Study Population

Subjects with KC were recruited from the outpatient clinic of the Department of Ophthalmology. Age-matched healthy participants were recruited from volunteers of the Medical University of Vienna. All participants were informed about the procedure and, prior to OCT imaging, their written informed consent was obtained in accordance with institutional and legal requirements. Before inclusion in the study, all subjects underwent Scheimpflug imaging (Pentacam HR, Oculus, Wetzlar, Germany) and slit-lamp examination. Inclusion criteria for the KC group were, first, topographic findings considered typical for KC such as corneal thinning, central or inferior steep zone, and asymmetrical bowtie with or without skewed axis astigmatism ; and second, slit-lamp examination findings such as Vogt’s striae, Fleischer ring, apical thinning, or apical scarring. Exclusion criteria from assessment and analysis of ET maps for keratoconic eyes and healthy eyes were previous ocular surgery, corneal cross-linking or trauma, associated corneal comorbidities such as corneal scaring. Forty-seven eyes of 47 subjects with KC and 20 eyes of 20 healthy, age-matched volunteers were included in the study according to the inclusion criteria. No exclusions of eyes due to lack of signal quality were necessary. Details of the study population’s characteristics are shown in Table 1 .

Conical Scanning PS-OCT System

For this study, a recently introduced custom-designed PS-OCT machine with a conical scanning optics design was used. Briefly, the system is based on swept source PS-OCT operating at 1,045 nm central wavelength with ~100 nm bandwidth. To increase the signal-to-noise ratio in the corneal periphery, a conical scanning pattern was used that is generated by specially designed scanning optics containing an aspherical lens. This allows for almost perpendicular beam incidence on the corneal surface and results in good signal quality over the entire cornea (limbus-to-limbus). The instrument provides an A-scan rate of 100 kHz and an axial resolution of ~6.3 μm in corneal tissue. In less than 2 seconds, 150 B-scans were recorded for 1 volume acquisition by using a raster scan pattern. The PS-OCT method used in this trial incorporates a PS detection unit and records B-scan images of 2 orthogonal linear polarization states (co- and cross-polarized light) simultaneously. The co-polarized channel contains signals from polarization preserving tissue such as epithelium or BL. In addition, the channel shows structures within the stroma that backscatter light that has at least a polarization component parallel to that of the incident light. The cross-polarized channel shows mainly signal from within the stroma that is generated by structures that backscatter light that has a polarization component orthogonal to that of the incident light. With the beam incident nearly perpendicular to the corneal surface, this cross-polarized light is mainly caused by depolarization. Note that for this study, only light backscattered from the cornea is considered. Light transmitted through the cornea can also change its polarization state because of the birefringent nature of the corneal fibrils. This effect can be observed mainly for light reflected at the backside of the cornea (endothelium) or at posterior structures of the eye such as lens or retina (although this effect is weak in a healthy cornea at near-perpendicular beam incidence). During OCT imaging, each participant was positioned in front of the device and was instructed to look at a fixation target. Each eye was scanned 3 times during a single visit. After assessment of image quality 1 volume was chosen for further processing.

Segmentation of Corneal Layers and Generation of Color-Scaled Thickness Maps

The corneal epithelium and BL boundaries were segmented B-scan-wise by using custom-designed software (LabView 2017 SP1, National Instruments, Austin, Texas) as described in detail elsewhere, which yields highly reproducible thickness maps of the corneal layers.

Briefly, the corneal surface is detected in the co-polarized channel as the first highly reflecting interface, and the B-scan images were flattened by removing pixels anterior to the surface of the cornea and by image cropping. The entire flattened 3-dimensional (3D) volume was further smoothed in both lateral directions by applying a floating average window of 15 (A-scans) × 5 (B-scans) pixels. Next, the interfaces of the epithelium and the top surface of BL that are visible mainly in the image data of the co-polarized channel were segmented based on these images. Finally, the posterior surface of BL was segmented using the information of the cross-polarized channel as the visibility of this boundary is not highly visible in the co-polarized channel. Depolarizing structures contained in the stroma yielded high signal in the cross-polarized channel, which allows for segmentation of the upper stroma interface by using image data of this channel. The individual intensity thresholds for detection of each interface in each eye were chosen empirically in 1 B-scan to compensate for differences in image quality and were set according to the visually best segmentation results. Apart from setting the thresholds, the segmentation was performed automatically with the same settings for the entire volume data ( Figure 1 ). Only the volume data with the best image quality and accurate segmentation were chosen for further data evaluation. Based on the segmentation lines, en face thickness maps of corneal layers (ET and BLT) were computed in an area of 11-mm-diameter and plotted on a color scale. For quantitative analysis, the maps were subdivided into 25 sectors ( Figure 2 ). However, parts of the cornea were obscured by the upper and lower eyelid. Thus, sectors in the superior region (sectors 23 and 24) and inferior regions (sectors 19 and 20) were excluded from data evaluation ( Figure 2 ).

Central B-scans of a keratoconic eye (A) and a healthy eye (B with and without segmentation (C/D, respectively) after flattening by customized software. The automated segmentation of the corneal stroma, epithelium, and BL was performed by B-scans. Due to the conical scanning optics design and the assessment by 2 orthogonal polarization channels, individual corneal layers could be detected and delineated in a field of view diameter of 11 mm. The central specular reflex (asterisk) was eliminated by intensity thresholding. Bar = 500 μm in an axial direction and 1 mm in a lateral direction. AS = anterior surface; BL = Bowman’s layer; PS = posterior surface.

Proposed evaluation grid of 25 sectors that was applied on the 11-mm diameter thickness map. Sectors 19, 20, 23, and 24 were not evaluated in ET and BLT maps because these areas were often covered by the upper or lower eyelid, respectively. 0 = central area; NI = nasal inferior quadrant; NS = nasal superior quadrant; TI = temporal inferior quadrant; TS = temporal superior quadrant.

The precision (repeatability) of corneal layer thickness measurements has been carefully studied in a previous study in healthy subjects. The repeatability study covered the whole measurement and data processing chain (subject realignment, measurement repetition, and data processing including threshold setting). The precision of ET and BLT measurements was best for the central sector (ET, 0.5 μm; BLT, 0.3 μm) and decreased gradually to 2.5 μm (ET) and 2.4 μm (BLT) for the sectors of the outermost ring. The precision values are better than the axial resolution (coherence length) of the present system (6.3 μm), which is enabled by the fact that a signal peak or edge can be determined with better precision than the separation of 2 signals which is determined by the signal width.

ET and BLT Thickness Map-Based Parameters

To determine differences between keratoconic eyes and healthy eyes, the mean ET and BLT measurements of the remaining 21 sectors of the 11-mm diameter map were evaluated. The sectors were further grouped into 5 regions: central circular area and 4 quadrants (temporal inferior, temporal superior, nasal inferior, and nasal superior) ( Figure 2 ). As an additional parameter, the mean thinnest and mean thickest sector of the corneal epithelium (minET and maxET) and BL (minBLT and maxBLT) were determined. To assess irregularities of the epithelium and BL, ratios between the thinnest and the thickest sectors of the corneal epithelium, R1ET, and between the ET of the inferior temporal quadrant and the superior nasal quadrant (R2ET) were calculated. Similar ratios were calculated for BL (R1BLT and R2BLT).

Statistical Analysis

Descriptive statistical analysis was performed for all eyes. Vertical mirrored symmetry superimposition was used so that nasal/temporal characteristics could be combined; and ET/BLT values of left eyes were reflected in the vertical axis and superimposed onto the ET and BLT values of right eyes. The ET and BLT of the central circular area and the temporal superior, temporal inferior, nasal superior, and nasal inferior quadrants and the minET, minBLT, maxET, and maxBLT are given as mean ± SD. R1ET, R1BLT, R2ET, and R2BLT are also given as mean ± SD. To explore normal distribution, a Shapiro-Wilk analysis was performed. To explore statistically significant differences, means of all parameters were compared between KC subjects and normal subjects using a 2-sample t test, and Bonferroni adjustments for multiple comparisons were made. A 2-sided probability value less than 0.05 was considered significant. Receiver operator characteristics (ROC) curve analyses were used to evaluate the diagnostic power of those ET- or BLT-based parameters, which showed statistically significant differences between keratoconic eyes and healthy eyes. To calculate the diagnostic power of these parameters regarding their efficacy to distinguish between KC and healthy eyes, the areas under the curve (AUC) were determined and compared.