The Role of Axial Resolution of Optical Coherence Tomography on the Measurement of Corneal and Epithelial Thicknesses

Abstract

Purpose.: Our purpose was to investigate the role of the axial resolution of optical coherence tomography (OCT) on the measurement of corneal and epithelial thickness by evaluating the repeatability and agreement among different OCT devices with different axial resolutions.

Methods.: Twenty right eyes of 20 healthy subjects (age: 22.3 ± 1.3 years) and 18 eyes of 18 patients (age: 25.7 ± 6.8 years) after laser in situ keratomileusis (LASIK) refractive surgery were tested. Each subject was imaged using four OCT devices: ultra-high resolution OCT (UHR-OCT), ultra-long scan depth OCT (UL-OCT), commercial RTVue, and Visante. The OCT images obtained from UHR-OCT, UL-OCT, and RTVue were processed with a custom automated algorithm for measuring the central corneal thickness (CCT) and central epithelial thickness (ET). CCT measurements from pachymetry maps that were generated by RTVue and Visante were also obtained.

Results.: For both groups, the CCT and ET measured by UHR-OCT and UL-OCT were highly correlated with RTVue when the automated image processing algorithm was used. The CCT measurements from the RTVue and Visante pachymetry were thicker than those measurements obtained from the automatic algorithm. The coefficient of repeatability was less than 4.9 μm in the healthy subjects and 7.9 μm in the LASIK patients, and the associated intraclass correlation coefficient (ICC) was greater than 0.992 in both groups for the CCT measurements. For the ET measurements using UHR-OCT, UL-OCT and RTVue, the coefficient of repeatability was less than 2.2 μm in the healthy subjects and 4.8 μm in the LASIK patients with an ICC that was greater than 0.84.

Conclusions.: The axial resolution of OCT may play a role in determining the precision with which the CCT and the ET can be measured, although it may not affect the measurement of results.

Introduction

Accurate measurement of the corneal thickness is an important part of the ocular evaluation. Measurement of the corneal thickness is essential to the preoperative assessment for refractive surgery such as laser in situ keratomileusis (LASIK). ^1,2 Overestimation of the corneal thickness may result in excessive ablation and increase the risk of keratoectasia. Conversely, underestimation may lead to the exclusion of potential refractive surgery. ^3,4 The central corneal thickness (CCT) is an essential part of the evaluation of various conditions, including keratoconus, ^5–7 refractive surgery–induced keratoectasia, ⁸ contact lens or other factor-induced corneal edema, ^9–12 and endothelial function. ^13,14 CCT is also an important factor in the measurement of intraocular pressure in glaucoma. ^15,16 The central epithelial thickness (ET) measurement is as important as the CCT measurement. The epithelial profile has been used as a diagnostic tool for keratoconus ^5,6,17–19 and as an indicator of the possible effects of contact lens wear. ^11,12,20 The epithelial thickness may also be used to monitor the development of corneal ectasia, in which the epithelium becomes thinner. ²¹ It has been suggested that the epithelial thickness is a significant factor in determining the outcomes of corneal refractive surgery and orthokeratology. ^20,22  

Many different types of ophthalmic devices are available to measure the CCT and the ET, such as ultrasound biomicroscopy, ¹⁸ confocal microscopy, ²³ and optical coherence tomography (OCT). ^9,10,24 OCT has been widely used as a noninvasive, noncontact optical imaging method for measuring corneal and epithelial thicknesses in vivo. ^9,10,25,26 Previous studies have demonstrated that OCT is a reliable monitoring tool in various conditions, including keratoconus, ^6,7 corneal tumor, ²⁷ and corneal edema. ^16–18 Tao et al. ²⁸ used ultrahigh-resolution OCT (UHR-OCT) to determine the topographical thicknesses of the corneal epithelium and Bowman's layer. Shen et al. ^29,30 utilized ultra-long scan depth optical coherence tomography (UL-OCT) to image the entire ocular surface shape and its interaction with a soft contact lens. Commercial OCT devices such as RTVue have also been used to measure corneal thickness ^16,31,32 and epithelial thickness. ²⁶ Different measurements of CCT and ET have been determined with various OCT devices. ^28,32,33 In addition, it can be speculated that more OCT devices with higher axial resolutions will be available for research and routine clinical use. One may assume that higher axial resolutions might result in better information for thickness measurements, with a precision that is higher than the resolution. However, it is unclear whether the axial resolution of OCT plays a role in the measurement of CCT and ET. This unanswered question may prevent researchers and clinicians from directly interpreting the measurements and comparing the results obtained with different devices. The purpose of this study was to investigate the role of axial resolution on the measurement of corneal and epithelial thicknesses by evaluating the measurement repeatability and the agreement among four OCT devices with different axial resolutions. 

Methods

This study was approved by the Office of Research Ethics, Wenzhou Medical College and was conducted in accordance with the tenets of the Declaration of Helsinki. Twenty healthy subjects (11 men and 9 women, mean ± standard deviation [SD] age: 22.3 ± 1.3 years) and 18 patients (9 men and 9 women, mean age ± SD: 25.7 ± 6.8 years) after LASIK refractive surgery were recruited from the Refractive Surgery Center of Wenzhou Medical College. The best-sphere refraction in the healthy eyes was −2.71 ± 2.57 D (range, 0 D to −9.00 D) and the astigmatic refractive error was −0.63 ± 0.14 D (range, 0 D to −0.75 D). For the patients after LASIK, the best-sphere refraction prior to the surgery was −5.09 ± 1.68 D (range, −3.50 D to −8.00 D) and the astigmatic refractive error was −0.46 ± 0.46 D (range, 0 D to −1.5 D). All of the subjects were able to fixate well upon the fixation target that was used. Informed consent was obtained from each subject prior to enrollment. Those individuals with previous ocular surgery (except for refractive surgery), a history of trauma, ocular pathology, recent treatment with ocular medication, and contact-lens wear were excluded. Each subject was imaged during two imaging sessions that occurred on two separate days within one week. 

The central cornea was imaged with four OCT devices, including two prototypes (UHR-OCT and UL-OCT) and two commercially available devices (RTVue OCT [Optovue, Inc., Fremont, CA] and Visante OCT [Carl Zeiss Meditec, Inc., Dublin, CA]). The configurations of the four OCT devices are listed in Table 1. As described previously, our UHR-OCT setup has a three-module super luminescent diode (SLD) light source with a broad bandwidth of 100 nm and a central wavelength of 840 nm. ³⁴ The scan width was set to 8 mm and the scan depth was calibrated as 1.85 mm in tissue. We measured the point spread function of the interference spectrogram near the zero delay line, and the calibrated axial resolution was approximately 3 μm in the cornea. ^34–38 As described previously, our UL-OCT setup has an SLD with a wavelength centered at 840 nm and a full width at the half-maximum bandwidth of 50 nm. ³⁹ The scan width was set to 8 mm, and the measured axial resolution for this setup was ∼7.5 μm in the cornea. ³⁹ The scan depth was calibrated as 5.6 mm in the tissue. Based on the manufacturer's guidelines, the RTVue uses an SLD with a wavelength centered at 830 nm and bandwidth of 50 nm that yields an axial resolution of 5 μm in the tissue. ¹⁶ Because the RTVue was primarily designed to image the retina, a high magnification corneal lens adapter (CAM-L) with a 1.96-mm scan depth in the tissue and a 6-mm scan width was added to the RTVue probe to image the cornea. For the RTVue, two scan modes, a pachymetry scan mode and a cross line scan mode, were used to image the cornea. Using the pachymetry scan mode, a pachymetry map was automatically generated and divided into three zones by octants and annular rings. Using the cross line scan mode, we could export the raw OCT files and use our custom automatic code to generate the images. The light source of the Visante OCT is an SLD with a center wavelength of 1310 nm that, according to the manufacturer, yields an axial resolution of 18 μm. The “pachymetry map” scan mode has a scan width of 10 mm with a depth of 3 mm in the tissue. The pachymetry map was generated automatically, and corneal thickness values were provided for four areas. In this study, the mean corneal thickness in the central zone was treated as the measured CCT for the Visante and RTVue. 

Table 1. 

 

Configurations of OCT Systems

OCT System Maker	UHR-OCT Custom	UL-OCT Custom	RTVue Optovue	Visante Zeiss
Wavelength, nm	840	840	830	1310
Axial resolution, μm	3	7.5	5	18
Scan depth in tissue, mm	1.85	5.6	1.96	3
Analyzed scan width, mm	1	1	Cross line: 1; pachymetry: 2	2
Scan mode	Horizontal line	Horizontal line	Cross line; pachymetry	Pachymetry map
Image pixel interval, μm	0.9	2.7	3.1

Calibration of the spectral domain OCT instrument has been described previously. ²⁸ Ten polymethyl methacrylate (PMMA) lenses of various thicknesses (113–330 μm) were used to calibrate four OCT devices, including the UHR-OCT setup, the UL-OCT setup, the RTVue, and the Visante. Using the Chott dispersion formula for PMMA in the Zemax software (Zemax, Software for Optical Design; Radiant Zemax, Redmond, WA), the indices of each PMMA lens at 840 and 1310 nm were 1.485 and 1.481, respectively. The central thickness of each PMMA lens was measured using both the aforementioned OCT systems and an electronic thickness gauge (ET-3; Rehder Development Co., Castro Valley, CA). The agreement between the measurements from these four OCT devices and the geometric thicknesses is shown in Figure 1. 

Figure 1. 

 

Four OCT systems calibrated with a set of 10 RGP lenses of known thicknesses ranging from 113 to 330 μm. The results strongly agreed with the thickness measurements obtained with an electronic thickness gauge.

All of the subjects were scheduled for OCT scans after 2:00 PM to minimize possible diurnal variations in the thickness of the cornea. ³³ The OCT images were collected during two sessions. Each subject was imaged twice on the first visit (visit 1) and once on the second (visit 2) by each of the four OCT devices using the same procedure to collect the same set of measurements. Both visits occurred during the same week. All of the OCT images were obtained by a single operator (LG). During the OCT imaging, each subject was positioned in front of each of the four OCT devices and was instructed to look at the fixation target. Under real-time video guidance, the operator adjusted the OCT device and centered the OCT scan at the pupil. Using the UHR-OCT and UL-OCT devices in the horizontal line scan mode and the RTVue in the cross line scan mode, the central cornea was imaged once the specular reflex was visualized in the horizontal and vertical orientations. Using the RTVue and Visante devices in the pachymetry scan mode, the central cornea was imaged when the corneal vertex reflex occurred. 

An automated image processing algorithm for segmenting the layers of the corneal images obtained from the UHR-OCT, UL-OCT, and RTVue devices was developed at Wenzhou Medical College. The utility and robustness of the automatic algorithm have been tested and described in our recently published study. ⁴⁰ Briefly, the algorithm filtered out background noise and artifacts including horizontal lines and specular reflections. Due to the specular reflex, axial scans from the central 0.5 mm of the scan width (128 axial lines for UHR-OCT and UL-OCT images, and 85 axial lines for the RTVue images) were removed. The ocular surface in the central region was defined as the first peak between the air and the tear film. The first peaks of the axial scans on the each side of the image within a range of 0.5 mm were aligned. The profiles from the central 0.5 mm on each side of the area that had been removed due to the specular reflex were averaged to create a longitudinal intensity profile (Fig. 2). From this profile, three corresponding interfaces were defined to determine the CCT and ET measurements. The first peak for each sampling line was defined as the boundary of the front surface of the cornea. The second peak corresponded to the location of the interface between the corneal epithelium and Bowman's layer. This peak was defined as one that had an intensity higher than the threshold and was located between 35 and 65 μm below the air–epithelium interface. ⁴¹ The final peak of each sampling line was defined as the one corresponding to the endothelium–aqueous interface. The corneal and epithelial thicknesses were determined from the distances between the first and last peaks and between the first and second peaks. A refractive index of 1.376 was used to calculate the corneal thickness from the UHR-OCT, UL-OCT, and RTVue setups. ⁴² The Visante device used a refractive index of 1.389 for corneal thickness calculations. ³² The means of the central corneal thicknesses (0 to 2 mm) taken from pachymetry maps were also included as the measured pachymetry CCTs from the Visante and RTVue devices. 

Figure 2. 

 

Central epithelium imaged by UHR-OCT and the longitudinal reflectivity profile. (A) Raw UHR-OCT image for the healthy subject. The strong reflectivity (range between two green lines) was removed. Axial-lines within a range of 1 mm (green curves) from the center of the cornea were processed to yield the average longitudinal reflectivity profile. (B) Longitudinal reflectivity profile for the image (A). The central epithelium thickness is the distance between peaks (a) and (b) and the central corneal thickness is the distance between peaks (a) and (c). (C) Raw UHR-OCT image for the LASIK patient. (D) Longitudinal reflectivity profile for the image (C). Scale bars: 500 μm.

The data analysis was performed using the SPSS statistical analysis software (version 16.0; SPSS Inc., Chicago, IL). Analysis of variance (ANOVA) and post hoc analysis were used to determine the significant differences among the measurements by the four OCT devices. Paired t-tests were used to determine whether between-visits and between-measurements pairwise differences existed (P < 0.05). Agreement between the four OCT devices was analyzed by the Bland and Altman method. ⁴³ The 95% limits of agreement (LoA) were defined as the mean difference ±1.96 SDs. The intrasession repeatability was measured using two OCT images obtained by the same operator during the first visit. The intrasession repeatability was evaluated in terms of the standard deviation (SD) of differences between the two measurements, the coefficient of repeatability, and the intraclass correlation coefficient (ICC). The coefficient of repeatability (CoR) was defined as 2 SDs of the difference between two measurements for the same subject, which were obtained by the same operator during the same visit; the coefficient of repeatability (CoR) (%) was defined as 2 SDs of the difference between the two measurements divided by the mean of each pair of values. The intersession repeatability was measured from 2 OCT images obtained by a single operator during 2 different visits. Then, the intersession SD, the coefficient of reproducibility, and the ICC were calculated. The coefficient of reproducibility was obtained during the repetition of the measurement during different visits and was defined as 2 SDs of the differences between the measurements divided by the average of the means of each pair of values. As proposed by Bartko and Carpenter, ⁴⁴ the ICC was determined based on a mixed-model analysis of variance. 

Results

The epithelium and Bowman's layers both in the healthy eyes and in the eyes after LASIK were clearly visualized in the corneal images (Fig. 3) that were obtained by these OCT devices, except for the Visante OCT. For all of the images obtained from the UHR-OCT, UL-OCT, and RTVue devices, the algorithm described above successfully measured the CCT and ET. The ranges of the CCT and ET values that were obtained from 20 normal subjects and 18 LASIK patients using the four OCT devices are displayed in Figure 4. 

Figure 3. 

 

The typical raw images acquired from UHR-OCT (A), UL-OCT (B), and RTVue (C) and the processed image from Visante (D). The images in the left column are from the normal subject and the images in the right column are from the LASIK patient. Scale bars: 500 μm.

Figure 4. 

 

The ranges of CCT values obtained using the four OCT devices for 20 healthy eyes (A) and 18 eyes after LASIK (B) and the ET values for the healthy subjects (C) and LASIK patients (D).

The mean CCTs, which were measured during the first visit using the UHR-OCT, UL-OCT, and RTVue devices with an automatic segmentation algorithm applied to each measurement, were not different from each other (Table 2). The mean CCT measurements derived from the RTVue and Visante pachymetry maps for both groups were thicker than those measurements obtained from the UHR-OCT, UL-OCT, and RTVue devices in conjunction with an automatic segmentation algorithm applied to each measurement (Tables 2, 3). The coefficient of repeatability of the healthy subjects ranged from 2.01 to 4.88 μm with an ICC that was greater than 0.997 for the CCT measurements from all of the OCT devices (Table 2). The coefficient of reproducibility was less than 6.0 μm with an ICC greater than 0.996 (Table 3). For the ET measurements from the UHR-OCT, UL-OCT, and RTVue devices, the coefficient of repeatability was less than 2.2 μm with an ICC greater than 0.890 (Table 4), and the coefficient of reproducibility ranged from 2.8 to 3.9 μm with an ICC greater than 0.74 (Table 5). For the LASIK patients, the coefficient of repeatability ranged from 3.5 to 7.8 μm with an ICC greater than 0.992 for the CCT measurements (Table 2), and the coefficient of reproducibility was less than 5.6 μm with an ICC greater than 0.992 (Table 3). The coefficient of repeatability for the ET measurements was less than 4.8 μm with an ICC greater than 0.84 (Table 4), and the coefficient of reproducibility ranged from 2.7 to 3.6 μm with an ICC greater than 0.86 (Table 5). In both groups, there were no significant differences between the various scans that were obtained either during the same visit or during different visits (all P > 0.05). Figure 5 shows the relationship between the coefficient of repeatability and the axial resolution of the OCT device when used to measure both the CCT and ET. In both groups, a higher axial resolution correlated with a lower coefficient of repeatability. 

Figure 5. 

 

Correlation between the axial resolution and the coefficient of repeatability of the OCT devices for the central corneal or epithelial thickness measurements. (A) Healthy subjects. (B) LASIK patients.

Table 2. 

 

Repeatability of the Four OCT Devices in Measuring Central Corneal Thickness

	Central Corneal Thickness, μm				ICC	CoR1, μm	CoR1, %	LoA, μm
	M1	M2	Mean	Difference (SD)
Healthy eyes (n = 20)
UHR-OCT	529.4 ± 32.0	529.5 ± 32.0	529.4 ± 32.0	−0.07 ± 1.00	0.998	2.00	0.38	−2.04 to 1.89
UL-OCT	527.9 ± 32.3	527.0 ± 32.4	527.4 ± 32.4	0.88 ± 1.98	0.998	3.96	0.75	−2.99 to 4.76
RTVue	527.1 ± 31.5	526.9 ± 31.7	527.0 ± 31.6	0.15 ± 1.56	0.999	3.12	0.59	−2.91 to 3.22
RTVue pachymetry	536.9 ± 32.9	537.0 ± 33.1	536.9 ± 33.0	−0.15 ± 1.79	0.999	3.58	0.66	−3.65 to 3.35
Visante	535.0 ± 32.9	534.1 ± 32.8	534.5 ± 32.8	0.95 ± 2.44	0.997	4.88	0.91	−3.83 to 5.73
Eyes after LASIK (n = 18)
UHR-OCT	444.3 ± 31.0	444.4 ± 31.3	444.4 ± 31.2	−0.05 ± 1.77	0.998	3.54	0.79	−3.51 to 3.41
UL-OCT	441.0 ± 31.4	440.8 ± 30.8	440.9 ± 31.1	0.15 ± 2.88	0.996	5.76	1.31	−5.49 to 5.79
RTVue	443.2 ± 31.3	442.7 ± 31.0	443.0 ± 31.2	0.51 ± 1.89	0.998	3.78	0.85	−3.20 to 4.22
RTVue pachymetry	452.1 ± 30.7	453.0 ± 31.2	452.5 ± 30.9	−0.94 ± 2.04	0.998	4.08	0.90	−4.95 to 3.06
Visante	454.5 ± 29.5	454.1 ± 30.2	454.3 ± 29.7	0.44 ± 3.91	0.992	7.82	1.72	−7.23 to 8.12

 M1, first measurement on visit 1; M2, second measurement on visit 1; ICC, intraclass correlation coefficient; CoR¹, coefficient of repeatability; LoA, 95% limits of agreement.

Table 3. 

 

Reproducibility of the Four OCT Devices in Measuring Central Corneal Thickness

	Central Corneal Thickness, μm				ICC	CoR2, μm	CoR2, %	LoA, μm
	M1	M3	Mean	Difference (SD)
Healthy eyes (n = 20)
UHR-OCT	529.4 ± 32.0	530.3 ± 32.5	529.8 ± 32.2	−0.89 ± 2.09	0.997	4.18	0.79	−4.98 to 3.20
UL-OCT	527.9 ± 32.3	528.7 ± 32.2	528.3 ± 32.3	−0.82 ± 2.00	0.998	4.00	0.76	−4.74 to 3.10
RTVue	527.1 ± 31.5	527.4 ± 32.1	527.2 ± 31.8	−0.31 ± 2.96	0.996	5.92	1.12	−6.12 to 5.50
RTVue pachymetry	536.9 ± 32.9	536.0 ± 33.2	536.4 ± 33.0	0.85 ± 3.00	0.996	6.00	1.12	−5.02 to 6.72
Visante	535.0 ± 32.9	534.9 ± 32.4	535.0 ± 32.6	0.10 ± 2.99	0.996	5.98	1.12	−5.76 to 5.96
Eyes after LASIK (n = 18)
UHR-OCT	444.3 ± 31.0	444.4 ± 32.0	444.4 ± 31.5	−0.10 ± 1.72	0.999	3.44	0.78	−3.47 to 3.27
UL-OCT	441.0 ± 31.4	441.3 ± 31.4	441.1 ± 31.4	−0.27 ± 2.28	0.997	4.56	1.03	−4.74 to 4.20
RTVue	443.2 ± 31.3	442.7 ± 31.1	443.0 ± 31.2	0.51 ± 1.89	0.998	3.78	0.86	−3.19 to 4.21
RTVue pachymetry	452.1 ± 30.7	452.4 ± 30.8	452.3 ± 30.7	−0.39 ± 1.50	0.999	3.00	0.66	−3.33 to 2.55
Visante	454.5 ± 29.5	454.8 ± 30.0	454.6 ± 29.7	−0.28 ± 2.82	0.996	5.64	1.24	−5.80 to 5.25

 M1, first measurement on visit 1; M3, first measurement on visit 2; ICC, intraclass correlation coefficient; CoR², coefficient of reproducibility; LoA, 95% limits of agreement.

Table 4. 

 

Repeatability of the Three OCT Devices in Measuring Central Epithelial Thickness

	Central Epithelial Thickness, μm				ICC	CoR1, μm	CoR1, %	LoA, μm
	M1	M2	Mean	Difference (SD)
Healthy eyes (n = 20)
UHR-OCT	54.0 ± 2.0	54.0 ± 2.0	54.0 ± 2.0	0.04 ± 0.62	0.95	1.24	2.31	−1.18 to 1.26
UL-OCT	55.1 ± 2.4	55.2 ± 2.3	55.1 ± 2.3	−0.14 ± 1.06	0.90	2.12	3.83	−2.21∼1.93
RTVue	54.1 ± 2.5	53.6 ± 2.3	53.8 ± 2.3	0.49 ± 1.12	0.89	2.24	4.15	−1.70∼2.68
Eyes after LASIK (n = 18)
UHR-OCT	53.2 ± 4.3	52.7 ± 4.2	53.0 ± 4.2	0.54 ± 1.73	0.92	3.46	6.55	−2.86∼3.94
UL-OCT	51.6 ± 4.2	51.3 ± 4.4	51.5 ± 4.1	0.30 ± 2.42	0.84	4.84	9.39	−4.44∼5.04
RTVue	53.3 ± 3.5	53.8 ± 4.0	53.5 ± 3.6	−0.51 ± 2.17	0.84	4.34	8.09	−4.75∼3.73

 M1, first measurement on visit 1; M2, second measurement on visit 1; ICC, intraclass correlation coefficient; CoR¹, coefficient of repeatability; LoA, 95% limits of agreement.

Table 5. 

 

Reproducibility of the Three OCT Devices in Measuring Central Epithelial Thickness

	Central Epithelial Thickness, μm				ICC	CoR2, μm	CoR2, %	LoA, μm
	M1	M3	Mean	Difference (SD)
Healthy eyes (n = 20)
UHR-OCT	54.0 ± 2.0	54.6 ± 2.9	54.3 ± 2.4	−0.55 ± 1.39	0.90	2.78	5.12	−3.28 to 2.18
UL-OCT	55.1 ± 2.4	54.8 ± 2.4	54.9 ± 2.3	0.27 ± 1.51	0.83	3.02	5.50	−2.69 to 3.23
RTVue	54.1 ± 2.5	54.1 ± 2.9	54.1 ± 2.5	0.03 ± 1.94	0.74	3.88	7.18	−3.77 to 3.83
Eyes after LASIK (n = 18)
UHROCT	53.2 ± 4.3	52.8 ± 4.7	53.0 ± 4.4	0.44 ± 1.34	0.96	2.68	5.05	−2.19 to 3.07
ULOCT	51.6 ± 4.2	51.0 ± 3.6	51.3 ± 3.8	0.60 ± 1.74	0.91	3.48	6.77	−2.81 to 4.01
RTVue	53.3 ± 3.5	53.3 ± 3.8	53.3 ± 3.5	0.00 ± 1.82	0.88	3.64	6.83	−3.57 to 3.57

 M1, first measurement on visit 1; M2, second measurement on visit 1; ICC, intraclass correlation coefficient; CoR², coefficient of reproducibility; LoA, 95% limits of agreement.

For both groups, the CCT measurements acquired with the UHR-OCT, UL-OCT, and RTVue devices using the automatic algorithm strongly correlated with each other, as did the ET measurements (Fig. 6). Both UHR-OCT and UL-OCT setups both demonstrated good agreement with the RTVue device for the CCT (the ICC = 0.998 and 0.992 for both UHR-OCT and UL-OCT for the healthy subjects and LASIK patients, respectively) and the ET measurements (ICC ranged from 0.75 to 0.90 for the healthy subjects and 0.62 to 0.74 for the LASIK patients), within the narrow (95%) limits of agreement and the high ICC (Tables 6 and 7). There were no significant differences in the CCT or ET measurements between the three devices (P > 0.05). 

Figure 6. 

 

Mean (± SD) of the central corneal thickness (CCT) and the epithelial thickness (ET) obtained from the custom automatic segmentation algorithm for the UHR-OCT, UL-OCT, and RTVue devices and CCT obtained from the pachymetry map for the RTVue and Visante devices. The CCT measurements from all four OCT devices correlated well, as did the ET measurements from all three OCT devices. (A) Healthy subjects. (B) LASIK patients.

Table 6. 

 

Agreement between the Four OCT Devices Regarding Central Corneal Thickness Measurements

Device or Methods Pairings	Mean Difference ± SD, μm	95% LoA, μm	ICC
Healthy eyes (n = 20)
UHR-OCT vs. UL-OCT	1.55 ± 2.11	−2.59 to 5.69	0.998
UHR-OCT vs. RTVue	2.34 ± 2.28	−2.13 to 6.82	0.998
UL-OCT vs. RTVue	0.79 ± 2.23	−3.57 to 5.15	0.998
RTVue vs. RTVue pachymetry	−9.79 ± 4.76	−19.12 to 0.46	0.990
RTVue pachymetry vs. Visante	1.85 ± 4.93	−7.82 to 11.52	0.998
Eyes after LASIK (n = 18)
UHR-OCT vs. UL-OCT	3.36 ± 4.02	−4.52 to 11.24	0.992
UHR-OCT vs. RTVue	1.13 ± 3.73	−6.18 to 8.44	0.993
UL-OCT vs. RTVue	−2.23 ± 1.97	−6.09 to 1.63	0.998
RTVue vs. RTVue pachymetry	−8.84 ± 3.19	−15.09 to −2.58	0.995
RTVue pachymetry vs. Visante	−2.44 ± 2.79	−7.91 to 3.03	0.996

 LoA; 95% limits of agreement; ICC, intraclass correlation coefficient.

Table 7. 

 

Agreement between the Three OCT Devices regarding Central Epithelial Thickness Measurements

Device Pairings	Mean Difference ± SD, μm	95% LoA, μm	ICC
Healthy eyes (n = 20)
UHR-OCT vs. UL-OCT	−1.07 ± 1.07	−3.17 to 1.04	0.90
UHR-OCT vs. RTVue	−0.08 ± 1.57	−3.16 to 2.99	0.77
UL-OCT vs. RTVue	0.98 ± 1.72	−2.40 to 4.36	0.75
Eyes after LASIK (n = 18)
UHR-OCT vs. UL-OCT	1.60 ± 3.41	−5.08 to 8.28	0.68
UHR-OCT vs. RTVue	−0.02 ± 3.49	−6.86 to 6.82	0.62
UL-OCT vs. RTVue	−1.62 ± 2.83	−7.17 to 3.93	0.74

The CCT measurements obtained from RTVue scans and estimated using our custom automatic algorithm were, on average, 9.8 μm thinner in the healthy subjects and 8.8 μm thinner in the LASIK patients than that were the CCTs estimated from the pachymetry maps (P < 0.05, Table 6). The CCT measurements obtained from the RTVue pachymetry maps agreed well with those obtained from the Visante pachymetry maps (Table 6). The difference between the CCT measurements acquired from the RTVue and Visante pachymetry maps was not statistically significant (P > 0.05) in either of the two groups. 

Discussion

The higher axial resolution of the OCT devices apparently yields better image quality, and more details of the structure are visualized, resulting in clear differentiation of the layers in the cornea. ^28,45–47 This notion may be the driving force behind the continuous improvement of the axial resolutions of OCT devices that has occurred within the last two decades. ⁴⁸ Drexler, ⁴⁹ Wojtkowski et al., ⁵⁰ and our group ⁵¹ have demonstrated the highest axial resolution (up to 1–2 μm) to date. As is evident from the images presented in the current study, the images obtained with a high-resolution (<5 μm) device are of higher quality than the images obtained with a lower resolution Visante device. Specifically, the Bowman's layer can be seen clearly in these high-resolution OCT images, and the ability to visualize the Bowman's layer may indicate the image quality, which is possibly related to the axial resolution of the OCT device. In addition, other factors such as polarization matching, image processing and scanning focus may affect the image quality. ⁵² High image qualities may allow for a more robust and accurate automatic image processing, including segmentation of the layers of interest. Previous work was performed using a manual or semimanual segmentation method, ^28,45 which is time consuming and subject to variation. ⁴⁵ Francoz et al. ⁴⁵ have used a commercial high resolution SD-OCT device to measure epithelial thickness using a manual method of segmentation. They acknowledged the variation in the measurement location when using a manual method, although their results were similar to results that have been published by others. ^28,53,54 Our automatic method has been validated using a manual method. Furthermore, the results in the present study were in good agreement with the previously published results from our group ^28,53 and others. ^55,56 Our results prove the concept that an automatic method for segmentation of an OCT image and subsequent measurement of the CCT and ET are entirely feasible for the advanced OCT devices tested in the present study. 

Although the optical resolution affects the image quality and detail during visualizations of the structure and/or layers of the cornea, as documented in the present study, the mean CCT or ET may not be influenced by the axial resolution of the OCT devices. Despite the range in the resolution (from 3 to 7.5 μm) of the three SD-OCT devices that we tested, the CCT and ET measurements that we obtained were similar when the same image-processing algorithms were applied, indicating that the devices can be used interchangeably. As expected, differences in the CCT were found between the OCT devices, mainly due to differences in the measurement areas and possibly due to differences in the image-processing algorithms. Using the same device (RTVue), there was a difference of approximately 9 μm between the two measurement protocols. However, when similar protocols for the measurement area (center 2-mm zone) from the pachymetry maps were applied, no differences were evident. These results indicate that the measurement protocol affects the results. Caution should be used when comparing the results obtained using different OCT devices. Researchers and manufacturers should establish a standard for measuring the definition and area so that others can directly compare the reported measurements from the different OCT devices. 

The axial resolution of the OCT devices may affect the repeatability of the CCT and ET measurements, as is evident from the present study. The linear regression results demonstrated that a higher optical resolution might yield a higher precision in determining the layer thickness, mainly due to an increase ability to resolve the layer boundaries (image quality). Although the averaged measurement may not be affected, the precision may differ among the various OCT devices with different resolutions. Information regarding measurement precision may be useful in calculating the sample size. Clearly, a better precision resulting from a high optical resolution should be used when studying diseases with a relatively small sample size. Oversampling during OCT data acquisition may further improve measurement precision. With the three tested SD-OCT devices, oversampling was evidently performed because pixel interval of the image was smaller than the optical resolution. Oversampling may help improve the image quality and overall boundary detection. However, no specific tests of this effect were conducted in the present study. A higher resolution device could have a higher noise-per-resolution element due to the apparently lower image contrast (signal-to-noise ratio). However, the finding that the coefficient of repeatability of the UHR-OCT setup was lower than that of the other OCT devices indicated that the higher noise did not appear to have a major role in the measurement variation using the method described in the present work. 

There are some limitations of the present study. First, we only tested three SD-OCT devices with different resolutions and one time domain OCT. Although the results are conclusive, further studies will focus on a wide range of different OCT devices including additional commercial devices. Second, we did not discover a method of exporting raw OCT images from the Visante device and thus did not use our custom algorithm to process the ET and CCT for comparison. Third, as technology advances, OCT devices with resolution higher than 3 μm will become available. For example, we recently developed an SD-OCT that has a resolution of 2 μm ^28,51 ; however, we did not include that OCT in the present study because it was not yet available for the collection of data. Other measurement errors within OCT may exist and these errors have been described elsewhere. ^28,29  

In conclusion, OCT devices with different axial resolutions appeared to yield similar results for CCT and ET measurements when the same image-processing algorithm was used. However, there were differences between the OCT devices with the various measurement protocols. It was found that the precision of measurement for both the CCT and ET was related to the optical resolutions of the OCT devices. 

Footnotes

 Supported by research grants from Zhejiang Provincial Program for the Cultivation of High-Level Innovative Health Talents (FL), Zhejiang Provincial Natural Science Foundation of China (Y2090674 [MS]), and The Development Program Project Grant from Wenzhou, China (Y20100192 and Y20100296 [AT]).

Footnotes

 Disclosure: L. Ge, None; Y. Yuan, None; M. Shen, None; A. Tao, None; J. Wang, None; F. Lu, None

Footnotes

 Section codes: axial resolution

Footnotes

  Proprietary interests: The authors have no proprietary interest in any materials or methods described within this article.