|Home | About | Journals | Submit | Contact Us | Français|
To determine the dynamic morphological development of the human fovea in-vivo utilizing portable spectral domain optical coherence tomography (SDOCT).
Prospective, observational case series.
31 prematurely born neonates, nine children and nine adults.
Sixty-two neonates were enrolled in this study. SDOCT imaging was performed after examination for retinopathy of prematurity (ROP) at the bedside in non-sedated infants ages 31-41 weeks post-menstrual-age PMA (PMA=gestational age in weeks + chronological age) and at outpatient follow-up ophthalmic examinations. Thirty-one neonates met eligibility criteria. Nine children and nine adults without ocular pathology served as control groups. Semi-automatic retinal layer segmentation was performed. Central foveal thickness (CFT), foveal to parafoveal (FP) ratio (CFT divided by thickness 1000 μm from the foveal center), and 3D thickness maps were analyzed.
In-vivo determination of foveal morphology, layer segmentation, analysis of sub-cellular changes, spatio-temporal layer shifting.
In contrast to the adult fovea, we observed several signs of immaturity in the neonates: a shallow foveal pit, persistence of inner retinal layers (IRL), and a thin photoreceptor layer (PRL) that was thinnest at the foveal center. Three-dimensional mapping showed displacement of retinal layers out of the foveal center as the fovea matured and the progressive formation of the inner/outer segment band in the opposite direction. The FP-IRL ratios decreased as IRL migrated prior to term and minimally after that, while FP-PRL ratios increased as PRL subcellular elements formed closer to term and into childhood. A surprising finding was the presence of cystoid macular edema in 58% of premature neonates which appeared to affect inner foveal maturation.
This study provides the first view into development of living cellular layers of the human retina and of subcellular specialization at the fovea in premature infant eyes using portable spectral domain optical coherence tomography. Our work establishes a framework of the timeline of human foveal development, allowing us to identify unexpected retinal abnormalities that may provide new keys to disease activity, and provide a method for mapping of foveal structures from infancy to adulthood that may be integral in future studies of vision and visual cortex development.
The human fovea is responsible for high spatial resolution and for central and color vision.1 It plays an important role in human visual cortex development, regulating calcarine fissure symmetry2 and cortical maps.3 Abnormal development of the fovea has been related to a decrease in gray matter volume at the occipital cortex4 and might be related to disruptions in visual pathways at the level of the optic chiasm.5 The neurosensory retina may be divided into the photoreceptor layer (outer retina) and the inner processing and transmitting layers (inner retina) (Fig. 1). Morphologically, the adult fovea is marked by a pit that is dense with photoreceptors, and devoid of overlying inner retinal layers (IRL), and vasculature. This facilitates light transmission to the photoreceptors without interference.6
The fovea is immature at birth.7,8 Histologic studies, originating with those of Mann in 1964,9 have reported: 1) persistence of IRL at the foveal center and absence of the normal foveal curvature in the fetal eye and 2) changes in the width and length of photoreceptor substructures over time. The process of foveal maturation has been described as a centrifugal migration of the IRL and a centripetal migration of the cone cells.7,8,9 Numerous factors may influence foveal development: prematurity,10 development of vasculature,11 foveal tissue elasticity, intraocular pressure, retinal stretching.12 Some of these may act synergistically. Also, eye growth throughout the neonatal period13,14 may stretch malleable foveal tissue to form the foveal pit.15 Studies of vascular endothelial growth factor (VEGF) expressed in the area of the future fovea suggest that hypoxia is present prior to the thinning of the retinal layers, although this does not explain the process of foveal formation.16
Investigations of foveal pit development have been based on postmortem fixed tissue from non-human primates17,18 or from rare human specimens,7,8,9 with advantages of sub-micron resolution imaging and immunocytochemical labeling. However, serial sectioning is time-consuming, specimens are obtained at one timepoint, the same tissue cannot be processed for a flat mount and serial sectioning, retinal detachment is a common fixation artifact and more severe deformation of structures occurs in fixed sections from the youngest human eyes. Measurements rely on rates of relative tissue shrinkage during processing and may not represent the configuration in-vivo. All of these factors limit study of the three-dimensional growth of the human macula, especially photoreceptors. A reliance on postmortem data precludes inputting the anatomic stage of foveal development for an individual when assessing relationships between ophthalmic maturity, retinal signaling and posterior visual pathways.
Data on microanatomy of the living human fovea in premature infants in a neonatal intensive care unit (NICU) are limited to color photographs, descriptions, drawings,19,20 and intravenous fluorescein angiography.21 Because these lack cross-sectional information regarding cellular development, human studies could not address the interrelationship between changing foveal structures or between these and later visual function, which is of particular importance in the premature infant eye.
By introducing the study of high-resolution ocular images from the NICU, we propose to define, in premature infants in-vivo, a baseline frame of reference for development of the first component of the human visual system, the retina. We also begin to define variations and pathologies involved in this development.
In this study, we have utilized spectral domain optical coherence tomography (SDOCT), a rapid, noninvasive, and noncontact imaging technique which uses a broadband infrared wavelength light (840 nm) to detect backscattered light from the retina using interferometric techniques to provide cross-sectional images of the retina (Fig. 1). This high-resolution, in-vivo imaging modality has improved understanding of adult retinal diseases and is currently an important tool in the diagnosis and management of retinal pathology.22,23 Imaging with OCT is based on the intrinsic reflectance of a tissue or the interface between tissue layers. It is particularly useful for retinal evaluation due to contrast between alternating layers of lower reflective cell nuclei and higher reflective axons, dendrites and melanosomes.24 Cross-sectional images can be readily oriented within a three-dimensional stack of captured scans to define unique retinal microanatomy in-vivo.25 We developed methods to image young children and infants during examination under anesthesia26,27 using a hand-held portable SDOCT system (Bioptigen Inc., Research Triangle Park, North Carolina). We subsequently analyzed the optical properties of the premature infant eye to determine methods to allow SDOCT imaging of these infants without sedation.28 For this study, we examine retinas from very young premature infants to examine the dynamic shift in substructures during a critical time in the early development of retinal signaling and visual processing and compare the organization of these substructures to that of children and adults.
All adults and the parents or guardians of all infants and children consented to the participation in this Institutional Review Board approved observational study of SDOCT imaging. Because this research was performed on subjects in a NICU, research imaging could not be performed at predetermined timepoints. Instead, imaging was performed only when the NICU staff approved the status of the infant for research imaging and at times when the infant eyes were already dilated for the conventional examination. Follow-up ended if infants were transferred to another hospital or discharged, unless there was a follow-up visit on site with study imaging in clinic. SDOCT imaging was performed after ophthalmoscopic retinopathy of prematurity (ROP) screening at the bedside following the Maldonado et. al. methodology for optimizing SDOCT in neonates.28 and using a portable hand-held SDOCT unit (Bioptigen Inc., Research Triangle Park, North Carolina). Sixty-two premature infants were enrolled as part of a prospective study comparing SDOCT findings to those from conventional ophthalmoscopic examination by a pediatric ophthalmologist for ROP. Only subjects with SDOCT imaging sessions including the fovea before 42 weeks post-menstrual-age PMA (PMA = gestational age in weeks + chronological age) and an ophthalmic ROP exam up to 42 weeks PMA or equivalent information (parents of two subjects who had outside ophthalmic follow-up were contacted by phone to identify whether laser treatment or ROP progression occurred after being discharged) were included. Nine subjects were excluded: two were discharged before imaging occurred, two were examined for systemic conditions with ocular pathology (neonatal hemochromatosis and cytomegalovirus infection), three did not have ROP screening up to 42 weeks PMA to determine maximum level of disease and two did not have imaging prior to 42 weeks. From the remaining 53 neonates, 21 progressed to advanced ROP (excluded if clinical diagnosis identified any stage 3 ROP, type 1 ROP or if the infant received laser treatment) and were excluded from the study. One additional subject who met all other inclusion criteria was excluded because no study images contained the fovea. Thus there were 31 premature infants included in this study. For information on comparative foveal characteristics with increasing age, a single imaging session was obtained in nine full-term born children (4 infants ages 1, 2, 6 and 9 months old; and 5 children ages 2, 3, 6, 6 and 15 years old) and 9 adults.
Cystoid macular edema (CME) was a surprising finding in 18 of the 31 premature infants. The CME distorted retinal layers and affected retinal thickness interfering in analysis of retinal development. Thus we divided our analysis into a primary analysis of retinal development in infants without any CME (Group 1, n=13) and separately analyzed the findings of the infants with CME (Group 2, n=18). This division into Groups 1 and 2 was by subject and not by eye, as CME was bilateral in all subjects.
SDOCT images were converted to a DICOM format and qualitatively graded by four experienced graders in OSIRIX medical imaging software (OSIRIX Foundation, Geneva, Switzerland). Graders were masked to other clinical data. The best three SDOCT volumes containing the fovea were evaluated for presence of pathology, and the grader selected the scan with best resolution and best foveal centration for segmentation and quantitative analysis.
The foveal scan was segmented semi-automatically using a custom program, DOCTRAP (Duke OCT Retinal Analysis Program) v2.1-SF, based in MATLAB (Mathworks, Natick, MA, USA). The foveal pit was selected manually. MATLAB was used to compute thickness values from segmentation at each A-scan. Axial pixel size in retina was 3.26 um/pixel, as reported by Bioptigen Inc. Lateral pixel size was determined by dividing the scan length on retina by the number of A-scans. Scan length on retina for the neonatal eye was calculated based on the axial length for age.13,14 Scan length ranged from 6.8-13.6mm, and number of A-scans ranged from 512 to 1260 to maintain a consistent A-scan density. To validate our assignment of retinal layers in premature infants, we followed the layers through their evolution from infancy through childhood and compared them to published light micrographs of premature infant and children’s eyes at corresponding ages.7-9 For this, we reviewed the literature searching Medline for the terms “fovea, foveal development, ocular development, visual development, foveal pit formation”, from 1950 to the 2010 and based assignment of layers on correlation between OCT retinal layers in adult primate and human histology.23,24,25 SDOCT graders evaluated qualitatively the scans for the presence or absence of retinal pathology, and retinal layers at the foveal center. To determine the reproducibility of this grading an inter-reader agreement analysis was performed.
A random number generator was used to select an eye from the first imaging session for each subject that was used for primary analysis as reported of Table Table11 and and2.2. Central foveal thickness was defined as the thickness of the entire retina from the inner aspect of the inner limiting membrane (ILM) to the inner aspect of the retinal pigment epithelium (RPE) at the foveal center. The inner retinal layers (IRL) included all retinal tissue from the inner aspect of the ILM to the outer border of the inner nuclear layer (INL). The outer retinal layers (Table 1) extended from the inner aspect of the outer plexiform layer (OPL) to the inner border of the RPE. Retinal layers from top to bottom (inner to outer) are labeled in Figure 1. The PRL in this manuscript extended from the outer aspect of OPL to the inner border of RPE.
Three-dimensional thickness maps were created using a custom, automated graph theory-based segmentation software (DOCTRAP Pediatric v2.1) based on Chiu et al.29 Automated segmentation was successful at identifying the inner border of the ILM, inner border of the IS/OS, and the inner border of the RPE. Then, semi-automated segmentation was used to correct automated segmentation errors and to delineate the inner border of the OPL.
A ratio of the thickness at the foveal center versus thickness 1000 μm peripheral to the fovea was computed for IRL and PRL. Ratios were compared for all study groups using measurement for one eye on the first imaging session using the Wilcoxon rank sum test (Table 2).
ImageJ (U. S. National Institutes of Health, Bethesda) software was used to sum SDOCT b-scans to improve visualization of images presented for this publication (Fig. 1, Fig. 2a, and 2b top row). No post-processing was needed to segment or grade images.
For statistical analysis and to avoid intra-eye variation, only one imaging session per subject per age group was used in the statistical analysis. We determined statistical differences in the median foveal thickness measurements of premature neonates versus adults using a Wilcoxon rank-sum test and the statistical differences between all age groups using a Kruskal-Wallis test of difference among medians, where p<0.05 was considered statistically significant.
Retinal imaging was possible from the earliest eye examination at 31 weeks PMA until time of discharge from the nursery in all 31 premature infants in this study (age range at imaging 31-65 weeks PMA). Half were female, 61% African-American, 32% White, and <6% Hispanic-Latino. The median gestational age at birth was 26 weeks PMA (range 24 to 29 weeks). Median birth weight was 825 g (450-1700). A total of 106 imaging sessions were performed in the 31 neonates and there were 1 to 8 imaging sessions per infant between 31 through 42 weeks (median = 3, with 53% imaged three or more times). Five of the infants remained in the nursery or returned for follow up imaging as late as 65 weeks PMA. The older comparison groups had a single imaging session and consisted of nine children ages 2 weeks to 15 years (33% female; 56% African-American, 33% White, 11% Hispanic-Latino) and 9 adults (ages 20-50 years; 62% females; 46% African-American, 38% White, 16% Hispanic-Latino).
Inter-grader agreement for detection of foveal features was 92% among pairs of SDOCT certified graders. Highest agreement was found on the presence of INL and IS/OS at the fovea which was 98% and 95%, respectively, and the presence of many other common retinal pathologies, including cystoid macular edema (91%). Lowest agreement was found on the presence of GCL at the fovea (62%) and qualitative scoring of presence of pre-cystic thickening of the INL (59%).
In review of SDOCT scans, we found cystoid macular edema (CME) distorting the fovea in 18 of the 31 subjects (58%) at one or more imaging sessions. CME was bilateral in the majority of cases except in only 4 imaging sessions of 4 subjects. The CME was an unexpected finding in 58% of subjects and varied in severity from a single rounded hyporeflective lesion (Fig. 3a) to more extensive edema that thickened inner layers and disturbed the foveal contour (Fig 3b) or produced an upward bulge at the foveal center (13 of 18 eyes, Fig 3c). CME was detected as early as 31 weeks PMA and was more prevalent after 37 weeks PMA (Fig. 3d). Because this was not a long term study, we did not have follow up to identify whether or when there was CME resolution in all subjects. At 42 weeks PMA, we observed persisting CME in both eyes of two subjects who also had CME documented at some point between 32 and 42 weeks. In both eyes of 6 other subjects, we observed CME resolution at 40 to 65 weeks; these infants had previously documented CME on SDOCT imaging at some timepoint between 35 and 43 weeks PMA.
Because the presence of CME distorted retinal morphology and grossly changed thickness measurements, we grouped premature infants into Group 1 (n=13, no CME at any visit) and Group 2 (n=18, CME at any visit) based on the SDOCT images. Although the infants with edema (Group 2) had both, lower gestational age and weight at birth (Table 1), these differences were not statistically significant. Half of the subjects in Group 1 and 72% of subjects in Group 2 reached stage 2 ROP (by study design, all were Type 2 ROP and none reached any stage 3 disease or required laser).
The foveal center was readily identified in both eyes of all 13 neonates in Group 1 as the site of deepest central depression in the stack of serial SDOCT scans. Retinal layers and central foveal thickness at the central foveal scan in all premature infant eyes were grossly different from those of children and adults, and these differences were most pronounced at the earliest ages (Fig. 1). When compared to the adult eye, the premature eye had a visibly shallower foveal depression, presence of many IRL at the foveal center, thinner retinal layers overall, and attenuation of PRL with absence of photoreceptor sub-layers relative to the adult (Fig. 1) The median central foveal thickness in Group 1 was significantly less than the thickness in adults, and it increased progressively from infancy through adulthood (Tables (Tables11 and and2).2). Three-dimensional thickness mapping created from volumetric stacks of OCT scans in these subjects demonstrated the pronounced regional variation in thickness in total retina, IRL, and photoreceptor layer (PRL), and how both regions and layers develop distinctly over time (Fig. 2a and 2b).
Median IRL thickness at the foveal center in neonates (Group 1) was significantly greater than IRL thickness in adults (Table (Table11,,22 and Fig. 1). Horizontal neuronal elements such as nerve fiber layer and plexiform layers stood out clearly as hyperreflective layers in contrast to the hyporeflective nuclei, even at this early age. The most immature foveas (31-33 weeks) were characterized by presence of the ganglion cell layer (GCL), inner plexiform layer (IPL), and inner nuclear layer (INL) as distinct measurable layers at the foveal center (Fig. 1 a-c). While the IPL and INL persisted in all foveas from the neonatal group, these condensed into a single thin hyperreflective band in children and adults (Fig. 1 b,d).
The thickness and number of inner retinal layers at the foveal center decreased over time as the premature infant eye matured and the foveal pit deepened. In four neonates from Group 1 with multiple imaging sessions, thickness of the IRL at the foveal center progressively decreased from 31 to 45 weeks PMA (Fig. 4a, available at http://aaojournal.org). Median central foveal thickness and median IRL thickness were significantly greater in Group 2 than in Group 1 (Table 2). Seven of seven Group 2 subjects that had an abnormal INL thickening without visible cystoid structures developed CME on later sessions.
As central foveal IRL thickness decreased, parafoveal IRL thickness increased, resulting in a decrease in the foveal to parafoveal IRL ratio (FP-IRL thickness ratio) with increasing age (Fig. 5a, available at http://aaojournal.org). The median FP-IRL thickness ratio in Group 1 neonates was significantly greater than that in adults (Tables (Tables11 and and22).
3D maps highlight regional ontogenetic changes in the retina over time. Both the FP-IRL ratios and the 3D maps of the IRL from subjects at 31 weeks PMA through 23 years of age are consistent with centrifugal cell migration of the IRL to form the foveal pit. These included data from the same eye of one premature infant from 34 to 43 weeks PMA (Figs. 2a and 2b). With increasing age there is a progressive increase in height of a parafoveal annulus of the retina in contrast to central and more peripheral retina, and the IRL is the major contributor to this increase (Fig. 2a: TL and IL) in contrast to the PRL which increases to a lesser height and over a longer period (Fig. 2a). The majority of the IRL migration occurred between 31-42 weeks with minimal refinement following. The IRL were also thicker closer to the optic nerve, consistent with nerve fiber layer (NFL) distribution.30
Although the majority of Group 1 followed a similar pattern of maturation, there was individual variation. For example, although the GCL was absent on most examinations from 31-42 weeks, there were 5 different subjects where GCL was present at either 33, 35 or 37 weeks PMA. IPL and INL were present in 12 of 13 Group 1.
Edema appeared to be associated with a possible delay in maturation of inner retinal layers, but did not appear to impact photoreceptor development. At 31-42 weeks PMA, the GCL persisted at the fovea in 15 of 18 (83%) of Group 2 where IPL and INL were present in all foveas from 31-42 weeks. FP- IRL thickness ratio was significantly higher in Group 2 than in Group 1 (Figure 5c, available at http://aaojournal.org and Table 2) and, this was consistent with and without cystoid structures present which might suggest thicker and likely less mature IRL at the foveal center.
The in-vivo PRL was notably thin across the infant retina with the thinnest portion located at the foveal center with a median of 29 microns (range 13-62) (Fig. 1). Median PRL thickness at the foveal center in Group 1 was significantly thinner than the in the adult group (Tables (Tables11 and and2).2). PRL at the foveal center increased in thickness progressively from infancy to adulthood (Table 1, Fig. 5b, available at http://aaojournal.org), and PRL thickness was significantly different across age groups (Group 1, Infants, Children, Adults; p< 0.001, Kruskal-Wallis test). Median PRL thickness in Group 2 was similar to PRL thickness in Group 1 (Tables (Tables11 and and2).2). Repeat imaging of four neonates without macular edema revealed progressive increase in PRL thickness at the foveal center with increasing age at almost every interval (Fig. 4c, available at http://aaojournal.org). Nevertheless, PRL had a more rapid increase in height after 38 weeks (Figs. (Figs.2b2b and 4c, available at http://aaojournal.org); and unlike the IRL, PRL axial growth was pronounced after birth in all regions and particularly in the cone-dense fovea (Fig. 2b). The PRL was thinnest at the foveal center in Group 1 which was the opposite of the central bulging PRL configuration in the adult (Fig. 1). This resulted in a significantly lower median FP-PRL thickness ratio for Group 1 versus the adults (Tables (Tables11 and and2,2, Fig. 5d, available at http://aaojournal.org). FP-PRL ratio was not different between groups despite a greater range of ratios for Group 2. Thus although there may be a delay in centrifugal migration of the inner retina in premature infant eyes with macular edema, we did not identify a definite delay in maturation of photoreceptor morphology when assessed at this early period of their life.
Photoreceptor subcellular structures were absent at the foveal center on SDOCT imaging in the premature group. These structures include external limiting membrane (ELM), inner segment to outer segment (IS/OS) junction and photoreceptor outer segments (POS) (Figure 1). The time course of development of these structures appeared to vary across infants. For instance, both, the ELM, and the IS/OS junction were absent in all eyes from 31 to 42 weeks PMA, and in both eyes of a term 2 month-old infant (~46 weeks PMA) but the IS/OS junction was visible at the foveal center in three of five premature infants between 43-48 weeks PMA and in one full-term infant at 45 weeks demonstrating variability in layer development between subjects. Also, another full-term infant, did not show IS/OS at 46 weeks PMA, but did show this band at 4 months of age (~54 weeks PMA).
The IS/OS band, first appeared as a poorly defined reflective band barely elevated from the RPE outside the fovea at 33 weeks PMA. The ring of the IS/OS and POS formation slowly increased in thickness as the outer segments appeared both peripheral and central to this zone, reaching the foveal center by 43 weeks (Fig. 2a). The distance from the foveal center to the edge of visible IS/OS band was comparable at any point when measured nasal versus temporal, and superior versus inferior; this was corroborated by the circular zone of OS absence on the 3D maps (Fig. 2a). From 37 to 42 weeks, the foveal center to IS/OS border mean distance was approximately 1380 microns (range 1320-1890) for Group 1, while in Group 2 this mean was 1654 (range 1030-2556) microns which was not significantly different from Group 1.
The main RPE reflex, a prominent well-defined hyperreflective band, was present in all subjects from 31 weeks PMA through adulthood (Fig. 1). A second subtle hyperreflective RPE band between the IS/OS junction and the main RPE reflex is believed to be the interface between photoreceptor outer segments and retinal pigment epithelium microvilli (OS/RPE).31 (Fig. 1). The OS/RPE band appeared to differentiate at the apical side of the RPE layer late in childhood, consistent with growth of apical microvilli of the RPE or a change in the interface with photoreceptor outer segments. It was not visible in any infants or children under age 10, and was visible with variable contrast only in mid-teens and adults.
Our results show that human foveal development continues after premature birth. This is the first report documenting the development of the human fovea in-vivo (PubMed Mesh search terms: optical coherence tomography, foveal development, prematurity, retinopathy of prematurity) without postmortem artifacts, illustrating the distinct differences in development between inner retinal layers and the photoreceptors, and showing the evolution of cellular structures in three dimensions over time. In contrast to previous studies, of non-human primate samples18 and postmortem human foveal specimens7,8,9, we could monitor retinal maturation in the same individual over time and could document the variation in the timing of foveal development across individuals (Figs 2a, 2b and 4, available at http://aaojournal.org).
Semi-automated segmentation of retinal layers was essential for three-dimensional mapping of the developing retina (Fig. 2). Utilizing this in-vivo mapping and quantifying the changing relationship between the fovea and peripheral retina (Fig. 5, available at http://aaojournal.org), we displayed the spatio-temporal shift in the cellular elements and structures during early development. The segmentation allowed quantitative characterization of foveal maturation from prematurity to adulthood with a reproducible imaging method (92% agreement between certified SDOCT graders).
Neuronal cell generation and differentiation are known to be complete at the fovea prior 31 weeks PMA as demonstrated in human retinal whole-mounts by mitosis cessation,32 thus cell migration and maturation are the primary events in the period of our imaging. A dynamic evaluation of cellular migration in-vivo was achieved by measuring the foveal to parafoveal ratio for the IRL and PRL thicknesses. IRL cell migration was in the opposite direction of PRL cell migration and elongation, with a relative time lag in the PRL growth (Fig. 5b, available at http://aaojournal.org). Timing of development of the PRL in premature infants was not always linked to the timing of development of IRL; therefore, some infants demonstrated highly developed IRL with delayed maturation of the PRL and others demonstrated the opposite. Imaging these stages of individual foveal microanatomy is likely to be useful for in-vivo studies of visual development, since the progress of retinal development in each infant may vary and may affect signaling and upstream visual pathways.33,34
Although the change in morphology of SDOCT retinal layers with age and in FP-IRL with age in the non-edematous eyes was generally consistent with timepoints of cellular redistribution reported on histologic studies,7,8,9,35 in-vivo analysis allowed a better sampling of the events of foveal development over time and across numerous individuals. However, none of the published postmortem specimens of infants less than one year of age were from infants who had survived for more than two days after birth, and this may explain why none of those studies describe macular edema during this period of development. Although we had longitudinal information, one weakness in our study was the inability to complete individual follow-up in all subjects because of frequent transfer to another hospital.
All retinal layers reported in histologic studies were observed on SDOCT except for the transient acellular layer of Chievitz.36 GCL and INL thickness reported on Table 1 appeared comparable to the 2-3 cells thickness of GCL and INL reported by Hendrickson et al at 34-36 weeks PMA.7 The short photoreceptor cells and absence of formed outer segments reported on histologic studies at 22-36 weeks, matches well with our finding of absent IS/OS band at 31-33 weeks PMA and gradual increase in height of both the photoreceptor nuclear layer and the outer segments during prematurity. In at least one case in-vivo foveal development varied from published histology: in the postmortem specimen, IRL were present at the foveal center and a there was a shallow pit in a 5 day old term infant, while in our study, many premature infants at 40-42 weeks already had a well-formed foveal pit with pronounced thinning or absence of IRL.
Although histology studies7,8,9 report that foveal development appeared complete by age 15 months, we found evidence of progressive development after that timepoint. The additional band that is thought to be the interface between outer segments and RPE was not visible in any young children while it was visible in a child at 15 years of age and in adults. To identify the exact timeline of development of this reflectivity, we will need more extensive study in young children and young adults. We theorize that the timing of development of this subtle interface is visible with in-vivo SDOCT imaging but not with light microscopy of sectioned tissue due to common photoreceptor-RPE disruption or separation with fixation of ocular tissue.
An unexpected finding in premature infants was the presence of CME (58% of premature subjects) at one or more timepoints from 32 to 43 weeks PMA. While visible on SDOCT images, CME was not detected on the clinical examination in our NICU. Macular thickening, edema or schisis have not been found previously in studies of neonatal ocular development, or in clinical trials of retinopathy of prematurity, except in a few cases of ROP with tractional retinal detachment extending at or near the macula37,38,39 and a report of cystoid maculopathy in three infants from a review of ocular histopathology of the infant for a period of 15 years.40 In the latter report, cystoid structures involved outer retinal layers in an infant who died at 11 days of life, the inner nuclear layer in one stillborn infant and splitting of NFL in another. Both stillborn infants had serious central nervous system abnormalities. The lack of previous reports of CME may be due to the previous lack of technology capable of capturing this information in earlier studies, and to the lack of human eye histopathology from subjects who lived several weeks after birth in a NICU. Although the CME bore some similarities to retinoschisis from traction which has been reported in ROP,30,41 the cystoid spaces were located only at the fovea and in the INL, with a rare cystoid structure in the PRL (Fig 3c) and were not associated with any traction. Although infants in this study were very premature, they did not develop any advanced ROP (see methods). We have imaged similar CME in the maculas of a term infant with liver failure and ascites, and documented resolution of the cystoid structures after liver transplant.42
Macular edema in premature infant eyes might reflect variability in remodeling of foveal architecture during development; however, we propose that this might relate to elevated VEGF levels in the eye analagous to adult CME that is responsive to anti-VEGF therapy. Although no macular edema was found in foveal development histology studies,7,8,9 increased VEGF mRNA expression had been found in the early developing fovea,16 and elevated intraocular VEGF levels have been identified in eyes undergoing surgery for advanced ROP.43,44 More importantly, VEGF elevation was found in the second phase of ROP progression from 32-34 weeks PMA.45 Thus it should not be surprising to find macular edema in premature infant eyes with areas of peripheral avascular retina, even in without advanced ROP. The high frequency of the CME among premature infants with non-advanced ROP points to the importance of understanding the cause of the edema, its predictive value, and whether this could impact visual development or acuity later in life. Inner retinal thickening in premature infants with CME might be the first evidence of lifelong thickening of inner retinal layers at the fovea as reported in older children and adults with a history of ROP.10,46,47 However, studies of children with a history of premature birth have shown normal results on extensive visual function and MRI testing48 or even advanced chromatic contrast sensitivity when compared to term birth subjects.49 Experts have disagreed on the presence of and extent of any difference in visual pathways after premature birth.50 Many studies were performed on surviving older populations to obtain imaging or visual function testing10,36,37,47. Considerable discussion has centered on isolating the contributions of visual experience and of ophthalmic disorders distinct from neuromotor or other impairments on adult visual function47-51 Retinal signaling delivered to the cortex may affect development of visual pathway and cortical mapping,3 and SDOCT imaging of the fovea would likely provide useful anatomic information on this actively changing site of sensory input and initial visual processing.
In conclusion, migration, redistribution and growth of subcellular structures occur in premature infants ex-utero. The development of inner versus outer retinal layers at the fovea may progress independently in some eyes. This dynamic process of foveal maturation and specialization can be studied in-vivo in premature infants with SDOCT imaging in non-sedated infants. Many premature infants have CME, an unanticipated finding that could be an ocular biomarker linked to additional systemic or ocular disease activities but which may have little impact on foveal development. An unanswered question is whether the morphology on SDOCT could be a useful predictor of future visual acuity. In order to link structure with function, future longer term studies of macular imaging of premature infants would be useful in conjunction age-appropriate testing of visual function.
Michelle McCall provided critical feedback and management for this study.
Financial Support: This research was made possible by the following grants: Angelica and Euan Baird; The Hartwell Foundation; Grant Number 1UL1 RR024128-01 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research. It’s contents are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH.
Conflict of Interest: Proprietary or commercial disclosure may be found after the references.
Financial disclosure(s): Proprietary or commercial disclosure may be found after the references.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.