In Vivo Retinal Pigment Epithelium Imaging using Transscleral Optical Imaging in Healthy Eyes

Abstract

Objective: To image healthy retinal pigment epithelial (RPE) cells in vivo using Transscleral OPtical Imaging (TOPI) and to analyze statistics of RPE cell features as a function of age, axial length (AL), and eccentricity.

Design: Single-center, exploratory, prospective, and descriptive clinical study.

Participants: Forty-nine eyes (AL: 24.03 ± 0.93 mm; range: 21.9-26.7 mm) from 29 participants aged 21 to 70 years (37.1 ± 13.3 years; 19 men, 10 women).

Methods: Retinal images, including fundus photography and spectral-domain OCT, AL, and refractive error measurements were collected at baseline. For each eye, 6 high-resolution RPE images were acquired using TOPI at different locations, one of them being imaged 5 times to evaluate the repeatability of the method. Follow-up ophthalmic examination was repeated 1 to 3 weeks after TOPI to assess safety. Retinal pigment epithelial images were analyzed with a custom automated software to extract cell parameters. Statistical analysis of the selected high-contrast images included calculation of coefficient of variation (CoV) for each feature at each repetition and Spearman and Mann-Whitney tests to investigate the relationship between cell features and eye and subject characteristics.

Main outcome measures: Retinal pigment epithelial cell features: density, area, center-to-center spacing, number of neighbors, circularity, elongation, solidity, and border distance CoV.

Results: Macular RPE cell features were extracted from TOPI images at an eccentricity of 1.6° to 16.3° from the fovea. For each feature, the mean CoV was < 4%. Spearman test showed correlation within RPE cell features. In the perifovea, the region in which images were selected for all participants, longer AL significantly correlated with decreased RPE cell density (R Spearman, Rs = -0.746; P < 0.0001) and increased cell area (Rs = 0.668; P < 0.0001), without morphologic changes. Aging was also significantly correlated with decreased RPE density (Rs = -0.391; P = 0.036) and increased cell area (Rs = 0.454; P = 0.013). Lower circular, less symmetric, more elongated, and larger cells were observed in those > 50 years.

Conclusions: The TOPI technology imaged RPE cells in vivo with a repeatability of < 4% for the CoV and was used to analyze the influence of physiologic factors on RPE cell morphometry in the perifovea of healthy volunteers.

Financial disclosures: Proprietary or commercial disclosure may be found after the references.

Keywords: AF, autofluorescence; AL, axial length; AO, adaptive optics; Adaptive Optics Transscleral Flood Illumination; BCVA, best-corrected visual acuity; CCS, center-to-center spacing; CoV, coefficient of variation; D, diopters; FOV, field of view; Healthy volunteers; High resolution retinal imaging; IOP, intraocular pressure; NIR, near-infrared; PRL, preferred retinal locus; QC, quality criterion; RE, refractive error; RPE, retinal pigment epithelium; Retinal Pigment Epithelium; SD, standard deviation; SLO, scanning laser ophthalmoscope; TOPI, transscleral optical imaging.

© 2022 by the American Academy of Ophthalmology. Published by Elsevier Inc.

Figures

Figure 1

Locations of the in vivo retinal pigment epithelium images in the right eye of P029 (male, 29 years). A, Spectralis infrared right eye fundus showing the 6 imaged zones. B, in vivo retinal pigment epithelial images in 6 areas (Z1: inferonasal; Z2: inferotemporal; Z3: superonasal; Z4: superotemporal; Z5: foveal center; Z6: nasal). For each 5° × 5° raw image, 1 1.6° × 1.6° subimage with flat-field correction is magnified. C, Example of iris pictures and low-resolution oblique-illuminated 30° × 30° infrared fundus recorded during Z3 examination. Scale bars = 200 μm.

Figure 2

Image processing pipeline. First step: each 5° × 5° raw retinal pigment epithelium image (A) is divided into 9 1.6° × 1.6° subimages (B). Each subimage is then processed by 2 subpipelines running in parallel. C–E, “Segmentation subpipeline”: (C) High-pass filtered image using an flat-field correction (FFC) with sigma = 10. D, Centers found using the Difference of Gaussians filter with sigma1 = 5 and sigma2 = 10. E, Voronoi-based cell segmentation. F–H “Vessel and haze detection subpipeline”: F, High-pass filtered image using an FFC with sigma = 3.1. G, High-pass filtered image using Butterworth filter (cutoff frequency w = 80 pix− 1). H, Vessel and haze mask. Final step: The mask of segmented cells (I) is generated by removing the vessels and haze mask (H) from the segmentation (E). CR = Contrast reversing.

Figure 3

Illustration of the retinal pigment epithelium images selected for the statistical analysis in function of the center-to-center spacing (CCS) and the quality criterion (QC) values. Images for which the CCS was below 12 μm were excluded so as not consider other cell types or structures (A, CCS = 11.66 μm, QC = 0.063, left eye, from P006 [male, 27 years] at 6.6° superotemporal; B, CCS = 9.64 μm, QC = 0.0761, right eye from P010 [female, 32 years] at 5.37° inferotemporal; C, CCS = 8.95 μm, QC = 0.157, right eye from P010 at 5.83° superotemporal). Images for which the QC was < 0.076 were excluded to not consider blurry images (D, CCS = 11.85 μm, QC = 0.0708, right eye from P039 [male, 33 years] at 3.11° superotemporal; G, CCS = 13.95 μm, QC = 0.0756, left eye from P010 at 4.39° inferonasal). The green square shows examples of retinal pigment epithelium images selected for the analysis (E, CCS = 12.23 μm, QC = 0.0817, right eye from P017 [male, 58 years] at 5.37° inferonasal; F, CCS = 12.02 μm, QC = 0.128, left eye from P039 at 4.39° superonasal; H, CCS = 13.98 μm, QC = 0.0765 right eye from P040 [female, 31 years] at 5.83° inferonasal; I, CCS = 14.55 μm, QC = 0.286, right eye from P014 [female, 29 years] at 5.37° superonasal).

Figure 4

Average cell density for participants as a function of subimage eccentricity in the nasal and temporal regions. Each point represents the average density of the inferior and superior quadrants in each region. Data collected at X = 0 are located at 1.6° in the inferior and superior quadrants. A, Group 1, 10 participants with consistently selected subimages from the fovea to the perifovea. Cell density decreases with eccentricity for 1 participant (P043 [male, 28 years], orange circle) in the nasal (Y = 31.93∗X + 3820; R2 = 44.6%, P = 0.0176) and temporal regions (Y = −42.81∗X + 3956; R2 = 78.8%, P < 0.0001) and for 1 participant (P013 [female, 32 years], blue square) in the nasal region. B, Group 2, 5 participants with consistently selected subimages in the perifovea. Cell density decreases with eccentricity for 1 participant (P017 [male, 58 years], orange diamond) in the nasal region (Y = 126.9∗X + 4211; R2 = 80.7%, P = 0.015) and for 2 participants in the temporal region (P019 [female, 25 years], blue circle: Y = −26.95∗X + 4424; R2 = 34.8%, P = 0.0436; P016 [male, 39 years], green dot: Y = −77.96∗X + 3986; R2 = 82.4%, P = 0.0124). C, Group 3, 14 participants with irregularly selected images in the perifovea. Cell density decreases with eccentricity for 1 participant (P025 [male, 52 years], black square) in the temporal region (Y = −68.27∗X + 3750; R2 = 99%, P = 0.0119). Raw data are available in Table S4 (available at www.ophthalmologyscience.org).

Figure 5

Linear regressions of retinal pigment epithelium cell features in function of eccentricity. Each point represents the mean ± standard deviation of average features from the 15 participants of groups 1 and 2 at each shared radial eccentricity in the perifoveal region. Between 3° and 8°, eccentricity has no influence on retinal pigment epithelium cell density (Y = −6.194∗X + 3804; R2 = 0.09%, P = 0.7829), area (Y = 0.5961∗X + 222.6; R2 = 0.22%, P = 0.6637), center-to-center spacing (Y = 0.1049∗X + 14.14; R2 = 3.59%, P = 0.0753), number of neighbors (Y = −0.01145∗X + 5.762; R2 = 1.35%, P = 0.2774), and circularity (Y = 0.002182∗X + 0.8851; R2 = 3.93%, P = 0.0626) and little influence on the other morphologic features: elongation (Y = −0.003524∗X + 0.6326; R2 = 5.20%, P = 0.0316; ∗), border distance coefficient of variation (Y = −0.001835∗X + 0.1639; R2 = 5.38%, P = 0.0287; ∗), and solidity (Y = 0.0002037∗X + 0.9506; R2 = 6.17%, P = 0.0189; ∗). Raw data are available in Table S5 (available at www.ophthalmologyscience.org).

Figure 6

Effect of axial length (AL) on retinal pigment epithelium cell features. A, Linear regressions of AL in function of cell density and circularity. Illustration of quantification in a subimage of the shorter eye (B, AL = 21.88 mm, left eye from P019 [female, 25 years] at 5.37° superotemporal) versus one of the longer eyes (C, AL = 26.51 mm, left eye from P029 [male, 29 years] at 4.39° superotemporal). CoV = coefficient of variation; OD = right eye; OS = left eye; SD = standard deviation.

Figure 7

Effect of age on retinal pigment epithelium cell features. (A) Linear regressions of age in function of cell density and circularity. Illustration of quantification in a subimage of one of the younger participants (B, 27 years, left eye from P002 [female] at 6.6° superotemporal) versus one of the older participants (C, 62 years, left eye from P007 [male] at 5.37° superotemporal). CoV = coefficient of variation; OD = right eye; OS = left eye; SD = standard deviation.