Vitreous opacities in infants born full-term and preterm by handheld swept-source optical coherence tomography

Abstract

Purpose

To compare vitreous opacity density in infants born at term and in infants born prematurely using an investigational handheld swept-source optical coherence tomography (SS-OCT).

Methods

Infants born at term underwent imaging once between 12 and 48 hours after birth; infants born prematurely were imaged at each routine retinopathy of prematurity (ROP) examination. Three masked, trained graders analyzed images. Semiautomated methods were used to quantify vitreous opacity density, which was correlated with ROP severity based on indirect ophthalmoscopy, other SS-OCT findings, and medical comorbidities.

Results

Between April 2018 and June 2019, 251 SS-OCT imaging sessions were performed on 78 infants (49% female; 36% preterm, with mean birth weight of 1018 ± 338 g and gestational age of 28.6 ± 3.2 weeks). All SS-OCT sessions produced images of adequate quality. Punctate vitreous opacities were present in 25 of 28 term infants (89%) and 41 of 50 premature infants (82%). Dice coefficient and F1 scores for intergrader agreement were 0.99 ± 0.03 and 0.77 ± 0.31, respectively. Vitreous opacity density was 0.118 ± 0.187 in prematurely born infants and 0.031 ± 0.118 in infants born at term (P = 0.009). In the former, vitreous opacity density was associated with ROP zone (P = 0.044) and stage (P = 0.031), intraventricular hemorrhage (P = 0.028), subchorionic hemorrhage (P = 0.026), and African American race (P = 0.023). In the latter, vitreous opacity density was associated with maternal diabetes (P = 0.049).

Conclusions

Our investigational handheld SS-OCT achieved high-quality vitreoretinal images. In our study cohort, punctate vitreous opacities were a frequent finding in infants born at term and those born prematurely, with increased density in those born prematurely, particularly those with severe ROP.

Worldwide, approximately 2.6 million newborns born prematurely are at risk of developing retinopathy of prematurity (ROP), a leading causes of preventable childhood blindness, with 32,200 infants newly blind each year.^1–2 ROP is characterized by aberrant growth of retinal vasculature leading to fibrosis and occasionally retinal detachment.³ The current gold standard screening and diagnostic modality for ROP involves binocular indirect ophthalmoscopy with scleral indentation. Widefield retinal photography is also well established.⁴ Retinal findings in ROP have been well defined through these screening criteria,⁵ but vitreous findings have received less attention.

Our group previously characterized vitreous findings by handheld spectral domain optical coherence tomography (SD-OCT) in the preterm infant population, identifying both vitreous bands and punctate hyperreflective vitreous opacities in infants at risk for ROP.⁶ The term vitreous opacity ratio was coined to describe the density of punctate hyperreflective vitreous opacities; in the current study, we refer to this value as “vitreous opacity density.” The vitreous opacity density and presence of tractional vitreous bands were both previously found to be associated with ROP severity, but without a comparison to control infants born at term.⁷ Finn and colleagues⁸ identified punctate hyperreflective vitreous opacities using portable SD-OCT in 15 of 15 eyes from healthy infants born at term, suggesting that vitreous opacities can be a normal finding in infant vitreous.

Handheld SD-OCT imaging is technically challenging in awake infants. The handheld swept source OCT (SS-OCT) offers several advantages, including faster scanning speeds (200 kHz), a ranging distance (diameter of retina viewed on a single scan) of 12 mm, a pupil-finding feature, and an on-screen OCT display, which increase vitreous image quality while maintaining retinal and choroidal image clarity.^9–10 This handheld SS-OCT has been shown to identify vitreous findings in awake infants born prematurely with higher sensitivity than handheld SD-OCT.¹⁰ To our knowledge, vitreous characteristics in infants, as determined by handheld SS-OCT, have not been well described. Using this newer device, this study quantifies and compares vitreous opacity density in infants born at term and those born prematurely.

Subjects and Methods

This is a prospective, observational cohort study involving infants undergoing imaging using an investigational handheld SS-OCT.^9–10 This study adhered to the tenets of the Declaration of Helsinki, and Approval was obtained from the University of Washington and Seattle Children’s Hospital institutional review boards to recruit newborns, both those born at term and those born prematurely, from the Maternal Infant Center and Neonatal Intensive Care Unit, respectively. Participating infants’ legal guardians provided informed consent. Demographic information was obtained from the patients’ medical record.

Newborns born at term were eligible if gestational age was >37 weeks, age was at least 12 hours at the time of imaging, the patient’s family could communicate in English, and there were no social barriers, such as drug addiction, guardianship issues, or behavioral concerns to interfere with the provision of other social services. Infants deemed too medically unstable for OCT imaging or examination were excluded. Newborns born at term underwent handheld SS-OCT imaging between January 2019 and June 2019.

Infants born prematurely were eligible if they met criteria for routine ROP screening (≤30 weeks gestation, ≤1500 g birth weight, or ≤2000 g birth weight with cardiorespiratory instability), and the patient’s family could communicate in English. Infants deemed too medically unstable by the supervising neonatologist for OCT imaging or examination were excluded. Enrolled infants underwent handheld SS-OCT imaging between April 2018 and May 2019. Dates of imaging differed between infants born at term and those born prematurely due to logistics of working around unrelated clinical study enrollment occurring in the Maternal Infant Center. Infants born prematurely at 30–37 weeks’ gestational age and not requiring routine ROP screening were excluded based on institutional review board assessment due to potential risk to a vulnerable premature population for unnecessary dilation and examination.

ROP Diagnosis

Pupillary dilation was achieved using Cyclomydril (phenylephrine hydrochloride 1% and cyclopentolate 0.2%; Alcon Laboratories, Fort Worth, TX) in infants. One of two pediatric ophthalmologists screened qualifying infants born prematurely using binocular indirect ophthalmoscopy, assisted by an eyelid speculum and scleral depression. Examiners were masked to the OCT findings. Decisions to perform follow-up examinations and/or treatment were informed by standard ROP guidelines independent of OCT findings.⁵ Newborns born at term did not undergo a binocular indirect ophthalmoscopic examination.

Imaging

Newborns born at term underwent imaging of both eyes once between 12–48 hours after birth, prior to discharge from the hospital. Infants born prematurely were scanned bilaterally the same day as routine screening ROP examinations. Following pupillary dilation, participants undergoing handheld SS-OCT imaging were placed in the supine position, unsedated. Sometimes a sucrose solution or pacifier were used to sooth infants. A trained imager then used the handheld SS-OCT scanner to obtain between 50–500 B-scans per volume (entire three-dimensional scan consisting of multiple cross-sectional B-scans), with the goal to acquire high quality foveal and optic nerve images, as previously described.^9–10 Scans in infants born prematurely were obtained either before or after the binocular indirect ophthalmoscopic examination with scleral depression, separated by at least 30 minutes.

Qualitative Grading of SS-OCT Vitreoretinal Findings

Two trained graders evaluated OCT volumes using ezDICOM software (ezDICOM medical viewer, Wolfgang Krug and Chris Rorden). Graders subjectively evaluated the best quality volumes for the following vitreoretinal findings: punctate hyperreflective vitreous opacities, vitreous bands, dome-shaped macula, cystoid macular edema (CME), epiretinal membrane (ERM), and subretinal fluid. Other volumes were also evaluated if findings were absent on the best-quality volume but present on a different high-quality volume. Punctate hyperreflective vitreous opacities were defined as at least one hyperreflective opacity within the vitreous area above the background signal.⁶ Vitreous bands were defined as coalescing linear opacities visible in at least three consecutive frames with intensity above the background signal. Vitreous bands were divided into tractional and nontractional bands, as previously defined.⁶ Dome-shaped macula was defined as a convex configuration of the chorioretinal macula,¹¹ excluding images where retina was not visible on both sides of the fovea. CME was defined as hyporeflective macular intraretinal spaces.¹² ERM was defined as a hyperreflective layer over the inner retina with intervening hyporeflective space and excluding the area overlying retinal vessels.¹³ Subretinal fluid was defined as a hyporeflective collection between the neurosensory retina and the retinal pigment epithelium. Disagreement between the two graders was mediated by a third, trained grader, masked to prior grading.

Quantification of Punctate Hyperreflective Vitreous Opacities

Semiautomated methods, previously published, were used to quantify vitreous opacity density.⁷ Briefly, among high-quality macular volumes noted to contain punctate hyperreflective vitreous opacities, foveal and parafoveal vitreous (3–5 evenly spaced B-scans, roughly 256 μm in diameter) was analyzed using the RadiAnt DICOM viewer (Medixant, Poznań, Poland), excluding vitreous overlying the optic nerve. For each B-scan, segmentation of the vitreoretinal border and identification of punctate hyperreflective vitreous opacities (Figure 1) were performed by two graders using a custom, open-sourced web browser–based application (https://github.com/uw-biomedical-ml/segmentations). An independent trained grader identified errors prompting repeat grading by original graders. Averaging the two graders, vitreous opacity density was calculated as follows: number of opacities ÷ vitreous area (in pixels) × 10,000.

FIG 1.

Representative examples of B-scans from infants born at term (A) and prematurely (B) showing punctate hyperreflective vitreous opacities (circled) and tracing of the vitreoretinal border (highlighted). The vitreous opacity density was calculated in a semiautomated fashion, using the following equation: vitreous opacity density = number of vitreous opacities ÷ vitreous area (in pixels) × 10,000.

Statistical Analysis

Intergrader agreement of qualitative analysis was assessed using the Cohen’s kappa (κ) statistic. For quantitative analysis, an F1 score and Dice coefficient determined intergrader agreement for counting opacities and vitreous segmentation, respectively.¹⁴ The mean vitreous opacity density of infants born at term was compared with the mean for infants born prematurely. Vitreous opacity density was also correlated with ROP presence, stage, and zone, as well as plus disease status, type 1 ROP,⁵ demographics, relevant comorbidities (anemia, intraventricular hemorrhage and subchorionic hemorrhage, which may suggest hemorrhage risk factors, and maternal diabetes, which may result in increased birth trauma due to macrosomia), and the aforementioned OCT findings. The χ² test and Fisher exact test were used for infant-based comparisons. A generalized linear mixed model was used for eye-based comparisons to account for multiple eye examinations and two eyes sampled per infant. Given that intraventricular hemorrhage and maternal subchorionic hemorrhage are both known risk factors for ROP,¹⁵ we performed a subanalysis controlling for ROP stage using a linear mixed model to independently assess the relationship between vitreous opacity density and these comorbidities. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC) and IBM SPSS Statistics for Windows, version 26.0 (IBM Corp, Armonk, NY). A P value of <0.05 was considered statistically significant.

Results

A total of 78 infants (36% born prematurely) underwent 251 eye imaging sessions (152 sessions in newborns born prematurely). All sessions yielded images of adequate quality to determine presence or absence of punctate hyperreflective vitreous opacities. Demographic and clinical characteristics are provided in Table 1. Based on clinical ROP screening examinations, 80 of 152 sessions (53%) resulted in a diagnosis of ROP. Twenty-two sessions (14%) showed stage 3 ROP, and fifteen (10%) showed pre-plus disease, while no eyes demonstrated plus disease.

Table 1.

Demographic and clinical characteristics

Patient characteristica	Preterm n = 28	Full-term n = 50
Birth weight, g, mean ± SD	1018 ± 338	3439 ± 491
Gestational age, weeks, mean ± SD	28.6 ± 3.2	39.3 ± 1.2
Sex
Male	16 (57)	24 (48)
Female	12 (43)	26 (52)
Ethnicity
White	15 (54)	30 (60)
African American	2 (7)	4 (8)
Hispanic	5 (18)	2 (4)
Asian	4 (14)	7 (14)
Declined	1 (4)	5 (10)
Multiple ethnicities	1 (4)	2 (4)
Clinical ROP statusb
ROP stage
0	68 (45)	N/A
1	29 (19)	N/A
2	29 (19)	N/A
3	22 (14)	N/A
ROP zone
I	8 (5)	N/A
II	82 (54)	N/A
III	58 (38)	N/A
Plus disease status
Plus	0 (0)	N/A
Pre-plus	15 (10)	N/A
No plus	137 (90)	N/A
SS-OCT findingsc
Punctate hyperreflective vitreous opacities	25 (89)	41 (82)
Tractional vitreous bands	18 (64)	5 (10)
Nontractional vitreous bands	6 (21)	1 (2)
CME	8 (29)	4 (8)
ERM	7 (25)	4 (8)
Dome-shaped macula	6 (21)	3 (6)
Subretinal fluid	0 (0)	15 (30)
Patient comorbidities
Anemia of prematurity	22 (79)	N/A
Intraventricular hemorrhage	5 (18)	N/A
Maternal comorbidities
Diabetes mellitus	3 (11)	8 (16)
Subchorionic hemorrhage	1 (4)	2 (4)
Anemia	N/A	1 (2)
Mode of Deliveryd
Spontaneous vaginal	N/A	9 (18)
Cesarean section	N/A	24 (48)
Nonspontaneous vaginal	N/A	16 (32)
Vacuum-assisted vaginal	N/A	1 (2)

CME, cystoid macular edema; ERM, epiretinal membrane;N/A, not applicable; SS-OCT, swept-source optical coherence tomography.

^aResults are no. (%), unless otherwise stated.

^bAs determined by indirect ophthalmoscopy; maximum disease shown.

^cSeen at least once among all imaging sessions.

^dBecause infants were imaged at least one month after birth, delivery mode data was not felt to be relevant for the preterm infant population so data is not reported.

SS-OCT Findings

Intergrader agreements for qualitative OCT image analysis of all vitreoretinal findings among preterm and full-term populations yielded a κ score of 0.630 (95% CI, 0.568–0.691) and 0.588 (95% CI, 0.505–0.671),¹⁴ respectively. OCT findings are shown in Table 2. Punctate hyperreflective vitreous opacities were seen in 116 of 152 imaging sessions (76%) from 25 of 28 infants (89%) born prematurely and 70 of 99 imaging sessions (70%) from 41 of 50 infants (82%) born at term (P = 0.120).

Table 2.

Associations between vitreous opacity density and demographic, clinical, and optical coherence tomography characteristics in preterm and full-term infants

Patient characteristic	Preterm vitreous opacity density, mean ± SD	P valuea	Full-term vitreous opacity density, mean ± SD	P valuea
Total VOR	0.118 ± 0.19		0.031 ± 0.12	0.009b
Demographics
Sex
Male	0.12 ± 0.20	0.457	0.01 ± 0.03	0.213
Female	0.12 ± 0.16		0.05 ± 0.17
Ethnicity
White	0.06 ± 0.11		0.01 ± 0.03
African American	0.29 ± 0.28	0.023	0.05 ± 0.10	0.149
Hispanic	0.07 ± 0.12		<0.01 ± 0.00
Asian	0.14 ± 0.18		0.13 ± 0.30
Gestational age		0.306		0.709
Birth weight		0.407		0.105
Clinical ROP statusc
ROP stage
0	0.08 ± 0.14		N/A
1	0.10 ± 0.18	0.031	N/A
2	0.23 ± 0.27		N/A
3	0.19 ± 0.17		N/A
ROP zone
I	0.40 ± 0.40		N/A
II	0.10 ± 0.16	0.044	N/A
III	0.11 ± 0.15		N/A
Plus disease status
No plus disease	0.12 ± 0.19	0.818	N/A
Pre-plus disease	0.11 ± 0.15		N/A
SS-OCT findings
CME
Yes	0.08 ± 0.18	0.487	<0.01 ± 0.00	0.564
No	0.13 ± 0.19		0.03 ± 0.12
ERM
Yes	0.06 ± 0.10	0.159	<0.01 ± 0.00	0.696
No	0.13 ± 0.19		0.03 ± 0.12
Dome-shaped macula
Yes	0.05 ± 0.09		0.05 ± 0.08
No	0.13 ± 0.20	0.198	0.03 ± 0.12	0.704
Subretinal fluid
Yes	N/A		0.01 ± 0.02	0.368
No	N/A		0.03 ± 0.12
Vitreous bands
Yes	0.20 ± 0.20	0.066	0.07 ± 0.10	0.666
No	0.09 ± 0.17		0.03 ± 0.12
Patient comorbidities
Anemia of prematurity	0.13 ± 0.19	0.323	N/A
Intraventricular hemorrhage	0.24 ± 0.25	0.028	N/A
Maternal comorbidities
Diabetes mellitus	0.04 ± 0.08	0.407	0.10 ± 0.28	0.049
Subchorionic hemorrhage	0.34 ± 0.24	0.026	0.01 ± 0.03	0.821
Anemia	N/A		<0.01 ± 0.00	0.774
Mode of deliveryd
Spontaneous vaginal	N/A		0.01 ± 0.02
Cesarean section	N/A		0.05 ± 0.17	0.868
Nonspontaneous vaginal	N/A		0.01 ± 0.03
Vacuum-assisted vaginal	N/A		<0.01 ± 0.00

CME, cystoid macular edema; DSM, dome-shaped macula; ERM, epiretinal membrane; ROP, retinopathy of prematurity; SD, standard deviation; SS-OCT, swept-source optical coherence tomography; VOR, vitreous opacity ratio (= vitreous opacity density).

^aGeneralized linear mixed model was used to account for nonindependence of multiple visits and eyes per infant; eye findings from each visit were compared to the vitreous opacity density from that visit.

^bComparing preterm and full-term infant populations.

^cAs determined by indirect ophthalmoscopy.

^dBecause infants were imaged at least 1 month after birth, delivery mode data was not analyzed in the preterm infant population.

Semiautomated analysis of vitreous opacity density and its correlation with ocular and systemic findings are shown in Table 2. Of 186 eye imaging sessions with known punctate vitreous opacities, 96% had adequate macular image quality for quantitative analysis. Dice coefficient and F1 scores for intergrader agreement were 0.99 ± 0.03 and 0.77 ± 0.31, respectively. Mean vitreous opacity density was greater for infants born prematurely (0.118 ± 0.187) compared with those born at term (0.031 ± 0.118; P = 0.009; 95% CI, 0.042–0.131). In infants born prematurely, vitreous opacity density was higher in eyes with more severe ROP stage (P = 0.031), more posterior ROP zone (P = 0.044), but not pre-plus disease (P = 0.818; there were no cases of plus disease). In neither population of infants was the vitreous opacity density correlated with other OCT findings (Table 2).

In infants born prematurely, higher vitreous opacity density was associated with intraventricular hemorrhage, maternal subchorionic hemorrhage, and African American race (P = 0.028, 0.026, and 0.023, resp.). In infants born at term, maternal diabetes was associated with higher vitreous opacity density (P = 0.049). Other demographic characteristics and comorbidities did not correlate with vitreous opacity density (Table 2).

In a subanalysis among infants born prematurely, controlling for ROP stage, the associations between vitreous opacity density and comorbidities were preserved. Eyes of infants with intraventricular hemorrhage had a mean vitreous opacity density of 0.208 ± 0.047 (95% CI, 0.108–0.305) compared with 0.082 ± 0.024 (95% CI, 0.034–0.131) in those without intraventricular hemorrhage (P = 0.028). Eyes of infants whose mothers had subchorionic hemorrhage had a mean vitreous opacity density of 0.345 ± 0.102 (95% CI, 0.134–0.556) compared with 0.095 ± 0.022 (95% CI, 0.050–0.141) in those without maternal subchorionic hemorrhage (P = 0.026).

We explored further the observed association between vitreous opacity density and maternal diabetes to identify a pathophysiologic mechanism to explain the association. We analyzed the relationship between vitreous opacity density and birth weight of infants born at term, including those with macrosomia, but found no such association (P = 0.105). Similarly, we did not detect an association between vitreous opacity density and vaginal mode of delivery in infants born at term (P = 0.868; Table 2). Because infants born prematurely are low birth weight and not imaged until at least 1 month after birth, we did not perform these analyses for this group.

Discussion

Of 251 eye imaging sessions using handheld SS-OCT, punctate hyperreflective vitreous opacities were visualized in 82% and 89% of infants born at term and prematurely, respectively. This high proportion in both populations agrees with prior studies.^6–8 Thus, punctate opacities may represent features of normal newborn vitreous, which is important to consider when evaluating infants using OCT. Nonetheless, this study identified higher vitreous opacity density in infants born prematurely compared to those born at term (P = 0.009), particularly those with greater ROP severity (P = 0.031), confirming prior work using handheld SD-OCT.⁷

The origins of punctate hyperreflective vitreous opacities are likely multifactorial. The vitreous is a complex gel composed primarily of water, with structure provided by collagen II fibrils, glycoproteins, proteoglycans, and hyaluronic acid molecules.¹⁶ Although largely acellular tissue in adults, during embryonic development the primary vitreous is filled with hyaloid vasculature that supports early anterior chamber and lens development. During the fourth month of normal embryogenesis, this vasculature regresses to produce optically clear vitreous. Nonetheless, hyaloid remnants likely continue after birth, with 95% of newborns born at term demonstrating Cloquet canal by handheld SD-OCT.¹⁷ Hyaloid remnants or associated hemorrhage may therefore result in newborn punctate hyperreflective vitreous opacities seen by OCT. In newborns born prematurely, the added density of vitreous opacities could be attributed to increased hyaloid remnants with associated hemorrhage, as well as protein, inflammatory cells, blood products, and/or astrocyte precursors shed from the ROP ridge.⁷ Sugioka and colleagues¹⁸ evaluated vitreous samples in patients with ROP and found higher levels of several different proteins compared to vitreous from infants without ROP. Scleral depression itself may also contribute to vitreous hemorrhage through local trauma combined with friable retinal neovascularization. Performing OCT imaging prior to scleral depression may minimize that effect. Legocki and colleagues⁷ showed that the density of vitreous opacities increased after panretinal photocoagulation for type 1 ROP. Our study confirmed previously described associations between vitreous opacity density and intraventricular hemorrhage,⁷ with new associations identified with maternal subchorionic hemorrhage and maternal diabetes. These conditions are associated with increased fetal stress, inflammation, and bleeding,^19–21 supporting both inflammatory and hemorrhagic etiologies. Histopathologic samples of post-mortem or post-enucleation eyes of patients with ROP could further elucidate the nature of the vitreous opacities, though vitreous loss during tissue processing may limit such analysis.

Prior studies identified SD-OCT-based ROP-severity biomarkers, including retinal vascular features of plus disease (VASO score),²² vitreous opacity density, and tractional vitreous bands.⁷ Altogether, these biomarkers may be combined into predictive models for disease. Standard ROP screening examinations, including binocular indirect ophthalmoscopy and widefield retinal photography, are challenged by subjectivity, need for high-level expertise, and repeated stress to infants.^23–25 In contrast, handheld OCT does not involve ocular contact and is well tolerated, with potential for more objective structural analysis and automation. In particular, the handheld SS-OCT prototype used for this study is fast, user friendly in awake infants, and offers high-quality vitreous, retina, and choroid images.¹⁰ This study demonstrates that this device is able to successfully detect ROP severity biomarkers, including vitreous opacity density, albeit with a limited sample of severe ROP. Therefore, this device may be feasible for future clinical application as an ROP screening tool, which requires further exploration.

There are important limitations to this study. Infants born prematurely were examined at the time of ROP screening examinations at least 1 month after birth, whereas infants born at term were examined immediately after birth, possibly leading to vitreoretinal differences related to proximity to delivery. It may have been helpful to include a comparison group of infants born prematurely but that did not qualify for ROP screening (ie, born at 30–37 weeks’ gestation), although this group is challenging to recruit. Another limitation was the small sample size for certain subgroups, notably infants of African American race and subchorionic hemorrhage, which deserve further study in larger populations. The small sample size limited subanalyses of possible confounders in this study; however we previously demonstrated that ROP stage was an independent risk factor for vitreous opacity density in infants born prematurely, regardless of postmenstrual age at the time of imaging.⁷ Finally, this study was limited by somewhat low intergrader agreements (κ = 0.63 for preterms and 0.588 for full-terms) for qualitative OCT findings, though this was mitigated by having a third, experienced, grader serve as a tie-breaker for disagreements and did not affect the primary conclusions of the study (intergrader agreements for vitreous opacity density were high). Study strengths include the use of a novel technology and inclusion of a control group of healthy newborns born at term.

In conclusion, handheld swept-source OCT shows high usability in awake infants. Punctate hyperreflective vitreous opacities were frequently seen in both newborns born at term and those born prematurely, but vitreous opacity density was significantly higher in infants born prematurely with more severe ROP. Vitreous opacity density could be explored as an ROP severity biomarker.