ∼4 μm axial resolution (in air). Light from SLD goes through a circulator and is divided into the sample and reference arms by a fiber coupler. 90% of the light energy goes to sample arm, while 10% goes to reference arm. The interference between the light from the sample arm and reference arm is detected by a customized spectrometer, which contains a 1200 lines/mm grating and a 2048-pixels line-scan chargecoupled (CCD) camera (EM4CL2014, e2v, UK). In the sample arm, the wires of the probe are linked to a custom electronic controller, which controls the current passing through the SMA spring and the electric driving signals of the hollow-core ultrasonic motor. A photograph of the fabricated probe is presented in Fig. 1(c). In fact, the glass tube is fixed to the stator with an ultraviolet (UV)-curing adhesive. Except for the distal part, the rest of the probe is covered with polyethylene terephthalate tube that is fixed to the glass tube, also with a UV-curing adhesive. An integrated three-axis translational stage, plus a twoaxis tilter stage, may be used to ensure good mounting. To test the imaging performance of the SD-OCT system, we first measure the sensitivity of the system. Traditionally, we place a mirror in a sample arm and insert a neutral density filter into the beam path to attenuate light intensity when we measure the sensitivity. However, it is not applicable for the probe, as the distance between the focal point and probe is too small to insert a filter. Therefore, a special sample is made by optical matching gel (Fibkey, China) and BK7 glass. Fresnel reflection laws give that the reflectivity of their interface is 0.067%, corresponding to an attenuation of 31.7 dB. The measured signalto-noise ratio (SNR) is 68.6 dB, as shown in Fig. 2(a), which means that the system sensitivity is 100.3 dB. We then measure the fall-off of the system by placing the same signal of the interface at different axial depths, as shown in Fig. 2(b). The measured 6 dB fall-off is ∼1.1 mm. We also measure the lateral resolution of the probe. To obtain the size of the light spot, we place a sharp-edge blade at the focal position and move the blade going across the light spot. We record the light transmission intensity (by PM120D, Thorlabs, Newton, New Jersey, U.S.) against the moving distance of the blade and then analyze it, as shown in Fig. 2(c). The lateral resolution is defined as a 20%–80% width of the slope region, corresponding to the full width at half-maximum diameter of the focal light spot [16]. We adjust the SMA spring and measure the lateral resolution at different working distances, as shown in Fig. 2(d). The working distance is measured by counting the pixel numbers between the sharp edge and the wall of glass tube in an OCT image, and converting it to the distance in air. We can see that the measured lateral resolutions fit well with the simulate curve of ZEMAX. The minimum lateral resolution is ∼4 μm (in air), corresponding to a DOF of 160 μm. As the working distance increases to 1 mm, the lateral resolution decreases to ∼10 μm. It is quite similar to a confocal foveated endomicroscope [17], which provides a large fieldof-view with high lateral resolution in the central region. Figure 2(e) shows the simulation result of the working distance changing with the distance between the fiber tip and the GRIN lens. With the fiber tip moving less than 1 mm, the working distance varies over 1.5 mm. Figure 2(f ) shows the relationship between the working distance and electric current passing through the SMA spring. As shown in Fig. 1(a), we may measure the spring length change with the change of spacing

4042


Fig. 2. (a) Measured signal from a sample made of optical matching gel and BK7 glass, corresponding to a sensitivity of 100.3 dB. (b) Measured fall-off performance of the endoscopic SD-OCT system. (c) Lateral resolution is measured by 20%–80% width of the fitting curve, which represents the light transmission intensity against the moving distance of a sharp edge. (d) Measured and simulated results of the lateral resolutions of the probe. (e) Simulation result of working distance changing with the distance between the fiber tip and GRIN lens. (f) Measured results of the working distance changing with the electric current passing through the SMA spring.

between the fiber tip and the GRIN lens surface. When we block the light from the reference arm, the fiber tip and GRIN lens surface actually form a common-path OCT system. Therefore, their spacing can be easily measured with an accuracy of ∼4 μm. The length of the spring has a fluctuation of less than 10% with room temperature changing between 20°C and 30°C. Compared with the sensitivity of the electrowetting lens to the vibration during imaging, our system is quite robust. As shown in Fig. 3, the performance of the focus-adjustable probe is tested by the imaging of a phantom, which is made by coiling translucent tape (Scotch, 3M, Saint Paul, Minnesota, U.S.) on a glass tube. The CCD runs at an integration time of 30 μs, and a circumferential image consists of ∼4000 A-scans. The rotation speed of the probe is slowed down to ∼8 r∕s to obtain images with more details of the sample. OCT image data are acquired and processed in the traditional method of interpolation and dispersion compensation. We use a distortion-free algorithm [18] to correct nonuniform rotation distortion of the images. From each image, we manually chose the clearest and brightest circular layer with the thickness of ∼DOF, as calculated according to the working distance. Thereafter, all the chosen layers are then simply put together based on the depth sequence to get the C-mode image. Figures 3(a)–3(c) show the circumferential images as the probe focuses in a shallow layer, middle layer, and deep layer, respectively. In Fig. 3(a), light is focused tightly, so the light scattering intensity and lateral resolution are high in this image, but the scanning range is limited. In Figs. 3(b) and 3(c), as the focal

Letter

Fig. 3. OCT images of a tape roll. The focal position is in the (a) shallow layer, (b) middle layer, and (c) deep layer. (d) Fused image of (a)–(c) (C-mode scanning image).

position goes deeper, the scanning range is enlarged, but the light scattering intensity and lateral resolution are decreased. To gather the advantages of each image, we fuse these three images and present a C-mode scanning image, as shown in Fig. 3(d). The axial imaging depth of Figs. 3(c) and 3(d) is ∼1.56 mm (without considering the refractive index of the sample). The electric current passing through the SMA spring in Figs. 3(a)–3(c) is 202, 192, and 182 mA, respectively. According to Fig. 2(f ), the working distances are estimated to be 0.19, 0.43, and 0.77 mm (in air). At an image depth of 200 μm in Figs. 3(a)–3(c), the image SNR is 41.3, 31.5, and 28.4 dB, respectively. Therefore, the largest SNR difference is 12.9 dB, which implies the SNR enhancement with our method. All the images are taken without a shadow effect caused by power wires, owing to the use of a homemade hollow-core ultrasonic motor. The dark areas in the deep layer may be caused by some dust or impurity in the shallow layer of the sample which hinders the light going deep. To demonstrate the probe’s feasibility of imaging biological tissues, SD-OCT imaging of a grape is implemented, as shown in Fig. 4. We dig a hole in a grape by using an annular blade with a diameter of 2.5 mm, and then we put the probe into the grape to acquire images. The CCD integration time, the rotation speed of the probe, and the image data processing are the same as those for phantom imaging. Images are taken as the focal position goes from the shallow layer, to the middle layer, and the deep layer, as shown in Figs. 4(a)–4(c). Figure 4(d) is the fused image of (a), (b), and (c). To better visualize and compare the fine structure of these OCT images, small regions of Figs. 4(a)–4(d), as indicated by dotted boxes, are zoomed in by four times and shown in Figs. 4(e)–4(h). White dashed lines in Figs. 4(e)–4(g) divide the images into three layers: a shallow layer, a middle layer, and a deep layer (from bottom to top). As we can see in Fig. 4(e), light is focused tightly in the shallow layer, where the image is bright and the microstructure can be seen clearly, owing to high transverse resolution and high

Letter


4043

Fig. 4. OCT images of a grape. The focal position is in the (a) shallow layer, (b) middle layer, and (c) deep layer. (d) Fused image of (a)–(c) (C-mode scanning image). (e)–(h) show the 4× zoomed-in images of the boxed regions in (a)–(d), respectively. The white dashed lines in (e)–(g) divide the images into three layers: (from bottom to top) the shallow layer, the middle layer, and the deep layer.

scattering intensity in this layer. However, the scattering signal decrease in the middle layer and structure of the deep layer are almost invisible. In Fig. 4(f ), the focal position is in the middle layer. Compared to (e), Fig. 4(f ) has a higher resolution and signal intensity in the middle layer, but the image is rough in the shallow layer, and the microstructure is unclear in the deep layer, as these regions are not in focus. Figure 4(g) has the largest imaging depth (∼1.2 mm, without considering refractive index of sample) and the most clear deep layer in these three images, but the microstructure is blurred in the shallow and middle layers, resulting from the light defocusing in the layers. In order to gather the advantages of these three images, we take one clear layer from each image and fuse them together, as shown in Figs. 4(d) and 4(h). As we can see, the fused image (C-mode scanning image) has the highest resolution, as well as the largest imaging depth. The working distances in Figs. 4(a)– 4(c) are about 0.26, 0.46, and 0.65 mm (in air). It is worth mentioning that all these images have an unobstructed 360 deg field-of-view, owing to the use of a homemade hollow-core motor. In summary, we proposed and fabricated a prototype focusadjustable endoscopic probe based on a two-way SMA spring and a homemade hollow-core ultrasonic motor. Superior to a traditional DOF extended probe, our novel probe has adjustable focus and, hence, a larger scanning range, with high resolution and no sensitivity loss. The focus-adjustable range is over 1.5 mm, with a sensitivity of 100.3 dB, and the best lateral resolution is ∼4 μm. With the use of hollow-core motor, the probe is also able to provide an unobstructed 360 deg field-ofview. SD-OCT imaging of the phantom and biological tissues with the probe is also demonstrated. To the best of our knowledge, this is the first demonstration of a focus-adjustable probe for C-mode scanning in endoscopic OCT. We believe that this novel probe will be very useful in future biomedical applications.

Funding. National Natural Science Foundation of China (NSFC) (61227807, 61575108); Tsinghua Initiative Scientific Research Program (2013THZ02-3). REFERENCES 1. K. K. Chu, C. Unglert, T. N. Ford, D. Cui, R. W. Carruth, K. Singh, L. Liu, S. E. Birket, G. M. Solomon, and S. M. Rowe, Biomed. Opt. Express 7, 2494 (2016). 2. J. Kim, J. Xing, H. S. Nam, J. W. Song, J. W. Kim, and H. Yoo, Opt. Lett. 42, 379 (2017). 3. D. Lorenser, X. Yang, and D. D. Sampson, Opt. Lett. 37, 1616 (2012). 4. N. Weber, D. Spether, A. Seifert, and H. Zappe, J. Opt. Soc. Am. A 29, 808 (2012). 5. B. Yin, K. K. Chu, C. P. Liang, K. Singh, R. Reddy, and G. J. Tearney, Opt. Express 24, 5555 (2016). 6. N. Zhang, T. H. Tsai, O. O. Ahsen, K. Liang, H. C. Lee, P. Xue, X. Li, and J. G. Fujimoto, Opt. Lett. 39, 186 (2014). 7. K. Aljasem, L. Froehly, A. Seifert, and H. Zappe, J. Microelectromech. Syst. 20, 1462 (2011). 8. D. Mukai, M. Brenner, S. Guo, T. Xie, and Z. Chen, Opt. Express 14, 3238 (2006). 9. D. Lorenser, S. C. Christian, A. Curatolo, and D. D. Sampson, Opt. Lett. 39, 548 (2014). 10. X. Hu, S. Zhang, Y. Liu, C. Qu, L. Lu, X. Ma, X. Zhang, and Y. Deng, Appl. Phys. Lett. 99, 213505 (2011). 11. S. Murali, K. S. Lee, and J. P. Rolland, Opt. Express 15, 15854 (2007). 12. L. Wu and H. Xie, IEEE J. Quantum Electron. 46, 1237 (2010). 13. E. Bo, Y. Luo, S. Chen, X. Liu, N. Wang, X. Ge, X. Wang, S. Chen, S. Chen, J. Li, and L. Liu, Optica 4, 701 (2017). 14. S. Kazuhiro, K. Kazuhiro, M. Shuichi, and Y. Yoshiaki, Biomed. Opt. Express 3, 2353 (2012). 15. J. Mo, G. M. De, and J. F. de Boer, Opt. Express 23, 4935 (2015). 16. T. Chen, N. Zhang, T. Huo, C. Wang, J. G. Zheng, T. Zhou, and P. Xue, J. Biomed. Opt. 18, 86011 (2013). 17. A. Shadfan, A. Hellebust, R. Richardskortum, and T. Tkaczyk, Biomed. Opt. Express 6, 2311 (2015). 18. X. Liu, Y. Huang, and J. U. Kang, Opt. Express 20, 16567 (2012).