Age- and Myopia-Related 3-Dimensional Remodeling of Choroidal Thickness in Children Authors: Luxiao Chen, Jixuan Yuan, Ziyao Wang, Zhe Pan, Dan Zhu, Zeyue Fan, Di Zhang, Ziyao Xia, Jost B. Jonas, Chong Li, Tien Yin Wong, Ya Xing Wang Published: 2026 ABSTRACT PurposeTo investigate the 3-dimensional (3D) topographic remodeling patterns of choroidal thickness (ChT) in the macular region in children, and their associations with age and myopia related ocular growth.DesignCross-sectional, population-based study.ParticipantsA total of 3508 children (mean age 13.6 ± 3.0 years, range 7–18 years, 47.7% male).MethodsSpectral-domain OCT with a 31-horizontal B-scan protocol was performed on right eyes. Choroidal thickness in each B-scan was automatically segmented and quantified using modified open-source deep learning algorithms (Choroidalyzer). For each eye, a 3D ChT map covering 6 × 6 mm macular region and composed of 12 × 12 grids centered on fovea was generated, and both mean thickness and grid-specific thickness values were calculated.Main Outcome MeasuresChT of each grid, cluster, and macular region; correlation with age, axial length (AL), and sex.ResultsThe mean ChT was 271.8 ± 56.1 μm, with significant variations across nonmyopia, low, moderate, and high myopia groups (P < 0.001). Age-related ChT changes showed a distinct temporal-nasal asymmetry pattern, with the nasal region showing the strongest negative correlation with age (r = –0.12), the central region demonstrating a weak correlation (r = –0.07, both P < 0.001), whereas the temporal region remained relatively stable (r = –0.02, P = 0.255), resulting in 4.3- and 2.6-fold (95% confidence interval: –0.4 to 8.9; –0.4 to 5.7) greater thinning in the nasal and central choroid than in the temporal region, respectively. Axial length-related ChT changes showed a concentric thinning pattern, characterized by more pronounced foveal thinning (r = –0.51, B = –35.2 μm/mm, both P < 0.001) compared with parafoveal (r = –0.48, B = –28.7 μm/mm, both P < 0.001) and perifoveal changes (r = –0.42, B = –22.3 μm/mm, both P < 0.001), corresponding to 1.6- and 1.3-fold (95% confidence interval: 1.4–1.7; 1.2–1.4) greater thinning in the fovea and parafovea than in the perifovea.ConclusionsUsing 3D topographic mapping, we provide population-based data on ChT in children, and identified 2 major spatial remodeling patterns: temporal–nasal asymmetry with age and concentric thinning with increasing axial length. These spatially and temporally distinct remodeling patterns likely reflect both physiological growth and myopia-related structural change, supporting the potential role of ChT as a biomarker for myopia development and progression.Financial Disclosure(s)Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article. The choroid has increasingly been recognized in playing a role in ocular physiology and in the pathogenesis of several major eye diseases, including age-related macular degeneration, glaucoma, and myopia.1, 2, 3 In myopia, choroidal thickness (ChT) is now seen as a key biomarker of myopia development and progression.4 To cite an example, progressive choroidal thinning is one of the major morphological changes during myopic axial elongation and has been linked to the risk of sight-threatening complications such as lacquer cracks, myopic patchy atrophies, and myopic macular neovascularization in pathological myopia.5, 6, 7 Choroidal thickness may also serve as a predictive indicator, for example, in assessing the effectiveness of orthokeratology treatment.8 These observations underscore the importance of precise characterization of ChT in children in the understanding of the physiology of eye growth and the pathophysiology of myopia. Although ChT is closely associated with refractive status and thus myopia development, distinguishing between normal physiological developmental changes in ChT from early abnormal alterations of ChT in myopia requires more detailed knowledge of the “normal” choroidal growth patterns in children and their key modifiers, particularly age and refractive status. Most previous studies, however, have relied on single-point or regional ChT measurements based on 2-dimensional (2D) OCT scans, which assume symmetry of the choroid. It might have led to an overlook of the variability of the true 3-dimensional (3D) structure of the ChT. Age-related and myopia-related effects have also often been analyzed separately, limiting insight into how these 2 major drivers of ocular growth jointly shape choroidal structure. A few 3D OCT studies of ChT have been conducted, but they involved relatively small samples (40 to 340 participants),9, 10, 11 and were underpowered to disentangle the highly interrelated effects of age and myopia. As a result, large, population-based normative 3D OCT data of the ChT in healthy children have been mostly lacking so far. To address this gap, we conducted a large-scale population-based study of 3508 children aged 7 to 18 years, using spectral-domain OCT with deep learning-based segmentation. Our study aimed to characterize the 3D topographic pattern of ChT in children, and their associations with age and refractive status as biomarkers of ocular growth and myopia development. By providing high-resolution, population-based 3D maps of pediatric ChT, our study aimed to establish normative reference data and a structural framework for the trajectory of choroidal remodeling across childhood growth and myopia progression. These data may inform future use of ChT as a clinically meaningful biomarker in myopia research and care. Methods Study Population The population-based study was conducted in Bayanhushu, Inner Mongolia, China, in May 2023. Using a school-based clustering sampling approach, we randomly selected 1 primary school, 1 secondary school, and 1 senior high school, and examined all students graded 1 to 12 in these schools. Among 4843 eligible students, 4596 attended the examination visit, yielding a response rate of 94.9%. Students with a refractive error of < –3.0 diopters in their worse eye were scheduled for annual follow-up. The present study refers to the measurements obtained at the baseline examination in May 2023. The study adhered to the tenets of the Declaration of Helsinki and was approved by the Ethics Committee of the Affiliated Hospital of Inner Mongolia Medical University. Written informed consents were obtained from at least 1 parent or legal guardian of each child. Clinical Examination and OCT Scanning Protocol All subjects underwent a series of comprehensive ophthalmic evaluations performed by trained technicians. These included measurement of uncorrected visual acuity and best-corrected visual acuity, anterior segment examination using slit lamp biomicroscopy, noncontact tonometry (Canon Full Auto Tonometer TX-20; Canon Inc), ocular biometry (IOL Master 5.5; Carl Zeiss Meditec, Inc), and spectral-domain OCT (Spectralis OCT; Heidelberg Engineering). Noncycloplegic auto-refractometry (TOPCON KR-1, Topcon Co) was conducted 3 times per each participant, and the mean value was used for statistical analyses. The macula region was scanned by OCT, centered on fovea, using a cube scan mode covering a 30° × 30° area comprising 31 continuous horizontal B scans. Each scan line consisted of 100 averaged scans. The axial (z) resolution was 3.87 μm/pixel. Data Preprocess For the statistical analysis, we included the data of only 1 eye (right eyes) per participant. Inclusion criteria were: (1) right eyes with a best-corrected visual acuity of ≥0.8 (decimals); (2) age between 7 and 18 years; (3) no history of intraocular diseases, intraocular surgery, or systemic diseases (such as arterial hypertension or diabetes mellitus); and (4) available horizontal OCT B-scan images with ChT measurements. Data from 3508 right eyes from 3508 eligible children were included for further analyses. We classified the myopia severity according to age-specific and sex-specific axial length (AL) percentiles. Participants were stratified by single-year age groups (e.g., 7, 8, 9) and by sex, and within each age-sex stratum, AL values were ordered, and percentile cutoffs were defined so that the proportions corresponded to the latest national estimates of myopia prevalence among Chinese primary and secondary school students.12 Accordingly, for each age-sex stratum, the lowest 48.1% of the AL distribution was categorized as nonmyopia, followed by 27.7% as low myopia, further 19.2% as moderate myopia, and the highest 5.0% as high myopia. Choroidal Image Analysis On the horizontal OCT images, the choroid scan was segmented by a modified open-source deep learning algorithm named Choroidalyzer based on U-Net architecture in Python software (v. 3.12.3; Python Software Foundation) to obtain binary masks of the choroid compartment, defined as the space between Bruch’s membrane (BM) and the choroid–sclera interface.13 The resulting binary masks were processed for thickness measurement using a custom-written algorithm based on the measurement methodology of DeepGPET.14 For each pixel along BM, a line perpendicular to BM was constructed and extended to the choroid-sclera interface. Choroidal thickness was defined as the perpendicular distance along this normal. A simple random sampling method without replacement was applied to randomly select 35 participants from the original participant list. An experienced ophthalmologist measured ChT manually on selected points by using ImageJ software (version 1.54g, National Institutes of Health; https://imagej.nih.gov/ij/). We assessed the agreement by applying Pearson's correlation analysis and Bland–Altman agreement analysis. The analyses on the remeasurements of ChT showed a high agreement between automatic and manual segmentations (Fig. S1, available at www.ophthalmologyscience.org). All these available measurements were converted into physical distance by penetration resolution scale, and bilinear interpolation was then performed to obtain point-wise choroid thickness at a resolution of 10 μm in the x-axis and y-axis in the MATLAB software (MATLAB 2025a; The MathWorks, Inc). As a result, a 6 × 6 mm squared and interpolated ChT map centered on the fovea was constructed. The map was divided into 12 × 12 equal squared grids with 500 μm in length. Grid-wise averaged ChT was calculated, and a color-coded thickness map was superimposed onto the en-face image centered at fovea. Two consecutive spatial analyses were performed. First, correlation analyses and regression analyses were performed between grid-wise averaged ChT measurements and physiological parameters (age and AL). Second, a cluster-related analysis was conducted based on the k-means clustering results of the previous grid-wise analysis using Euclidean distance. The silhouette score, measuring how well each data point fits within its assigned cluster compared to other clusters, was then used to evaluate the quality of clustering results. Statistical Analysis All statistical analyses of this study were performed by Python software. Data were presented as mean (standard deviation) for continuous variable and count (proportion) for categorical variable. In these spatial analyses, Shapiro–Wilk test was used to determine normality, and Levene’s test was applied to confirm homogeneity of variance when comparing metrics between different groups. Analysis of variance (ANOVA) and post hoc Tukey honestly significant difference test were applied in homogeneity and Welch’s ANOVA and Games–Howell post hoc test else. The chi-square test was used for comparing categorical variables. Student’s t test with Bonferonni’s correction was conducted for grid-wise comparison between sex-defined groups. When comparing matching groups, repeated measures ANOVA with Geisser–Greenhouse correction was applied with paired t test with Bonferroni correction as post hoc analysis. Spearman’s correlation and point-biserial analysis of correlation analysis were for continuous and categorical independent variables, respectively. Univariate and stepwise removal multivariate regression were further used to evaluate the correlation between age, AL and sex, and cluster-averaged ChT. Only independent variables with P < 0.05, and variance inflation factor (VIF) <10 were selected for constructing final models. A 2-sided P value of <0.05 was considered statistically significant. Results Demographic Characteristics The study included 3508 eyes of 3508 subjects (1674 boys, 47.7%) with mean age of 13.6 ± 3.0 years (range: 7–18 years). The mean refractive error was –2.1 ± 2.2 diopters (range: -11.0 to +5.5 diopters), and the mean AL was 24.2 ± 1.2 mm (range: 20.5–29.0 mm). Ocular and systemic parameters stratified by myopia severity were summarized in Table 1.Table 1Demographic Characteristics of Recruited SubjectsParametersTotal (N = 3508)Nonmyopia (n = 1684)Low Myopia (n = 973)Moderate Myopia (n = 649)High Myopia (n = 202)P ValueSystemic parameters Boys, no. (%)1674 (47.7)795 (47.2)466 (47.9)311 (47.9)102 (50.5)0.844 Age, yrs13.6 (3.0)13.6 (3.0)13.6 (3.0)13.6 (3.0)13.7 (3.0)0.993 Height, cm158.5 (13.1)157.5 (13.1)159.0 (13.2)160.0 (12.8)161.0 (12.4)<0.001 Weight, kg55.0 (16.3)53.8 (15.9)55.5 (16.4)56.6 (16.5)58.0 (18.0)<0.001 SBP, mmHg111.8 (15.9)111.3 (16.0)111.9 (16.0)112.0 (15.2)114.1 (16.6)0.125 DBP, mmHg70.2 (8.7)69.9 (8.7)70.3 (8.7)70.2 (8.6)71.8 (9.6)0.028Ocular parameters K-mean43.5 (1.5)43.9 (1.4)43.2 (1.3)42.9 (1.4)42.4 (1.3)<0.001 Astig, diopter1.2 (0.6)1.2 (0.6)1.2 (0.6)1.3 (0.7)1.4 (0.7)<0.001 IOP, mmHg16.2 (2.9)16.2 (2.9)16.2 (2.9)16.2 (2.8)16.4 (2.6)0.779 ACD, mm3.6 (0.3)3.5 (0.2)3.6 (0.2)3.7 (0.2)3.8 (0.2)<0.001 AL, mm24.2 (1.2)23.3 (0.6)24.5 (0.5)25.4 (0.6)26.4 (0.8)<0.001 SE, diopter–2.1 (2.2)–0.9 (1.4)–2.3 (1.8)–3.9 (2.0)–5.6 (2.3)<0.001ACD = anterior chamber depth; AL = axial length; Astig = astigmatism degree; DBP = diastolic blood pressure; IOP = intraocular pressure; SBP = systolic blood pressure; SE = spherical equivalent.Data are presented as mean (standard deviation) for continuous variables and number (%) for sex. P values are for comparisons among the nonmyopia, low, moderate, and high myopia groups, with significance set at P < 0.05 (2-tailed). Table 1Demographic Characteristics of Recruited SubjectsParametersTotal (N = 3508)Nonmyopia (n = 1684)Low Myopia (n = 973)Moderate Myopia (n = 649)High Myopia (n = 202)P ValueSystemic parameters Boys, no. (%)1674 (47.7)795 (47.2)466 (47.9)311 (47.9)102 (50.5)0.844 Age, yrs13.6 (3.0)13.6 (3.0)13.6 (3.0)13.6 (3.0)13.7 (3.0)0.993 Height, cm158.5 (13.1)157.5 (13.1)159.0 (13.2)160.0 (12.8)161.0 (12.4)<0.001 Weight, kg55.0 (16.3)53.8 (15.9)55.5 (16.4)56.6 (16.5)58.0 (18.0)<0.001 SBP, mmHg111.8 (15.9)111.3 (16.0)111.9 (16.0)112.0 (15.2)114.1 (16.6)0.125 DBP, mmHg70.2 (8.7)69.9 (8.7)70.3 (8.7)70.2 (8.6)71.8 (9.6)0.028Ocular parameters K-mean43.5 (1.5)43.9 (1.4)43.2 (1.3)42.9 (1.4)42.4 (1.3)<0.001 Astig, diopter1.2 (0.6)1.2 (0.6)1.2 (0.6)1.3 (0.7)1.4 (0.7)<0.001 IOP, mmHg16.2 (2.9)16.2 (2.9)16.2 (2.9)16.2 (2.8)16.4 (2.6)0.779 ACD, mm3.6 (0.3)3.5 (0.2)3.6 (0.2)3.7 (0.2)3.8 (0.2)<0.001 AL, mm24.2 (1.2)23.3 (0.6)24.5 (0.5)25.4 (0.6)26.4 (0.8)<0.001 SE, diopter–2.1 (2.2)–0.9 (1.4)–2.3 (1.8)–3.9 (2.0)–5.6 (2.3)<0.001ACD = anterior chamber depth; AL = axial length; Astig = astigmatism degree; DBP = diastolic blood pressure; IOP = intraocular pressure; SBP = systolic blood pressure; SE = spherical equivalent.Data are presented as mean (standard deviation) for continuous variables and number (%) for sex. P values are for comparisons among the nonmyopia, low, moderate, and high myopia groups, with significance set at P < 0.05 (2-tailed). Demographic Characteristics of Recruited Subjects ACD = anterior chamber depth; AL = axial length; Astig = astigmatism degree; DBP = diastolic blood pressure; IOP = intraocular pressure; SBP = systolic blood pressure; SE = spherical equivalent. Data are presented as mean (standard deviation) for continuous variables and number (%) for sex. P values are for comparisons among the nonmyopia, low, moderate, and high myopia groups, with significance set at P < 0.05 (2-tailed). The mean ChT was 272±56 μm (range: 123–429 μm). In multivariate analysis, it decreased significantly with both older age and longer AL (age: B = –1.9 μm/year; P < 0.001; AL: B = –27.2 μm/mm; P < 0.001). Similar ChT decline was also observed in most grids with increasing age and myopia degree (Fig. S2, Table S1; available at www.ophthalmologyscience.org). ChT with Age: Temporal-Nasal Asymmetry Pattern Grid-wise partial correlation and multivariate analysis revealed a distinct temporal–nasal asymmetry in age-related ChT change (Fig. 1).Figure 1Grid-wise and cluster-based analysis of choroidal thickness (ChT) and age showing temporal–nasal asymmetry pattern. A–D, Standardized grid-wise ChT estimates across age groups: 7 to 9 years (A), 10 to 12 years (B), 13 to 15 years (C), and 16 to 18 years (D), derived from linear regression models. Red indicates thicker choroid; blue indicates thinner choroid. I = inferior; N = nasal; S = superior; T = temporal. E–F, K-means clustering of age-related ChT changes (E) and the modified vertical subregions used for analysis (F): temporal, central, and nasal. G, LOWESS-smoothed ChT trajectories across age for 3 subregions. Data points represent mean with 95% confidence interval (CI) per year of age. Temporal ChT remains stable, whereas central and nasal regions show progressive thinning with increasing age. Grid-wise and cluster-based analysis of choroidal thickness (ChT) and age showing temporal–nasal asymmetry pattern. A–D, Standardized grid-wise ChT estimates across age groups: 7 to 9 years (A), 10 to 12 years (B), 13 to 15 years (C), and 16 to 18 years (D), derived from linear regression models. Red indicates thicker choroid; blue indicates thinner choroid. I = inferior; N = nasal; S = superior; T = temporal. E–F, K-means clustering of age-related ChT changes (E) and the modified vertical subregions used for analysis (F): temporal, central, and nasal. G, LOWESS-smoothed ChT trajectories across age for 3 subregions. Data points represent mean with 95% confidence interval (CI) per year of age. Temporal ChT remains stable, whereas central and nasal regions show progressive thinning with increasing age. Using k-means clustering on these results, the macular region was divided into temporal, central, and nasal clusters (average silhouette score = 0.54). The temporal choroid was thickest (312 ± 58 μm), followed by the central (277 ± 63 μm) and nasal regions (227 ± 57 μm) (P < 0.001). Temporal ChT showed minimum changes with age (B = –0.5 μm/year, P = 0.248), whereas central and nasal ChT exhibited significant thinning (central: B = –2.1 μm/year; nasal: B = –3.3 μm/year, both P < 0.001), after adjusting AL and sex. Nasal and central thinning occurred 4.3- and 2.6-fold (95% confidence interval: –0.4 to 8.9; –0.4 to 5.7) faster than temporal thinning, respectively (Table 2).Table 2Regression Analysis of Choroidal Thickness (ChT) with Age by ClustersThickness (μm)ChT with Age (Univariate)ChT with Age (Multivariate)BPRR (95% CI)BPRR (95% CI)Temporal312 ± 58–0.30.349-–0.50.248-Central277 ± 63–2.1<0.0016.7 (–7.5, 20.8)–2.1<0.0012.6 (–0.4, 5.7)Nasal227 ± 57–2.4<0.0017.8 (–8.7, 24.3)–3.3<0.0014.3 (–0.4, 8.9)B = unstandardized coefficient; ChT = choroidal thickness; CI = confidence interval RR = rate ratio.Thickness is presented as mean ± standard deviation. Multivariate analysis was corrected with axial length and sex. Rate ratio was calculated with temporal as reference. Table 2Regression Analysis of Choroidal Thickness (ChT) with Age by ClustersThickness (μm)ChT with Age (Univariate)ChT with Age (Multivariate)BPRR (95% CI)BPRR (95% CI)Temporal312 ± 58–0.30.349-–0.50.248-Central277 ± 63–2.1<0.0016.7 (–7.5, 20.8)–2.1<0.0012.6 (–0.4, 5.7)Nasal227 ± 57–2.4<0.0017.8 (–8.7, 24.3)–3.3<0.0014.3 (–0.4, 8.9)B = unstandardized coefficient; ChT = choroidal thickness; CI = confidence interval RR = rate ratio.Thickness is presented as mean ± standard deviation. Multivariate analysis was corrected with axial length and sex. Rate ratio was calculated with temporal as reference. Regression Analysis of Choroidal Thickness (ChT) with Age by Clusters B = unstandardized coefficient; ChT = choroidal thickness; CI = confidence interval RR = rate ratio. Thickness is presented as mean ± standard deviation. Multivariate analysis was corrected with axial length and sex. Rate ratio was calculated with temporal as reference. ChT Pattern with AL: Concentric Thinning Pattern With increasing AL, a concentric pattern of choroidal thinning was observed, with each grid showing significant negative correlation and regression coefficient (P < 0.000347 for Bonferroni correction in each grid, Fig. 2).Figure 2Grid-wise and cluster-based analysis of choroidal thickness (ChT) and axial length (AL) showing concentric thinning pattern. A–D, Standardized grid-wise ChT estimates across refractive groups: nonmyopia (A), low (B), moderate (C), and high myopia (D). Red indicates thicker choroid; blue indicates thinner. E–F, K-means clustering of AL-related ChT changes (E) and corresponding modified concentric subregions (F): fovea, parafovea, and perifovea. G, LOWESS curves showing ChT versus AL for 3 foveal subregions. Choroidal thinning accelerates with increasing AL, most prominently in the central fovea. I = inferior; N = nasal; S = superior; T = temporal. Grid-wise and cluster-based analysis of choroidal thickness (ChT) and axial length (AL) showing concentric thinning pattern. A–D, Standardized grid-wise ChT estimates across refractive groups: nonmyopia (A), low (B), moderate (C), and high myopia (D). Red indicates thicker choroid; blue indicates thinner. E–F, K-means clustering of AL-related ChT changes (E) and corresponding modified concentric subregions (F): fovea, parafovea, and perifovea. G, LOWESS curves showing ChT versus AL for 3 foveal subregions. Choroidal thinning accelerates with increasing AL, most prominently in the central fovea. I = inferior; N = nasal; S = superior; T = temporal. Similar clustering analysis divided the macular choroid into fovea, parafovea, and perifovea regions (average silhouette score = 0.55). The magnitude of thinning varied across these regions, with decrease of 35.2 μm (12.8%), 28.7 μm (10.6%), and 22.3 μm (8.2%) with every mm increase in AL in the fovea, parafovea, and perifovea respectively (all P < 0.001), after adjusting age and sex (Table 3, Fig. 2). Foveal and parafoveal thinning was 1.6- and 1.3-fold (95% confidence interval: 1.4–1.7; 1.2–1.4) greater than perifovea (Table 3).Table 3Regression Analysis of Choroidal Thickness with Axial Length by ClustersThickness (μm)ChT with AL (Univariate)ChT with AL (Multivariate)BPRR (95% CI)BPRR (95% CI)Fovea275 ± 68–28.7<0.0011.5 (1.4, 1.7)–35.2<0.0011.6 (1.4, 1.7)Parafovea270 ± 58–23.3<0.0011.3 (1.1, 1.4)–28.7<0.0011.3 (1.2, 1.4)Perifovea272 ± 52–18.6<0.001-–22.3<0.001-AL = axial length; B = unstandardized coefficient; ChT = choroidal thickness; CI = confidence interval; RR = rate ratio.Thickness is presented as mean ± standard deviation. Multivariate analysis was corrected with age and sex. Rate ratio was calculated with perifovea as reference. Table 3Regression Analysis of Choroidal Thickness with Axial Length by ClustersThickness (μm)ChT with AL (Univariate)ChT with AL (Multivariate)BPRR (95% CI)BPRR (95% CI)Fovea275 ± 68–28.7<0.0011.5 (1.4, 1.7)–35.2<0.0011.6 (1.4, 1.7)Parafovea270 ± 58–23.3<0.0011.3 (1.1, 1.4)–28.7<0.0011.3 (1.2, 1.4)Perifovea272 ± 52–18.6<0.001-–22.3<0.001-AL = axial length; B = unstandardized coefficient; ChT = choroidal thickness; CI = confidence interval; RR = rate ratio.Thickness is presented as mean ± standard deviation. Multivariate analysis was corrected with age and sex. Rate ratio was calculated with perifovea as reference. Regression Analysis of Choroidal Thickness with Axial Length by Clusters AL = axial length; B = unstandardized coefficient; ChT = choroidal thickness; CI = confidence interval; RR = rate ratio. Thickness is presented as mean ± standard deviation. Multivariate analysis was corrected with age and sex. Rate ratio was calculated with perifovea as reference. In addition, we also analyzed ChT pattern with sex, which displayed thicker choroid in males than females in the inferior-central region (Fig. S3, Table S2, available at www.ophthalmologyscience.org). Discussion In this large, population-based cohort of school-aged children, we combined spectral-domain OCT with deep learning–based segmentation to generate fine-grained 3D topographic maps of ChT. Extending beyond prior work that relied on 2D or single-point measurements and typically examined age or myopia in isolation, we simultaneously mapped the full macular ChT landscape and jointly assessed the 2 major drivers of ocular growth—chronological age and myopia-related axial elongation. We identified 2 primary, spatially heterogeneous remodeling patterns—a temporal–nasal asymmetry predominantly associated with age and a concentric thinning pattern associated with myopic eye growth. By moving beyond simple associative averages to spatially resolved, growth- and myopia-sensitive maps, our findings help clarify the dynamic role of the choroid in childhood eye development and provide a structural framework for future applications that use ChT as a biomarker for myopia development and progression. To stratify myopia risk, we used an age- and sex-specific AL percentile framework rather than fixed refractive-error cut-offs, to better reflect the fact that physiological ocular growth is ongoing throughout childhood and to reduce potential risk misclassification in younger children. This approach rests on 2 assumptions. First, recent epidemiological evidence suggests that myopia prevalence in China has been relatively stable since around 2010,15 making it feasible to use a consistent population reference distribution for percentile-based grouping while limiting major cohort effects. Second, although ocular dimensions increase with age, children in a given percentile of AL at age 7 are broadly comparable to those within the same percentile at age 18 in terms of relative positioning within the population distribution, such that percentile rank may better capture an individual’s risk trajectory than absolute values.16,17 Although AL percentiles do not directly define myopia on their own, they provide a robust statistical framework to stratify individuals by relative ocular growth status for analyses of developmental patterns and associated anatomical changes. A key finding of this study is the temporal-nasal asymmetry of ChT change with age. This vertically distributed ChT pattern also showed features suggestive of a radial distribution centered on the optic nerve head,18 which may reflect aspects of early choroidal vasculogenesis and subsequent regional choroidal remodeling. This discovery offers crucial insights into the highly contradictory reports within the field of pediatric ChT research. Some previous studies reported thickening with age around 4-17 years old (primarily in European and Australian populations10,19,20), with others reporting thinning about 6 to 19 years old (primarily in Asian populations9,21,22), and some finding no correlation in children aged 3-6 years.23 These differences may stem from multiple factors, beginning with ethnic variance. Our study, based on a Chinese pediatric cohort, observed ChT thinning with age (mainly in the nasal and central regions), which is broadly consistent with previous findings in Asian populations.9,21,22 More importantly, we propose that the methodological differences of prior studies, particularly the limitations in measurement location, is a core reason for these conflicting conclusions. Past research has often relied on single-point subfoveal ChT10,19,21,23 or fixed regional averages (e.g., the concentric pattern by Nagasawa et al24). In contrast, our 3D grid-based mapping demonstrates that age-related ChT changes are highly heterogeneous: the temporal region remained almost unchanged (B = –0.5 μm/year), whereas the central and nasal regions thinned significantly (B = –2.1 and –3.3 μm/year, respectively). Therefore, studies reporting stable or thickened ChT10,19,20 may have had measurements diluted or dominated by the relatively stable temporal region. Conversely, studies reporting thinning9,21,22 may have better captured the changes we observed in the central and nasal zones. This highlights the limitations of relying on single-region or point measurement (like subfoveal thickness4,25,26) when assessing complex biological processes. Our 3D analysis suggests these seemingly contradictory reports may not be mutually exclusive but rather collectively reflect the complex and spatially heterogeneous remodeling of the choroid during maturation. Despite ethnic and methodological considerations, asymmetrical ocular enlargement can partially account for this asymmetry from a biological perspective.19 The second major finding of our study is the concentric pattern of ChT thinning associated with axial elongation, with the thinning rate at the fovea (–35.2 μm/mm) being faster than in the periphery (–22.3 μm/mm). This result aligns with the broad consensus that AL is a primary negative determinant of ChT in children10,27, 28, 29 and young adults,30 although a few studies found no clear association27,29 or only in limited areas.31 Our 3D map builds upon this consensus by providing critical details. This fovea-centered thinning pattern is also supported by other studies that have confirmed eccentricity-dependent changes in ChT.11,21,28,32 The potential mechanism for this concentric pattern may be related to mechanical stretching of the posterior pole caused by axial elongation. The fovea endures the greatest expansion stress, leading to the most severe mechanical thinning. This might also partially explain why some studies have specifically noted the nasal region as the most common site for myopic degeneration33 or the area of most significant thinning,11,34 as the nasal area is adjacent to the optic disc and fovea, which also bear significant tension. Our data contrast with 1 longitudinal study suggesting greater peripheral choroidal thinning,35 a discrepancy possibly due to different definitions of periphery. Clinically, the significant foveal thinning we observed is a structural signature of myopic ocular elongation. This strongly supports why thinning of the subfoveal ChT is considered a predictor for myopia progression, myopic maculopathy,25 and choroidal atrophy,36 as this region is structurally the most vulnerable and susceptible to pathological challenges. In our study, we also found thicker ChT in males than females, especially in the inferior-central region. The literature on sex-based ChT differences is similarly divided. Some studies found no sex difference,9,10,37,38 some reported thicker choroid in males,34,39 and 1 study observed significantly greater ChT in females by ultra-widefield-OCT testing.5 Our grid-based analysis provides spatial profiles of these differences. This sex-differential ChT is possibly related to modulation by sexual hormone, regional anatomical difference, and basal metabolic rate. Future research on exploring physiological factors with inferior sex-differential ChT distribution is warranted. The strengths of this study include its large population-based sample, objective segmentation using automated deep learning algorithms, and 3D grid-based topographic analysis, which is more refined than traditional single-point or ETDRS-grid measurements. However, this study also has limitations. First, it is cross-sectional. Although it reveals strong associations with age and AL, it cannot infer causality or true longitudinal change. Future longitudinal studies are needed to confirm the trajectory of these spatial patterns. Second, the population of this study was restricted in Bayanhushu, while national estimates of myopia prevalence were used to define myopia severity in this cohort, which might not reflect the true prevalence in this area. Third, we did not control potential confounders such as diurnal choroidal variation, hydration status, caffeine intake, or near work, all of which could affect transient ChT measurements. Next, we only investigated the absolute ChT rather than choroidal volume, which has been shown to remain unchanged despite choroidal thinning in highly myopic eyes.40 Additionally, body height and weight differed across myopia-severity groups after adjusting for age and sex. Both parameters were independently associated with ChT in multivariable models, consistent with previous literature.10 However, these associations and related structure underlying were not further examined in the current study and warrant further investigation. In conclusion, by applying 3D OCT mapping to a large cohort of healthy school-aged children, we revealed two major spatial patterns of ChT change during ocular growth: a temporal–nasal asymmetry with age and a concentric thinning pattern associated with axial elongation and myopia progression. These topographic signatures move ChT beyond a simple associative marker, positioning it as a dynamic, region-specific indicator of physiological maturation and myopia-related structural stress. Understanding these spatial dynamics provides a comprehensive reference for normal choroidal development and a foundation for identifying early pathological changes in pediatric myopia.