Introduction
The cornea and lens are important elements of the refractive components of the eye. When certain pathologic conditions occur in the cornea or lens, decreased transparency and increased light scattering can occur [
1]. These changes in corneal health reflect changes in corneal density (CD). Therefore, CD can assist in the diagnosis of multiple keratopathy, the monitoring of corneal conditions after refractive surgery [
2], and the prediction of visual acuity [
3] as well as the visual quality [
4]. The clarity and density of the lens are also crucial for maintaining high-quality vision. As cataract is the main cause of blindness in less developed countries [
5], objective and quantifiable evaluation of lens density (LD) has become an essential clinical practice. As the number of patients undergoing implantable collamer lens (ICL) implantation has grown in recent years, whether their LD has undergone procedure-induced changes has attracted much attention [
6,
7]. Long-term follow-up of LD can reflect the state of aqueous fluid and the risk of anterior capsule opacification after ICL implantation [
8,
9]. Consequently, it is of great clinical significance to establish an objective and convenient way to dynamically observe changes in the density of the cornea and lens.
Pentacam, based on a rotating Scheimpflug photography system, obtains multiple images of the ocular anterior segment and computes a three-dimensional corneal map. It can provide an evenly focused representation of the cornea and a section of the lens as compared to slit-lamp photographs [
10]. The technique also allows for three imaginary perpendicular planes: the lens, the image, and the subject. Therefore, Pentacam has an extended depth of focus and provides images with sharp resolution [
11]. Moreover, Pentacam can provide a series of biological parameters of the anterior segment, including a built-in system that can measure the optical density of the cornea and lens based on the pixel intensity of the obtained images. Its densitometry measurements have been reported to exhibit good accuracy and repeatability [
12].
CASIA2 is the latest swept-source anterior segment optical coherence tomography (AS-OCT) instrument. It can facilitate the observation and analysis of the cornea, anterior chamber, and intraocular lens using a wavelength of 1310 nm with a speed of 50,000 axial scans per second [
13], boasting a faster scanning velocity, a broader scanning scope, and higher resolution of images. For CASIA2, each 3D image is composed of 512 A-scans and 128 B-scans; a depth of 11 mm and a width of 16 mm are achieved in the scanning scope [
14]. CASIA2 has been compared with many other anterior segment imaging devices, such as Spectralis, Anterion, and Visante, to evaluate its repeatability, reproducibility, and inter-device agreement [
15‐
17]. As one of the widely used measurement tools for the anterior segment, Pentacam has often been chosen as the gold standard for agreement evaluation with CASIA2 [
18], and studies have been conducted to compare the anterior chamber depth (ACD) [
19], corneal thickness, corneal curvature [
20], and angle-to-angle distance [
14] using these two devices in both healthy participants and patients with certain ophthalmic diseases. Good agreement was found in the measurement of the most anterior-segment biometrics.
However, no studies have compared the accuracy of corneal and lens densitometry between Pentacam and CASIA2. This study aimed to investigate the feasibility of corneal and lens densitometry measurements using CASIA2 and its agreement with Pentacam in myopic participants, in the hope of providing suggestions for ophthalmologists in the assessment and follow-up of patients undergoing refractive surgery and ICL implantation.
Discussion
Densitometry of the cornea and lens is of great clinical significance. It can reflect the overall condition and local lesions of the examined areas. The measurement of CD and LD has undergone several upgrades, from slit-lamp examination and ultrasound biomicroscopy to Schiempflug imaging.
This is the first study to compare the measurement of CDs and LDs between CASIA2 and Pentacam. Prior studies have reported the results of corneal densitometry using Scheimpflug optical assessment. Otri et al. reported that when measured by Pentacam, the normal density value of the total cornea was 12.99 ± 2.58%grayscale, with the range of 4.7–22.0 in healthy corneas [
1]. Dhubhghaill reported that densitometry values of the cornea were lowest in the central zone (16.76 ± 1.87%grayscale) and highest in the periphery (27.36 ± 7.47%grayscale) in healthy participants, while the surrounding 2–6 mm annulus had very similar densitometry values to the central zone [
21]; notably, these results are in accordance with our findings from Pentacam data. The Pentacam-Scheimpflug image system has been used to investigate lens densitometry in many published studies [
26] [
27]. Weiner et al. reported that lens densitometry based on Scheimpflug imaging was highly repeatable in eyes without cataracts [
26]. In a prospective study on LD conducted by Bayrak et al., the average LD of healthy participants measured by Pentacam in the control group was 8.3 ± 0.9% grayscale [
27]. The range of LDs in our study approximated to what were reported in previous studies.
In this study, we found good agreement in the measurements of the average density of the cornea and lens using CASIA2 and Pentacam. However, the density values of the two devices were not directly interchangeable and comparable, even after the percentage conversion. The two sets of data generated by CASIA2 and Pentacam were of a linear correlation relationship with random bias. Thus, consistency ICC instead of absolute agreement ICC was used in the study. Except for cornea 0–2 mm, the ICC values of all other zones of the cornea and lens were above 0.7, suggesting a good level of consistency. The weaker agreement of the central corneal density was caused by the paradoxically high CD in the 0–2 mm cornea obtained from CASIA2. CD has been shown to be higher in the periphery compared to the central by Pentacam measurement [
21,
28]. We believed the abnormally high central CD was an artifact related to the central position of CASIA2’s light source and its reflection. In the cross-sectional image generated by CASIA2, a bright line with strong reflection along the optical path was clearly visible, passing through the central corneal (Fig.
1). This would greatly affect the results of CD 0–2 mm, which was calculated based on the grayscale of image. Therefore, CASIA2 measurement of central corneal density was rather overestimated and deviated from its actual value, hence it must be considered with caution.
In all zones of the cornea and lens, the density values obtained by CASIA2 were higher than those obtained by Pentacam, which could be caused by the different approaches of the two devices for anterior segment imaging. CASIA2 is a swept-source OCT (SS-OCT), a high-resolution tomographic and biomicroscopic device used for in vivo imaging and measurement of ocular structures in the anterior segment. SS-OCT relies on backscattered light compared to a reference beam and employs Fourier transformation for image reconstruction [
29]. Pentacam is a Scheimpflug imaging-based device that uses a rotating camera to capture multiple images of the anterior segment and generate 3D images [
30]. Although the densitometry analyses of the two devices are both based on the grayscale of the captured images, their principles of imaging differ greatly. As a result, there were noticeable differences in terms of image quality and density values of the cornea and lens obtained from CASIA2 and Pentacam. The densitometry analysis software for these two devices also provides different user experiences. CASIA2’s densitometry boasts customization of the scanning scope width, automated density calculation and analyses, and simultaneous measurement of the cornea, lens, and anterior chamber under one scanning mode. In comparison, Pentacam’s densitometry slightly fails in flexibility, simplicity, and precision because of the unchangeable preset parameters of CD measurement and manual operation of LD measurement, which requires users to choose one or more Schiempflug images with relatively good quality and to select the zones of the lens either by entering parameters or shifting the marquee.
In Bland-Altman analyses, LOA was constructed to help determine whether the agreement between two methods was sufficiently close for them to be interchangeable. Based on previous reports of corneal density in pathological conditions, total cross-sectional CD increased to 58.4 ± 19.5%grayscale during acute infection and returned to 33.4 ± 17.3%grayscale after complete resolution, whilst the healthy cornea had a density of 12.99 ± 2.58%grayscale [
1]. There was a large gap in CD values between healthy and infected corneas. In other conditions, such as keratoconus accompanied with Down syndrome, the difference in the CD between patients and healthy controls was smaller (19.35 ± 2.92% vs. 15.78 ± 2.67%) [
31]. We believed this was related to the nature of each keratopathy. In diseases where change in density was one of the defining pathologic manifestations, the possibility of misdiagnosing keratopathy due to inter-equipment errors in CD was relatively low. In other diseases where the change in CD was more subtle, typically more comprehensive examinations were required to arrive at a diagnosis. Therefore, we believed the LOA of CD was considered clinically acceptable in the study, if interpreted with other examination results and clinical evidence. In terms of LD, the prediction interval was quite narrow, and the strong linear correlation between two sets of data facilitated the inter-equipment transformation of densitometry readings.
In addition, although Pentacam is one of the most widely used devices in the assessment of ocular anterior segment, it is restricted by pupil diameter and its limited depth of scanning scope when measuring the LD. While CASIA2 has a scanning scope of 11 mm in depth and 16 mm in width [
14], its lens densitometry analysis takes into account the periphery of the lens. Since the opacity of the lens usually starts in the peripheral lens after ICL implantation [
32], peripheral LD is considered of great clinical value. In addition, CASIA2 could reflect more minor and earlier changes in CD and LD; hence, its superior sensitivity could remind ophthalmologists of trivial pathological signs that may otherwise be overlooked. Although the absolute numeric density values of the two devices were not interchangeable owing to differences in imaging principles, the two sets of data presented the same tendency and a certain degree of linear association. Therefore, we believed that CASIA2’s densitometry could indicate the relative level of CD and LD and distinguish obvious abnormal density values in pathological conditions. Nevertheless, the literature on CD and LD measured by CASIA2 is rather scant. We suggested that CASIA2’s densitometry results be mathematically transformed before they were subjected to interpretation based on knowledge or understandings from Pentacam’s densitometry results, until further studies establish more accurate diagnostic standards that are tailored for CASIA2’s densitometry.
In this study, the inter-device agreement between CASIA2 and Pentacam densitometry did not appear to be affected by age. Intriguingly, we found that the agreement of CD measurements was poorer in the high myopia subgroup. This can be partly explained by poorer fixed vision during examination in the high myopia group, which could result in a larger scope of the central cornea affected by the light beam of CASIA2, resulting in a greater systematic bias. On the other hand, since the CD values were higher in the high myopia group than in the mild-to-moderate myopia group in our findings, the increase in measurement values could cause the two sets of data to diverge due to a scaling effect, resulting in a lower ICC and weaker agreement. Similar scaling effects have been noted in previous studies comparing AS-OCT measurements between the Spectralis, CASIA2, and Cirrus [
15,
33]. One hypothesis is that the systematic effect originates from how OCT devices account for corneal refraction, which is a parameter used to scale the corresponding OCT B-scans [
15].
This study had a few limitations. First, all densitometric measurements were obtained without mydriasis. This limited the scanning scope of the lens, and only the central and anterior regions of the lens were included in the density calculations. Second, analysis of CD at different depths and layers was not included in this study. The average density values of the entire corneal layer were adopted and compared between the two devices. Third, the study included both eyes of participants in the analyses to maximize the sample size and retain more information. However, the correlation between fellow eyes could potentially result in underestimated p values and narrower confidence intervals [
34]. Despite the use of generalized linear model and hierarchical regression method to account for such correlation, the influence of including both eyes in the study should be noted by readers. Finally, the study enrolled healthy participants who had no ocular conditions other than myopia. The measurement of density using CASIA2 in patients with other ocular diseases, such as cataract, requires further evaluation.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.