Longitudinal Changes in Choroidal Vascularity in Myopic and Non-Myopic Children

Abstract

Purpose: The purpose of this study was to evaluate longitudinal changes in choroidal vascular characteristics in childhood, and their relationship with eye growth and refractive error.

Methods: Analysis of high-resolution optical coherence tomography (OCT) scans, collected over an 18-month period as part of the Role of Outdoor Activity in Myopia (ROAM) study, was conducted in 101 children (41 myopic, 60 non-myopic, age 10–15 years). OCT images were automatically analyzed and binarized using a deep learning software tool. The output was then used to compute changes in macular choroidal vascularity index (CVI), choroidal luminal, and stromal thickness over 18-months. Associations of these variables with refractive error and axial length were analyzed.

Results: CVI decreased significantly, whereas luminal and stromal thickness increased significantly over 18 months (all P < 0.001). The magnitude of change was approximately double in stromal tissue compared to luminal tissue (luminal β = 2.6 µm/year; 95% confidence interval [CI] = −1.0 to 4.1 µm/year; stromal β = 5.2 µm/year; 95% CI = 4.0, 6.5 µm/year). A significant interaction between baseline axial length and change in CVI over time (P = 0.047) was observed, with a greater CVI reduction in those with shorter axial lengths. Significant associations were observed between the change in CVI, luminal thickness, stromal thickness, and change in axial length over time (all P < 0.05).

Conclusions: Faster axial eye growth was associated with smaller reductions in CVI, and less increase in choroidal luminal and stromal thickness. The changes in choroidal vascularity, particularly in the stromal component, may thus be a marker for eye growth.

Translational Relevance: This knowledge of the longitudinal changes in choroidal vascularity in childhood and their relationship with eye growth may assist clinicians in the future to better predict eye growth and myopia progression in childhood.

Introduction

The potential role of the choroid in regulating human eye growth and refractive error development has been examined in a number of recent studies. Cross-sectional studies have shown that a thinning of the choroid is associated with childhood myopia, and the magnitude of thinning associated with axial length, with longer eyes exhibiting a larger magnitude of thinning.^1–4 Longitudinal studies have shown that in myopic eyes or in eyes showing a myopic shift,^5–8 a thinning of the choroid, or less thickening of the choroid (compared to emmetropic eyes or eyes showing slower eye growth) is typically observed. The rate of eye growth has also been found to be negatively associated with choroidal thickness changes, with slower eye growth being associated with more thickening of the choroid over time in childhood.⁵ 

Whereas most previous studies of the in vivo human choroid have focused on measures of tissue thickness, advancements in optical coherence tomography (OCT) imaging technology and image analysis methods, including deep learning-based methods, have allowed to expand the quantification of OCT images providing more rigorous examination of the human choroid and its association with myopia. Additional metrics, such as the choroidal vascularity index (CVI),⁹ which is a quantitative measurement of the vascularity status of the choroid, can now be included as part of a more comprehensive analysis of the choroidal changes in myopia, complementing structural metrics such as thickness. 

To date, there have been several cross-sectional studies which have examined detailed choroidal vascular characteristics in childhood myopia with varying results.^10–13 Some studies observed significant associations between axial length and one or more choroidal metrics including: CVI, choroidal luminal area, and choroidal stromal area.^10,11 

Although there have been a few recent cross-sectional studies examining CVI in childhood, there is a particular lack of understanding of the longitudinal changes in CVI, and how these changes are related to eye growth in childhood. Observing how CVI changes over time in childhood will provide useful information for a better understanding of the pathophysiology and progression of myopia in children. In this study, we have examined the longitudinal changes in CVI (and other vascular metrics, including choroidal luminal and stromal thickness) across an 18-month period in myopic and non-myopic children and explored the associations of these changes with axial eye growth. 

Methods

This prospective observational study analyzed the longitudinal changes in CVI, choroidal luminal, and stromal thickness and axial length occurring over an 18-month period, in myopic and non-myopic children aged between 10 and 15 years. The data and high-resolution OCT images analyzed in this study were collected as part of the Role of Outdoor Activity in Myopia (ROAM) study. This was an observational, longitudinal study examining the normal eye growth and choroidal thickness of myopic and non-myopic children, and a detailed description of the data collection protocols and analyses of the choroidal thickness data from this study has been published previously.^2,5 Approval from the Queensland University of Technology human research ethics committee was obtained before commencement of this study, and written informed consent was provided by all participating children and their parents. All participants were treated in accordance with the tenets of the Declaration of Helsinki. 

Subjects

Participants were classified as myopic or non-myopic based upon the subjective, noncycloplegic spherical equivalent refractive error (SEQ) of their right eye. Myopic participants had an SEQ of −0.75 diopters (D) or less and non-myopic participants had an SEQ between +1.00 and −0.50 D. All participants had no known systemic comorbidities and had best corrected visual acuity of 0.00 logMAR or better in both eyes. One hundred one children were enrolled and had data captured at the baseline study visit (mean age of 13.1 ± 1.4 years, 52% girls, 60 non-myopes, and 41 myopes). Of the 101 participants, by the final study visit, 3 children were lost to follow-up and 4 had begun orthokeratology contact lens wear. Therefore, 94 children (97% of participants enrolled at baseline), including 59 non-myopes and 35 myopes completed all 4 study visits. The mean age of these 94 children was 13.0 ± 1.3 years, and 50% were girls. 

OCT Image Data Set

In the ROAM study, OCT imaging and measures of axial length were collected over 4 visits conducted every 6 months over an 18-month period. All measurements were collected between 2 and 5 PM to reduce the confounding effects of diurnal variations on choroidal thickness and axial length.¹⁴ At each visit, participants in this study had chorioretinal OCT scans of their right eye captured with the Heidelberg Spectralis spectral domain OCT device (Heidelberg Engineering, Heidelberg, Germany). High resolution images were captured using enhanced depth imaging (EDI), with a six-line, foveal centered star scan captured for each participant, and each OCT image represents the average of 30 individual scans captured from the same retinal location using the instrument's automatic real-time tracking and follow-up features (Fig. 1A). Four children (two myopes and two non-myopes) could not maintain stable fixation to have all 6 radial scans captured, so instead had a single horizontal scan (30 degrees and the average of 30 individual scans) captured and analyzed at each visit. Ocular biometry measures, including axial length, central corneal thickness, anterior chamber depth, and lens thickness were also assessed at each visit using the Lenstar LS 900 Optical biometer (Haag-Streit AG, Switzerland). 

Figure 1.

 

(A) Overview of the six-line star scan OCT scanning protocol. (B) The data derived from the OCT images is divided into meridians (temporal [T], superior-temporal [ST], superior [S], superior-nasal [SN], nasal [N], inferior-nasal [IN], inferior [I], and inferior-temporal [IT]) and regions (foveal, parafoveal, and perifoveal). (C) Original OCT image obtained using enhanced depth imaging (EDI) mode (vertical scan line). (D) OCT image illustrating segmentation of RPE (blue) and CSI (red). (E) Output image after binarization, where black pixels represent the vascular luminal area and white pixels represent the stromal area. The ROI between the two boundary lines has been binarized using the deep learning network.

Image Analysis

Following each study visit, the OCT images were exported from the instrument for further analysis, to derive choroidal vascularity data across the central 6 mm macular region, using custom written software. To ensure that analyses were performed across the same 6 mm region of all scans across all subjects in the study, the transverse magnification of the scans were adjusted to account for ocular magnification factors, based on the individual refractive error and ocular biometry measures of each subject.¹⁵ 

The CVI quantifies the ratio of the luminal area (the hypo-reflective regions in the choroid in the OCT scans) to the total choroidal area in a binarized OCT image. CVI measures for the subjects in this study were obtained using a recently developed deep learning software tool which has been trained and validated in a previous study¹⁶ and was utilized to binarize the choroidal region of each of the captured OCT images. The resulting binarized image output was then used to compute the CVI values. The methodology used for the binarization has been described in detail previously¹⁶ but a short summary will be provided here. Initially, the exported OCT images were segmented to delineate the outer surface of the retinal pigment epithelium (RPE) and the chorioscleral interface (CSI). This was done by an automated method¹⁷ and an experienced masked human observer² who checked the automated segmentation and manually corrected any errors (Fig. 1D). The region between the two segmentation lines (region from the posterior boundary of the RPE to the CSI) was then flattened using the RPE as a reference landmark, to form the region of interest (ROI) for binarization analysis. The ROI of each image was the input for a trained and fully automated deep learning network, which output a binarized image through a process known as two class classification. Within the output binarized image, the dark pixels represent vascular luminal areas, and the white pixels represent interstitial stromal areas (Fig. 1E). This fully automated deep learning process was trained using similar OCT images to that used in the current study which had been previously binarized by a trained observer using a local binarization method with custom window size determination and has been shown to have accuracy above 96%.¹⁶ 

The CVI was calculated for each A-scan across the central 6 mm within the ROI of each OCT image. The CVI data from all six scans were then used to calculate the mean CVI across the foveal region (central 1 mm), the parafoveal region (1 mm to 3 mm), and the perifoveal region (3 mm to 6 mm; Fig. 1B). Within each region, the mean CVI within 8 meridians (temporal, superior-temporal, superior, superior-nasal, nasal, inferior-nasal, inferior, and inferior-temporal) was also determined. 

Following this process, the choroidal thickness data (the axial distance between the RPE and the CSI) and CVI data from each scan were used to calculate the choroidal luminal thickness and stromal thickness. Luminal thickness was obtained by multiplying the CVI (range = 0 to 1) by the thickness measurement (luminal thickness = CVI * thickness), stromal thickness was obtained by multiplying difference of CVI from 1, by thickness (stromal thickness = (1-CVI) * thickness). 

Data Analysis

Using linear mixed model (LMM) analyses in IBM SPSS Statistics version 28 (Armonk, NY, USA), longitudinal changes in CVI over the 18-month study period were examined. This model analyzed the effect of visit time in months (from the initial visit) as a continuous variable on CVI, using a first order autoregressive covariance structure (which assumes a higher correlation for measurements taken closer together in time). Categorical predictor variables (refractive error group, gender, choroidal measurement region, and choroidal measurement meridian) were included in the model as fixed factors, and continuous predictor variables (age at baseline, baseline axial length, and change in axial length) were included as covariates. Analyses of refractive error group and axial length (baseline axial length and change in axial length) were conducted as separate LMMs, as these factors tend to be highly correlated. LMM analyses were repeated to analyze the longitudinal changes of choroidal luminal thickness and stromal thickness over the 18-month study period. 

Results

Choroidal Vascularity Index

The LMM analyses of the data across the 18-month study period revealed significant variations in CVI with both measurement region and meridian (both P < 0.001; Tables 1, 2). CVI was observed to be highest in the perifoveal region (mean = 63.7%, 95% confidence interval [CI] = 62.9 to 64.5%) and lowest in foveal region (mean = 61.5%, 95% CI = 60.7 to 62.3%). CVI was highest in the nasal meridian (mean = 63.6%, 95% CI = 62.8 to 64.4%) and lowest in temporal meridian (mean = 61.3%, 95% CI = 60.5 to 62.1%). Maps of the mean CVI values, illustrating the topographical variations in CVI at the baseline visit are shown in Figure 2. A significant effect of age at baseline was observed (β = −0.63%/year, 95% CI = −1.21 to −0.06%/year, P = 0.048), indicative of a small reduction in mean CVI with increasing age in this population of children. There were no significant main effects of sex or refractive group, and no significant interactions between these factors and visit time (all P > 0.05). The similarity of the topographical variations in CVI between refractive groups can be observed in Figures 2B and 2C, respectively. 

Table 1.

 

Mean CVI (%) and 95% Confidence Interval (CI) Across Regions

 

Table 2.

 

Mean CVI (%) and 95% Confidence Interval (CI) Across Meridians

 

Figure 2.

 

Map of mean CVI values at baseline for all subjects who had 6 radial OCT scans captured (n = 97) across the central 6 mm macular region (A), for the non-myopic children (n = 58) (B) and for the myopic children (n = 39) (C).

CVI was found to decrease significantly over time (P < 0.001, β = −0.66%/year, 95% CI = −0.81 to −0.50%/year). For all subjects considered together, CVI showed a mean decrease of 1.05 ± 0.12% over the 18-month study period (Fig. 3A). There was no significant interaction between visit time and region or meridian (both P >0.05), indicating a similar change in CVI with time across the macular area. Although a trend was observed for the change in CVI over time to be slightly less in the myopes (mean change of −0.84 ± 0.21%) compared to the non-myopes (mean change = −1.17 ± 0.15%; Fig. 3B), the differences between the groups did not reach significance (refractive group by visit time interaction P = 0.762). 

Figure 3.

 

Mean change in CVI over the 18-month study period for all subjects (A) and for the myopic and non-myopic participants considered separately (B). Error bars represent the standard error of the mean.

Further analyses were conducted to examine the association between the changes in CVI and the changes in axial length over time. These analyses revealed a significant interaction between the baseline axial length and the change in CVI over time (P = 0.047, β = 0.14%/ mm, 95% CI = 0.00006 to 0.28%/ mm) with a greater reduction in CVI seen in those with a shorter axial length (Fig. 4A). Mixed model analysis also revealed a significant interaction between the change in CVI and the change in axial length over time (P = 0.025, β = 1.2%/ mm, 95% CI = 0.29 to 2.12%/ mm; Fig. 4B). This shows that, in children with a higher rate of axial eye growth (i.e. greater myopia progression), there was less reduction in CVI over time and in some cases an increase in CVI. 

Figure 4.

 

Relationship between the rate of change in CVI (%) and baseline axial length (A) and between the rate of change in CVI (%) and rate of axial eye growth (B).

Choroidal Luminal and Stromal Thickness

Analysis of the choroidal luminal and stromal thickness data revealed significant main effects of measurement region and meridian (all P < 0.05). Luminal thickness was significantly lower in the perifoveal region (mean = 193.8 µm, 95% CI = 185.3 to 202.3 µm) compared to the fovea and parafovea, and the superior meridian showed the highest luminal thickness (mean = 219.2 µm, 95% CI = 210.7 to 227.7 µm), and the nasal meridian the lowest (mean = 179.2 µm, 95% CI = 170.7 to 187.7 µm). Similarly, stromal thickness was also lowest in the perifoveal region (mean = 112.8 µm, 95% CI = 106.1 to 119.5 µm) and highest in the foveal region (mean = 131.3 µm, 95% CI = 124.6 to 138.0 µm). The superior meridian exhibited the highest stromal thickness (mean = 129.8 µm, 95% CI = 123.1 to 136.6 µm) and the nasal meridian the lowest (mean = 107.1 µm, 95% CI = 100.4 to 113.8 µm; Tables 3, 4). 

Table 3.

 

Mean Luminal and Stromal Thickness and 95% Confidence Interval (CI) Across the Macular Regions

 

Table 4.

 

Mean Luminal and Stromal Thickness and 95% Confidence Interval (CI) Across Meridians

 

There were no significant effects of sex for either luminal or stromal thickness (both P > 0.05). A significant main effect of refractive group was observed for both luminal and stromal thickness (P < 0.05), with the myopic children exhibiting a significantly thinner luminal choroid (mean luminal thickness difference = 31.4 µm, 95% CI = 14.5 to 48.3 µm) and stromal choroid (mean stromal thickness difference = 15.3 µm, 95% CI = 2.2 to 28.4 µm) than the non-myopic children. 

A significant effect of age at baseline was observed for stromal thickness (P = 0.048, β = 4.8 µm/year, 95% CI = 0.04 to 9.6 µm/year), indicative of an increase in stromal thickness with increasing age. However, this relationship was not significant for luminal thickness (P > 0.05). For all subjects considered together, both luminal and stromal thickness were found to increase significantly over time (both P < 0.001), with the rate of change in stromal thickness observed to be approximately double that observed in the luminal thickness over the 18-month study period. Figure 5 illustrates the changes in luminal thickness (β = 2.6 µm/year, 95% CI = −1.0 to 4.1 µm/year) and stromal thickness (β = 5.2 µm/year, 95% CI = 4.0 to 6.5 µm/year) over time. Although a trend was observed for the non-myopes to show a greater increase in thickness than the myopes for both luminal and stromal thickness (Fig. 6), this difference did not reach statistical significance (refractive group by visit time interaction P > 0.05). 

Figure 5.

 

Mean change in luminal and stromal thickness in all subjects over the 18-month study period. Error bars represent the standard error of the mean.

Figure 6.

 

Change in mean luminal thickness (A) and stromal thickness (B) over time in the two refractive groups. Error bars represent the standard error of the mean.

A significant interaction between the baseline axial length and the mean luminal and stromal thickness was observed (luminal P = 0.03, β = −12.5 µm/ mm, 95% CI = −20.7 to −4.3 µm/ mm; stromal P = 0.023, β = −7.4 µm/ mm, 95% CI = −13.8 to −1 µm/ mm) with a thinner luminal and stromal choroid seen in those with a longer axial length (Fig. 7). A highly significant association was also found between the change in axial length over time and the changes in choroidal luminal thickness (P < 0.001, β = −17.8 µm/ mm, 95% CI = −26.5 to −9.1 µm/ mm) and with the changes in stromal thickness over time (P < 0.001, β = −16.0 µm/ mm, 95% CI = −22.7 to −9.2 µm/ mm; Fig. 8). This suggests that children with more rapid axial eye growth (i.e. greater myopia progression) exhibited less increase and, in some cases, a decrease in choroidal luminal and stromal thickness over time, albeit with some variability between participants given the relatively wide confidence intervals associated with these relationships. 

Figure 7.

 

Relationship between baseline axial length and mean choroidal stromal thickness (orange symbols) and mean choroidal luminal thickness (blue symbols).

Figure 8.

 

Relationship between the rate of change in axial length over time and the change in mean choroidal stromal thickness (A), change in mean choroidal luminal thickness (B), and change in mean overall choroidal thickness (C) over the course of the study.

Discussion

In this paper, we have explored the longitudinal changes in choroidal vascular characteristics, including the CVI, choroidal luminal thickness, and stromal thickness occurring over an 18-month period, in myopic and non-myopic children. The topographical variations in these parameters were also examined over the same time frame. Associations among changes in measurements, topographical variations, myopia, and axial eye growth were also explored. In this novel longitudinal analysis, significant changes in CVI, choroidal luminal, and stromal thickness occurred across 18 months, and significant interactions among these changes and baseline axial length and axial eye growth were observed. 

In previous reports from this same study population, significant changes in choroidal thickness were found that were also associated with the rate of eye growth over the same study period.⁵ Choroidal thickness was observed to be significantly thinner in myopic children, which is consistent with differences in choroidal luminal and stromal thickness between refractive groups in our current analyses. However, there was no difference in CVI between refractive groups. Read et al. (2015)⁵ also observed that choroidal thickness increased significantly over time with those with faster eye growth showing less increase or a decrease in thickness. This was similar to the observations in our analysis of choroidal luminal and stromal measures. 

Longitudinal Changes in CVI

This study provides the first longitudinal assessment of changes in CVI during childhood. In this population of healthy myopic and non-myopic children, CVI, was observed to decrease significantly over time, suggesting that normal eye growth over time is accompanied by a reduction in CVI. A significant interaction between the baseline axial length and the change in CVI over time and between the change in CVI and the change in axial length over time was found, with a lesser reduction in CVI seen in those with a longer axial length, and in those with more rapid eye growth. 

Only a small number of studies have examined topographical variations in CVI in children, with previous studies examining only a limited number of meridians. In children, Wu et al.¹¹ observed that in eyes with longer axial length, CVI in both vertical and horizontal meridians were significantly lower, with CVI in the horizontal meridian showing a larger difference between longer and shorter eyes. Similarly, Wang et al.,¹² observed that in young patients, CVI in the horizontal meridian differed more compared to the vertical meridian with increasing degree of myopia. Wang et al.,¹² further observed that CVI in the temporal and nasal quadrant in the central 1 mm and 3 mm regions showed a strong positive association to the degree of myopia. Our current study used a more detailed 6-line star scanning protocol, that revealed significant variations in CVI with both measurement region and meridian. However, there was no significant interaction between time and region or meridian, which suggests that CVI changes in a similar manner across the different macular regions over time. 

Longitudinal Changes in Luminal and Stromal Thickness

Choroidal luminal and stromal thicknesses were observed to increase significantly over time, with the rate of change in stromal thickness observed to be approximately double of that observed in the luminal thickness over the 18-month study period, suggesting that an increase in choroidal stromal thickness is a feature of normal eye growth in childhood. As the choroid increased in thickness with eye growth, an increase in stromal components may have resulted to provide additional structural support. These findings suggest that although both luminal and stromal thickness varied significantly with time, the changes in overall vascularity might have been driven more by changes in stromal than by luminal tissue. Our analyses also demonstrated a highly significant association between the change in axial length over time and the changes in choroidal luminal and stromal thickness. Children with faster axial eye growth exhibited less increase and, in some cases, a decrease in choroidal luminal and stromal thickness over time. Consistent with previous studies of choroidal thickness in childhood,¹⁵ substantial between subject variations in choroidal measures were found in this study, as evidenced by the relatively large confidence intervals associated with these parameters. 

Some previous cross-sectional studies have reported on the association between structural changes in the choroid and childhood myopia with varying results. Wu et al.¹¹ observed that, in anisomyopic children, eyes with longer axial length were associated with lower CVI and choroidal thickness, whereas Wang et al.¹² observed a negative linear regression between SEQ and CVI. Aşıkgarip et al.¹³ found no relationship between axial length and CVI but a strong relationship between SEQ and CVI (lower SEQ associated with lower CVI). In the current study, we observed that the CVI of the two refractive groups were not significantly different at baseline, although in Figure 4, it can also be observed that eyes with shorter baseline axial length tend to show a greater reduction in CVI over time. 

Our results suggest that progression of myopia is associated with progressive changes in multiple choroidal structures over time, possibly more prominently within the stromal region. Looking at the changes in the choroid in correlation to time-based factors, such as rate of eye growth might be more appropriate than with categorical factors (myopes versus non-myopes) especially because myopia is well understood to be a progressive condition. With reference to Figure 8 it can be observed that the correlation between change in mean stromal thickness is stronger than with overall choroidal thickness (R² values are 0.178 and 0.166, respectively). Thus, structural changes within the choroid when examined together may form a more accurate predictive matrix for eye growth. Stromal changes appear to be a stronger predictor for eye growth compared to changes in CVI (R² value = 0.079), luminal thickness (R² value = 0.116), or overall choroidal thickness. 

Childhood myopia has been linked to increased risk of sight-threatening retinal complications later in life.¹⁸ An accurate predictive matrix based on automated methods of image analysis can help in simultaneously increasing awareness of the potential high-risk patients without the increased requirement of trained manpower. This will allow clinicians to make better decisions on the extent of myopic control required in pediatric patients. The findings of this study offer new insight into the role of different components of the choroid in childhood myopic pathogenesis. This knowledge may also help researchers develop automated tools which can more accurately predict the relationship between the human choroid and eye growth. 

Evidence from past animal^19,20 and human studies^21–23 have established a relationship between thinner choroids with myopia and a weaker sclera. Troilo et al.²⁴ suggested that the choroid may control scleral growth by acting as a diffusion barrier for growth factors and thus thinner choroids are associated with more rapid eye growth. We hypothesize the slower increase and, in some cases, a decrease in the thickness of choroidal stromal tissue leads to faster eye growth due to less inhibition of movement of growth factors across the posterior ocular tissues. Further investigation will be useful in establishing the causative nature of this relationship. 

Strengths, Limitations, and Future Directions

Some of the strengths of the study and data collection methods include the low attrition of enrolled subjects, well designed OCT imaging protocol for all visits, and limitations of confounding factors, such as diurnal variation. The longitudinal analysis of choroidal vascularity variation is novel in childhood myopia. The use of a validated, automated algorithm to analyze EDI-OCT scans is the major strength of this study as this reduced observer bias and provided a more comprehensive understanding of choroidal components compared with other studies which used traditional manual image processing and analysis, that could potentially bias outcomes. It should be noted that CVI measures derived from this automated network do require the prior delineation of the two choroidal boundaries which are generated by automated segmentation with manual correction. There were several limitations to our study,² the sample size was relatively small with unmatched numbers of myopic and non-myopic children. This might have influenced some results, such as the relationship between baseline axial length and change in CVI over time and the effect of age at baseline on stromal thickness which were of borderline significance, which were of borderline significance. Further studies involving a larger cohort are required to confirm these observations. The time lapse between follow-up study visits were short and there was a lack of cycloplegic refraction measures. A noncycloplegic refraction may reduce the reliability of refraction measures in children,²⁵ and, for this reason, the majority of analyses were conducted between choroidal measures and axial eye growth which are less likely to be substantially influenced by cycloplegia.²⁶ We also note that the myopic participants in the study all had clinically established myopic refractive errors, and so the lack of cycloplegia is unlikely to have influenced their refractive group categorization. Future longitudinal studies on vascularity variation in myopic and non-myopic children should involve a larger number of well-matched study subjects with longer time between study visits and more reliable measurements of refractive error. Future studies can involve larger data sets using similar deep learning processes to provide further insight into how choroidal vascular structures change across a longer time frame. 

The exact interplay between structural choroidal changes and how they result in myopic progression requires future studies. To provide a better understanding of both changes in choroidal blood flow as well as structural changes, future studies can consider collecting both structural and functional scans, such as OCT-angiography. The rate of myopic progression within the myopic group in the ROAM study was lower than that of other similar studies. In one study of Chinese myopic children, the myopic progression was about double of those in the current myopic group.⁸ In future studies, larger cohorts with a higher number of faster progressing myopes may provide new insights. 

Conclusions

Choroidal characteristics in childhood myopia have been well-studied, although most research has focused on tissue thickness measures and used cross-sectional study designs. In this research, high resolution OCT images, coupled with validated deep learning-based image processing methods were used to provide an improved understanding of the vascular structure of the choroid in childhood myopia. The topographical distribution of choroidal vascularity was examined, as well as longitudinal changes over time. 

Significant decreases in CVI and increases in luminal and stromal thickness over the 18-month study period were observed. The magnitude of change over time was approximately double in stromal tissue thickness compared to luminal tissue thickness. This suggests that during normal eye growth in childhood, the documented increase in overall choroidal thickness⁵ is primarily driven by increases in choroidal stromal tissue, which leads to an associated decrease in CVI over time. Although normal eye growth was associated with a reduction in CVI over time, those exhibiting faster axial eye growth (i.e. more rapid myopia progression) showed a smaller reduction and sometimes an increase in CVI over time. 

Acknowledgments

The Role of Outdoor Activity in Myopia Study was funded by an Australian Research Council “Discovery Early Career Research Award” (DE120101434). 

Supported by the Alexandra Health Fund Ltd through the Science-Translational and Applied Research (STAR) grant, STAR191102. 

Disclosure: E. Ho, None; S.A. Read, None; D. Alonso-Caneiro, None; K. Neelam, None