Structure-Function Associations of Choroidal Thickness With Retinal Sensitivity in Myopia

Abstract

Purpose

The purpose of this study was to investigate the focal relationship between choroidal thickness and retinal sensitivity in myopic eyes.

Methods

Participants underwent swept-source optical coherence tomography (SS-OCT) imaging and microperimetry testing. Choroidal thicknesses were obtained by segmenting the SS-OCT scans using a deep-learning approach. Retinal sensitivity was measured at 33 locations using scotopic microperimetry, and corresponding focal choroidal thickness at these locations were computed. Focal structure-function associations between retinal sensitivity and choroidal thickness across 15 retinal eccentricities were evaluated, and adjusted for retinal thickness, signal strength, age, axial length, and gender.

Results

The analysis included 280 eyes from 155 participants (mean age = 25.8 ± 3.0 years) with myopia with a mean axial length of 26.56 ± 0.99 mm and refractive error of −6.87 ± 2.05 diopters (D). Mean retinal sensitivity was 27.68 ± 1.28 decibels (dB). Mean choroid thickness was 217.85 ± 66.66 µm, with choroidal thickness in the nasal quadrant significantly thinner than in the other quadrants. Choroidal thicknesses were correlated with retinal sensitivity in 14 of 15 retinal regions, including the global eccentricity zone (r = 0.261, P < 0.001). Significant associations were similarly found with retinal thickness (r = 0.305, P < 0.001), age (r = 0.182, P < 0.05), and axial length (r = −0.402, P < 0.001). Focal structure-function models further substantiated these relationships, demonstrating a significant effect size (β = 0.78, 95% confidence interval [CI] = 0.39–1.16, P < 0.001) after adjusting for retinal thickness and other demographic factors.

Conclusions

Significant associations between choroidal thickness measurements and microperimetry were observed in a cross-sectional cohort of myopic eyes, suggesting a structure-function relationship with retinal sensitivity at the choroid. Further studies will be required to evaluate this in other cohorts and in progression.

The rising prevalence of myopia has become a significant ophthalmic issue worldwide, particularly in East and Southeast Asian populations.¹ Projections estimated that by 2050 nearly half of the world's population, or approximately 5 billion people, will have myopia, with about 1 billion expected to develop high myopia.^2,3 High myopia is characterized by globe elongation and a refractive error of −5 diopters (D) or less, accompanied by structural changes in the peripheral retina, including alterations in the retinal and choroidal layers.^3–9 These changes have been associated with sight-threatening complications, such as myopic maculopathy, choroidal neovascularization, and visual field defects, which can result in irreversible blindness.^6,7,10–12

The choroid, a vascular layer situated between the retina and the sclera, plays a vital role in maintaining retinal health by supplying oxygen and nutrients to the outer retina.^13,14 Many factors, including age, axial length, and refractive error, have been associated with choroidal thickness, and studies indicate that thinning of this layer is linked to the progression of myopia.^4,8,15–18 Choroidal thinning may be associated with decreased choroidal blood supply and insufficient blood flow to the optic nerve head,⁵ which has been linked to visual field defects^19,20 and a higher risk of developing myopic maculopathy.^6,7,21 Because the progression of myopia has been associated with changes in choroidal thickness and visual function, investigating these relationships can provide insights on how structural alterations in the eye impact visual function in myopia.

Although associations between myopia progression and changes in choroidal thickness have been reported,^4,8,15–20 there have been limited studies relating structural choroidal changes with functional outcomes, specifically retinal sensitivity. Previous studies have often used methodologies that either superimpose data across the entire macular region or have focused on regions such as defined by the Early Treatment Diabetic Retinopathy Study (ETDRS) grid.^8,20 Another study⁴ relied on single-point manual measurements of choroidal thickness, which may not capture the full extent of changes. The association of retinal sensitivity with focal choroidal thickness remains relatively unexplored.

The objective of this study was to evaluate the relationships between retinal sensitivity obtained with microperimetry and choroidal thickness from swept-source optical coherence tomography (SS-OCT) imaging. Using focal choroidal thickness regions mapped from microperimetry stimuli, this study aims to provide a more comprehensive understanding of the structure-function association in the choroid. This approach can enhance our understanding of the contribution of choroidal characteristics toward the development of potential biomarkers for myopia management.

Methods

Study Population

This observational study included data from participants with myopia enrolled in the Atropine Treatment Long-Term Assessment Study recruited at the Singapore National Eye Centre, a tertiary eye care institute in Singapore. Details of the study have been previously reported.^10,22,23 Participants with any diagnoses of glaucoma, retinopathies, or pathological signs arising from myopia were excluded. The study adhered to the ethical guidelines of the Declaration of Helsinki and received approval from the SingHealth Centralized Institutional Review Board (No. 2020/2249). All participants provided written consent prior to their participation.

Ocular Examinations

All participants were interviewed to obtain detailed clinical histories, covering ocular and medical conditions, surgical history, eye complaints, and family history. The participants also underwent comprehensive ophthalmology examinations, including visual acuity tests using the logarithm of the minimum angle of resolution chart (The Lighthouse, New York, NY, USA), and examination of both the anterior and posterior segments of their eyes. Cycloplegia was induced with tropicamide 1% and cyclopentolate 1%; a second dose was administered if dilation was insufficient 15 minutes later. Cycloplegic autorefraction was performed using the Topcon Auto Kerato-Refractometer (Topcon Corp., Tokyo, Japan) to calculate the mean spherical equivalent of refractive error from three measurements. Ocular biometry, including axial length, anterior chamber depth, and corneal curvature, was measured with the Carl Zeiss IOL Master (Carl Zeiss Meditec, Jena, Germany), averaging 5 readings to ensure consistency within 0.05 mm.

Assessment of Retinal Sensitivity on Microperimetry

Retinal sensitivity was assessed using scotopic microperimetry by mapping visual field responses directly onto real-time fundus images with the NIDEK MP-3 Microperimeter (NIDEK CO., Ltd., Gamagori, Japan) which provided a 45 degree non-mydriatic view of the fundus (Fig. 1A). The MP-3 used a 4–2 full threshold staircase strategy to measure perimetric values, with a stimulus intensity range from 0 to 34 decibels (dB) and a background luminance of 31.4 apostilbs (asb). It has a maximum stimulus intensity of 10,000 asb. The testing was conducted across 33 points arranged in a 2-degree by 2-degree grid within a 20-degree diameter of the fovea, using a Goldmann size III stimulus, projected for 200 ms. The eye-tracking system compensated for ocular movements and monitored fixation stability during pupil dilation.

Figure 1.

Illustration of structure-function measurements. (A) Color fundus image with microperimetry test locations; (B) a cross-sectional view of the retina in 3D; and (C) choroidal thickness (CT) and retinal thickness (RT) measurements, where choroidal thickness represents the thickness of the choroid layer (blue boundary) and retinal thickness denotes the thickness of the retina (green boundary).

Generation of Choroidal and Retinal Thickness Maps

The SS-OCT was performed using the ZEISS PLEX Elite 9000 (Carl Zeiss Meditec, Dublin, CA, USA). This OCT imaging used a 12 × 12 mm scanning protocol, featuring a light-source wavelength of 1050 nm, a scanning speed of 100,000 A-scans per second, and an A-scan depth of 3.0 mm. Each SS-OCT scan volume included 500 A-scans and 500 B-scans, covering a 12 × 12 × 3 mm volume (Fig. 1B), with a resolution of 500 × 1536 pixels for each B-scan. All OCT datasets were reviewed by trained graders blinded to the participant characteristics. Only scans with a signal strength of 6 or higher were included. Scans were excluded if they had poor clarity, inconsistent signal intensity, significant motion artifacts, or missing refractive error data.

The choroidal layer, defined as the distance from the outer surface of the retinal pigment epithelium (RPE) to the choroid-sclera interface, was automatically segmented for each eye using a SA-Net architecture²⁴ that integrates a UNet-based architecture with a multi-task learning approach. This method segments the choroid from three-dimensional OCT images by aggregating spatial context from adjacent cross-sectional slices.

The retinal layer, defined as the distance from the inner limiting membrane (ILM) to the outer surface of the RPE, comprises distinct layers, including the retinal nerve fiber layer (RNFL), the combined ganglion cell and inner plexiform layers (GCL-IPLs), the inner nuclear and outer plexiform layers (INL-OPLs), the outer nuclear layer (ONL), and the interface between the inner segment/outer segment (IS/OS) and the RPE. These layers were segmented using a semi-supervised cross-teaching framework that integrates convolutional neural networks and transformer architectures, as detailed in a previous study.²⁵ All segmentation results were visually inspected, and poor-quality segmentations were excluded to ensure the reliability of the subsequent analysis. An example of a cross-sectional B-scan with the obtained choroidal and retinal segmentations is presented in Figure 1C.

Figure 2 outlines the processing steps involved in integrating choroidal thickness and retinal thickness with microperimetry data. The workflow begins with the segmentation of the choroid and retinal layers, followed by the generation of enface thickness maps from OCT scans. Figure 2A illustrates the process of generating a choroidal thickness map. Morphological operations were applied as a post-processing step. The final thickness profile was obtained by measuring the distance between the segmented layer boundaries and mapping it as a thickness profile.

Figure 2.

The processing steps for integrating choroidal thickness with microperimetry data. Choroidal thickness was mapped to microperimetry focal points within a 6 × 6 mm field of view centered on the fovea, approximately equivalent to 20 degrees. (A) Choroidal segmentation was performed on OCT images, followed by morphological post-processing. Thickness profiles were generated by measuring the distance between the inner and outer choroidal boundaries. (B) To align thickness maps with microperimetry data, the enface OCT image was registered to the fundus photograph using the Scale-Invariant Feature Transform (SIFT) algorithm, with blood vessels serving as landmarks. Microperimetry test locations were then interpolated onto the thickness maps. (C) Mean choroidal thickness was integrated at 33 microperimetry focal points.

Focal Structure-Function Measurements

For each microperimetry test, the MP-3 provides an accompanying color fundus photograph indicating the locations on the retina where focal sensitivities were tested (see Fig. 1A). To align the choroidal and retinal thickness maps with the retinal sensitivity test locations, we first registered the enface OCT image with this fundus photograph. The registration process involved two key steps (see Fig. 2B). First, blood vessels were extracted from both images using a UNet model²⁶ as landmarks. Next, the Scale-Invariant Feature Transform (SIFT) algorithm²⁷ detected and matched key points between the modalities, with descriptors matched using a Brute Force Matcher and Lowe’s ratio test. A homography matrix was then computed via Random Sample Consensus (RANSAC)^28,29 to align the OCT image with the fundus photograph.

The choroidal thickness was mapped to microperimetry focal points, following the microperimetry data pattern^30,31 within a 20-degree diameter of the fovea, equivalent to approximately 6 × 6 mm, as detailed in Figure 2C. We registered the enface OCT image using the fundus image as a reference to align the spatial information (Fig. 3A), and manually verified the registration accuracy by visually assessing the alignment through overlaying corresponding data. Examples of the image registrations are provided in Supplementary Figure S1. The enface choroidal thickness map was generated from the registered OCT image, and the resulting multimodal imaging overlay enabled direct topographical correlation of choroidal thickness within the boundaries of the microperimetry focal points (Fig. 3B). To account for potential magnification effects due to axial length differences in myopic eyes, we adjusted the lateral scale of OCT images using the axial length-based correction method described by Sampson et al.³² The true size of all loci in each OCT image was modified accordingly. This adjustment ensures more accurate and reliable correlations between retinal sensitivity and choroidal thickness, particularly across various retinal loci.

Figure 3.

Image co-registration and overlay. (A) Overlaid registered enface OCT image on MP paired with color fundus image; (B) integrated images of the registered choroidal thickness map of the macular (6 × 6 mm) on microperimetry focal points, (C) 33 mean choroidal thickness measurements, corresponding to 33 retinal sensitivity focal points.

Relationships Between Retinal Sensitivity and Choroidal Thickness

We investigated the association between retinal sensitivity and choroidal thickness across 15 retinal eccentricities, as depicted in Figure 4. The analysis was performed at specific eccentricities from the fovea (2 degrees, 4 degrees, 6 degrees, 8 degrees, and 10 degrees), within cumulative eccentricity regions (≤2 degrees, ≤4 degrees, ≤6 degrees, ≤8 degrees, and ≤10 degrees), and across retinal subfields: foveal (F), nasal (N), superior (S), temporal (T), and inferior (I). For each microperimetry (MP) point, corresponding choroidal thickness values were averaged to compute the mean thickness (Fig. 3C). Retinal sensitivity values were converted to a linear scale by taking the anti-log of their logarithmic decibel values.³³ The mean focal choroidal thickness and retinal sensitivity values were then calculated within each zone, and correlations between retinal sensitivity and choroidal thickness were evaluated based on these averaged values. This method enabled the relationship between retinal sensitivity and choroidal thickness across different retinal eccentricity zones to be evaluated.

Figure 4.

Testing protocols. Thirty-three micro-perimetry test locations points were divided into 15 eccentricity groups according to (A) degree from the fovea (2 degrees, 4 degrees, 6 degrees, 8 degrees, and 10 degrees); (B) within each of the eccentricity regions (≤ 2 degrees, ≤ 4 degrees, ≤ 6 degrees, ≤ 8 degrees, and ≤ 10 degrees); (C) within retinal subfield regions: subfovea (F), nasal (N), superior (S), temporal (T), and inferior (I).

Statistical Analysis

Statistical analyses were conducted to examine the relationship between retinal sensitivities and choroidal thickness. Demographic and characteristic data were summarized as mean (SD) for continuous variables and as number and proportion for categorical variables. Pearson's correlation was used to investigate factors associated with choroidal thickness, including retinal sensitivity, signal strength, axial length, age, and retinal thickness, with clustering at the participant level to adjust for inter-eye correlations. This analysis was performed using bootstrap resampling (1000 iterations) to assess the stability of the correlation estimates. The association between choroidal thickness and gender was assessed using ANOVA. Factors associated with retinal sensitivity were evaluated using univariate and multivariate mixed models, also clustered at the participant level to adjust for inter-eye correlations. In univariate analyses, the models were performed separately for each variable. Multivariate regression modeling was performed with retinal sensitivity as the dependent variable and choroidal thickness, axial length, signal strength, age, gender, and retinal thickness as the independent variables. A P value less than 0.05 was considered statistically significant. All data processing and analysis were carried out with Python software version 3.7.

Results

Two hundred eighty eyes from 155 patients with myopia were included in this study. The subjects had a mean age of 25.8 ± 3.0 years and a mean refractive error of −6.87 ± 2.05 D. Mean axial length and signal strength was 26.56 ± 0.99 mm and 9.14 ± 0.46, respectively. Participant characteristics, demographics, and corresponding mean values for retinal sensitivity and choroidal thickness are detailed in Table 1. The mean retinal sensitivity and retinal thickness were 27.68 ± 1.28 dB, and 299.75 ± 12.15 µm, respectively, whereas the global choroidal thickness was 217.85 ± 66.66 µm with choroidal thickness in the nasal subfield (181.36 ± 62.95 µm) significantly thinner than in the other quadrants, with mean differences of 32.68 µm, 57.05 µm, and 58.05 µm compared to the inferior, temporal, and superior subfields, respectively (P < 0.001, Tukey honestly significant difference [HSD] post hoc test).

Table 1.

Characteristics and Demographics of the Study Subjects

Characteristic	Mean (±SD)
Number of subjects, eyes	155, 280
Age, y	25.8 (±3.0)
Gender, F (%)	67 (43)
Signal strength	9.14 (±0.46)
Axial length, mm	26.56 (±0.99)
Refractive error, D	–6.87 (±2.05)
Retinal sensitivity, dB	27.68 (±1.28)
Retinal thickness, µm
Global	299.75 (±12.15)
Subfoveal	242.62 (±26.30)
Nasal	324.81 (±14.29)
Temporal	290.123 (±11.48)
Superior	309.89 (±13.08)
Inferior	300.74 (±12.91)
Choroidal thickness, µm
Global	217.85 (±66.66)
Subfoveal	215.67 (±72.648)
Nasal	181.36 (±62.95)
Temporal	238.29 (±68.74)
Superior	239.33 (±68.38)
Inferior	213.67 (±70.88)

SD, standard deviation.

We examined the association of focal choroidal thickness with retinal sensitivity in all retinal eccentricity regions along with other factors (Table 2). Retinal sensitivity, retinal thickness, axial length, and age were correlated with the choroidal thickness (P value < 0.05), except subfoveally. Signal strength was not correlated with choroidal thickness. The correlations between global choroidal thickness and global retinal sensitivity, global retinal thickness, axial length, and age, as well as the correlations between mean choroidal thickness and mean retinal sensitivity across the retinal subfields (nasal, temporal, inferior, and superior) are illustrated in the scatter plots shown in Figure 5. The plots indicate that the relationship between retinal sensitivity and choroidal thickness varies depending on the retinal eccentricity region.

Table 2.

Association of the Choroidal Thickness With Clinical and Imaging Factors

	Retinal Sensitivity†		Signal Strength†		Axial Length†		Age†		Gender‡		Retinal Thickness†
RE	r (95% CI)	P Value*	r (95% CI)	P Value*	r (95% CI)	P Value*	r (95% CI)	P Value*	r	P Value*	r (95% CI)	P Value*
2 degrees	0.204 (0.08 to 0.32)	<0.001	0.049 (−0.08 to 0.17)	0.414	−0.416 (−0.50 to −0.33)	<0.001	0.187 (0.06 to 0.32)	0.002	2.577	0.109	0.215 (0.12 to 0.31)	<0.001
4 degrees	0.256 (0.14 to 0.36)	<0.001	0.048 (−0.09 to 0.17)	0.425	−0.402 (−0.48 to −0.31)	<0.001	0.184 (0.05 to 0.31)	0.002	3.682	0.056	0.346 (0.24 to 0.44)	<0.001
6 degrees	0.307 (0.20 to 0.42)	<0.001	0.056 (−0.06 to 0.18)	0.351	−0.396 (−0.48 to −0.31)	<0.001	0.174 (0.05 to 0.30)	0.004	3.773	0.053	0.314 (0.19 to 0.43)	<0.001
8 degrees	0.248 (0.15 to 0.35)	<0.001	0.042 (−0.09 to 0.16)	0.481	−0.381 (−0.47 to −0.29)	<0.001	0.172 (0.04 to 0.30)	0.004	3.605	0.059	0.297 (0.18 to 0.41)	<0.001
10 degrees	0.240 (0.12 to 0.35)	<0.001	0.044 (−0.08 to 0.16)	0.464	−0.363 (−0.45 to −0.27)	<0.001	0.171 (0.04 to 0.30)	0.004	3.260	0.072	0.225 (0.10 to 0.34)	<0.001
≤ 2 degrees	0.160 (0.05 to 0.28)	0.008	0.048 (−0.07 to 0.19)	0.425	−0.417 (−0.50 to −0.34)	<0.001	0.187 (0.06 to 0.32)	0.002	2.393	0.123	0.145 (0.03 to 0.25)	0.016
≤ 4 degrees	0.202 (0.09 to 0.32)	<0.001	0.048 (−0.07 to 0.17)	0.428	−0.413 (−0.50 to −0.33)	<0.001	0.186 (0.05 to 0.31)	0.002	3.015	0.084	0.234 (0.13 to 0.34)	<0.001
≤ 6 degrees	0.245 (0.13 to 0.35)	<0.001	0.050 (−0.07 to 0.18)	0.407	−0.410 (−0.49 to −0.32)	<0.001	0.184 (0.06 to 0.31)	0.002	3.208	0.074	0.285 (0.18 to 0.40)	<0.001
≤ 8 degrees	0.255 (0.14 to 0.37)	<0.001	0.049 (−0.08 to 0.18)	0.420	−0.406 (−0.50 to −0.32)	<0.001	0.182 (0.06 to 0.31)	0.002	3.316	0.070	0.304 (0.20 to 0.40)	<0.001
≤ 10 degrees	0.261 (0.15 to 0.36)	<0.001	0.048 (−0.08 to 0.17)	0.424	−0.402 (−0.49 to −0.31)	<0.001	0.182 (0.05 to 0.31)	0.002	3.345	0.069	0.305 (0.20 to 0.40)	<0.001
F	0.070 (−0.05 to 0.2)	0.249	0.034 (−0.09 to 0.17)	0.572	−0.409 (−0.49 to −0.32)	<0.001	0.181 (0.04 to 0.31)	0.003	2.428	0.120	0.058 (−0.06 to 0.17)	0.337
N	0.224 (0.11 to 0.34)	<0.001	0.032 (−0.09 to 0.16)	0.594	−0.398 (−0.48 to −0.31)	<0.001	0.214 (0.08 to 0.34)	<0.001	0.728	0.394	0.218 (0.12 to 0.31)	<0.001
T	0.278 (0.15 to 0.38)	<0.001	0.058 (−0.07 to 0.19)	0.332	−0.358 (−0.45 to −0.26)	<0.001	0.131 (0.00 to 0.26)	0.011	7.103	0.008	0.284 (0.15 to 0.42)	<0.001
S	0.158 (0.05 to 0.27)	0.008	0.024 (−0.09 to 0.15)	0.688	−0.361 (−0.45 to −0.27)	<0.001	0.152 (0.03 to 0.28)	0.028	1.438	0.231	0.286 (0.18 to 0.40)	<0.001
I	0.193 (0.10 to 0.29)	0.001	0.031 (−0.09 to 0.16)	0.609	−0.356 (−0.45 to −0.27)	<0.001	0.170 (0.03 to 0.30)	0.005	6.059	0.014	0.241 (0.12 to 0.36)	<0.001

RE, Retinal eccentricity; refers to the distance from the center of the fovea, with the fovea itself being at 0 degrees eccentricity; F, subfovea; N, nasal; T, temporal; S, superior; I, inferior; r, Pearson correlation coefficient; P P value; f, ANOVA f-statistic; 95% CI, 95% confidence interval.

^*Bolded values of P value indicate statistically significance.

^†Association using Pearson correlation with 1000 bootstraps.

^‡Association using ANOVA test.

Figure 5.

Association between choroidal thickness and retinal sensitivity across various retinal eccentricities. Panels A1 to A4 depict the correlations between global choroidal thickness and (A1) global retinal sensitivity, (A2) global retinal thickness, (A3) axial length, and (A4) age, respectively. Panels B1 to B4 illustrate the correlations between mean choroidal thickness and mean retinal sensitivity in different retinal subfields: (B1) nasal, (B2) temporal, (B3) inferior, and (B4) superior. Retinal sensitivity values were converted to a linear scale by taking the anti-log of their logarithmic decibel values.³³

Next, we identified the factors associated with retinal sensitivity using both univariate and multivariate mixed models. Each variable was analyzed separately in univariate analyses (Table 3). Greater choroidal thickness was significantly associated with greater retinal sensitivity, except at the subfovea. Retinal thickness was significantly associated with retinal sensitivity in 6 of the 15 retinal eccentricity regions, whereas age and signal strength were significantly associated in 1 of the 15 regions. The axial length and gender were not associated with retinal sensitivity. The multivariate mixed models, which included choroidal thickness adjusted for age, signal strength, axial length, gender, and retinal thickness, are presented in Table 4, and showed that retinal sensitivities and choroidal thicknesses remained significantly associated.

Table 3.

Univariate Mixed Models for Retinal Sensitivity

	Choroidal Thickness		Signal Strength		Axial Length		Age		Gender†		Retinal Thickness
RE	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*
2 degrees	0.73 (0.25 to 1.21)	0.003	−19.30 (−85.11 to 46.5)	0.570	22.37 (−14.08 to 58.83)	0.229	8.25 (−3.78 to 20.28)	0.179	−7.12 (−80.76 to 66.52)	0.850	1.57 (−0.50 to 3.64)	0.137
4 degrees	0.77 (0.39 to 1.16)	<0.001	−58.60 (−109.14 to −8.05)	0.020	−3.83 (−32.33 to 24.66)	0.792	5.33 (−4.38 to 15.05)	0.282	22.41 (−36.72 to 81.54)	0.458	3.46 (1.63 to 5.30)	<0.001
6 degrees	0.96 (0.57 to 1.36)	<0.001	27.66 (−23.16 to 78.48)	0.290	−10.89 (−39.65 to 17.87)	0.458	7.54 (−2.29 to 17.36)	0.133	−8.55 (−68.66 to 51.57)	0.780	2.69 (0.73 to 4.65)	0.007
8 degrees	0.59 (0.26 to 0.92)	<0.001	−1.20 (−38.15 to 35.7)	0.950	0.66 (−21.46 to 22.78)	0.953	4.80 (−2.86 to 12.46)	0.220	4.59 (−42.09 to 51.28)	0.847	2.04 (0.27 to 3.80)	0.024
10 degrees	0.57 (0.24 to 0.89)	0.001	3.30 (−30.55 to 37.16)	0.850	−0.24 (−20.65 to 20.17)	0.982	4.17 (−2.91 to 11.24)	0.248	8.08 (−34.98 to 51.15)	0.713	−0.16 (−1.67 to 1.44)	0.885
≤ 2 degrees	0.45 (0.00 to 0.89)	0.049	0.26 (−58.47 to 58.99)	0.990	23.57 (−9.96 to 57.10)	0.168	6.07 (−2.37 to 14.52)	0.159	−10.07 (−79.54 to 59.4)	0.776	0.16 (−1.47 to 1.79)	0.845
≤ 4 degrees	0.53 (0.13 to 0.92)	0.009	−20.77 (−71.16 to 29.63)	0.420	15.86 (−13.91 to 45.63)	0.296	6.53 (−2.42 to 15.48)	0.153	−0.01 (−62.17 to 62.15)	1.000	0.97 (−0.84 to 2.77)	0.295
≤ 6 degrees	0.63 (0.26 to 1.01)	0.001	−8.50 (−54.65 to 37.66)	0.720	10.10 (−17.76 to 37.96)	0.477	7.08 (−2.51 to 16.67)	0.148	−2.11 (−60.69 to 56.48)	0.944	1.27 (−0.67 to 3.22)	0.200
≤ 8 degrees	0.62 (0.27 to 0.98)	0.001	−7.16 (−48.76 to 34.44)	0.740	9.11 (−16.69 to 34.91)	0.489	7.13 (−3.05 to 17.30)	0.170	−0.76 (−55.38 to 53.87)	0.978	1.58 (−0.36 to 3.53)	0.111
≤ 10 degrees	0.61 (0.27 to 0.95)	<0.001	−5.45 (−43.53 to 32.63)	0.780	8.38 (−15.82 to 32.58)	0.497	8.11 (−3.25 to 19.47)	0.162	0.67 (−50.86 to 52.20)	0.980	1.39 (−0.53 to 3.31)	0.156
F	0.26 (−0.32 to 0.83)	0.377	19.37 (−63.69 to 102.43)	0.650	23.34 (−19.15 to 65.84)	0.282	9.04 (−5.16 to 23.23)	0.212	−17.15 (−103.87 to 69.56)	0.698	0.15 (−1.33 to 1.63)	0.843
N	0.52 (0.08 to 0.96)	0.021	−5.19 (−54.33 to 43.96)	0.840	6.34 (−22.71 to 35.38)	0.669	4.73 (−5.19 to 14.65)	0.350	18.01 (−42.27 to 78.28)	0.558	1.61 (−0.31 to 3.52)	0.100
T	0.89 (0.53 to 1.25)	<0.001	5.16 (−39.64 to 49.95)	0.820	−6.45 (−32.38 to 19.47)	0.626	9.17 (0.33 to 18.00)	0.042	15.51 (−38.80 to 69.82)	0.576	3.80 (1.66 to 5.94)	0.001
S	0.35 (0.03 to 0.68)	0.034	−15.09 (−54.15 to 23.98)	0.450	9.46 (−14.44 to 33.36)	0.438	4.37 (−3.93 to 12.66)	0.302	21.94 (−28.40 to 72.27)	0.393	1.75 (0.00 to 3.49)	0.050
I	0.37 (0.09 to 0.66)	0.011	−23.25 (−60.54 to 14.04)	0.220	−1.66 (−23.49 to 20.16)	0.881	2.59 (−4.90 to 10.08)	0.498	−20.94 (−66.25 to 24.37)	0.365	1.66 (0.04 to 3.28)	0.045

Mixed models were clustered at the individual level to adjust for inter-eye correlations.

^*Bolded values of P value indicate statistically significance.

^†Female gender was set as the reference.

Table 4.

Multivariate Mixed Models for Retinal Sensitivity With Choroidal Thickness Adjusted for Age, Axial Length, Gender, and Retinal Thickness

	Choroidal Thickness		Signal Strength		Axial Length		Age		Gender†		Retinal Thickness
RE	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*	β (95% CI)	P Value*
2 degrees	1.03 (0.48 to 1.57)	<0.001	4.07 (−61.81 to 69.96)	0.904	64.99 (22.33 to 107.65)	0.003	6.33 (−5.5 to 18.16)	0.294	−75.23 (−154.86 to 4.4)	0.064	1.21 (−0.87 to 3.29)	0.255
4 degrees	0.70 (0.25 to 1.14)	0.001	−55.87 (−106.31 to −5.42)	0.030	14.01 (−18.61 to 46.62)	0.400	0.83 (−8.48 to 10.14)	0.861	−5.30 (−67.34 to 56.74)	0.867	2.47 (0.53 to 4.4)	0.012
6 degrees	1.08 (0.63 to 1.53)	<0.001	35.62 (−14.36 to 85.61)	0.162	36.89 (3.8 to 69.98)	0.029	3.64 (−5.76 to 13.03)	0.448	−57.70 (−120.25 to 4.85)	0.071	1.73 (−0.24 to 3.71)	0.086
8 degrees	0.66 (0.29 to 1.03)	<0.001	4.18 (−32.9 to 41.26)	0.825	25.56 (−0.05 to 51.17)	0.050	2.67 (−4.73 to 10.07)	0.480	−24.82 (−73.93 to 24.29)	0.322	1.56 (−0.24 to 3.36)	0.089
10 degrees	0.67 (0.31 to 1.03)	<0.001	8.96 (−25.10 to 43.02)	0.606	16.71 (−6.82 to 40.24)	0.164	2.95 (−4.02 to 9.92)	0.407	−14.80 (−60.8 to 31.19)	0.528	−0.54 (−2.08 to 1.0)	0.494
≤2 degrees	0.74 (0.24 to 1.24)	0.003	20.43 (−39.29 to 80.16)	0.502	57.70 (18.56 to 96.84)	0.004	6.93 (−4.23 to 18.09)	0.223	−62.89 (−136.9 to 11.11)	0.096	0.02 (−1.59 to 1.63)	0.98
≤4 degrees	0.73 (0.28 to 1.18)	0.001	−8.41 (−59.51 to 42.69)	0.747	43.14 (8.64 to 77.63)	0.014	5.36 (−4.58 to 15.3)	0.291	−43.66 (−109.74 to 22.42)	0.195	0.61 (−1.19 to 2.41)	0.505
≤6 degrees	0.83 (0.41 to 1.25)	<0.001	2.38 (−43.98 to 48.74)	0.920	41.89 (9.97 to 73.82)	0.010	4.93 (−4.33 to 14.18)	0.297	−47.37 (−108.93 to 14.19)	0.132	0.68 (−1.25 to 2.62)	0.488
≤8 degrees	0.80 (0.40 to 1.19)	<0.001	2.11 (−39.55 to 43.77)	0.921	39.22 (9.80 to 68.63)	0.009	4.42 (−4.18 to 13.03)	0.314	−44.18 (−101.31 to 12.96)	0.13	0.97 (−0.97 to 2.91)	0.328
≤10 degrees	0.78 (0.39 to 1.16)	<0.001	2.96 (−35.11 to 41.03)	0.879	36.27 (8.71 to 63.82)	0.010	4.10 (−4.05 to 12.24)	0.324	−39.25 (−93.17 to 14.67)	0.154	0.76 (−1.15 to 2.67)	0.437
F	0.54 (−0.10 to 1.19)	0.099	46.98 (−39.23 to 133.19)	0.285	57.62 (6.16 to 109.09)	0.028	8.77 (−5.29 to 22.84)	0.222	−70.39 (−163.9 to 23.11)	0.14	0.05 (−1.43 to 1.52)	0.952
N	0.58 (0.08 to 1.07)	0.023	−0.46 (−50.77 to 49.84)	0.986	21.79 (−11.52 to 55.11)	0.200	2.64 (−7.06 to 12.34)	0.594	−3.07 (−66.27 to 60.14)	0.924	1.26 (−0.65 to 3.17)	0.195
T	0.90 (0.50 to 1.29)	<0.001	9.07 (−34.4 to 52.54)	0.682	26.55 (−2.58 to 55.69)	0.074	5.85 (−2.67 to 14.37)	0.178	−37.10 (−94.92 to 20.72)	0.209	2.77 (0.58 to 4.96)	0.013
S	0.35 (0.00 to 0.71)	0.049	−12.33 (−51.83 to 27.16)	0.540	19.11 (−7.94 to 46.16)	0.166	3.12 (−5.04 to 11.29)	0.453	5.11 (−48.25 to 58.46)	0.851	1.43 (−0.36 to 3.22)	0.117
I	0.43 (0.10 to 0.75)	0.007	−16.34 (−54.27 to 21.58)	0.398	18.23 (−7.29 to 43.75)	0.162	0.97 (−6.38 to 8.32)	0.797	−42.98 (−91.99 to 6.03)	0.086	1.47 (−0.19 to 3.13)	0.082

Mixed models were clustered at the individual level to adjust for inter-eye correlations.

^*Bolded values of P value indicate statistically significance.

^†Female gender was set as the reference.

Supplementary Figure S2 illustrates the choroidal thickness maps in eyes with long versus short axial lengths. Eyes with longer axial lengths show a thinner choroid and lower retinal sensitivity compared with those with shorter axial lengths, demonstrating the structural and functional differences associated with axial elongation. We further quantified the proportion of variance in retinal sensitivity independently explained by choroidal thickness, signal strength, age, gender, and axial length. Univariate analysis was performed to assess the contribution of each predictor, whereas multivariate analysis calculated the total fixed-effect variance. The results, including the proportion of fixed-effect variance explained by each factor, are detailed in Supplementary Table S1.

Discussion

In this study, our results show that choroidal thickness in myopic eyes is significantly associated with retinal sensitivity across various retinal loci. This association suggests that choroidal thickness may be associated with visual function based on retinal sensitivities. Our findings align with previous studies,^34,35 which highlighted choroidal thickness as an important factor influencing visual acuity and retinal sensitivity. Furthermore, we observed that choroidal thickness is associated with age, axial length, and retinal thickness, consistent with findings in the literature.^4,8,15–18 Retinal sensitivity showed distinct regional patterns: whereas univariate analysis revealed no significant associations with axial length and only one significant association in 15 regions for age and signal strength, multivariate analysis identified axial length as an independent predictor in 8 of 15 retinal regions after adjustment. To our knowledge, this is the first study to assess the structure-function relationship in myopia using focal choroidal thickness measurements aligned with standardized microperimetry testing. A key strength of this study is the use of deep-learning models to measure choroidal and retinal thickness, as well as the mapping from microperimetry stimuli. This approach eliminates variability inherent in manual measurements, reduces potential bias, and ensures the reproducibility of results. By investigating these focal relationships, our work not only contributes novel mechanistic insights into myopic retinal dysfunction but also establishes a framework for artificial intelligence (AI)-driven, high-resolution, structural-functional analyses in future research.

Choroidal thinning has been associated with a reduced supply of oxygen and nutrients to the retina, which may contribute to visual field defects and an increased risk of myopic maculopathy and choroidal neovascularization.^4–8 This reduced blood flow weakens the retina's ability to meet its metabolic needs, putting significant stress on photoreceptors and other retinal structures. Over time, this stress can accelerate retinal damage, potentially resulting in more severe conditions which are associated with considerable vision loss in patients with myopia.^36–38 Furthermore, choroidal thinning is linked to a broader spectrum of degenerative eye diseases, highlighting the essential role of the choroid in preserving retinal health and function.¹³ Previous research has typically used broader sectoral approaches, as seen in Wu et al.,²⁰ who aggregated data across the macula into inner and outer rings using a 3 × 3 mm macular scan with the MP-1 microperimeter. Similarly, da Silva et al.⁸ used the ETDRS grid to average choroidal thickness over large sectors and relied on single-point manual measurements for the outer retinal layers, which may miss focal changes. Furthermore, their study cohort was relatively small, comprising only 37 patients divided into 2 groups—the healthy controls and the patients with high myopia. They found that choroidal thinning correlated with decreased retinal sensitivity using a 7 × 7 mm macular scan and the MP-3 microperimeter across the superior, nasal, inferior, and temporal sectors. Another study by Zaben et al.⁴ identified partial correlations in specific macular regions using a 6 × 6 mm macular scan protocol and the MAIA microperimeter. However, these studies did not provide the granularity necessary to detect variations at individual test points. Our study assessed retinal sensitivity at 33 locations using an MP-3 microperimeter, with mean choroidal thickness and retinal thickness at these points automatically calculated by deep-learning models. This approach provided a more precise representation of focal differences in retinal sensitivity, choroidal thickness, and retinal thickness, avoiding the pitfalls of averaging over larger sectors or relying solely on single-point measurements. Our results are consistent with previous findings using the ETDRS map.⁸ However, unlike their study, we did not find significant associations in the subfoveal sector. This discrepancy may be explained by the presence of diffuse chorioretinal atrophy in their cohort, whereas our study included participants with non-pathological myopia. Additionally, correlations in the temporal and inner superior sectors were similarly noted by Zaben et al.,⁴ whereas Wu et al.²⁰ observed significant correlations in the outer ring of the macular regions. Our study identified significant associations between retinal sensitivity and choroidal thickness in 14 regions and demonstrated that choroidal thickness varies significantly by location within the eye. The thickest choroidal thickness was observed in the superior macula, whereas the thinnest choroidal thickness was in the nasal region, consistent with previous studies,^4,8 even those utilizing manual single-point measurements.

Myopia progression is associated with retinal thinning, and factors such as age, gender, axial length, and refractive error influence both choroidal and retinal thickness.^{15,20,39–41} In our analysis of 15 retinal eccentricity regions, we found significant correlations between retinal sensitivity and choroidal thickness at most testing points, excluding the subfovea, even after including age, signal strength, axial length, gender, and retinal thickness as predictors. The strongest correlation was at 6 degrees (r = 0.307, P < 0.001), followed by 4 degrees, 8 degrees, 10 degrees, and 2 degrees, with the global choroidal thickness (≤ 10 degrees) showing r = 0.261 and a P value < 0.001. The retinal subfield analysis revealed the highest correlation in the temporal region (r = 0.278), followed by the nasal and inferior regions, with the superior region reporting the lowest correlation (r = 0.158). There was no significant correlation in the central retina subfield (0 degrees). Our findings suggest that different models may be required for specific retinal eccentricity regions. This is particularly evident in the central region, which had the lowest mean retinal sensitivity (Supplementary Table S2) and showed no correlation with either choroidal thickness or retinal thickness. This indicates that central retinal sensitivity may be less sensitive to differences in choroidal thickness or retinal thickness compared with peripheral regions, offering insights into the complexities of this relationship. The lack of significant correlations in the subfoveal region may be attributed to several factors. Biologically, the foveola is optimized for high-acuity vision and contains a very high density of cone photoreceptors, with recent studies reporting peak densities ranging from approximately 147,000 to over 215,000 cones/mm².⁴² This dense packing may confer functional redundancy, reducing sensitivity to subtle structural changes. Sampling limitations may also have contributed, as the study cohort primarily consisted of young individuals (mean age = 25.8 ± 3.0 years) without pathological myopia (mean refractive error = −6.87 ± 2.05 D), thereby limiting variability in the central retinal and choroidal structures. Consequently, the subfoveal region tends to exhibit stable OCT measurements, limiting the detectability of structure–function correlations. The central retina may therefore be less responsive to variations in retinal and choroidal thickness compared with peripheral regions.⁴³ These results provide insights into the complexity of structure-function relationships in the retina and indicate that central retinal sensitivity may be less influenced by choroidal or retinal thickness in this population.

This study has several limitations that should be considered. First, the time of day during OCT imaging was not controlled, which may have introduced minor variability in choroidal thickness measurements due to known diurnal fluctuations.⁴⁴ Second, the findings are based on data from a single cohort, limiting the generalizability of the results to other cohorts, ethnicities, or imaging devices. Third, only a single OCT scan was acquired per eye, and reproducibility was not assessed. Fourth, the study did not include cases with pathological conditions, leaving variations in choroidal thickness in disease states to be further investigated. However, the data presented here could serve as normative references for future comparisons with pathological cases. Finally, the cross-sectional design of this study precludes establishing causal relationships between observed changes in choroidal thickness, retinal sensitivity, and other factors, such as retinal thickness, underscoring the need for longitudinal studies to confirm these associations.

In conclusion, we found significant associations between choroidal thicknesses segmented from OCT and retinal sensitivities from MP, suggesting a structure-function relationship at the choroid. Longitudinal studies are necessary to further explore the relationship among choroidal thickness, retinal sensitivity, and myopia progression.