1. Introduction

Laser-based imaging modalities like scanning laser ophthalmoscopy (SLO) and optical coherence tomography (OCT) became indispensable for noninvasive in vivo diagnostics of the eye, especially the retina. The multimodal combination of these techniques offers several possibilities, which have been thoroughly published over the past decades [1,2].

Simultaneously, corneal confocal microscopy (CCM) and OCT became valuable tools for imaging the anterior segment of the living eye. Different modalities of CCM enabled cellular examination and characterization of the central and peripheral cornea, tarsal and palpebral conjunctiva or lid tissues [3]. Despite its advantageous value, CCM suffers from a small field of view, limited depth resolution and the lack of precise knowledge of image position and orientation inside the cornea. To overcome some of these limitations, especially the latter one, we applied the multimodal retinal imaging approach to corneal imaging. In order to achieve this, we adapted a dedicated lens module to a commercially available multimodal imaging platform for the posterior segment of the eye combining SLO and OCT. The lens module shifts the en face imaging plane from the retina to the cornea in the same way as it was previously described for single modal imaging systems [4,5].

OCT-guided CCM can potentially improve the diagnostic value and simplify the alignment procedure during contact measurements of the cornea due to the real-time information about CCM image position and orientation gained by OCT. As a matter of course, the axial resolution of OCT has been dramatically increased in the past years and thickness measurements of single corneal layers are available by ultrahigh-resolution OCT [6,7], but it does not provide cellular en face information of the cornea. While the OCT provides corneal thickness and depth data directly, its combination with CCM provides cellular en face information of distinct corneal layers, which are not available using the OCT technology. Addressing this lack of information, the OCT-guided corneal imaging concept, presented in this paper, offers the possibility of thickness measurements and simultaneous cellular en face imaging of the same region with a single device.

2. Materials and methods

The imaging system consists of a customized, modular lens adapter in conjunction with a modified SPECTRALIS OCT2 platform (Heidelberg Engineering GmbH, Heidelberg, Germany) with a high magnification objective (HMO, Heidelberg Engineering GmbH, Heidelberg, Germany) attached. This platform combines SLO and OCT. The lens adapter is a completely redesigned version of the so-called Rostock Cornea Module (RCM, Heidelberg Engineering GmbH, Heidelberg, Germany) and was presented recently as RCM 2.0 [5,8]. It utilizes a piezo actuator for closed-loop focal plane control, including through focusing, of up to 500 µm. Equipped with a water immersion objective lens (Achroplan 63x/0.95 W; Zeiss, Jena, Germany), the RCM 2.0 creates a flat imaging plane into the cornea. A sterile cap (TomoCap) is used to contact the cornea. Immersion gel is applied between the objective lens and TomoCap as well as in between TomoCap and cornea in order to reduce surface reflections. The TomoCap remains in a fixed position during focal plane changes but can be displaced beforehand in order to adjust the initial focus. This is especially useful for imaging of the topmost cell layers, where a certain distance between the TomoCap and corneal surface is necessary. Otherwise, if the focal plane is almost at the TomoCap surface, strong reflections arise from the refractive index difference and impede imaging. Thus, imaging of superficial cells requires a specific immersion gel layer between TomoCap and cornea.

CCM and OCT modalities share a common beam path through the SPECTRALIS. Therefore, the optical path length changes with the addition of the RCM 2.0 optics. In order to have an interference signal, the OCT reference arm has to be adjusted accordingly, see Fig. 1. That was realized by a modification to the axial scanner inside the SPECTRALIS OCT2.

 

Fig. 1 Simplified schematic of an SLO-OCT-combination and reference arm readjustment. The original position of the mirror in the reference arm (A) is moved (B) to account for the change in optical path length induced by the additional optics of the RCM 2.0. Red lines show different beam paths (dashed – OCT, dotted – SLO). Both beam paths are overlapped (dotdashed line) in such a way that they have the same focus. D – detector. LS – light source.

Download Full Size | PDF

Since the SPECTRALIS OCT2 is designed for the retina, it uses a refractive index of n_retina = 1.36 for the axial OCT scale. This needs to be rescaled according to the average refractive index of the cornea n_cornea = 1.376 [9]. OCT-guided CCM was performed simultaneously with an 805x805 μm² field of view at a wavelength of 815 nm for CCM (en face, xy-image) and 760x1898 μm² at a central wavelength of 880 nm with a bandwidth of 80 nm for the OCT B-scan (sagittal image, xz-section). Regarding the CCM image, this is an increase of almost a factor of five compared to the current HRT/RCM 2.0 combination with its 350x350 µm² [5]. This increased field of view is due to different optics and an increased scanning angle of 30°. The SPECTRALIS system offers high speed (HS) and high resolution (HR) modes, which affect the framerates and pixel resolutions of both modalities (HS: CCM with 768x768 pixels at 8.9 fps and OCT with 768x496 pixels at 90 fps; HR: CCM with 1536x1536 pixels at 4.8 fps and OCT with 1536x496 pixels at 50 fps). The OCT images are cropped in the axial direction, as the corneal thickness is only a fraction of the OCT imaging depth. Please note the anisotropic scale in the OCT image and also the slightly different x-scales between OCT and CCM image. Both may be corrected by a rescaling done in a post-process. However, it is better to address the latter by a scanning angle calibration.

Due to a sufficient refractive index change at the cornea’s anterior and posterior surface, the total corneal thicknesses can directly be measured in the OCT image. Inside the cornea, the refractive index change is very small. Structures like the Bowman’s membrane can be identified using the SPECTRALIS OCT2 combined with the Anterior Segment Module. However, due to small depth of focus induced by the RCM 2.0, the intensity decreases rapidly for out-of-focus structures. Therefore, particular corneal layer interfaces can only be visualized in the OCT channel, if CCM imaging plane is near the interface. Although this enables direct measurements of the epithelial thickness, this does not apply to thin layers as for instance the Bowman’s membrane. Hence, direct thickness measurements of corneal substructures are challenging. Anyhow, they are feasible by various techniques. In case of epithelial thickness, these are namely high-frequency ultrasound [10], high-resolution OCT [11–13] or confocal microscopy through focusing [14]. But while the first two methods do not offer a corresponding high-resolution en face image at the thickness measurement position, the latter one does. Nevertheless, due to post-processing and 3D stack reconstruction, this method is very time consuming and complex [5,15]. Another solution to overcome this problem is to exploit oblique CCM sections in conjunction with their corresponding OCT scans. The OCT image contains the angle α between corneal surface and CCM imaging plane. With this additional information, the layer thickness can be calculated. Figure 2 demonstrates the calculation of the layer thickness t by trigonometric relations and the approximation of a plane corneal surface for a small arc length. The OCT imaging plane lies in the paper plane and the CCM imaging plane goes into it. The projected distance d and the angle α are measured in the CCM and OCT image, respectively. The thickness t is then simply the product of the distance d and the sine of α.

 

Fig. 2 Exemplary determination of the epithelial thickness t by an oblique cornea section. The OCT imaging plane lies in the paper plane while the CCM imaging plane (gray) goes into the paper. The angle α is measured in the OCT image and the distance d in the CCM image.

Download Full Size | PDF

The following results are measurements performed on a 51-year-old healthy male subject. The study was conducted in accordance with the Declaration of Helsinki and it was explained in detail to the subject. Informed consent was obtained before any investigative procedures were conducted. In this feasibility study, no further subjects were examined.

3. Results

Figure 3 to 5 combine simultaneously taken CCM and OCT images of a cornea. Since both modalities share the same optical path in the RCM 2.0, they also have approximately the same focal point. Due to the tight focus of the RCM 2.0 objective lens, the most intense backscattered OCT signal originates from the OCT focus, which is essentially the place of the CCM image plane. This is the reason, why the CCM image plane is visible in the OCT image and why its movement in the OCT image can be easily traced while changing the focus or in through focusing experiments during image stack acquisition (Fig. 3(A)). In the presented case of Fig. 3, the OCT image displays the cornea’s anterior (here in contact with the TomoCap) and posterior interface. This allows corneal thickness measurements in the OCT image. In Fig. 3 the total cornea thickness is 593 µm and the CCM imaging plane is 476 µm below the corneal surface. Compared to the published values of the central corneal thickness [16], the measured value seems to be too large. Corneal thickness measurements by Scheimpflug photography (Pentacam, Oculus, Germany) provided 548 µm centrally and 600 µm in 3 mm distance from the apex. Presumably, the Fig. 3 was taken paracentral. The OCT image also reveals a so-called mirror artifact arising from the Fourier transformation [17]. This artifact is identified by the opposite direction of curvature. Furthermore, the TomoCap surface and the CCM image plane do not appear to be flat. This is most likely due to an improper optical path length calibration. Hence, the cornea shows a wrong curvature in the OCT images. This problem can be addressed by image post-processing or by software calibration of the imaging system’s optical path length. Beyond that, a locally dependent intensity scaling may help to increase the visibility of the surfaces in OCT images.

 

Fig. 3 Sagittal OCT (A) and en face CCM (B) image of the posterior stroma. The OCT image shows the anterior corneal surface in contact with the TomoCap (1), a mirror artifact (2), the CCM imaging plane (3) and the posterior corneal surface (4). The area in the green rectangle is magnified and depicted in a pseudocolor image (C) for visual purpose. The same area was used for the averaged intensity profile (D). Direct measurements provide a corneal thickness of 593 µm and a CCM imaging plane depth of 476 µm. Please note the anisotropic scale in the OCT images (A) and (C).

Download Full Size | PDF

 

Fig. 4 OCT (A) and CCM (B) image showing the cornea before contact with the TomoCap. The OCT image exhibits a mirror artifact (1), the contact surface of the TomoCap (2), the intersection of the CCM imaging plane with the corneal surface (3), the CCM imaging plane also visible outside the cornea due to the immersion gel (4), and the corneal surface (5). The CCM image shows an oblique section starting from outside the cornea (middle), crossing the epithelium and showing some structures of the Bowman’s membrane (top). Due to the low frame rate of the HR mode, the CCM image shows wavy movement distortions (6).

Download Full Size | PDF

 

Fig. 5 Sagittal OCT (A) and CCM (B) image of an oblique section. A part of the OCT image was rectified (see text) and stretched to isotropic scaling (C) in order to measure the angle between the corneal surface (1) and CCM imaging plane (2). The CCM image shows the anterior stroma (3), the Bowman’s membrane with the subbasal nerve plexus (4), the epithelium (5) and the superficial cells of the epithelium (6). The oblique imaging angle α and the projected layer distances d_E and d_B result in an epithelial thickness of 46 µm and a Bowman’s membrane thickness of 15.3 µm.

Download Full Size | PDF

Figure 3(B) reveals central surface reflections caused by suboptimal coatings of the RCM 2.0 objective lens. Also, the image corners are obscured. An improved optical system may help to overcome these problems.

Figure 4 shows the corneal surface in the OCT image even in regions where it is invisible in the CCM image, while also showing the CCM imaging plane outside the cornea in the OCT image due to immersion gel. This feature can be exploited for easier alignment in the first stage of a subject examination. By considering two perpendicular OCT scans the operator is able to align the CCM imaging plane to the cornea’s apex before the TomoCap touches the cornea. Please note that Fig. 4 was captured in HR mode with the lower framerate. Hence, movement distortions (Fig. 4(B): arrows denoted with 6) are more likely to occur as seen in the CCM image. In order to reduce eye movements, a dedicated curved contact cap was proposed in [5].

An oblique CCM image and the corresponding OCT image are shown in Fig. 5. The epithelial and Bowman’s membrane thicknesses were exemplary determined in the same way as proposed in section 2. Therefore, the OCT image (Fig. 5(A)) was rectified to correct the wrong curvatures and stretched to obtain an isotropic resolution (Fig. 5(C)). To rectify this image region, its rows are shifted horizontally until their maximum intensity values have the same horizontal position. After this post-process, the CCM imaging plane appears as a straight vertical line in the OCT image. The angle between the corneal surface and CCM imaging plane was approximated to (10.5 ± 0.5)°. The projected image distances in the CCM image (Fig. 5(B)), anterior to posterior epithelium surface and anterior to posterior Bowman’s membrane surface, are d_E = (250 ± 1) µm and d_B = (84 ± 1) µm, respectively. This results in an epithelial thickness of (46 ± 5) µm. The uncertainty was analytically calculated from trigonometric relations with the approximated single values uncertainties and assuming an undetermined corneal radius of curvature between 7.5 mm and infinity, which corresponds to completely applanated. In the same way, the Bowman’s membrane thickness is calculated to 15.3 µm with an uncertainty of at least ± 1.5 µm. For improvements the before mentioned software calibration and edge detection and automated line fits are necessary, making a post-process obsolete. Nevertheless, compared to other values of the epithelial thickness ranging from 42 µm to 55 µm [18–20], the determined value lies almost in the middle of this range. Also, the determined value of the Bowman’s membrane thickness is in good agreement with values obtained by other methods, e.g. confocal microscopy through focusing [14] or using a sub-micrometer axial resolution OCT [7].

4. Discussion and conclusions

The presented preliminary results demonstrate how CCM benefits from simultaneous OCT imaging. A variety of publications dealing with bimodal retinal imaging by means of SLO and OCT have been reported in the literature. The SLO is usually used to support the OCT imaging, e.g. for image stabilization purposes. To the best of our knowledge, this combination has not been reported in case of corneal imaging. In contrast to retinal imaging, the approach presented here utilizes OCT information to support CCM imaging instead. Thus, it is possible to assign any CCM image with its correct imaging depth and orientation. Furthermore, the possibility of detecting the corneal surface in the OCT image, even before the TomoCap touches the cornea and before it is visible in the actual CCM image, greatly simplifies the alignment procedure in the initial step of a subject examination. In this context, an additional perpendicular OCT scan line would allow to pre-align the CCM to a desired position, e.g. cornea’s apex, and improve location-based diagnosis.

As a further substantial benefit, this system achieves an almost five times larger field of view while maintaining or even increasing the in-plane resolution depending on the acquisition mode (HS or HR) used, but currently at the cost of a decreased framerate compared to the HRT/RCM 2.0 system.

The presented imaging method using CCM and OCT simultaneously is just an initial step, but it may be helpful in experimental as well as clinical usage upon further improvement and optimization. This includes e.g. (1) OCT reference arm adjustment, (2) elimination of parasitic lens reflections, (3) utilization of the full 2D scanning range in order to get rid of the obscured corners in the CCM images, (4) the calibration of optical path length to get flat imaging planes and (5) locally dependent intensity scaling. Furthermore, a verification of the thickness measurements by comparing to other techniques is necessary and phantom studies are recommended for that purpose. Additionally, the inhomogeneity of the cornea and its refractive index [21] must be taken into consideration in order to further increase the precision of corneal measurements. Finally, thorough evaluations in human studies will answer the question for usability and advantage compared to established systems. In conclusion, even though there are many challenges that have to be addressed in the future, our initial results can be pathbreaking for the development of an improved novel corneal diagnostic.