1.

Introduction

Endoscopic optical coherence tomography (EOCT) is receiving increasing attention because this imaging technique has the potential to visualize subsurface structures inside cavities and hollow organs. EOCT devices can be classified as either forward or side imaging. In general, side imaging OCT probes are two-dimensional (2-D) in nature. These probes commonly consist of a fiber, a lens for focusing, and a prism for lateral deflection of the light. The combination of rotational side imaging and back-and-forth translation of the probe is used to generate three-dimensional (3-D) images.^{1, 2, 3} Side imaging probes are used for imaging hollow organs, such as the intravascular system or the gastrointestinal tract.^{3 , 4} In comparison, forward imaging probes are more technically challenging because the generation of a 3-D image cannot be realized by the back-and-forth translation. To achieve 3-D information of the investigated structures, scanning is realized by, e.g., a piezoelectric transducer, galvanometer scanners, or a microelectromechanical system.^{5, 6, 7} That makes it very difficult to maintain the overall diameter small and still achieve a large working distance (WD) and a broad field of view (FOV). While being immune against normal vibrations, forward imaging probes are needed in different application areas, such as gynecology, urology, or otorhinolaryngology.^{8, 9, 10} The intention of our research was to develop a suitable EOCT device especially for imaging the tympanic membrane (TM) and behind it. Because the TM lies at the end of the auditory canal, EOCT is intended to detect diseases of the TM and the middle ear, such as otitis media or cholesteatoma.^{10, 11, 12} Cholesteatomas show characteristic findings in OCT imaging that differentiate them from healthy tissues.¹⁰ The detection of otitis media by otoscopy is still difficult, even for a specialist. Chronic cases of otitis media are accompanied by a bacterial biofilm behind the TM that could be visible with OCT, but cannot be detected with a regular otoscope.¹² OCT allows depth-resolved imaging of the TM and structures behind it and creates new opportunities in disease detection.¹³ It could be possible to distinguish between the acute and chronic form and thereby avoid the overuse of antibiotics that are not necessary in a chronic case and can lead to resistant bacteria.¹² Also swelling and inflammation of otitis media could be visible in earlier stages. To avoid direct contact to the inflamed tissues and allow imaging of the entire TM, an EOCT system with a large WD and a broad FOV is necessary. Because of the position of the TM, a forward imaging probe is more feasible. Furthermore, we decided to develop a rigid probe because the generation of a 3-D image can be executed outside the body. There are no moving parts inside the auditory canal that are required for 3-D imaging, allowing us to achieve a smaller probe diameter. Additionally, rigid optics designs still have a better optical quality in contrast to fiber-optic bundles,¹⁴ so the axial and lateral resolution can be kept in a sufficient region to allow the visualization of microstructures of the investigated tissues.

2.

Materials and Methods

2.1.

Endoscopic Optics

The optics of the EOCT scanning unit consisted of a collimator for light coupling; two galvanometer scanners for 3-D imaging; a triplet (Edmund Optics, Barrington, New Jersey, NT45-251), a gradient-index (GRIN) relay lens (GRINTECH GmbH, Jena, Germany, GT-IFRL-200-100-10-NC), a GRIN objective lens (GRINTECH GmbH, Jena, Germany, GT-IFRL-200-010-50-NC) for OCT imaging; and glass fibers for illumination (Fig. 1). Additionally, a dichroic mirror, which is reflective for infrared and transparent for visible light, allowed the coupling of the camera beam path into the sample arm for online video image guidance in the visible (VIS) spectral range [Fig. 2(b)]. In the camera beam path the light from the sample arm was focused by a lens on the CMOS video camera sensor. That opened the opportunity to match the focus for VIS imaging to the focus of the OCT beam path. The collimator and the triplet were chosen for using the whole numerical aperture (NA) of 0.1 of the GRIN relay lens. An enhancement of the NA is possible but would result in an undesired shortening of the GRIN relay lens length, and the combination of multiple relay lenses would lead to a worsening of the optical aberrations. The 1.0 pitch GRIN relay lens allows image guidance over a long distance and is about 100 mm in length and 2 mm in diameter. The lens was chosen for its sufficient length for imaging the TM inside the auditory canal (the length of which is approximately 30 mm), as well as its small diameter. A GRIN objective lens with an NA of 0.5 was placed into the sample arm for the realization of both a large WD and a broad FOV. Additional VIS illumination was achieved by the use of glass fibers. For this, an adapter was designed that arranged the glass fibers around the GRIN relay lens in a ring shape (Fig. 1). The glass fibers were supplied with VIS light by a high power SMD LED (surface-mounted device light emitting diode), which was installed on the housing of the scanner head. The GRIN optics and the glass fibers were housed in a metal tube for protection, which gave an overall diameter of 3.5 mm. To improve the OCT image quality, we corrected the optics for dispersion using glass compensation in the reference arm. Therefore, a similar GRIN relay lens was inserted. To avoid additional surfaces, this relay lens further acts as the reference mirror by carrying a gold layer on the end surface (Fig. 2). The achieved depth resolution after dispersion compensation was approximately 7.6 μm compared with 6.4 μm without the additional GRIN optics.

Fig. 1

Schematic of the endoscopic configuration. (a) A triplet is used to focus the incident light inside the GRIN relay lens. This lens guides the light to its distal end and focuses it on the surface of the GRIN objective lens and the GRIN objective lens focuses the beam on the sample. (b) The whole optics and the illumination are housed in a metal tube. (c) Top view of the distal end of the EOCT device.

Fig. 2

Endoscopic optical coherence tomography (EOCT) spectral domain setup. (a) Superluminescent light-emitting diode (SLED), standalone spectrometer, and fiber coupler (FC). (b) Optics of the scanning unit of the EOCT device with integrated Michelson interferometer. The light from the collimator (C) is split by the beam splitter (BS) into reference and sample arm. GS: galvanometer scanner, DM: dichroic mirror, CMOS: video camera sensor, M: mirror, DC: dispersion compensation, RM: reference mirror, L1: triplet, L2, L3: lenses, GRIN: gradient index lens. Orange: OCT beam path, blue: VIS beam path. (c) Fan-shaped optical scanning pattern.

3.

Experimental Results

3.1.

Specifications of the Optics

EOCT is required to detect diseases of the TM and the middle ear. To avoid direct contact with the inflamed tissues and allow imaging of the entire TM, EOCT systems must have a large WD and a broad FOV. Thus, we designed gradient index optics for forward EOCT imaging. Our scanner head had a maximum WD of $\sim 10\mathrm{mm}$ with a corresponding FOV of $\sim 10\mathrm{mm}$ [Fig. 3(c) and 3(f)], thus allowing the entire TM, which is generally 10 mm in diameter, to be imaged in one view. The lateral resolution was dependent on the WD; the shorter the WD, the better the lateral resolution. We measured the lateral resolution by OCT with the United States Air Force (USAF) resolution test target, finding that a WD of $\sim 10\mathrm{mm}$ resulted in a lateral resolution of $\sim 28\mathrm{\mu m}$ , WD of $\sim 7\mathrm{mm}$ in a lateral resolution of $\sim 22\mathrm{\mu m}$ , and WD of $\sim 5\mathrm{mm}$ in a lateral resolution of $\sim 20\mathrm{\mu m}$ [Fig. 3(a)–3(c)]. Similarly, we found that the FOV was dependent on the WD. We determined the FOV by OCT with a millimeter paper test chart, detecting a FOV of $\sim 10\mathrm{mm}$ at a WD of $\sim 10\mathrm{mm}$ , $\sim 7\mathrm{mm}$ at a WD of $\sim 7\mathrm{mm}$ , and $\sim 5\mathrm{mm}$ at a WD of $\sim 5\mathrm{mm}$ [Fig. 3(d)–3(f)]. The data correspond to our calculated theoretical lateral resolutions: $\sim 25\mathrm{\mu m}$ at 10 mm, $\sim 21\mathrm{\mu m}$ at 7 mm, and $\sim 18\mathrm{\mu m}$ at 5 mm (Fig. 4). The theoretical data were determined with the lens equation, treating the GRIN lenses as ideal lenses. If a better lateral resolution than 28 μm is necessary, an enhancement is possible by adjusting the distance between the triplet and the GRIN relay lens in the sample arm. This change in distance resulted in a shortening of the WD. Furthermore, a smaller FOV that could be beneficial for smaller TMs, such as the TMs of children, could be achieved through the distance change.

Fig. 3

Images taken for measuring the lateral resolution ( $\mathrm{\Delta }x$ ) and field of view (FOV) in dependence of the adjusted working distance (WD). (a)–(f) The three-dimensional OCT-data cube was averaged in axial direction. As the information is concentrated in a few planes this is similar to an enface image of the deformed plane. (a)–(c) USAF resolution test target. (d)–(f) millimeter paper. (a) $\mathrm{\Delta }x$ $\sim 20\mu m$ in a WD of $\sim 5\mathrm{mm}$ with corresponding FOV of $\sim 5\mathrm{mm}$ (d). (b)  $\mathrm{\Delta }x$ $\sim 22\mu m$ in a WD of $\sim 7\mathrm{mm}$ with a FOV of $\sim 7\mathrm{mm}$ (e) and (c) $\mathrm{\Delta }x$ of $\sim 28\mu m$ in a WD of $\sim 10\mathrm{mm}$ with a FOV of $\sim 10\mathrm{mm}$ (f). Unfortunately, some fissures of the distal end of the gradient index (GRIN) relay lens are visible in the boundary area of the FOVs.

Fig. 4

Measured and theoretical lateral resolution of the OCT beam path in different working distances. The measured resolution (red line) was determined with the USAF resolution test target. The theoretical curve (blue line) was determined from the lense equation.

3.2.

Image Correction

In the case of our endoscopic optics, the use of the fan-shaped optical scan pattern caused a field curvature. The position of the beam intersection is dependent of the image height, so that an object point that was far away from the optical axis resulted in an image point, which was shifted in axial direction. That means that the image was not generated on a plane, but on a curved surface. In the favorable case of imaging, the TM the cone form of the TM is faced in the same direction, much like the field of curvature caused by the fan-shaped scanning optics, achieving good image quality for that special case. Nevertheless, a slight error remained and the images had to be corrected (Fig. 5). At first, they were rescaled to achieve equal resolutions in $X$ - and $Y$ -direction and afterwards they were transferred to a linear scan by means of a digital unwrap image process. In this process, a selected region was unwrapped from a circular strip, like the surface of the millimeter paper in Fig. 5(a), to a rectangular strip [Fig. 5(b)] by the use of a coordinate transformation and a linear interpolation in both $X$ - and $Y$ -direction to compute the pixel location. First of all, we tested the correction on a millimeter paper and then on a human TM [illumination and camera beam 5(c) and 5(e)].

Fig. 5

Image correction. Optical coherence tomography (OCT) cross-sectional image of a millimeter paper (a) and the coressponding corrected image (b) and an OCT cross-sectional image of a TM (c) with the corrected image (d). The green boxes show the coordinate transformation.

3.3.

TM Imaging Ex vivo

We tested the performance of our EOCT device by 3-D imaging a human TM ex vivo. Representative OCT images are shown in Fig. 6. Typical regions of the TM, such as the manubrium of malleus (MM), were detected and a 3-D OCT image of the whole TM was recorded. A cross-section of the TM in the region of the umbo, the deepest point in the cone of the TM and a part of the MM were visible [Fig. 6(a) and 6(b)]. We also tested the additional illumination and camera beam path [Fig. 6(d)], finding that our system is suitable to image the whole TM inside the auditory canal with one view. Different regions of the TM such as the umbo or the MM could be distinguished in the VIS 2-D enface image, thus providing additional video image guidance during the investigation. The cone form of the TM was clearly visible when acquiring OCT images of the whole TM [Fig. 6(e)].

Fig. 6

Representative images of a tympanic membrane (TM) ex vivo acquired with the endoscopic setup. Optical coherence tomography (OCT) cross-sectional images through (a) the umbo, the deepest point of the TM, and (b) a part of the manubrium of malleus (MM). (c) top view OCT image of the TM and (d) top view camera picture of the TM acquired with the integrated VIS illumination, the umbo (*) and the MM (**) can be identified. (e) 3-D OCT image of the TM. Red and green frame belong to the position of (a) and (b). Scale bars correspond to a distance of 1 mm (Video 1, MPEG, 1.23 MB) [URL: http://dx.doi.org/10.1117/1.JBO.17.7.071302.1] 10.1117/1.JBO.17.7.071302.1

3.4.

TM Imaging In vivo

In a recent study, we tested the performance of our EOCT device by imaging a human TM in vivo. Therefore, a healthy human volunteer was positioned on a headrest and the examiner inserted the distal end of the endoscopic setup into the auditory canal. A camera image with the executed scan pattern is visible [Fig. 7(a)]. The corresponding B-scans in $X$ - and $Y$ -direction showed the suitability to distinguish between different regions of the TM. In the X-scan, a part of the MM was discernible, and in the Y-scan, some other ossicular structures and some bony structures behind the TM could be detected [Fig. 7(b) and 7(c)]. Additional a VIS 2-D enface image was acquired showing the feasibility of the camera beam path in the in vivo situation [Fig. 7(d)]. Thus, our EOCT device is suitable for imaging the TM in vivo.

Fig. 7

Representative images of a tympanic membrane (TM) in vivo acquired with the endoscopic setup. (a) 2-D VIS camera image showing the scan pattern. OCT cross-sectional images through (b) a part of the manubrium of malleus in X-scan direction and (c) a part of rest of the ossicular chain and some bony structures in Y-scan direction. (d) Top view camera picture of the TM acquired with the integrated VIS illumination, the umbo (*) and the MM (**) can be identified. Scale bars correspond to a distance of 1 mm.

4.

Discussion

We developed a suitable EOCT device because there is a clinical need for the detection of diseases concerning the TM and the middle ear. The system should achieve a large WD to avoid direct contact to the inflamed tissues and a broad FOV to allow acquiring images of the whole TM with one recording. For this, we designed a gradient index optics for forward EOCT imaging. In comparison to the previous work reported so far, this is the first forward imaging endoscopic system achieving a broad FOV while maintaining the overall diameter of the probe small, achieving a ratio of approximately 1:3 (overall diameter:FOV). The smallest reported forward imaging probe has an outer diameter of 0.82 mm and was developed for ophthalmic OCT.¹⁵ The optical setup consists of two counteracting gradient index (GRIN) lenses. This probe achieves a WD of approximately 0.78 mm and has a maximum tilt angle of 15.3-deg, not sufficient for the aforementioned application. The EOCT probe with the broadest FOV that has been presented thus far achieves a FOV of 6 mm, but has a large oval diameter of $7.5\times 3\mathrm{mm}$ .⁷ This system was developed for arthroscopic OCT. The optical layout mainly consists of a scan lens. The forward imaging probe with the largest WD reported so far achieves an adjustable WD up to 7.5 mm, and was developed for example for thoracoscopic imaging of pleural cancer.^{16 , 17} Unfortunately, the FOV is limited by the diameter of the selected lens and is 4.5 mm by a nearly similar overall diameter of the probe. In this optical setup, a GRIN relay lens was chosen. A further system mentioned in the literature for side imaging with large WD could easily be adapted to forward imaging by taking away the prism, but show other drawbacks solved with our design. The works of Yu et al.¹⁸ and Guo et al.¹⁹ describe this system that operates at a WD of 65 mm with a lateral scanning range of 5 mm. The setup is realized in a “double-barreled” configuration with the video endoscope separated from the OCT device. That leads to an unnecessary enlargement of the overall diameter. In general, the WD can be increased arbitrarily but result in a worsening of the numerical aperture and thus of the amount of light back scattered from a target. A compromise has to be found to achieve a feasible WD and FOV, and anyhow a sufficient numerical aperture and overall diameter of the probe. In our EOCT device, the camera beam path follows the same way as the OCT beam path. That allows keeping the diameter of the probe much smaller and because of the suitable NA for OCT a sufficient image quality was achieved.

5.

Conclusion

In summary, we developed a new design for 3-D forward imaging endoscopic OCT. The system allows images to be acquired at a large WD with a large FOV. This is done by maintaining the outer diameter for the distal end small, and consequently enables access to cavities not accessible with bulky optics. When applying this to imaging the TM, a rigid GRIN optics design allows the development of a probe with an outer diameter not larger than 3.5 mm to gain access through the auditory canal. In addition, a camera beam path and illumination via glass fibers were incorporated into the probe, enabling video endoscopy inside cavities. Ex vivo and in vivo investigations with the EOCT device were performed. Thus, the developed device is suitable for acquiring images of the TM. An application in other medical or industrial fields is also feasible.