Open Access
Add to BibSonomyAbstract
High-speed scanning in optical coherence tomography (OCT) often comes with either compromises in image quality, the requirement for post-processing of the acquired images, or both. We report on distortion-free OCT volumetric imaging with a dual-axis micro-electro-mechanical system (MEMS)-based handheld imaging probe. In the context of an imaging probe with optics located between the 2D MEMS and the sample, we report in this paper on how pre-shaped open-loop input signals with tailored non-linear parts were implemented in a custom control board and, unlike the sinusoidal signals typically used for MEMS, achieved real-time distortion-free imaging without post-processing. The MEMS mirror was integrated into a compact, lightweight handheld probe. The MEMS scanner achieved a 12-fold reduction in volume and 17-fold reduction in weight over a previous dual-mirror galvanometer-based scanner. Distortion-free imaging with no post-processing with a Gabor-domain optical coherence microscope (GD-OCM) with 2 μm axial and lateral resolutions over a field of view of 1 × 1 mm2 is demonstrated experimentally through volumetric images of a regular microscopic structure, an excised human cornea, and in vivo human skin.
© 2016 Optical Society of America
1. Introduction
Optical Coherence Tomography (OCT) is a non-invasive, high-resolution imaging technique based on low-coherence interferometry [1]. Handheld probes for OCT based on different types of scanning devices were developed in recent years for medical imaging and nondestructive testing [2–7]. Some commercial probes include a significant part of the OCT system and have weights ranging from 1.5 kg (Envisu; Bioptigen) to 2.2 kg (iVue; Optovue); thus, rather than being held in the hand, they are positioned on platforms in front of the areas of interest on the patient’s body. More recently, compact handheld probes based on Micro-Electro-Mechanical Systems (MEMS) have been developed with limited scan rate (e.g., 25 or 30 frames/s by Thorlabs and Wasatch Photonics, respectively). Lighter and simpler OCT handheld probes based on uni-dimensional (1D) galvanometer-based scanners (GSs) [6,7] were developed for applications such as dentistry; however the functionality offered by these probes is limited - they can perform only B-scans (i.e., transversal sections into the sample), while the clinician manually moves the probe to perform the third-axis lateral scan. As a consequence, probes with a 1D scanner are generally not sufficient in numerous clinical and industrial applications requiring near real-time tri-dimensional (3D) reconstructions of a sample.
To address this need, bi-dimensional (2D) GSs [4] or 2D MEMS [3, 5, 8, 9] have been investigated. In these systems, two main types of distortion appear, with a third additional type of distortion that appears concurrently with the first two. The first type of distortion (type I) occurs when there are no imaging optics between the scanning mirror and the volume being imaged; in this case, gross geometrical distortions arise [9]. The second type of distortion (type II), which may arise even in systems with imaging optics between the scanner and the volume being imaged, occurs when the two axes of scan are not in the same plane, as is the case with 2D GSs [10]. The third type of distortion (type III) is caused by a non-linear movement of the MEMS, which occurs either by excitation of its resonance or by a non-linear scanning pattern [5, 8]. Distortion of type III, which is of smaller scale than the first two types, becomes significant once distortions of type I and II are mitigated.
GSs have a mature technology and can be driven in closed-loop with triangular input signals that, as we have demonstrated, produce the most distortion-free OCT images, as well as the highest duty cycles (i.e., highest time efficiency of the scanning process) [11]; however, they are rather bulky, add a significant weight to the probe, and introduce astigmatism in the imaging beam (type II distortion), causing an apparent curvature of interfaces in the OCT images that has to be corrected in post-processing. Auxiliary relay optic of one GS to the other may be used to correct the type II distortion [12, 13], but at the expense of increased size, cost, and weight of 2D GSs-based probes.
On the other hand, MEMS scanners are small, light, and can achieve dual-axis scanning with a single surface that can be placed at the pupil of the optical system to eliminate type II distortion. The main drawback of open-loop operation of MEMS mirrors reported in the literature for OCT imaging is that, due to their characteristic resonance, sinusoidal signals at resonance have been typically used. To compensate for the non-linear scan pattern, often oversampling is required (e.g., 4x in [5]). Furthermore, the resulting images acquired with OCT [5, 8] and also with other imaging techniques, such as confocal microscopy [14], are affected by type III distortion and require post-processing. Given the large size of the OCT data, post-processing introduces considerable computational complexity, and one must also carefully account for how noise in the original images may affect image quality after post-processing. While with continually improving hardware and numerical methods real-time imaging may be achieved, a solution without post-processing and no additional hardware, as reported here, is always preferred.
In this paper, we report on a MEMS-based scanning device to produce distortion-free images with a Gabor-Domain Optical Coherence Microscope (GD-OCM). To our knowledge, this is the first report of distortion-free OCT imaging with no post-processing. To achieve these results, we employ imaging optics between the scanner and the volume being imaged to avoid distortion of type I, and novel to this paper we simultaneously place a dual-axis MEMS at the pupil location to avoid distortion of type II and design a pre-shaped input signal with tailored non-linear portions to eliminate distortion of type III, as described in the following section.
2. Distortion-free imaging in GD-OCM wih a dual-axis MEMS scanner
In the GD-OCM system shown in Fig. 1, imaging optics between the scanner and the volume being imaged is employed to avoid distortion of type I [15]. A dual-axis MEMS is placed at the pupil location to avoid distortion of type II and is driven with a pre-shaped input signal to eliminate distortion of type III.
GD-OCM produces 3D reconstructions of a sample with high invariant microscopic resolution throughout the volume by combining Fourier domain OCT with dynamic refocusing of a biomimetic liquid lens-based microscope with no moving parts [17]. An implementation of the system used in the reported experiments provides a volumetric resolution of 2 µm (depth and lateral resolution) across a field of view of 1 × 1 mm2, and an imaging depth of ~1.6 mm. The source is a superluminescent diode with a center wavelength of 840 nm and 100 nm bandwidth (BroadLighter D-840-HP-I, Superlum). A custom spectrometer with a high-speed CMOS line camera (spl4096-140km, Basler Inc.) with a line period of 15 µs was used to acquire the spectral information [18]. A parallelized multi-Graphic Processing Unit (GPU) architecture allows for near real-time visualization of the sample [19].
In this work, the scanning probe based on two single-axis GSs in our prior work was replaced with a single dual-axis MEMS scanner (Mirrorcle Technologies, Inc.) with an aperture of 2.4 mm and a resonant frequency of 542 Hz. A custom handheld probe was developed for the MEMS scanner and integrated in the GD-OCM setup (Fig. 2(b)) [16], achieving a volume reduction of the scanner of approximately 12 times over the prior 2D GSs-based scanner (i.e., from 150 × 100 × 115 mm3 to 75 × 40 × 50 mm3 – Fig. 2(a) to Fig. 2(b), respectively). The weight of the scanner part of the probe was also reduced by 17 times, from 2.3 kg to 0.13 kg.
Fig. 2 Handheld probe for GD-OCM with a 2D GS (a) and with a 2D MEMS scanner (b), the latter also rendered in (c).
As we documented in detail for GSs [11], sinusoidal scanning produces the most distorted images in OCT compared to other common driving signals, such as triangular and sawtooth; this holds for any oscillatory scanner, including MEMS. For the latter, as they are resonant scanners, sinusoidal scanning is usually used; however, since computationally expensive post- processing correction of type III distortion is required [5, 8], sinusoidal scanning is not considered in our approach. Efforts have been made to drive MEMS with triangular signals, as is the case with the Mirrorcle micro-scanner used in this study; however, such signals excite resonances of the MEMS if driven in open loop (Fig. 3(b1)). The movement on one axis of the micro-mirror in open loop can be approximated as a classical second order oscillator [20], where J is the moment of inertia of the micro-mirror on the oscillatory axis considered, c is the damping coefficient, k is the stiffness of the torsion springs that support the mirror, and T(t) is the magneto-electric active torque applied in order to produce the desired movement. This dynamic equation of the device can also be written as
where and are the undamped angular frequency ω0 = 2πf0 (where f0 is the natural frequency) and the damping ratio of the system, respectively. The solution of Eq. (1), which gives the angular position θ(t) of the micro-mirror, can be derived exactly. For example, for the first T/4 part of the triangular input signal, which is a ramp signal with a constant slope p = 4θm/T (Fig. 3(a1)), the solution iswhere the phase of the periodic function is . One may observe from this solution the oscillatory behavior of the output, which has an asymptotic tendency towards the θ = p(t-τ) line and produces the type III distortions that affect the imaging process. The time interval delay τ = 2ξ/ω0 with respect to an ideal output pt (which should be identical to the input - Fig. 3(a1)) becomes relevant for larger aperture micro-mirrors (e.g., 4.2 mm), but for the 2.4 mm aperture mirror used here the oscillatory term is dominant and must be corrected, while the delay can be neglected, as can be seen in Figs. 3(b1) and 3(c1).Fig. 3 Input (a) and output (b) scanning signals of the fast axis of the MEMS, with highlighted portion of (b) shown in (c): (1) purely triangular and (2) custom pre-shaped. Notations: T, scan time period, equal to 1/fs, where fs is the scan frequency of the MEMS scanner for (1); Tf, scan time period for (2); ta, active scan time (for which the scanning is performed with constant speed); Δt, duration of the stop-and-turn part of the input signal; θm, scan amplitude; ƞ = 2ta/T, ideal duty cycle of the scanning process.
The step response of the MEMS is characterized by an overshoot of ~80% and a settling time of 19 ms. As pointed out above, a purely triangular input (Fig. 3(a1)) excites the resonant frequency of the MEMS and produces ringing in the position of the micro-mirror (Fig. 3(b1)), which persists throughout the entire period of the triangular wave and introduces type III distortion that affects the entire scan. To study these aspects experimentally, stand-alone, custom electronics were developed to drive the MEMS in real-time at low cost without CPU intervention, test its different relevant scanning inputs, and generate optimized input signals. This is critical for achieving distortion-free images, since non real-time operating systems (such as Windows or Linux) usually introduce jitter due to interrupts and other background processes - this is unacceptable as deviations from the ideal driving signal directly lead to artifacts in the images and tend to excite the resonance of the MEMS.
Also, a module with a laser diode and a position-sensing device (PSD) was developed to determine the current position of the MEMS mirror on the two oscillatory axes for validation. Figure 3(b1) presents the output signal, i.e., the current angular position of the mirror on the fast axis of the MEMS, which, as expected, exhibits higher frequency oscillations superposed to the signal characterized by the scan frequency fs = 1/T, as shown in detail in Fig. 3(c1). When driven with such signals, the MEMS scanner produces severely distorted GD-OCM images (type III distortion) as shown in Figs. 4(a1) and 4(b1).
Fig. 4 GD-OCM images obtained with the MEMS-based handheld scanning probe of a 1 × 1 × 0.6 mm3 volume of a regular microscopic structure – en face view (a) and volumetric view (b). The images were obtained: (1) with a purely triangular and (2) with a pre-shaped input signal.
To solve this problem, an input-shaping [21, 22] low-pass filter was applied to the driving command of the MEMS to effectively cancel the ringing in the response of the mirror. For the signal shown in Fig. 3(a2), a 2nd order Butterworth filter with a cut-off frequency of 200 Hz was used. In contrast, the MEMS manufacturer recommends a 6th order Bessel low-pass filter. A 2nd order filter is significantly simpler to implement than a 6th order one and, most importantly, it is three times faster to calculate in the microcontroller board, allowing faster scan frequencies. For real-time imaging this is a critical step. We chose to implement a software filter rather than a hardware filter because the software solution offers the flexibility to readily change the filter’s cutoff frequency and therefore can accommodate different MEMS devices that may be adopted for different applications and specifications.
Although small oscillations are still present, as shown in the detail in Fig. 3(c2), they are reduced to an acceptable threshold for the resolution required by the imaging application. An overshoot of duration ∆t was introduced to the signal to maintain the same duration ta of the linear portion, thus decreasing the effective duty cycle ƞ from the ideal 100% (Fig. 3(a1)) to 86% for the filtered signal (Fig. 3(a2)); the scan frequency also decreased to 1/Tf, where Tf is the period of the input-shaped signal. This aspect has been optimized for the signals in this study for a desired scan frequency fs and amplitude θm. Further investigations may readily extend this analysis to other scanning regimes, with different scan frequencies and amplitudes. For example, the time interval Δt of the optimized input (Fig. 3(a2)) has to be set large enough to avoid non-linearity within the desired linear parts of duration ta, which would produce distortions at the margins of the scan field [11]. On the other hand, the time interval also has to be minimized to avoid excessive reduction of the duty cycle. The scan frequency used on the fast axis in the experiments was 57 Hz. A 57 kHz A-scan acquisition rate was thus achieved; real-time imaging using the GPU platform produced 1 × 1 × 0.6 mm3 volumetric images in less than 2 minutes.
3. Experimental results
The result of applying the filtered signals described in Section 2 is shown in the GD-OCM images in Figs. 4(a2) and 4(b2) for a regular microscopic structure. As compared to the outcome of purely triangular signals (Figs. 3(b1) and 3(c1)), which produce severely distorted images (type III distortion) as in Figs. 4(a1) and 4(b1), the pre-shaped filtered signals (Figs. 3(b2) and 3(c2)) reveal the real structures of the sample and yield distortion-free imaging without post-processing throughout the volume being imaged (Figs. 4(a2) and 4(b2)).
The effectiveness of the new MEMS-based scanning system was also validated for clinical applications through distortion-free GD-OCM imaging of human skin in vivo and excised human cornea, shown in Figs. 5 and 6, respectively.
Fig. 5 (a, c) 2D and (b, d) 3D views of a 1 × 1 × 0.6 mm3 image of human skin acquired with GD-OCM in less than 2 minutes. The plane of the en face view (c) is shown in red in the side view (a) and in the 3D image (b).
Fig. 6 (a) Volumetric image acquired with GD-OCM of an excised human cornea. The 1 × 1 mm2 en face views in (b-d) correspond to increasing depths inside the tissue sample. In particular, the transition between the epithelium and the stroma in correspondence of Bowman’s layer is shown in (b). Stromal keratocytes in (c) and endothelial cells in (d) are readily visible.
The cellular resolution of the distortion-free MEMS-based GD-OCM system was demonstrated in skin (Fig. 5(c)) and in corneal endothelium (Fig. 6(d)).
4. Conclusion
In summary, we have presented pre-shaped open-loop input signals for a MEMS-based handheld scanning probe coupled to a custom GD-OCM system. Samples of 1 × 1 × 0.6 mm3 were imaged in less than 2 minutes with an invariant resolution of 2 µm. While prior art with MEMS sinusoidal scanning for OCT or confocal microscopyc [5, 6, 14] required post-processing of the acquired images, we have demonstrated distortion-free GD-OCM images by implementing optimized input signals to drive the dual-axis MEMS in the context of having relay optics between the 2D MEMS and the sample. The lack of relay optics in itself would lead to gross distortions that are eliminated here by design. Compactness of the handheld probe and distortion-free images with cellular resolution are key requirements for on-site use in clinical applications. In addition to the image quality improvement, the 2D MEMS-based scanner attachment for the GD-OCM microscope achieved a volume reduction of the scanner of approximately 12 times over the prior 2D GSs-based scanner. The weight of the scanner part of the probe is also reduced by 17 times, thus facilitating handheld operation in situ for clinical and industrial applications. While testing the probe in handheld free-motion, we occasionally encountered artifacts due to hand tremor of the operator, which however, due to its low frequency, did not excite the MEMS resonance.
The advantages of MEMS versus GSs are expected to expand, as MEMS technology continues to improve, including optimized control structures and driving signals that are also subject of future work in our groups. As MEMS devices reach mass production, they also have the potential to reduce the cost per scan axis to at least an order of magnitude below GSs. Applications in real-time medical imaging, including for skin and cornea [23, 24], as well as for nondestructive testing in industry [25, 26], are poised to benefit from the improved image quality and robustness achieved with the 2D MEMS scanner developed herein. Future work includes the generalization of this study to a broad range of frequencies and amplitudes, ergonomic designs of handheld probes, and the development of a robotic arm to move the probe around the patient, thus avoiding hand tremor artifacts [5].
Acknowledgments
This research was funded by the National Science Foundation under Grants No. IIP-1346453 and IIP-1534701. We acknowedge the NYSTAR Foundation and Center for Emerging and Innovative Sciences, and the Romanian National Authority for Scientific Research (CNDI–UEFISCDI project PN-II-PT-PCCA-2011-3.2-1682). We thank our collaborators in immunology - David Topham, - in opthalmology - Philippe Gain, Gilles Thuret, and Holly Hindman, - and in dermatology - Sherrif Ibrahim and Alice Pentland - for sharing their passion in their respective fields that has fueled our efforts towards achieving distortion-free images without the need for image processing. We also thank them all for seed support they have provided over the past year towards system integration and testing. We thank NVIDIA® for the donation of GeForce® GTX Titan GPUs to LighTopTech Corp. and the University of Rochester.
References and links
1. W. Drexler, M. Liu, A. Kumar, T. Kamali, A. Unterhuber, and R. A. Leitgeb, “Optical coherence tomography today: speed, contrast, and multimodality,” J. Biomed. Opt. 19(7), 071412 (2014). [CrossRef] [PubMed]
2. A. M. Zysk, F. T. Nguyen, A. L. Oldenburg, D. L. Marks, and S. A. Boppart, “Optical coherence tomography: a review of clinical development from bench to bedside,” J. Biomed. Opt. 12(5), 051403 (2007). [CrossRef] [PubMed]
3. D. Nankivil, G. Waterman, F. LaRocca, B. Keller, A. N. Kuo, and J. A. Izatt, “Handheld, rapidly switchable, anterior/posterior segment swept source optical coherence tomography probe,” Biomed. Opt. Express 6(11), 4516–4528 (2015). [CrossRef] [PubMed]
4. W. Jung, J. Kim, M. Jeon, E. J. Chaney, C. N. Stewart, and S. A. Boppart, “Handheld optical coherence tomography scanner for primary care diagnostics,” IEEE Trans. Biomed. Eng. 58(3), 741–744 (2011). [CrossRef] [PubMed]
5. C. D. Lu, M. F. Kraus, B. Potsaid, J. J. Liu, W. Choi, V. Jayaraman, A. E. Cable, J. Hornegger, J. S. Duker, and J. G. Fujimoto, “Handheld ultrahigh speed swept source optical coherence tomography instrument using a MEMS scanning mirror,” Biomed. Opt. Express 5(1), 293–311 (2014). [CrossRef] [PubMed]
6. V.-F. Duma, G. Dobre, D. Demian, R. Cernat, C. Sinescu, F. I. Topala, M. L. Negrutiu, Gh. Hutiu, A. Bradu, and A. G Podoleanu, “Handheld scanning probes for optical coherence tomography,” Rom. Rep. Phys. 67(4), 1346–1358 (2015).
7. R. Cernat, T. S. Tatla, J. Pang, P. J. Tadrous, A. Bradu, G. Dobre, G. Gelikonov, V. Gelikonov, and A. G. Podoleanu, “Dual instrument for in vivo and ex vivo OCT imaging in an ENT department,” Biomed. Opt. Express 3(12), 3346–3356 (2012). [CrossRef] [PubMed]
8. M. Strathman, Y. Liu, E. G. Keeler, M. Song, U. Baran, J. Xi, M.-T. Sun, R. Wang, X. Li, and L. Y. Lin, “MEMS scanning micromirror for optical coherence tomography,” Biomed. Opt. Express 6(1), 211–224 (2015). [CrossRef] [PubMed]
9. D. Wang, P. Liang, S. Samuelson, H. Jia, J. Ma, and H. Xie, “Correction of image distortions in endoscopic optical coherence tomography based on two-axis scanning MEMS mirrors,” Biomed. Opt. Express 4(10), 2066–2077 (2013). [CrossRef] [PubMed]
10. Y. Li, “Beam deflection and scanning by two-mirror and two-axis systems of different architectures: a unified approach,” Appl. Opt. 47(32), 5976–5985 (2008). [CrossRef] [PubMed]
11. V.-F. Duma, P. Tankam, J. Huang, J. Won, and J. P. Rolland, “Optimization of galvanometer scanning for optical coherence tomography,” Appl. Opt. 54(17), 5495–5507 (2015). [CrossRef] [PubMed]
12. G. Sharafutdinova, J. Holdsworth, and D. van Helden, “Improved field scanner incorporating parabolic optics. Part 1: simulation,” Appl. Opt. 48(22), 4389–4396 (2009). [CrossRef] [PubMed]
13. D. Kessler, “Afocal beam relay,” United States patent US8531750 B2 (September 10, 2013).
14. C. L. Arrasmith, D. L. Dickensheets, and A. Mahadevan-Jansen, “MEMS-based handheld confocal microscope for in-vivo skin imaging,” Opt. Express 18(4), 3805–3819 (2010). [CrossRef] [PubMed]
15. S. Murali, K. P. Thompson, and J. P. Rolland, “Three-dimensional adaptive microscopy using embedded liquid lens,” Opt. Lett. 34(2), 145–147 (2009). [CrossRef] [PubMed]
16. C. Canavesi, A. Cogliati, A. Hayes, A. P. Santhanam, P. Tankam, and J. P. Rolland, “Gabor-domain optical coherence microscopy with integrated dual-axis MEMS scanner for fast 3D imaging and metrology,” Proc. SPIE 9633, 96330O (2015). [CrossRef]
17. J. P. Rolland, P. Meemon, S. Murali, K. P. Thompson, and K.-S. Lee, “Gabor-based fusion technique for Optical Coherence Microscopy,” Opt. Express 18(4), 3632–3642 (2010). [CrossRef] [PubMed]
18. K. S. Lee, K. P. Thompson, and J. P. Rolland, “Broadband astigmatism-corrected Czerny-Turner spectrometer,” Opt. Express 18(22), 23378–23384 (2010). [CrossRef] [PubMed]
19. P. Tankam, A. P. Santhanam, K. S. Lee, J. Won, C. Canavesi, and J. P. Rolland, “Parallelized Multi-Graphics Processing Unit Framework For High-Speed Gabor-Domain Optical Coherence Microscopy,” J. Biomed. Opt. 19(7), 071410 (2014). [CrossRef] [PubMed]
20. V.-F. Duma, “Optimal scanning function of a galvanometer scanner for an increased duty cycle,” Opt. Eng. 49(10), 103001 (2010). [CrossRef]
21. B. C. Kuo and F. Golnaraghi, Automatic Control Systems, 8th ed. (John Wiley, 2003).
22. B. Borovic, A. Q. Liu, D. Popa, H. Cai, and F. L. Lewis, “Open-loop versus closed-loop control of MEMS devices: choices and issues,” J. Micromech. Microeng. 15(10), 1917–1924 (2005). [CrossRef]
23. K. S. Lee, K. P. Thompson, P. Meemon, and J. P. Rolland, “Cellular resolution optical coherence microscopy with high acquisition speed for in-vivo human skin volumetric imaging,” Opt. Lett. 36(12), 2221–2223 (2011). [CrossRef] [PubMed]
24. P. Tankam, Z. He, Y.-J. Chu, J. Won, C. Canavesi, T. Lepine, H. B. Hindman, D. J. Topham, P. Gain, G. Thuret, and J. P. Rolland, “Assessing microstructures of the cornea with Gabor-domain optical coherence microscopy: pathway for corneal physiology and diseases,” Opt. Lett. 40(6), 1113–1116 (2015). [CrossRef] [PubMed]
25. P. Meemon, J. Yao, K.-S. Lee, K. P. Thompson, M. Ponting, E. Baer, and J. P. Rolland, “Optical Coherence Tomography Enabling Non Destructive Metrology of Layered Polymeric GRIN Material,” Sci. Rep. 3, 1709 (2013). [CrossRef]
26. P. Tankam, J. Won, C. Canavesi, I. Cox, and J. P. Rolland, “Optical assessment of soft contact lens edge-thickness,” Optometry Vis. Sci. 93(8) (2016, to be published).
